{"id":"https://openalex.org/W4394841768","doi":"https://doi.org/10.1109/tc.2024.3389507","title":"Raptor-T: A Fused and Memory-Efficient Sparse Transformer for Long and Variable-Length Sequences","display_name":"Raptor-T: A Fused and Memory-Efficient Sparse Transformer for Long and Variable-Length Sequences","publication_year":2024,"publication_date":"2024-04-16","ids":{"openalex":"https://openalex.org/W4394841768","doi":"https://doi.org/10.1109/tc.2024.3389507"},"language":"en","primary_location":{"id":"doi:10.1109/tc.2024.3389507","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2024.3389507","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-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/A5036278941","display_name":"H. Wang","orcid":"https://orcid.org/0009-0003-8329-0145"},"institutions":[{"id":"https://openalex.org/I37461747","display_name":"Wuhan University","ror":"https://ror.org/033vjfk17","country_code":"CN","type":"education","lineage":["https://openalex.org/I37461747"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hulin Wang","raw_affiliation_strings":["School of Computer Science, Wuhan University, Hubei, China"],"raw_orcid":"https://orcid.org/0009-0003-8329-0145","affiliations":[{"raw_affiliation_string":"School of Computer Science, Wuhan University, Hubei, China","institution_ids":["https://openalex.org/I37461747"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005205340","display_name":"Donglin Yang","orcid":"https://orcid.org/0000-0002-3913-3623"},"institutions":[{"id":"https://openalex.org/I4210127875","display_name":"Nvidia (United States)","ror":"https://ror.org/03jdj4y14","country_code":"US","type":"company","lineage":["https://openalex.org/I4210127875"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Donglin Yang","raw_affiliation_strings":["Nvidia Corp., Santa Clara, CA, USA"],"raw_orcid":"https://orcid.org/0000-0002-3913-3623","affiliations":[{"raw_affiliation_string":"Nvidia Corp., Santa Clara, CA, USA","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038795287","display_name":"Yaqi Xia","orcid":"https://orcid.org/0009-0006-8101-785X"},"institutions":[{"id":"https://openalex.org/I37461747","display_name":"Wuhan University","ror":"https://ror.org/033vjfk17","country_code":"CN","type":"education","lineage":["https://openalex.org/I37461747"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yaqi Xia","raw_affiliation_strings":["School of Computer Science, Wuhan University, Hubei, China"],"raw_orcid":"https://orcid.org/0009-0006-8101-785X","affiliations":[{"raw_affiliation_string":"School of Computer Science, Wuhan University, Hubei, China","institution_ids":["https://openalex.org/I37461747"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013177309","display_name":"Zheng Zhang","orcid":"https://orcid.org/0000-0001-6599-9976"},"institutions":[{"id":"https://openalex.org/I37461747","display_name":"Wuhan University","ror":"https://ror.org/033vjfk17","country_code":"CN","type":"education","lineage":["https://openalex.org/I37461747"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zheng Zhang","raw_affiliation_strings":["School of Computer Science, Wuhan University, Hubei, China"],"raw_orcid":"https://orcid.org/0000-0001-6599-9976","affiliations":[{"raw_affiliation_string":"School of Computer Science, Wuhan University, Hubei, China","institution_ids":["https://openalex.org/I37461747"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102931585","display_name":"Qigang Wang","orcid":"https://orcid.org/0009-0008-8533-4726"},"institutions":[{"id":"https://openalex.org/I4210156165","display_name":"Lenovo (China)","ror":"https://ror.org/04srd9d93","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210156165"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qigang Wang","raw_affiliation_strings":["AI Lab, Lenovo Research, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0008-8533-4726","affiliations":[{"raw_affiliation_string":"AI Lab, Lenovo Research, Beijing, China","institution_ids":["https://openalex.org/I4210156165"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100728569","display_name":"Jianping Fan","orcid":"https://orcid.org/0000-0003-2290-1785"},"institutions":[{"id":"https://openalex.org/I4210156165","display_name":"Lenovo (China)","ror":"https://ror.org/04srd9d93","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210156165"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jianping Fan","raw_affiliation_strings":["AI Lab, Lenovo Research, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-2290-1785","affiliations":[{"raw_affiliation_string":"AI Lab, Lenovo Research, Beijing, China","institution_ids":["https://openalex.org/I4210156165"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101677080","display_name":"Xiaobo Zhou","orcid":"https://orcid.org/0009-0004-9500-3390"},"institutions":[{"id":"https://openalex.org/I204512498","display_name":"University of Macau","ror":"https://ror.org/01r4q9n85","country_code":"MO","type":"education","lineage":["https://openalex.org/I204512498"]}],"countries":["MO"],"is_corresponding":false,"raw_author_name":"Xiaobo Zhou","raw_affiliation_strings":["IOTSC & Department of Computer and Information Sciences, University of Macau, Macau S.A.R., China"],"raw_orcid":"https://orcid.org/0009-0004-9500-3390","affiliations":[{"raw_affiliation_string":"IOTSC & Department of Computer and Information Sciences, University of Macau, Macau S.A.R., China","institution_ids":["https://openalex.org/I204512498"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5063911669","display_name":"Dazhao Cheng","orcid":"https://orcid.org/0000-0003-2869-7623"},"institutions":[{"id":"https://openalex.org/I37461747","display_name":"Wuhan University","ror":"https://ror.org/033vjfk17","country_code":"CN","type":"education","lineage":["https://openalex.org/I37461747"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dazhao Cheng","raw_affiliation_strings":["School of Computer Science, Wuhan University, Hubei, China"],"raw_orcid":"https://orcid.org/0000-0003-2869-7623","affiliations":[{"raw_affiliation_string":"School of Computer Science, Wuhan University, Hubei, China","institution_ids":["https://openalex.org/I37461747"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":{"value":2495,"currency":"USD","value_usd":2495},"apc_paid":null,"fwci":2.3912,"has_fulltext":false,"cited_by_count":12,"citation_normalized_percentile":{"value":0.89441369,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":96,"max":99},"biblio":{"volume":"73","issue":"7","first_page":"1852","last_page":"1865"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11034","display_name":"Digital Filter Design and Implementation","score":0.9955999851226807,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11034","display_name":"Digital Filter Design and Implementation","score":0.9955999851226807,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.9940999746322632,"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"}},{"id":"https://openalex.org/T10299","display_name":"Photonic and Optical Devices","score":0.9821000099182129,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/variable","display_name":"Variable (mathematics)","score":0.5800561308860779},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5180618762969971},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4679967761039734},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3857644200325012},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.34201011061668396},{"id":"https://openalex.org/keywords/electrical-engineering","display_name":"Electrical engineering","score":0.15118461847305298},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.11666107177734375},{"id":"https://openalex.org/keywords/voltage","display_name":"Voltage","score":0.05725923180580139}],"concepts":[{"id":"https://openalex.org/C182365436","wikidata":"https://www.wikidata.org/wiki/Q50701","display_name":"Variable (mathematics)","level":2,"score":0.5800561308860779},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5180618762969971},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4679967761039734},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3857644200325012},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.34201011061668396},{"id":"https://openalex.org/C119599485","wikidata":"https://www.wikidata.org/wiki/Q43035","display_name":"Electrical engineering","level":1,"score":0.15118461847305298},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.11666107177734375},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.05725923180580139},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tc.2024.3389507","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2024.3389507","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.6200000047683716,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[{"id":"https://openalex.org/G267310179","display_name":null,"funder_award_id":"Macao S.A.R. (FDCT) Project 0078/2023/AMJ","funder_id":"https://openalex.org/F4320321655","funder_display_name":"Science and Technology Development Fund"},{"id":"https://openalex.org/G3830727860","display_name":null,"funder_award_id":"62341410","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G4803673267","display_name":null,"funder_award_id":"2023YFE0205700","funder_id":"https://openalex.org/F4320335777","funder_display_name":"National Key Research and Development Program of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321655","display_name":"Science and Technology Development Fund","ror":"https://ror.org/044vr6g03"},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":38,"referenced_works":["https://openalex.org/W1591801644","https://openalex.org/W2027377866","https://openalex.org/W2064675550","https://openalex.org/W2896457183","https://openalex.org/W2939507640","https://openalex.org/W2962985038","https://openalex.org/W2963323070","https://openalex.org/W2963748441","https://openalex.org/W2970972587","https://openalex.org/W2973727699","https://openalex.org/W2979826702","https://openalex.org/W2984864519","https://openalex.org/W3015468748","https://openalex.org/W3081168214","https://openalex.org/W3129831491","https://openalex.org/W3130716829","https://openalex.org/W4214686755","https://openalex.org/W4285261371","https://openalex.org/W4321636575","https://openalex.org/W4321636820","https://openalex.org/W4384648639","https://openalex.org/W4384705353","https://openalex.org/W4385245566","https://openalex.org/W6635446068","https://openalex.org/W6713134421","https://openalex.org/W6751349269","https://openalex.org/W6761748628","https://openalex.org/W6766978945","https://openalex.org/W6767997687","https://openalex.org/W6769823692","https://openalex.org/W6771626834","https://openalex.org/W6771915120","https://openalex.org/W6771917389","https://openalex.org/W6776048684","https://openalex.org/W6778883912","https://openalex.org/W6781533629","https://openalex.org/W6838322825","https://openalex.org/W6854308872"],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W4391375266","https://openalex.org/W1979597421","https://openalex.org/W2007980826","https://openalex.org/W2051487156","https://openalex.org/W2061531152","https://openalex.org/W3002753104","https://openalex.org/W2077600819","https://openalex.org/W2142036596","https://openalex.org/W2072657027"],"abstract_inverted_index":{"Transformer-based":[0],"models":[1],"have":[2],"made":[3],"significant":[4],"advancements":[5],"across":[6],"various":[7],"domains,":[8],"largely":[9],"due":[10,34,77],"to":[11,16,35,49,54,78,84,123,136,161],"the":[12,51,58,74,79,89,117,120,142,178],"self-attention":[13],"mechanism\u2019s":[14],"ability":[15],"capture":[17],"contextual":[18],"relationships":[19],"in":[20,95],"input":[21,92],"sequences.":[22,114],"However,":[23],"processing":[24,128,159],"long":[25,111,129],"sequences":[26,93,130],"remains":[27],"computationally":[28],"expensive":[29],"for":[30,73,109,127,173],"Transformer":[31],"models,":[32],"primarily":[33],"theO(n2)":[36],"complexity":[37],"associated":[38,180],"with":[39,181],"self-attention.":[40],"To":[41,140],"address":[42,141],"this,":[43],"sparse":[44,59,75,121,167],"attention":[45,144],"has":[46],"been":[47],"proposed":[48],"reduce":[50,124],"quadratic":[52],"dependency":[53],"linear.":[55],"Nevertheless,":[56],"deploying":[57],"transformer":[60,76,106,122],"efficiently":[61],"encounters":[62],"two":[63],"major":[64],"obstacles:":[65],"1)":[66],"Existing":[67],"system":[68],"optimizations":[69,135],"are":[70],"less":[71],"effective":[72],"algorithm\u2019s":[80],"approximation":[81],"properties":[82],"leading":[83],"fragmented":[85,143],"attention,":[86],"and":[87,97,112,149,183,211,222],"2)":[88],"variability":[90],"of":[91,119,220],"results":[94,205],"computation":[96,186],"memory":[98],"access":[99],"inefficiencies.":[100],"We":[101],"present":[102],"Raptor-T,":[103],"a":[104],"cutting-edge":[105],"framework":[107],"designed":[108],"handling":[110],"variable-length":[113,174],"Raptor-T":[115,146,170,208],"harnesses":[116],"power":[118],"resource":[125],"requirements":[126],"while":[131],"also":[132],"implementing":[133],"system-level":[134],"accelerate":[137],"inference":[138],"performance.":[139],"issue,":[145],"employs":[147],"fused":[148],"memory-efficient":[150],"Multi-Head":[151],"Attention.":[152],"Additionally,":[153],"we":[154,191],"introduce":[155],"an":[156,199,214],"asynchronous":[157],"data":[158],"method":[160],"mitigate":[162],"GPU-blocking":[163],"operations":[164],"caused":[165],"by":[166],"attention.":[168],"Furthermore,":[169],"minimizes":[171],"padding":[172,182],"inputs,":[175],"effectively":[176],"reducing":[177],"overhead":[179],"achieving":[184,213],"balanced":[185],"on":[187,198],"GPUs.":[188],"In":[189],"evaluation,":[190],"compare":[192],"Raptor-T\u2019s":[193],"performance":[194,218],"against":[195],"state-of-the-art":[196],"frameworks":[197],"NVIDIA":[200],"A100":[201],"GPU.":[202],"The":[203],"experimental":[204],"demonstrate":[206],"that":[207],"outperforms":[209],"FlashAttention-2":[210],"FasterTransformer,":[212],"impressive":[215],"average":[216],"end-to-end":[217],"improvement":[219],"3.41X":[221],"3.71X,":[223],"respectively.":[224]},"counts_by_year":[{"year":2026,"cited_by_count":5},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":3}],"updated_date":"2026-08-31T07:19:18.363097","created_date":"2025-10-10T00:00:00"}