{"id":"https://openalex.org/W4403220287","doi":"https://doi.org/10.1145/3640457.3688037","title":"Toward 100TB Recommendation Models with Embedding Offloading","display_name":"Toward 100TB Recommendation Models with Embedding Offloading","publication_year":2024,"publication_date":"2024-10-08","ids":{"openalex":"https://openalex.org/W4403220287","doi":"https://doi.org/10.1145/3640457.3688037"},"language":"en","primary_location":{"id":"doi:10.1145/3640457.3688037","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3640457.3688037","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3640457.3688037","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"18th ACM Conference on Recommender Systems","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/10.1145/3640457.3688037","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5032599348","display_name":"Intaik Park","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Intaik Park","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0002-6958-5997","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092183287","display_name":"Ehsan Ardestani","orcid":"https://orcid.org/0000-0003-1267-6887"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ehsan Ardestani","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0003-1267-6887","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111879392","display_name":"D. Reeves","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Damian Reeves","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0009-0008-7653-5665","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016976996","display_name":"Sarunya Pumma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sarunya Pumma","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0002-4662-3129","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049571614","display_name":"Henry Ling-Hei Tsang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Henry Tsang","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0002-5835-1474","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111357208","display_name":"Levy Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Levy Zhao","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0009-0007-5384-7204","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Jian He","orcid":"https://orcid.org/0000-0002-3947-6183"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jian He","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0002-3947-6183","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111357209","display_name":"Joshua Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Joshua Deng","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0009-0004-9012-3254","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038941467","display_name":"Dennis Van der Staay","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dennis Van der Staay","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0009-0006-8254-4977","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013273553","display_name":"Yu Guo","orcid":"https://orcid.org/0000-0001-7243-2089"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu Guo","raw_affiliation_strings":["Meta, USA"],"raw_orcid":"https://orcid.org/0000-0001-7243-2089","affiliations":[{"raw_affiliation_string":"Meta, USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5107823285","display_name":"Paul Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paul Zhang","raw_affiliation_strings":["AI Infra, Meta, USA"],"raw_orcid":"https://orcid.org/0009-0004-1063-1372","affiliations":[{"raw_affiliation_string":"AI Infra, Meta, USA","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":11,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.6895,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.78067687,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":null,"issue":null,"first_page":"841","last_page":"843"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10203","display_name":"Recommender Systems and Techniques","score":0.9991999864578247,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10203","display_name":"Recommender Systems and Techniques","score":0.9991999864578247,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.9775000214576721,"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/T10028","display_name":"Topic Modeling","score":0.967199981212616,"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":0.6747342348098755},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.6004107594490051},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.21879112720489502}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6747342348098755},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.6004107594490051},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.21879112720489502}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3640457.3688037","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3640457.3688037","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3640457.3688037","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"18th ACM Conference on Recommender Systems","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3640457.3688037","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3640457.3688037","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3640457.3688037","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"18th ACM Conference on Recommender Systems","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4403220287.pdf"},"referenced_works_count":6,"referenced_works":["https://openalex.org/W2164705534","https://openalex.org/W2794670651","https://openalex.org/W3197720002","https://openalex.org/W3201621211","https://openalex.org/W4296591817","https://openalex.org/W4387302734"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W2081900870","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052","https://openalex.org/W4402327032"],"abstract_inverted_index":{"Training":[0],"recommendation":[1],"models":[2],"become":[3],"memory-bound":[4],"with":[5,100],"large":[6,28],"embedding":[7,20,34],"tables,":[8],"and":[9,22,40,48,63,75,77,84],"fast":[10],"GPU":[11,74],"memory":[12,32],"is":[13,65,79],"scarce.":[14],"In":[15],"this":[16],"paper,":[17],"we":[18],"explore":[19],"caches":[21,46],"prefetch":[23],"pipelines":[24],"to":[25,81,96],"effectively":[26],"leverage":[27],"but":[29],"slow":[30],"host":[31],"for":[33],"tables.":[35],"We":[36],"introduce":[37],"Locality-Aware":[38],"Sharding":[39],"iterative":[41],"planning":[42],"that":[43,57],"automatically":[44],"size":[45,99],"optimally":[47],"produce":[49],"effective":[50],"sharding":[51],"plans.":[52],"Embedding":[53,90],"Offloading,":[54],"a":[55],"system":[56],"combines":[58],"all":[59],"of":[60,69,86],"these":[61],"components":[62],"techniques,":[64],"implemented":[66],"on":[67],"top":[68],"Meta\u2019s":[70],"open-source":[71],"libraries,":[72],"FBGEMM":[73],"TorchRec,":[76],"it":[78],"used":[80],"improve":[82],"scalability":[83],"efficiency":[85],"industry-scale":[87],"production":[88],"models.":[89],"Offloading":[91],"achieved":[92],"37x":[93],"model":[94,98],"scale":[95],"100TB":[97],"only":[101],"26%":[102],"training":[103],"speed":[104],"regression.":[105]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}