{"id":"https://openalex.org/W4408029418","doi":"https://doi.org/10.1145/3710848.3710878","title":"Mario: Near Zero-cost Activation Checkpointing in Pipeline Parallelism","display_name":"Mario: Near Zero-cost Activation Checkpointing in Pipeline Parallelism","publication_year":2025,"publication_date":"2025-02-28","ids":{"openalex":"https://openalex.org/W4408029418","doi":"https://doi.org/10.1145/3710848.3710878"},"language":"en","primary_location":{"id":"doi:10.1145/3710848.3710878","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3710848.3710878","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3710848.3710878","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 30th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3710848.3710878","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5061661957","display_name":"Weijian Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Weijian Liu","raw_affiliation_strings":["SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5046708261","display_name":"Mingzhen Li","orcid":"https://orcid.org/0000-0002-4115-9072"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Mingzhen Li","raw_affiliation_strings":["SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I4210165038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006750799","display_name":"Guangming Tan","orcid":"https://orcid.org/0000-0002-6361-5948"},"institutions":[{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]},{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guangming Tan","raw_affiliation_strings":["SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I4210165038"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101648065","display_name":"Weile Jia","orcid":"https://orcid.org/0000-0001-8539-8326"},"institutions":[{"id":"https://openalex.org/I4210165038","display_name":"University of Chinese Academy of Sciences","ror":"https://ror.org/05qbk4x57","country_code":"CN","type":"education","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210165038"]},{"id":"https://openalex.org/I4210090176","display_name":"Institute of Computing Technology","ror":"https://ror.org/0090r4d87","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weile Jia","raw_affiliation_strings":["SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China"],"affiliations":[{"raw_affiliation_string":"SKLP, Institute of Computing Technology, CAS, Beijing, China and University of Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I4210165038"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5061661957"],"corresponding_institution_ids":["https://openalex.org/I4210090176","https://openalex.org/I4210165038"],"apc_list":null,"apc_paid":null,"fwci":2.5207,"has_fulltext":false,"cited_by_count":2,"citation_normalized_percentile":{"value":0.87133966,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"197","last_page":"211"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10917","display_name":"Smart Grid Security and Resilience","score":0.9980000257492065,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10917","display_name":"Smart Grid Security and Resilience","score":0.9980000257492065,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems 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/T11598","display_name":"Internet Traffic Analysis and Secure E-voting","score":0.9972000122070312,"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/T10772","display_name":"Distributed systems and fault tolerance","score":0.9968000054359436,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/pipeline","display_name":"Pipeline (software)","score":0.7755143642425537},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7395133376121521},{"id":"https://openalex.org/keywords/zero","display_name":"Zero (linguistics)","score":0.7190381288528442},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.7094301581382751},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.7027566432952881},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.32481616735458374}],"concepts":[{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.7755143642425537},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7395133376121521},{"id":"https://openalex.org/C2780813799","wikidata":"https://www.wikidata.org/wiki/Q3274237","display_name":"Zero (linguistics)","level":2,"score":0.7190381288528442},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.7094301581382751},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.7027566432952881},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.32481616735458374},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3710848.3710878","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3710848.3710878","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3710848.3710878","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 30th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3710848.3710878","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3710848.3710878","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3710848.3710878","source":null,"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 30th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4408029418.pdf"},"referenced_works_count":30,"referenced_works":["https://openalex.org/W1493774699","https://openalex.org/W2734941459","https://openalex.org/W2883283076","https://openalex.org/W2914304175","https://openalex.org/W2969388332","https://openalex.org/W2972087877","https://openalex.org/W3012479151","https://openalex.org/W3012514909","https://openalex.org/W3086105743","https://openalex.org/W3120480394","https://openalex.org/W3132107458","https://openalex.org/W3157864729","https://openalex.org/W3167436278","https://openalex.org/W3204998121","https://openalex.org/W3205803342","https://openalex.org/W3206832494","https://openalex.org/W4220967350","https://openalex.org/W4282974849","https://openalex.org/W4310285541","https://openalex.org/W4384705429","https://openalex.org/W4386348101","https://openalex.org/W4386709668","https://openalex.org/W4386763939","https://openalex.org/W4386768656","https://openalex.org/W4393145114","https://openalex.org/W4393406906","https://openalex.org/W4394923298","https://openalex.org/W4395117922","https://openalex.org/W4396712969","https://openalex.org/W4404400583"],"related_works":["https://openalex.org/W2045183646","https://openalex.org/W2162409446","https://openalex.org/W2109463584","https://openalex.org/W2504075107","https://openalex.org/W3037187668","https://openalex.org/W2354058185","https://openalex.org/W3014500535","https://openalex.org/W2033862586","https://openalex.org/W1595672120","https://openalex.org/W4230999561"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,73,186],"have":[3],"to":[4,10,41,66,150],"be":[5,94],"trained":[6],"in":[7,29,97,104],"parallel":[8],"due":[9],"their":[11],"large":[12],"number":[13],"of":[14,167,173],"parameters":[15],"and":[16,43,79,85,138,175,184,202],"significant":[17],"memory":[18,45,81,114],"footprint.":[19],"Among":[20],"various":[21],"parallelism":[22,25],"techniques,":[23],"pipeline":[24,38,59,68,127,148,161,176,195,217],"is":[26],"widely":[27],"adopted":[28],"inter-node":[30],"scenarios":[31],"with":[32,77,135],"minimal":[33],"communication":[34],"overhead.":[35],"However,":[36],"state-of-the-art":[37,194],"schemes":[39,196],"lead":[40],"extra":[42],"imbalanced":[44],"footprints,":[46],"leaving":[47],"room":[48,131],"for":[49,120,132,164,169,214],"further":[50,133],"improvement.":[51],"In":[52],"this":[53],"paper,":[54],"we":[55,144,157],"propose":[56],"Mario,":[57,165],"a":[58,146,211],"optimizer":[60],"that":[61,188],"automatically":[62],"tessellates":[63],"activation":[64,91],"checkpointing":[65,174],"existing":[67,193],"schemes,":[69],"enabling":[70],"training":[71],"larger":[72],"(or":[74],"longer":[75],"sequences)":[76],"less":[78],"balanced":[80],"footprint":[82,115],"across":[83],"GPUs":[84],"improved":[86],"GPU":[87],"utilization.":[88],"First,":[89],"the":[90,98,105,126,141],"recomputation":[92],"can":[93,190],"effectively":[95],"overlapped":[96],"bubbles":[99],"by":[100,204],"moving":[101],"it":[102],"earlier":[103],"execution":[106,152],"process,":[107],"thereby":[108],"improving":[109],"overall":[110],"efficiency.":[111],"With":[112],"eliminated":[113],"through":[116],"checkpointing,":[117],"Mario":[118,189],"allows":[119],"preposing":[121],"more":[122,130],"forward":[123],"computation":[124],"into":[125],"bubbles,":[128],"making":[129],"overlapping":[134],"greater":[136],"flexibility,":[137],"thus":[139],"exploiting":[140],"bubbles.":[142],"Then":[143],"design":[145],"lightweight":[147],"simulator":[149],"model":[151],"behavior":[153],"w/o|w/":[154],"Mario.":[155],"Finally,":[156],"introduce":[158],"an":[159],"automatic":[160],"scheduler":[162],"specifically":[163],"capable":[166],"searching":[168],"near":[170],"optimal":[171],"combination":[172],"configurations":[177],"within":[178],"minutes.":[179],"Experimental":[180],"results":[181],"on":[182,206],"GPT3":[183],"LLaMA2":[185],"show":[187],"speed":[191],"up":[192],"(w/o|w/":[197],"checkpointing)":[198],"including":[199],"1F1B,":[200],"Chimera,":[201],"Interleave":[203],"1.16\u00d7|1.57\u00d7":[205],"average.":[207],"This":[208],"work":[209],"paves":[210],"new":[212],"direction":[213],"effective":[215],"low-cost":[216],"training.":[218]},"counts_by_year":[{"year":2025,"cited_by_count":2}],"updated_date":"2025-12-28T23:10:05.387466","created_date":"2025-10-10T00:00:00"}