{"id":"https://openalex.org/W4309201220","doi":"https://doi.org/10.48550/arxiv.2211.07715","title":"Fast DistilBERT on CPUs","display_name":"Fast DistilBERT on CPUs","publication_year":2022,"publication_date":"2022-10-27","ids":{"openalex":"https://openalex.org/W4309201220","doi":"https://doi.org/10.48550/arxiv.2211.07715"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2211.07715","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2211.07715","pdf_url":"https://arxiv.org/pdf/2211.07715","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,"listed_in":[],"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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2211.07715","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5056650480","display_name":"Haihao Shen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shen, Haihao","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079726843","display_name":"Ofir Zafrir","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zafrir, Ofir","raw_affiliation_strings":["Intel Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Labs","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100746742","display_name":"Bo Dong","orcid":"https://orcid.org/0000-0001-9189-9506","observed_orcids":["https://orcid.org/0000-0001-9189-9506"]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dong, Bo","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091105633","display_name":"Hengyu Meng","orcid":"https://orcid.org/0009-0005-8961-3982","observed_orcids":["https://orcid.org/0009-0005-8961-3982"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meng, Hengyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101818638","display_name":"Xinyu Ye","orcid":"https://orcid.org/0000-0003-0077-7058","observed_orcids":["https://orcid.org/0000-0003-0077-7058"]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ye, Xinyu","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100407640","display_name":"Zhe Wang","orcid":"https://orcid.org/0000-0002-8333-4748","observed_orcids":["https://orcid.org/0000-0002-8333-4748"]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wang, Zhe","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100752793","display_name":"Yi Ding","orcid":"https://orcid.org/0009-0009-2986-8501","observed_orcids":["https://orcid.org/0009-0009-2986-8501","https://orcid.org/0009-0004-2461-264X"]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ding, Yi","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027881296","display_name":"Hanwen Chang","orcid":"https://orcid.org/0000-0002-9641-8457","observed_orcids":["https://orcid.org/0000-0002-9641-8457"]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chang, Hanwen","raw_affiliation_strings":["Intel Corporation"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Corporation","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067212276","display_name":"Guy Boudoukh","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Boudoukh, Guy","raw_affiliation_strings":["Intel Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Labs","institution_ids":["https://openalex.org/I1343180700"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5088026371","display_name":"Moshe Wasserblat","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1343180700","display_name":"Intel (United States)","ror":"https://ror.org/01ek73717","country_code":"US","type":"company","lineage":["https://openalex.org/I1343180700"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wasserblat, Moshe","raw_affiliation_strings":["Intel Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Intel Labs","institution_ids":["https://openalex.org/I1343180700"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9771000146865845,"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":0.9771000146865845,"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/T10028","display_name":"Topic Modeling","score":0.01489999983459711,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.0032999999821186066,"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/knowledge-distillation","display_name":"knowledge distillation","score":0.6809999942779541},{"id":"https://openalex.org/keywords/quantization","display_name":"quantization","score":0.8920000195503235},{"id":"https://openalex.org/keywords/transformer-inference","display_name":"transformer inference","score":0.6359999775886536},{"id":"https://openalex.org/keywords/cpu-inference","display_name":"CPU-only inference","score":0.8709999918937683},{"id":"https://openalex.org/keywords/sparse-operators","display_name":"sparse operators","score":0.578000009059906},{"id":"https://openalex.org/keywords/inference-throughput","display_name":"inference throughput","score":0.8100000023841858},{"id":"https://openalex.org/keywords/question-answering","display_name":"question answering","score":0.6690000295639038},{"id":"https://openalex.org/keywords/squad","display_name":"SQuAD","score":0.7950000166893005},{"id":"https://openalex.org/keywords/onnx-runtime","display_name":"ONNX Runtime","score":0.8999999761581421}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8328582048416138},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.7020425200462341},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.57161945104599},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.48408782482147217},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.47484996914863586},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.47202152013778687},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.44350534677505493},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.41431325674057007},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.22462499141693115},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.1946050524711609},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.13422101736068726},{"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/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2211.07715","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2211.07715","pdf_url":"https://arxiv.org/pdf/2211.07715","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,"listed_in":[],"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"},{"id":"doi:10.48550/arxiv.2211.07715","doi":"https://doi.org/10.48550/arxiv.2211.07715","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2211.07715","pdf_url":null,"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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2211.07715","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2211.07715","pdf_url":"https://arxiv.org/pdf/2211.07715","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,"listed_in":[],"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":[],"sustainable_development_goals_aurora":[{"score":0.4699999988079071,"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":0,"referenced_works":[],"related_works":["https://openalex.org/W2058965144","https://openalex.org/W2164382479","https://openalex.org/W2146343568","https://openalex.org/W98480971","https://openalex.org/W2097707447","https://openalex.org/W4206178588","https://openalex.org/W3094491777","https://openalex.org/W3214715529","https://openalex.org/W4287635093","https://openalex.org/W2031349701"],"abstract_inverted_index":{"Transformer-based":[0],"language":[1,11],"models":[2,31,83],"have":[3],"become":[4],"the":[5,19,109,125],"standard":[6],"approach":[7],"to":[8,22,52,63,150,154],"solving":[9],"natural":[10],"processing":[12],"tasks.":[13],"However,":[14,56],"industry":[15],"adoption":[16],"usually":[17],"requires":[18],"maximum":[20],"throughput":[21,130],"comply":[23],"with":[24,99],"certain":[25],"latency":[26],"constraints":[27,135],"that":[28],"prevents":[29],"Transformer":[30,82,95],"from":[32],"being":[33],"used":[34,51],"in":[35],"production.":[36],"To":[37],"address":[38],"this":[39,70],"gap,":[40],"model":[41,119],"compression":[42,58],"techniques":[43,59],"such":[44],"as":[45],"quantization":[46],"and":[47,65,79,92,104,129,136,152],"pruning":[48],"may":[49],"be":[50],"improve":[53],"inference":[54,96],"efficiency.":[55],"these":[57],"require":[60],"specialized":[61],"software":[62],"apply":[64],"deploy":[66],"at":[67,166],"scale.":[68],"In":[69],"work,":[71],"we":[72],"propose":[73],"a":[74,116],"new":[75],"pipeline":[76,113],"for":[77,102],"creating":[78,115],"running":[80],"Fast":[81,117],"on":[84,124],"CPUs,":[85],"utilizing":[86],"hardware-aware":[87],"pruning,":[88],"knowledge":[89],"distillation,":[90],"quantization,":[91],"our":[93,112],"own":[94],"runtime":[97,146],"engine":[98],"optimized":[100],"kernels":[101],"sparse":[103],"quantized":[105],"operators.":[106],"We":[107],"demonstrate":[108],"efficiency":[110],"of":[111],"by":[114,148],"DistilBERT":[118],"showing":[120],"minimal":[121],"accuracy":[122],"loss":[123],"question-answering":[126],"SQuADv1.1":[127],"benchmark,":[128],"results":[131,139],"under":[132],"typical":[133],"production":[134],"environments.":[137],"Our":[138],"outperform":[140],"existing":[141],"state-of-the-art":[142],"Neural":[143],"Magic's":[144],"DeepSparse":[145],"performance":[147,156],"up":[149,153],"50%":[151],"4.1x":[155],"speedup":[157],"over":[158],"ONNX":[159],"Runtime.":[160],"Source":[161],"code":[162],"is":[163],"publicly":[164],"available":[165],"https://github.com/intel/intel-extension-for-transformers.":[167]},"counts_by_year":[],"updated_date":"2026-09-30T09:27:27.532063","created_date":"2025-10-10T00:00:00"}