{"id":"https://openalex.org/W7172425507","doi":"https://doi.org/10.1109/2575-8411.2026.00068","title":"BCD-Megatron: A Cost-Effective Training System for Large Language Models","display_name":"BCD-Megatron: A Cost-Effective Training System for Large Language Models","publication_year":2026,"publication_date":"2026-06-01","ids":{"openalex":"https://openalex.org/W7172425507","doi":"https://doi.org/10.1109/2575-8411.2026.00068"},"language":"en","primary_location":{"id":"doi:10.1109/2575-8411.2026.00068","doi":"https://doi.org/10.1109/2575-8411.2026.00068","is_oa":true,"landing_page_url":"https://doi.org/10.1109/2575-8411.2026.00068","pdf_url":null,"source":{"id":"https://openalex.org/S4363608224","display_name":"IEEE International Conference on Distributed Computing Systems (ICDCS)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE 46th International Conference on Distributed Computing Systems (ICDCS)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.1109/2575-8411.2026.00068","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5144213301","display_name":"Zeyu Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I135714990","display_name":"North University of China","ror":"https://ror.org/047bp1713","country_code":"CN","type":"education","lineage":["https://openalex.org/I135714990"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeyu Liu","raw_affiliation_strings":["North University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"North University of China","institution_ids":["https://openalex.org/I135714990"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144337739","display_name":"Yan Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan Li","raw_affiliation_strings":["Lenovo Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Lenovo Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144313508","display_name":"Yunquan Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I135714990","display_name":"North University of China","ror":"https://ror.org/047bp1713","country_code":"CN","type":"education","lineage":["https://openalex.org/I135714990"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yunquan Zhang","raw_affiliation_strings":["North University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"North University of China","institution_ids":["https://openalex.org/I135714990"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129737523","display_name":"Boyang Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"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":"Boyang Zhang","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5157309114","display_name":"Guoyong Jiang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guoyong Jiang","raw_affiliation_strings":["Xidian University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xidian University","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144384320","display_name":"Xin Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"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":"Xin Zhang","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144359175","display_name":"Limin Xiao","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Limin Xiao","raw_affiliation_strings":["Lenovo Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Lenovo Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128101247","display_name":"Weifeng Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weifeng Zhang","raw_affiliation_strings":["Lenovo Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Lenovo Research","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5010944123","display_name":"Daning Cheng","orcid":"https://orcid.org/0000-0002-9217-6980","observed_orcids":["https://orcid.org/0000-0002-9217-6980"]},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"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":"Daning Cheng","raw_affiliation_strings":["Institute of Computing Technology, Chinese Academy of Sciences"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Computing Technology, Chinese Academy of Sciences","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210090176"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.49084603,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"653","last_page":"663"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.22110000252723694,"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":0.22110000252723694,"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":0.1454000025987625,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.1340000033378601,"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/large-language-models","display_name":"large language models","score":0.800000011920929},{"id":"https://openalex.org/keywords/training-system","display_name":"training system","score":0.41999998688697815},{"id":"https://openalex.org/keywords/cost-effective-training","display_name":"cost-effective training","score":0.7070000171661377},{"id":"https://openalex.org/keywords/distributed-training","display_name":"distributed training","score":0.6809999942779541}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6588000059127808},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.47940000891685486},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4453999996185303},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.414900004863739},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.36629998683929443},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3133000135421753},{"id":"https://openalex.org/C2776857766","wikidata":"https://www.wikidata.org/wiki/Q7832987","display_name":"Training system","level":2,"score":0.2757999897003174},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.24979999661445618},{"id":"https://openalex.org/C129792486","wikidata":"https://www.wikidata.org/wiki/Q1050419","display_name":"Language identification","level":3,"score":0.2442999929189682},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.23819999396800995}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/2575-8411.2026.00068","doi":"https://doi.org/10.1109/2575-8411.2026.00068","is_oa":true,"landing_page_url":"https://doi.org/10.1109/2575-8411.2026.00068","pdf_url":null,"source":{"id":"https://openalex.org/S4363608224","display_name":"IEEE International Conference on Distributed Computing Systems (ICDCS)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE 46th International Conference on Distributed Computing Systems (ICDCS)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1109/2575-8411.2026.00068","doi":"https://doi.org/10.1109/2575-8411.2026.00068","is_oa":true,"landing_page_url":"https://doi.org/10.1109/2575-8411.2026.00068","pdf_url":null,"source":{"id":"https://openalex.org/S4363608224","display_name":"IEEE International Conference on Distributed Computing Systems (ICDCS)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2026 IEEE 46th International Conference on Distributed Computing Systems (ICDCS)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":16,"referenced_works":["https://openalex.org/W1540764732","https://openalex.org/W2000769684","https://openalex.org/W2023901033","https://openalex.org/W2072925128","https://openalex.org/W2117686388","https://openalex.org/W2950798207","https://openalex.org/W3034587794","https://openalex.org/W3081168214","https://openalex.org/W3173195499","https://openalex.org/W3187255235","https://openalex.org/W3205803342","https://openalex.org/W4220741164","https://openalex.org/W4321636575","https://openalex.org/W4327694855","https://openalex.org/W4386768656","https://openalex.org/W4402671897"],"related_works":["https://openalex.org/W4415798025","https://openalex.org/W7197005704","https://openalex.org/W6930787056","https://openalex.org/W7167063453","https://openalex.org/W7169765031","https://openalex.org/W7172423201","https://openalex.org/W7196928916","https://openalex.org/W4415798907","https://openalex.org/W7148327510","https://openalex.org/W7169774661"],"abstract_inverted_index":{"Training":[0],"large":[1],"language":[2],"models":[3],"typically":[4],"demands":[5],"substantial":[6,125],"GPU":[7,87,162],"memory":[8,88],"and":[9,22,32,61,71,135,151,155],"significant":[10],"financial":[11],"investment,":[12],"posing":[13],"a":[14,29,38,44,74,106],"major":[15],"barrier":[16,146],"for":[17],"many":[18],"small-to-medium":[19],"research":[20],"groups":[21],"companies.":[23],"In":[24],"this":[25],"paper,":[26],"we":[27],"present":[28],"full-parameter":[30,158],"pretraining":[31],"fine-tuning":[33],"framework,":[34],"BCD-Megatron,":[35],"built":[36],"on":[37,53,118],"block-wise":[39],"update":[40],"strategy,":[41],"together":[42],"with":[43],"set":[45],"of":[46,76,105,111,157],"engineering":[47],"optimizations,":[48],"enabling":[49],"efficient":[50],"model":[51,67,108,149],"training":[52,80,103,117,133,159],"both":[54],"lowcost":[55],"consumer":[56],"GPUs":[57],"(e.g.,":[58],"RTX":[59],"4090)":[60],"high-performance":[62],"computing":[63],"clusters.":[64,163],"By":[65],"partitioning":[66],"parameters":[68,77],"into":[69],"blocks":[70],"updating":[72],"only":[73],"subset":[75],"in":[78],"each":[79],"iteration,":[81],"our":[82,99,128],"approach":[83,129],"markedly":[84],"reduces":[85,101],"the":[86,95,102,124,144,153],"requirement.":[89],"Experimental":[90],"results":[91],"show":[92],"that,":[93],"under":[94],"same":[96],"hardware":[97,145],"configuration,":[98],"method":[100],"cost":[104,126],"7B-parameter":[107],"to":[109,137,147],"33%":[110],"that":[112],"required":[113],"by":[114],"conventional":[115],"fullparameter":[116],"high-end":[119],"clusters":[120],"(A100/A800).":[121],"Moreover,":[122],"despite":[123],"reduction,":[127],"achieves":[130],"comparable":[131],"wall-clock":[132],"time":[134],"accuracy":[136],"traditional":[138],"methods.":[139],"This":[140],"progress":[141],"significantly":[142],"lowers":[143],"large-scale":[148],"development":[150],"demonstrates":[152],"practicality":[154],"cost-effectiveness":[156],"using":[160],"low-cost":[161]},"counts_by_year":[],"updated_date":"2026-11-03T10:50:19.383937","created_date":"2026-08-05T00:00:00"}