{"id":"https://openalex.org/W4388032098","doi":"https://doi.org/10.48550/arxiv.2310.18313","title":"FP8-LM: Training FP8 Large Language Models","display_name":"FP8-LM: Training FP8 Large Language Models","publication_year":2023,"publication_date":"2023-10-27","ids":{"openalex":"https://openalex.org/W4388032098","doi":"https://doi.org/10.48550/arxiv.2310.18313"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2310.18313","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.18313","pdf_url":"https://arxiv.org/pdf/2310.18313","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2310.18313","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5080495327","display_name":"Houwen Peng","orcid":"https://orcid.org/0000-0001-8544-8952","observed_orcids":["https://orcid.org/0000-0001-8544-8952"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng, Houwen","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100368801","display_name":"Kan Wu","orcid":"https://orcid.org/0000-0002-8306-5066","observed_orcids":["https://orcid.org/0000-0002-8306-5066"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Kan","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009248953","display_name":"Yixuan Wei","orcid":"https://orcid.org/0000-0003-1775-7301","observed_orcids":["https://orcid.org/0000-0003-1775-7301"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Yixuan","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079805574","display_name":"Guoshuai Zhao","orcid":"https://orcid.org/0000-0003-4392-8450","observed_orcids":["https://orcid.org/0000-0003-4392-8450"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Guoshuai","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016817843","display_name":"Yuxiang Yang","orcid":"https://orcid.org/0000-0002-1837-3628","observed_orcids":["https://orcid.org/0000-0002-1837-3628","https://orcid.org/0009-0001-6187-3097"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Yuxiang","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100349457","display_name":"Ze Liu","orcid":"https://orcid.org/0000-0002-5121-7547","observed_orcids":["https://orcid.org/0000-0002-5121-7547"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Ze","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102245054","display_name":"Yifan Xiong","orcid":"https://orcid.org/0000-0001-9056-1386","observed_orcids":["https://orcid.org/0000-0001-9056-1386","https://orcid.org/0009-0001-9575-0247","https://orcid.org/0000-0002-1921-7424"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Yifan","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029661948","display_name":"Ziyue Yang","orcid":"https://orcid.org/0000-0002-1658-0260","observed_orcids":["https://orcid.org/0000-0002-1658-0260","https://orcid.org/0009-0004-6196-9293"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Ziyue","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012017422","display_name":"Bolin Ni","orcid":"https://orcid.org/0009-0000-7160-5523","observed_orcids":["https://orcid.org/0009-0000-7160-5523"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ni, Bolin","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109632338","display_name":"Jingcheng Hu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Jingcheng","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019706917","display_name":"Ruihang Li","orcid":"https://orcid.org/0009-0008-5164-5470","observed_orcids":["https://orcid.org/0009-0008-5164-5470"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Ruihang","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104189549","display_name":"Miaosen Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Miaosen","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100369870","display_name":"Chen Li","orcid":"https://orcid.org/0000-0002-7508-7222","observed_orcids":["https://orcid.org/0000-0002-7508-7222"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Chen","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Ning, Jia","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ning, Jia","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102754531","display_name":"Ruizhe Wang","orcid":"https://orcid.org/0000-0002-2455-9074","observed_orcids":["https://orcid.org/0000-0002-2455-9074","https://orcid.org/0009-0004-4169-6785","https://orcid.org/0000-0002-8713-1892"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ruizhe","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039340081","display_name":"Zheng Zhang","orcid":"https://orcid.org/0000-0002-4141-2795","observed_orcids":["https://orcid.org/0000-0002-4141-2795"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zheng","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100384089","display_name":"Shuguang Liu","orcid":"https://orcid.org/0000-0003-3564-8336","observed_orcids":["https://orcid.org/0000-0003-3564-8336"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Shuguang","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104754746","display_name":"Joe Chau","orcid":"https://orcid.org/0009-0006-2669-3847","observed_orcids":["https://orcid.org/0009-0006-2669-3847"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chau, Joe","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091049278","display_name":"Hu Han","orcid":"https://orcid.org/0000-0001-6010-1792","observed_orcids":["https://orcid.org/0000-0001-6010-1792","https://orcid.org/0000-0001-5104-6146"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Han","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5111090174","display_name":"Peng Cheng","orcid":"https://orcid.org/0000-0002-4453-2274","observed_orcids":["https://orcid.org/0000-0002-4453-2274","https://orcid.org/0000-0003-4014-4757"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Peng","raw_affiliation_strings":["Microsoft Azure and Microsoft Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft Azure and Microsoft Research","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":129,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11697","display_name":"Numerical Methods and Algorithms","score":0.551800012588501,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.551800012588501,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.16300000250339508,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.12080000340938568,"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/fp8","display_name":"FP8","score":0.8209999799728394},{"id":"https://openalex.org/keywords/large-language-model-training","display_name":"large language model training","score":0.8330000042915344},{"id":"https://openalex.org/keywords/mixed-precision-training","display_name":"mixed-precision training","score":0.8429999947547913},{"id":"https://openalex.org/keywords/memory-efficiency","display_name":"memory efficiency","score":0.7170000076293945},{"id":"https://openalex.org/keywords/training-cost-reduction","display_name":"training cost reduction","score":0.7160000205039978},{"id":"https://openalex.org/keywords/instruction-tuning","display_name":"instruction tuning","score":0.6840000152587891},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.9520000219345093}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7917765378952026},{"id":"https://openalex.org/C121158502","wikidata":"https://www.wikidata.org/wiki/Q4652161","display_name":"AKA","level":2,"score":0.6750888824462891},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5286940336227417},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.47694364190101624},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.46914592385292053},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.45092886686325073},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4434935748577118},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4292401075363159},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38174164295196533},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.1149868369102478},{"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},{"id":"https://openalex.org/C161191863","wikidata":"https://www.wikidata.org/wiki/Q199655","display_name":"Library science","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2310.18313","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.18313","pdf_url":"https://arxiv.org/pdf/2310.18313","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2310.18313","doi":"https://doi.org/10.48550/arxiv.2310.18313","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2310.18313","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2310.18313","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.18313","pdf_url":"https://arxiv.org/pdf/2310.18313","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4388032098.pdf","grobid_xml":"https://content.openalex.org/works/W4388032098.grobid-xml"},"referenced_works_count":26,"referenced_works":["https://openalex.org/W2736601468","https://openalex.org/W2763421725","https://openalex.org/W2794325560","https://openalex.org/W2805206884","https://openalex.org/W2946609015","https://openalex.org/W2946659172","https://openalex.org/W2963341956","https://openalex.org/W2964121744","https://openalex.org/W2965373594","https://openalex.org/W2973727699","https://openalex.org/W3032816972","https://openalex.org/W3118781290","https://openalex.org/W4221167110","https://openalex.org/W4223908421","https://openalex.org/W4226146865","https://openalex.org/W4226399820","https://openalex.org/W4229005866","https://openalex.org/W4297812065","https://openalex.org/W4322718191","https://openalex.org/W4327810158","https://openalex.org/W4366341216","https://openalex.org/W4377121468","https://openalex.org/W4388979610","https://openalex.org/W7133220558","https://openalex.org/W7133225605","https://openalex.org/W7133239970"],"related_works":["https://openalex.org/W4415035937","https://openalex.org/W7083676559","https://openalex.org/W4407386106","https://openalex.org/W4403747399","https://openalex.org/W4404408104","https://openalex.org/W4406961439","https://openalex.org/W7105202352","https://openalex.org/W4387294240","https://openalex.org/W4399198049","https://openalex.org/W4416557041"],"abstract_inverted_index":{"In":[0],"this":[1],"paper,":[2],"we":[3,49],"explore":[4],"FP8":[5,53,66,106,155,186],"low-bit":[6],"data":[7,36],"formats":[8,37],"for":[9,57,75,149],"efficient":[10],"training":[11,32,58,74,97,108,147,157,188],"of":[12,65,98,137],"large":[13,150],"language":[14],"models":[15],"(LLMs).":[16],"Our":[17,185],"key":[18],"insight":[19],"is":[20,159,190],"that":[21],"most":[22],"variables,":[23],"such":[24,169],"as":[25,170],"gradients":[26],"and":[27,42,71,84,174],"optimizer":[28,82],"states,":[29,83],"in":[30,87,117,182],"LLM":[31,171],"can":[33,162],"employ":[34],"low-precision":[35,187],"without":[38],"compromising":[39],"model":[40,100],"accuracy":[41],"requiring":[43],"no":[44],"changes":[45],"to":[46,68,166],"hyper-parameters.":[47],"Specifically,":[48],"propose":[50],"a":[51,113],"new":[52],"automatic":[54],"mixed-precision":[55,70,107,156],"framework":[56,61,109,131,189],"LLMs.":[59,76],"This":[60,143],"offers":[62],"three":[63],"levels":[64],"utilization":[67],"streamline":[69],"distributed":[72,85],"parallel":[73],"It":[77,161],"gradually":[78],"incorporates":[79],"8-bit":[80],"gradients,":[81],"learning":[86,176],"an":[88],"incremental":[89],"manner.":[90],"Experiment":[91],"results":[92],"show":[93],"that,":[94],"during":[95],"the":[96,127,135,146],"GPT-175B":[99],"on":[101],"H100":[102],"GPU":[103],"platform,":[104],"our":[105,154],"not":[110],"only":[111],"achieved":[112],"remarkable":[114],"39%":[115],"reduction":[116],"real":[118],"memory":[119],"usage":[120],"but":[121],"also":[122],"ran":[123],"75%":[124],"faster":[125],"than":[126],"widely":[128],"adopted":[129],"BF16":[130],"(i.e.,":[132],"Megatron-LM),":[133],"surpassing":[134],"speed":[136],"Nvidia":[138],"Transformer":[139],"Engine":[140],"by":[141],"37%.":[142],"largely":[144],"reduces":[145],"costs":[148],"foundation":[151],"models.":[152],"Furthermore,":[153],"methodology":[158],"generic.":[160],"be":[163],"seamlessly":[164],"applied":[165],"other":[167],"tasks":[168],"instruction":[172],"tuning":[173],"reinforcement":[175],"with":[177],"human":[178],"feedback,":[179],"offering":[180],"savings":[181],"fine-tuning":[183],"expenses.":[184],"open-sourced":[191],"at":[192],"{https://github.com/Azure/MS-AMP}{aka.ms/MS.AMP}.":[193]},"counts_by_year":[{"year":2026,"cited_by_count":55},{"year":2025,"cited_by_count":54},{"year":2024,"cited_by_count":19},{"year":2023,"cited_by_count":1}],"updated_date":"2026-11-11T12:16:26.678958","created_date":"2023-11-01T00:00:00"}