{"id":"https://openalex.org/W4321466385","doi":"https://doi.org/10.1145/3572848.3577500","title":"Dynamic N:M Fine-Grained Structured Sparse Attention Mechanism","display_name":"Dynamic N:M Fine-Grained Structured Sparse Attention Mechanism","publication_year":2023,"publication_date":"2023-02-21","ids":{"openalex":"https://openalex.org/W4321466385","doi":"https://doi.org/10.1145/3572848.3577500"},"language":"en","primary_location":{"id":"doi:10.1145/3572848.3577500","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3572848.3577500","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3572848.3577500","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 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3572848.3577500","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5037216840","display_name":"Zhaodong Chen","orcid":"https://orcid.org/0000-0001-9601-4586"},"institutions":[{"id":"https://openalex.org/I154570441","display_name":"University of California, Santa Barbara","ror":"https://ror.org/02t274463","country_code":"US","type":"education","lineage":["https://openalex.org/I154570441"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhaodong Chen","raw_affiliation_strings":["UC Santa Barbara, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Santa Barbara, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I154570441"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003935082","display_name":"Zheng Qu","orcid":"https://orcid.org/0000-0001-6574-0649"},"institutions":[{"id":"https://openalex.org/I154570441","display_name":"University of California, Santa Barbara","ror":"https://ror.org/02t274463","country_code":"US","type":"education","lineage":["https://openalex.org/I154570441"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zheng Qu","raw_affiliation_strings":["UC Santa Barbara, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Santa Barbara, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I154570441"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087088688","display_name":"Yuying Quan","orcid":null},"institutions":[{"id":"https://openalex.org/I154570441","display_name":"University of California, Santa Barbara","ror":"https://ror.org/02t274463","country_code":"US","type":"education","lineage":["https://openalex.org/I154570441"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuying Quan","raw_affiliation_strings":["UC Santa Barbara, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Santa Barbara, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I154570441"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100324294","display_name":"Liu Liu","orcid":"https://orcid.org/0000-0003-0792-8146"},"institutions":[{"id":"https://openalex.org/I165799507","display_name":"Rensselaer Polytechnic Institute","ror":"https://ror.org/01rtyzb94","country_code":"US","type":"education","lineage":["https://openalex.org/I165799507"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu Liu","raw_affiliation_strings":["Rensselaer Polytechnic Institute, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Rensselaer Polytechnic Institute, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I165799507"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048052285","display_name":"Yufei Ding","orcid":"https://orcid.org/0000-0002-8716-5793"},"institutions":[{"id":"https://openalex.org/I154570441","display_name":"University of California, Santa Barbara","ror":"https://ror.org/02t274463","country_code":"US","type":"education","lineage":["https://openalex.org/I154570441"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yufei Ding","raw_affiliation_strings":["UC Santa Barbara, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Santa Barbara, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I154570441"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100385336","display_name":"Yuan Xie","orcid":"https://orcid.org/0000-0003-2093-1788"},"institutions":[{"id":"https://openalex.org/I4210095624","display_name":"Alibaba Group (United States)","ror":"https://ror.org/00rn0m335","country_code":"US","type":"company","lineage":["https://openalex.org/I4210095624","https://openalex.org/I45928872"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuan Xie","raw_affiliation_strings":["ALIBABA DAMO Academy, Santa Barbara, CA, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ALIBABA DAMO Academy, Santa Barbara, CA, USA","institution_ids":["https://openalex.org/I4210095624"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":3.0316,"has_fulltext":true,"cited_by_count":26,"citation_normalized_percentile":{"value":0.94017907,"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":"369","last_page":"379"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":1.0,"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":1.0,"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":0.9965999722480774,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.9965000152587891,"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/speedup","display_name":"Speedup","score":0.8722854852676392},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8263435363769531},{"id":"https://openalex.org/keywords/sparse-matrix","display_name":"Sparse matrix","score":0.5786346793174744},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5467568039894104},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.5234726667404175},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.4065156579017639},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.35912424325942993}],"concepts":[{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8722854852676392},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8263435363769531},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.5786346793174744},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5467568039894104},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5234726667404175},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4065156579017639},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35912424325942993},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"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/C6557445","wikidata":"https://www.wikidata.org/wiki/Q173113","display_name":"Agronomy","level":1,"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/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3572848.3577500","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3572848.3577500","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3572848.3577500","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 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3572848.3577500","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3572848.3577500","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3572848.3577500","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 28th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4321466385.pdf","grobid_xml":"https://content.openalex.org/works/W4321466385.grobid-xml"},"referenced_works_count":6,"referenced_works":["https://openalex.org/W2963807318","https://openalex.org/W2980186997","https://openalex.org/W3012178976","https://openalex.org/W3131922516","https://openalex.org/W3173365702","https://openalex.org/W3208099998"],"related_works":["https://openalex.org/W1509211761","https://openalex.org/W2391299576","https://openalex.org/W2133693067","https://openalex.org/W1531488649","https://openalex.org/W1585350690","https://openalex.org/W2366027386","https://openalex.org/W2582456645","https://openalex.org/W2479014312","https://openalex.org/W1583465708","https://openalex.org/W2074226157"],"abstract_inverted_index":{"Transformers":[0],"are":[1],"becoming":[2],"the":[3,17,21,36,41,44,51,83,97,114,125,131,137,158,164,168,215,221,236,259,270],"mainstream":[4],"solutions":[5],"for":[6],"various":[7,244],"tasks":[8,242],"like":[9],"NLP":[10],"and":[11,55,123,160,166,204,223,229],"Computer":[12],"vision.":[13],"Despite":[14],"their":[15],"success,":[16],"high":[18],"complexity":[19],"of":[20,153,214,266],"attention":[22,37,45,99,198,217,261],"mechanism":[23,38],"hinders":[24],"them":[25],"from":[26,243,249,269],"being":[27],"applied":[28],"to":[29,34,50,60,74,90,102,121,192,207,251],"latency-sensitive":[30],"tasks.":[31],"One":[32],"opportunity":[33],"accelerate":[35],"is":[39,111,119,210,255],"leveraging":[40],"sparsity":[42,118,225],"in":[43,140,181],"weight":[46],"matrix.":[47],"However,":[48],"due":[49],"dilemma":[52],"between":[53],"\"dynamic\"":[54],"\"fine-grained\",":[56],"previous":[57,176],"studies":[58],"fail":[59],"achieve":[61,230],"speedup":[62,180],"on":[63,113,128,238,256],"GPUs":[64],"under":[65,226],"moderate":[66],"sequence":[67,183,247],"lengths.":[68,184],"They":[69],"also":[70],"require":[71],"costly":[72],"retraining":[73],"recover":[75],"accuracy.":[76],"In":[77],"this":[78,92],"paper,":[79],"we":[80],"present":[81],"DFSS,":[82],"first":[84,152],"GPU-friendly":[85],"dynamic":[86,115,146],"fine-grained":[87,104,132],"pruning":[88,122],"mechanism,":[89],"address":[91],"dilemma.":[93],"DFSS":[94,178,209],"dynamically":[95],"prunes":[96,163],"full":[98,197,216,260],"score":[100],"matrix":[101,127,149,171],"N:M":[103,117],"structured":[105],"sparse":[106,126,170],"pattern.":[107],"Our":[108],"key":[109,161],"insight":[110],"that":[112,156],"side,":[116,133],"friendly":[120],"encoding":[124],"GPU.":[129,240],"On":[130,241],"it":[134],"always":[135],"preserves":[136],"dominant":[138],"entries":[139],"each":[141],"row.":[142],"We":[143,200,219],"develop":[144],"a":[145,188,211,264],"sampled":[147],"dense-dense":[148],"multiplication":[150],"kernel,":[151],"its":[154,253],"kind,":[155],"multiplies":[157],"query":[159],"matrices,":[162],"result,":[165],"encodes":[167],"compressed":[169],"without":[172],"overhead.":[173],"Compared":[174],"with":[175,196,246,258],"studies,":[177],"achieves":[179],"arbitrary":[182],"It":[185],"only":[186,263],"takes":[187],"few":[189],"fine-tuning":[190],"epochs":[191,268],"reach":[193],"on-par":[194],"accuracy":[195,254],"mechanism.":[199,218],"provide":[201],"both":[202],"theoretical":[203],"empirical":[205],"evidence":[206],"demonstrate":[208],"good":[212],"approximation":[213],"evaluate":[220],"1:2":[222],"2:4":[224],"different":[227],"settings":[228],"1.38":[231],"~":[232],"1.86\u00d7":[233],"speedups":[234],"over":[235],"full-attention":[237],"A100":[239],"domains":[245],"lengths":[248],"384":[250],"4096,":[252],"par":[257],"after":[262],"couple":[265],"finetuning":[267],"dense":[271],"pre-trained":[272],"model.":[273]},"counts_by_year":[{"year":2026,"cited_by_count":9},{"year":2025,"cited_by_count":10},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":2}],"updated_date":"2026-08-12T21:12:35.861297","created_date":"2025-10-10T00:00:00"}