{"id":"https://openalex.org/W7148570079","doi":"https://doi.org/10.1109/asru65441.2025.11434664","title":"Efficient Deployment of Large Speech Recognition Models on GPU","display_name":"Efficient Deployment of Large Speech Recognition Models on GPU","publication_year":2025,"publication_date":"2025-12-06","ids":{"openalex":"https://openalex.org/W7148570079","doi":"https://doi.org/10.1109/asru65441.2025.11434664"},"language":null,"primary_location":{"id":"doi:10.1109/asru65441.2025.11434664","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asru65441.2025.11434664","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)","raw_type":"proceedings-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/A5074415592","display_name":"Yuekai Zhang","orcid":null},"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":true,"raw_author_name":"Yuekai Zhang","raw_affiliation_strings":["Nvidia Corporation"],"affiliations":[{"raw_affiliation_string":"Nvidia Corporation","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132769168","display_name":"Shuang Yu","orcid":null},"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":"Shuang Yu","raw_affiliation_strings":["Nvidia Corporation"],"affiliations":[{"raw_affiliation_string":"Nvidia Corporation","institution_ids":["https://openalex.org/I4210127875"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5103287856","display_name":"Junjie Lai","orcid":null},"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":"Junjie Lai","raw_affiliation_strings":["Nvidia Corporation"],"affiliations":[{"raw_affiliation_string":"Nvidia Corporation","institution_ids":["https://openalex.org/I4210127875"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5074415592"],"corresponding_institution_ids":["https://openalex.org/I4210127875"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.87577898,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"4"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.36730000376701355,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.36730000376701355,"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/T10860","display_name":"Speech and Audio Processing","score":0.17409999668598175,"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/T10057","display_name":"Face and Expression Recognition","score":0.04399999976158142,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.4959999918937683},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.31929999589920044},{"id":"https://openalex.org/keywords/signal-processing","display_name":"Signal processing","score":0.2980000078678131},{"id":"https://openalex.org/keywords/voice-activity-detection","display_name":"Voice activity detection","score":0.29100000858306885},{"id":"https://openalex.org/keywords/speaker-recognition","display_name":"Speaker recognition","score":0.2888000011444092},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.2851000130176544}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7175999879837036},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.5047000050544739},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4959999918937683},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4499000012874603},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3441999852657318},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.31929999589920044},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.2980000078678131},{"id":"https://openalex.org/C204201278","wikidata":"https://www.wikidata.org/wiki/Q1332614","display_name":"Voice activity detection","level":3,"score":0.29100000858306885},{"id":"https://openalex.org/C133892786","wikidata":"https://www.wikidata.org/wiki/Q1145189","display_name":"Speaker recognition","level":2,"score":0.2888000011444092},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.2851000130176544},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2685999870300293},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C117623542","wikidata":"https://www.wikidata.org/wiki/Q621974","display_name":"Automatic target recognition","level":3,"score":0.25870001316070557}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/asru65441.2025.11434664","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asru65441.2025.11434664","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":12,"referenced_works":["https://openalex.org/W2963242190","https://openalex.org/W3203407300","https://openalex.org/W4312120634","https://openalex.org/W4319862255","https://openalex.org/W4385245566","https://openalex.org/W4385822293","https://openalex.org/W4387321091","https://openalex.org/W4391021542","https://openalex.org/W4391021675","https://openalex.org/W4402111558","https://openalex.org/W4402112192","https://openalex.org/W4415433449"],"related_works":[],"abstract_inverted_index":{"Large":[0],"automatic":[1],"speech":[2,58,87],"recognition":[3],"(ASR)":[4],"models":[5,40,85],"have":[6],"achieved":[7],"remarkable":[8],"progress":[9],"in":[10,16,120],"recent":[11],"years,":[12],"but":[13],"their":[14],"deployment":[15,51,107,114],"production":[17,121],"faces":[18],"significant":[19],"challenges":[20],"due":[21],"to":[22,74,90],"large":[23,38],"model":[24,76],"sizes":[25],"and":[26,48,57,86,111],"autoregressive":[27],"decoding":[28],"methods.":[29],"This":[30],"paper":[31],"presents":[32],"comprehensive":[33],"solutions":[34,108],"for":[35,83],"efficiently":[36],"deploying":[37],"ASR":[39],"on":[41,65],"GPUs":[42],"using":[43],"NVIDIA":[44,66],"Triton":[45],"Inference":[46],"Server":[47],"TensorRT-LLM.":[49],"Our":[50],"framework":[52,69],"supports":[53],"both":[54],"encoder-decoder":[55],"architectures":[56],"LLMs.":[59,88],"With":[60],"a":[61],"modular":[62],"design":[63],"based":[64],"Triton,":[67],"the":[68],"can":[70],"be":[71],"easily":[72],"extended":[73],"other":[75],"architectures.":[77],"We":[78],"implement":[79],"optimized":[80],"TensorRT-LLM":[81],"engines":[82],"Whisper":[84,94],"Compared":[89],"existing":[91],"implementations,":[92],"our":[93],"TensorRTLLM":[95],"solution":[96],"achieves":[97],"more":[98],"than":[99],"$\\mathbf{5":[100],"0":[101],"\\%}$":[102],"throughput":[103],"improvement.":[104],"The":[105],"complete":[106],"are":[109],"open-sourced":[110],"provide":[112],"one-click":[113],"through":[115],"docker-compose,":[116],"facilitating":[117],"rapid":[118],"adoption":[119],"environments.":[122],"<sup":[123,126],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[124,127,129,131],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">1</sup>":[125],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">2</sup><sup":[128],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">1</sup>https://github.com/k2-fsa/sherpa/tree/master/triton/whisper<sup":[130],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">2</sup>https://github.com/k2-fsa/sherpa/tree/master/triton/speech_llm":[132]},"counts_by_year":[],"updated_date":"2026-04-03T16:44:17.987007","created_date":"2026-04-03T00:00:00"}