{"id":"https://openalex.org/W4385488663","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191067","title":"A Scalable GPT-2 Inference Hardware Architecture on FPGA","display_name":"A Scalable GPT-2 Inference Hardware Architecture on FPGA","publication_year":2023,"publication_date":"2023-06-18","ids":{"openalex":"https://openalex.org/W4385488663","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191067"},"language":"en","primary_location":{"id":"doi:10.1109/ijcnn54540.2023.10191067","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191067","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn54540.2023.10191067","pdf_url":null,"source":{"id":"https://openalex.org/S4363607707","display_name":"International Joint Conference on Neural Networks (IJCNN)","issn_l":"2161-4407","issn":["2161-4407"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["jufo-1","norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","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/A5092582455","display_name":"Anil Yemme","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I59270414","display_name":"Indian Institute of Science Bangalore","ror":"https://ror.org/04dese585","country_code":"IN","type":"education","lineage":["https://openalex.org/I59270414"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Anil Yemme","raw_affiliation_strings":["Indian Institute of Science,Dept. of Electronics Systems Engineering,Bengaluru,India","Dept. of Electronics Systems Engineering, Indian Institute of Science, Bengaluru, India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Indian Institute of Science,Dept. of Electronics Systems Engineering,Bengaluru,India","institution_ids":["https://openalex.org/I59270414"]},{"raw_affiliation_string":"Dept. of Electronics Systems Engineering, Indian Institute of Science, Bengaluru, India","institution_ids":["https://openalex.org/I59270414"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5055106245","display_name":"Shayan Srinivasa Garani","orcid":"https://orcid.org/0000-0002-2459-1445","observed_orcids":["https://orcid.org/0000-0002-2459-1445"]},"institutions":[{"id":"https://openalex.org/I59270414","display_name":"Indian Institute of Science Bangalore","ror":"https://ror.org/04dese585","country_code":"IN","type":"education","lineage":["https://openalex.org/I59270414"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Shayan Srinivasa Garani","raw_affiliation_strings":["Indian Institute of Science,Dept. of Electronics Systems Engineering,Bengaluru,India","Dept. of Electronics Systems Engineering, Indian Institute of Science, Bengaluru, India"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Indian Institute of Science,Dept. of Electronics Systems Engineering,Bengaluru,India","institution_ids":["https://openalex.org/I59270414"]},{"raw_affiliation_string":"Dept. of Electronics Systems Engineering, Indian Institute of Science, Bengaluru, India","institution_ids":["https://openalex.org/I59270414"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I59270414"],"apc_list":null,"apc_paid":null,"fwci":1.07,"has_fulltext":true,"cited_by_count":6,"citation_normalized_percentile":{"value":0.79357256,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":89,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.5282999873161316,"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"}},"topics":[{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.5282999873161316,"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"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.3292999863624573,"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/T13292","display_name":"Embedded Systems and FPGA Applications","score":0.061500001698732376,"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/gpt-2","display_name":"GPT-2","score":0.9919999837875366},{"id":"https://openalex.org/keywords/fpga-implementation","display_name":"FPGA implementation","score":0.7020000219345093},{"id":"https://openalex.org/keywords/hardware-architecture","display_name":"hardware architecture","score":0.722000002861023},{"id":"https://openalex.org/keywords/transformer-inference","display_name":"transformer inference","score":0.6729999780654907},{"id":"https://openalex.org/keywords/edge-computing","display_name":"edge computing","score":0.8859999775886536},{"id":"https://openalex.org/keywords/attention-mechanism","display_name":"attention mechanism","score":0.6380000114440918},{"id":"https://openalex.org/keywords/scalable-hardware-design","display_name":"scalable hardware design","score":0.40700000524520874},{"id":"https://openalex.org/keywords/latency","display_name":"latency","score":0.34200000762939453},{"id":"https://openalex.org/keywords/virtex-7","display_name":"Virtex-7","score":0.7009999752044678},{"id":"https://openalex.org/keywords/hardware-acceleration","display_name":"hardware acceleration","score":0.3449999988079071}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8108346462249756},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6340900659561157},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.6316273212432861},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.6192693710327148},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.526984691619873},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4990572929382324},{"id":"https://openalex.org/C52173422","wikidata":"https://www.wikidata.org/wiki/Q766483","display_name":"Opcode","level":2,"score":0.4310725927352905},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.42677462100982666},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.41824740171432495},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.40840578079223633},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.3704991936683655},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.37012046575546265},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32078081369400024},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2807084321975708},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.14647254347801208},{"id":"https://openalex.org/C153349607","wikidata":"https://www.wikidata.org/wiki/Q36649","display_name":"Visual arts","level":1,"score":0.0},{"id":"https://openalex.org/C142362112","wikidata":"https://www.wikidata.org/wiki/Q735","display_name":"Art","level":0,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/ijcnn54540.2023.10191067","doi":"https://doi.org/10.1109/ijcnn54540.2023.10191067","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ijcnn54540.2023.10191067","pdf_url":null,"source":{"id":"https://openalex.org/S4363607707","display_name":"International Joint Conference on Neural Networks (IJCNN)","issn_l":"2161-4407","issn":["2161-4407"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["jufo-1","norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2023 International Joint Conference on Neural Networks (IJCNN)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.5378999710083008,"id":"https://metadata.un.org/sdg/9","display_name":"Industry, innovation and infrastructure"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":13,"referenced_works":["https://openalex.org/W1614298861","https://openalex.org/W1632114991","https://openalex.org/W2064675550","https://openalex.org/W2194775991","https://openalex.org/W2525332836","https://openalex.org/W2608093348","https://openalex.org/W2626778328","https://openalex.org/W2799715319","https://openalex.org/W2896457183","https://openalex.org/W2963341956","https://openalex.org/W3037932933","https://openalex.org/W3162542754","https://openalex.org/W3196923642"],"related_works":["https://openalex.org/W4411725728","https://openalex.org/W7162414827","https://openalex.org/W7161142939","https://openalex.org/W4396653497","https://openalex.org/W4406265955","https://openalex.org/W7162443579","https://openalex.org/W7204039812","https://openalex.org/W7167262144","https://openalex.org/W7163904164","https://openalex.org/W4413120711"],"abstract_inverted_index":{"Transformer-based":[0],"architectures":[1,10,17,118],"using":[2],"attention":[3],"mechanisms":[4],"are":[5],"a":[6,35,108,128],"class":[7],"of":[8,110,131,149,153,189],"learning":[9,38],"for":[11,119],"sequence":[12,37],"processing":[13],"tasks.":[14],"These":[15],"include":[16],"such":[18],"as":[19],"the":[20,26,51],"generative":[21],"pretrained":[22],"transformer":[23,42],"(GPT)":[24],"and":[25,50,69,103,113,151,164],"bidirectional":[27],"encoder":[28],"representations":[29],"from":[30],"transformers":[31],"(BERT).":[32],"GPT-2":[33,44,111,132],"is":[34,45,74,81,92,136,185],"popular":[36],"architecture":[39,135],"that":[40],"uses":[41],"architecture.":[43,178],"trained":[46],"on":[47,83,94,122,138,176],"text":[48,67],"prediction,":[49],"network":[52],"parameters":[53],"obtained":[54],"during":[55],"this":[56,126,183],"training":[57,80],"process":[58],"can":[59],"be":[60],"used":[61],"in":[62,78,182],"various":[63],"other":[64],"tasks":[65],"like":[66,97],"classification":[68],"premise-hypothesis":[70],"testing.":[71],"Edge":[72],"computing":[73],"an":[75],"recent":[76],"trend":[77],"which":[79],"done":[82,93],"cloud":[84],"or":[85],"server":[86],"with":[87],"multiple":[88],"GPUs,":[89],"but":[90],"inference":[91,134,144],"edge":[95,123],"devices":[96],"mobile":[98],"phones":[99],"to":[100,115,172,187],"reduce":[101],"latency":[102,152],"improve":[104],"privacy.":[105],"This":[106],"necessitates":[107],"study":[109],"performance":[112],"complexity":[114],"distill":[116],"hardware-based":[117],"their":[120],"usability":[121],"devices.":[124],"In":[125],"paper,":[127],"single":[129],"layer":[130],"based":[133],"implemented":[137],"Virtex-7":[139],"xc7vx485tffg1761-2":[140],"FPGA":[141],"board.":[142],"The":[143,179],"engine":[145],"has":[146],"model":[147],"dimensionality":[148],"128":[150],"1.637":[154],"ms":[155],"while":[156],"operating":[157],"at":[158],"142.44":[159],"MHz,":[160],"consuming":[161],"85.6K":[162],"flip-flops":[163],"96.8K":[165],"lookup":[166],"tables,":[167],"achieving":[168],"1.73x":[169],"speedup":[170],"compared":[171],"previously":[173],"reported":[174],"work":[175],"transformer-based":[177],"approach":[180],"proposed":[181],"paper":[184],"scalable":[186],"models":[188],"higher":[190],"dimensionality.":[191]},"counts_by_year":[{"year":2025,"cited_by_count":5},{"year":2024,"cited_by_count":1}],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2023-08-03T00:00:00"}