{"id":"https://openalex.org/W4224267386","doi":"https://doi.org/10.1145/3470496.3527423","title":"Accelerating attention through gradient-based learned runtime pruning","display_name":"Accelerating attention through gradient-based learned runtime pruning","publication_year":2022,"publication_date":"2022-05-31","ids":{"openalex":"https://openalex.org/W4224267386","doi":"https://doi.org/10.1145/3470496.3527423"},"language":"en","primary_location":{"id":"doi:10.1145/3470496.3527423","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3470496.3527423","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3470496.3527423","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 49th Annual International Symposium on Computer Architecture","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/10.1145/3470496.3527423","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100414942","display_name":"Zheng Li","orcid":"https://orcid.org/0000-0001-5909-3545"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zheng Li","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082499242","display_name":"Soroush Ghodrati","orcid":"https://orcid.org/0000-0001-5514-8027"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Soroush Ghodrati","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070172290","display_name":"Amir Yazdanbakhsh","orcid":"https://orcid.org/0000-0001-8199-7671"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Amir Yazdanbakhsh","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084514143","display_name":"Hadi Esmaeilzadeh","orcid":"https://orcid.org/0000-0002-8548-1039"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hadi Esmaeilzadeh","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5006712350","display_name":"Mingu Kang","orcid":"https://orcid.org/0000-0001-8104-5136"},"institutions":[{"id":"https://openalex.org/I2803209242","display_name":"University of California System","ror":"https://ror.org/00pjdza24","country_code":"US","type":"education","lineage":["https://openalex.org/I2803209242"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mingu Kang","raw_affiliation_strings":["University of California"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California","institution_ids":["https://openalex.org/I2803209242"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":5.8267,"has_fulltext":true,"cited_by_count":46,"citation_normalized_percentile":{"value":0.96552644,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":100},"biblio":{"volume":null,"issue":null,"first_page":"902","last_page":"915"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9995999932289124,"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":0.9995999932289124,"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":0.9975000023841858,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9966999888420105,"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":0.7953252792358398},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6801087260246277},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.6131144165992737},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5727248787879944},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5639500021934509},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.49079737067222595},{"id":"https://openalex.org/keywords/differentiable-function","display_name":"Differentiable function","score":0.43948492407798767},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.41631871461868286},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.400771826505661},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.18755069375038147},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.11846518516540527}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7953252792358398},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6801087260246277},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6131144165992737},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5727248787879944},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5639500021934509},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.49079737067222595},{"id":"https://openalex.org/C202615002","wikidata":"https://www.wikidata.org/wiki/Q783507","display_name":"Differentiable function","level":2,"score":0.43948492407798767},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.41631871461868286},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.400771826505661},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.18755069375038147},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.11846518516540527},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"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/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.0},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","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/C6557445","wikidata":"https://www.wikidata.org/wiki/Q173113","display_name":"Agronomy","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3470496.3527423","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3470496.3527423","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3470496.3527423","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 49th Annual International Symposium on Computer Architecture","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3470496.3527423","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3470496.3527423","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3470496.3527423","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 49th Annual International Symposium on Computer Architecture","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1057515859","display_name":null,"funder_award_id":"2021-AH-3039","funder_id":"https://openalex.org/F4320306087","funder_display_name":"Semiconductor Research Corporation"},{"id":"https://openalex.org/G137895390","display_name":"Collaborative Research: SHF: Medium: Spatial Multi-Tenant Neural Acceleration for Next Generation Datacenters","funder_award_id":"2107598","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G2178652442","display_name":null,"funder_award_id":"CCF#2107598","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G2225644610","display_name":null,"funder_award_id":"HR0011-18-C-0020","funder_id":"https://openalex.org/F4320332180","funder_display_name":"Defense Advanced Research Projects Agency"},{"id":"https://openalex.org/G5781175801","display_name":null,"funder_award_id":"CNS#1822273","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G6064698370","display_name":null,"funder_award_id":"R01EB028350","funder_id":"https://openalex.org/F4320332161","funder_display_name":"National Institutes of Health"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320306087","display_name":"Semiconductor Research Corporation","ror":"https://ror.org/047z4n946"},{"id":"https://openalex.org/F4320332161","display_name":"National Institutes of Health","ror":"https://ror.org/01cwqze88"},{"id":"https://openalex.org/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"},{"id":"https://openalex.org/F4320332195","display_name":"Samsung","ror":"https://ror.org/04w3jy968"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4224267386.pdf","grobid_xml":"https://content.openalex.org/works/W4224267386.grobid-xml"},"referenced_works_count":29,"referenced_works":["https://openalex.org/W1503398984","https://openalex.org/W2048266589","https://openalex.org/W2233797083","https://openalex.org/W2442974303","https://openalex.org/W2515287984","https://openalex.org/W2518511512","https://openalex.org/W2541839172","https://openalex.org/W2551895583","https://openalex.org/W2605347906","https://openalex.org/W2613989746","https://openalex.org/W2790925711","https://openalex.org/W2883929540","https://openalex.org/W2888062646","https://openalex.org/W2935331687","https://openalex.org/W2963748441","https://openalex.org/W2964137095","https://openalex.org/W2979691890","https://openalex.org/W2979826702","https://openalex.org/W2980200167","https://openalex.org/W2998655561","https://openalex.org/W3107491168","https://openalex.org/W3159727696","https://openalex.org/W3177828909","https://openalex.org/W3207622241","https://openalex.org/W4240168186","https://openalex.org/W4242577057","https://openalex.org/W4251054771","https://openalex.org/W6600336938","https://openalex.org/W6811370846"],"related_works":["https://openalex.org/W3049463507","https://openalex.org/W2936497627","https://openalex.org/W4288365749","https://openalex.org/W3013624417","https://openalex.org/W4287826556","https://openalex.org/W4287598411","https://openalex.org/W3098382480","https://openalex.org/W3198458223","https://openalex.org/W4288267738","https://openalex.org/W3126642501"],"abstract_inverted_index":{"Self-attention":[0],"is":[1,49,61,76,94],"a":[2,19,32,36,56,92,102,132,149],"key":[3],"enabler":[4],"of":[5,39,59,111],"state-of-art":[6],"accuracy":[7,137,199],"for":[8,22,80,154,172],"various":[9],"transformer-based":[10],"Natural":[11],"Language":[12],"Processing":[13],"models.":[14,180],"This":[15,114],"attention":[16,66],"mechanism":[17],"calculates":[18],"correlation":[20],"score":[21],"each":[23],"word":[24,45],"with":[25,43,158],"respect":[26],"to":[27,64,122],"the":[28,44,78,81,108,112,119,125,128,197],"other":[29],"words":[30,40],"in":[31],"sentence.":[33],"Commonly,":[34],"only":[35,50],"small":[37],"subset":[38],"highly":[41],"correlates":[42],"under":[46],"attention,":[47],"which":[48,84],"determined":[51],"at":[52],"runtime.":[53],"As":[54],"such,":[55],"significant":[57],"amount":[58],"computation":[60,86,139],"inconsequential":[62],"due":[63],"low":[65],"scores":[67,82],"and":[68,127,138,177,191],"can":[69],"potentially":[70],"be":[71,88],"pruned.":[72],"The":[73],"main":[74],"challenge":[75],"finding":[77],"threshold":[79,93,126],"below":[83],"subsequent":[85],"will":[87],"inconsequential.":[89],"Although":[90],"such":[91],"discrete,":[95],"this":[96,144],"paper":[97],"formulates":[98],"its":[99],"search":[100],"through":[101],"soft":[103],"differentiable":[104],"regularizer":[105],"integrated":[106],"into":[107],"loss":[109],"function":[110],"training.":[113],"formulation":[115],"piggy":[116],"backs":[117],"on":[118,185],"back-propagation":[120],"training":[121],"analytically":[123],"co-optimize":[124],"weights":[129],"simultaneously,":[130],"striking":[131],"formally":[133],"optimal":[134],"balance":[135],"between":[136],"pruning.":[140],"To":[141],"best":[142],"utilize":[143],"mathematical":[145],"innovation,":[146],"we":[147],"devise":[148],"bit-serial":[150],"architecture,":[151],"dubbed":[152],"LeOPArd,":[153],"transformer":[155,179],"language":[156],"models":[157],"bit-level":[159],"early":[160],"termination":[161],"microarchitectural":[162],"mechanism.":[163],"We":[164],"evaluate":[165],"our":[166],"design":[167],"across":[168],"43":[169],"back-end":[170],"tasks":[171],"MemN2N,":[173],"BERT,":[174],"ALBERT,":[175],"GPT-2,":[176],"Vision":[178],"Post-layout":[181],"results":[182],"show":[183],"that,":[184],"average,":[186],"LeOPArd":[187],"yields":[188],"1.9\u00d7and":[189],"3.9\u00d7speedup":[190],"energy":[192],"reduction,":[193],"respectively,":[194],"while":[195],"keeping":[196],"average":[198],"virtually":[200],"intact":[201],"(<":[202],"0.2%":[203],"degradation).":[204]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":12},{"year":2024,"cited_by_count":12},{"year":2023,"cited_by_count":16},{"year":2022,"cited_by_count":2}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}