{"id":"https://openalex.org/W4291653336","doi":"https://doi.org/10.1109/tvlsi.2022.3197282","title":"An Algorithm\u2013Hardware Co-Optimized Framework for Accelerating N:M Sparse Transformers","display_name":"An Algorithm\u2013Hardware Co-Optimized Framework for Accelerating N:M Sparse Transformers","publication_year":2022,"publication_date":"2022-08-15","ids":{"openalex":"https://openalex.org/W4291653336","doi":"https://doi.org/10.1109/tvlsi.2022.3197282"},"language":"en","primary_location":{"id":"doi:10.1109/tvlsi.2022.3197282","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2022.3197282","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"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 Very Large Scale Integration (VLSI) Systems","raw_type":"journal-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2208.06118","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5066888122","display_name":"Chao Fang","orcid":"https://orcid.org/0000-0003-3430-1189"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chao Fang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0003-3430-1189","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041431476","display_name":"Aojun Zhou","orcid":"https://orcid.org/0000-0002-4742-8624"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Aojun Zhou","raw_affiliation_strings":["CUHK-Sensetime Joint Laboratory, The Chinese University of Hong Kong (CUHK), Hong Kong, Shatin, China"],"raw_orcid":"https://orcid.org/0000-0002-4742-8624","affiliations":[{"raw_affiliation_string":"CUHK-Sensetime Joint Laboratory, The Chinese University of Hong Kong (CUHK), Hong Kong, Shatin, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100696999","display_name":"Zhongfeng Wang","orcid":"https://orcid.org/0000-0002-7227-4786"},"institutions":[{"id":"https://openalex.org/I881766915","display_name":"Nanjing University","ror":"https://ror.org/01rxvg760","country_code":"CN","type":"education","lineage":["https://openalex.org/I881766915"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhongfeng Wang","raw_affiliation_strings":["School of Electronic Science and Engineering, Nanjing University, Nanjing, China"],"raw_orcid":"https://orcid.org/0000-0002-7227-4786","affiliations":[{"raw_affiliation_string":"School of Electronic Science and Engineering, Nanjing University, Nanjing, China","institution_ids":["https://openalex.org/I881766915"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":5.8515,"has_fulltext":false,"cited_by_count":86,"citation_normalized_percentile":{"value":0.97569611,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":89,"max":100},"biblio":{"volume":"30","issue":"11","first_page":"1573","last_page":"1586"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9894999861717224,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9894999861717224,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.9747999906539917,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9740999937057495,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7020137906074524},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5956673622131348},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5004477500915527},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.4790045917034149},{"id":"https://openalex.org/keywords/sparse-matrix","display_name":"Sparse matrix","score":0.46482375264167786},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.4633197486400604},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4468371868133545},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.4195987284183502},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.3821234405040741},{"id":"https://openalex.org/keywords/gaussian","display_name":"Gaussian","score":0.1378690004348755}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7020137906074524},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5956673622131348},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5004477500915527},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4790045917034149},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.46482375264167786},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4633197486400604},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4468371868133545},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4195987284183502},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.3821234405040741},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.1378690004348755},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","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/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/tvlsi.2022.3197282","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tvlsi.2022.3197282","pdf_url":null,"source":{"id":"https://openalex.org/S37538908","display_name":"IEEE Transactions on Very Large Scale Integration (VLSI) Systems","issn_l":"1063-8210","issn":["1063-8210","1557-9999"],"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 Very Large Scale Integration (VLSI) Systems","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:2208.06118","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2208.06118","pdf_url":"https://arxiv.org/pdf/2208.06118","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,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2208.06118","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2208.06118","pdf_url":"https://arxiv.org/pdf/2208.06118","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,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2250357272","display_name":null,"funder_award_id":"62174084","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G4941541027","display_name":null,"funder_award_id":"62104097","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":53,"referenced_works":["https://openalex.org/W2490345425","https://openalex.org/W2625457103","https://openalex.org/W2809679348","https://openalex.org/W2896457183","https://openalex.org/W2906043559","https://openalex.org/W2923014074","https://openalex.org/W2946355854","https://openalex.org/W2952682304","https://openalex.org/W2977634443","https://openalex.org/W2980282514","https://openalex.org/W2981124343","https://openalex.org/W2995463996","https://openalex.org/W2998135987","https://openalex.org/W3012178976","https://openalex.org/W3017024317","https://openalex.org/W3034100311","https://openalex.org/W3037132819","https://openalex.org/W3038838661","https://openalex.org/W3041223772","https://openalex.org/W3043034704","https://openalex.org/W3047848469","https://openalex.org/W3085139254","https://openalex.org/W3094502228","https://openalex.org/W3105966348","https://openalex.org/W3118337360","https://openalex.org/W3118709344","https://openalex.org/W3127067080","https://openalex.org/W3134012069","https://openalex.org/W3158912502","https://openalex.org/W3159481202","https://openalex.org/W3159727696","https://openalex.org/W3162542754","https://openalex.org/W3164217046","https://openalex.org/W3179312164","https://openalex.org/W3196923642","https://openalex.org/W3206453033","https://openalex.org/W3207622241","https://openalex.org/W3211907200","https://openalex.org/W4230989867","https://openalex.org/W4240168186","https://openalex.org/W4251575795","https://openalex.org/W4287208846","https://openalex.org/W4287363917","https://openalex.org/W4287667694","https://openalex.org/W4385245566","https://openalex.org/W6739901393","https://openalex.org/W6768817161","https://openalex.org/W6771852177","https://openalex.org/W6780125659","https://openalex.org/W6789946416","https://openalex.org/W6790503700","https://openalex.org/W6794457956","https://openalex.org/W6803607599"],"related_works":["https://openalex.org/W2058965144","https://openalex.org/W2164382479","https://openalex.org/W2146343568","https://openalex.org/W98480971","https://openalex.org/W2150291671","https://openalex.org/W2013643406","https://openalex.org/W2027972911","https://openalex.org/W2157978810","https://openalex.org/W3138386522","https://openalex.org/W2499279132"],"abstract_inverted_index":{"The":[0],"Transformer":[1],"has":[2],"been":[3],"an":[4,38,78,96,207],"indispensable":[5],"staple":[6],"in":[7],"deep":[8],"learning.":[9],"However,":[10],"for":[11,134,249],"real-life":[12],"applications,":[13],"it":[14,61],"is":[15,37,125],"very":[16],"challenging":[17],"to":[18,23,41,56,82,111,128,151,197,225],"deploy":[19],"efficient":[20,146],"Transformers":[21,87,119],"due":[22],"the":[24,65,131,185,242],"immense":[25],"parameters":[26],"and":[27,68,84,145,170,228,234],"operations":[28],"of":[29,115,209],"models.":[30,73],"To":[31],"relieve":[32],"this":[33],"burden,":[34],"exploiting":[35],"sparsity":[36,54,92,102],"effective":[39],"approach":[40],"accelerate":[42,86],"Transformers.":[43,159,250],"Newly":[44],"emerging":[45],"Ampere":[46],"graphics":[47],"processing":[48],"units":[49],"(GPUs)":[50],"leverage":[51],"a":[52,101,113,138,143,164,180],"2:4":[53],"pattern":[55],"achieve":[57,152],"model":[58,122],"acceleration,":[59],"while":[60],"can":[62,220],"hardly":[63],"meet":[64],"diverse":[66],"algorithm":[67,97],"hardware":[69,139,147],"constraints":[70],"when":[71,155],"deploying":[72,156],"By":[74],"contrast,":[75],"we":[76,99,141],"propose":[77,100],"algorithm\u2013hardware":[79],"co-optimized":[80],"framework":[81],"flexibly":[83],"efficiently":[85],"by":[88],"utilizing":[89],"general":[90],"N:M":[91,116,157,200],"patterns.":[93],"First,":[94],"from":[95,137,187],"perspective,":[98,140],"inheritance":[103],"mechanism":[104],"along":[105],"with":[106,174,214],"inherited":[107],"dynamic":[108],"pruning":[109],"(IDP)":[110],"obtain":[112],"series":[114],"sparse":[117,158,201],"candidate":[118],"rapidly.":[120],"A":[121],"compression":[123],"scheme":[124],"further":[126],"proposed":[127],"significantly":[129],"reduce":[130],"storage":[132],"requirement":[133],"deployment.":[135],"Second,":[136],"present":[142],"flexible":[144],"architecture,":[148],"namely,":[149],"STA,":[150],"significant":[153],"speedup":[154],"STA":[160,219],"features":[161],"not":[162],"only":[163],"computing":[165],"engine":[166],"unifying":[167],"both":[168],"sparse\u2013dense":[169],"dense\u2013dense":[171],"matrix":[172],"multiplications":[173],"high":[175,215],"computational":[176],"efficiency":[177],"but":[178],"also":[179],"scalable":[181],"softmax":[182],"module":[183],"eliminating":[184],"latency":[186],"intermediate":[188],"off-chip":[189],"data":[190],"communication.":[191],"Experimental":[192],"results":[193],"show":[194],"that,":[195],"compared":[196,224],"other":[198],"methods,":[199],"Transformers,":[202],"generated":[203],"using":[204],"IDP,":[205],"achieves":[206],"average":[208],"6.7%":[210],"improvement":[211],"on":[212],"accuracy":[213],"training":[216],"efficiency.":[217],"Moreover,":[218],"achieve$14.47\\times":[221],"$and$11.33\\times":[222],"$speedups":[223],"Intel":[226],"i9-9900X":[227],"NVIDIA":[229],"RTX":[230],"2080":[231],"Ti,":[232],"respectively,":[233],"perform$2.00":[235],"\\,\\,\\sim":[236],"19.47":[237],"\\times":[238],"$faster":[239],"inference":[240],"than":[241],"state-of-the-art":[243],"field-programmable":[244],"gate":[245],"array":[246],"(FPGA)-based":[247],"accelerators":[248]},"counts_by_year":[{"year":2026,"cited_by_count":14},{"year":2025,"cited_by_count":28},{"year":2024,"cited_by_count":30},{"year":2023,"cited_by_count":13},{"year":2022,"cited_by_count":1}],"updated_date":"2026-08-02T14:50:37.381335","created_date":"2022-08-15T00:00:00"}