{"id":"https://openalex.org/W4388032243","doi":"https://doi.org/10.1145/3613424.3614307","title":"PockEngine: Sparse and Efficient Fine-tuning in a Pocket","display_name":"PockEngine: Sparse and Efficient Fine-tuning in a Pocket","publication_year":2023,"publication_date":"2023-10-28","ids":{"openalex":"https://openalex.org/W4388032243","doi":"https://doi.org/10.1145/3613424.3614307"},"language":"en","primary_location":{"id":"doi:10.1145/3613424.3614307","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614307","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614307","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614307","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5047194942","display_name":"Ligeng Zhu","orcid":"https://orcid.org/0000-0002-6969-1706"},"institutions":[{"id":"https://openalex.org/I1310837484","display_name":"Association for Asian Studies","ror":"https://ror.org/05txqkd02","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1310837484"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Ligeng Zhu","raw_affiliation_strings":["EECS, MIT, USA"],"raw_orcid":"https://orcid.org/0000-0002-6969-1706","affiliations":[{"raw_affiliation_string":"EECS, MIT, USA","institution_ids":["https://openalex.org/I1310837484"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055638481","display_name":"Lanxiang Hu","orcid":"https://orcid.org/0000-0003-0641-3677"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lanxiang Hu","raw_affiliation_strings":["UCSD, USA"],"raw_orcid":"https://orcid.org/0000-0003-0641-3677","affiliations":[{"raw_affiliation_string":"UCSD, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065001783","display_name":"Ji Lin","orcid":"https://orcid.org/0000-0001-6053-4344"},"institutions":[{"id":"https://openalex.org/I1310837484","display_name":"Association for Asian Studies","ror":"https://ror.org/05txqkd02","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1310837484"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ji Lin","raw_affiliation_strings":["EECS, MIT, USA"],"raw_orcid":"https://orcid.org/0000-0001-6053-4344","affiliations":[{"raw_affiliation_string":"EECS, MIT, USA","institution_ids":["https://openalex.org/I1310837484"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100624432","display_name":"Wei-Ming Chen","orcid":"https://orcid.org/0000-0001-7177-4167"},"institutions":[{"id":"https://openalex.org/I1310837484","display_name":"Association for Asian Studies","ror":"https://ror.org/05txqkd02","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1310837484"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wei-Ming Chen","raw_affiliation_strings":["EECS, MIT, USA"],"raw_orcid":"https://orcid.org/0000-0001-7177-4167","affiliations":[{"raw_affiliation_string":"EECS, MIT, USA","institution_ids":["https://openalex.org/I1310837484"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100685804","display_name":"Wei-Chen Wang","orcid":"https://orcid.org/0000-0002-9435-6598"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei-Chen Wang","raw_affiliation_strings":["EECS, MIT, United States"],"raw_orcid":"https://orcid.org/0000-0002-9435-6598","affiliations":[{"raw_affiliation_string":"EECS, MIT, United States","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040877128","display_name":"Chuang Gan","orcid":"https://orcid.org/0000-0003-4031-5886"},"institutions":[{"id":"https://openalex.org/I1341412227","display_name":"IBM (United States)","ror":"https://ror.org/05hh8d621","country_code":"US","type":"company","lineage":["https://openalex.org/I1341412227"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Chuang Gan","raw_affiliation_strings":["EECS, MIT-IBM Watson AI Lab, USA"],"raw_orcid":"https://orcid.org/0000-0003-4031-5886","affiliations":[{"raw_affiliation_string":"EECS, MIT-IBM Watson AI Lab, USA","institution_ids":["https://openalex.org/I1341412227"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5070926896","display_name":"Song Han","orcid":"https://orcid.org/0000-0002-4186-7618"},"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":"Song Han","raw_affiliation_strings":["EECS, MIT, USA and NVIDIA, USA"],"raw_orcid":"https://orcid.org/0000-0002-4186-7618","affiliations":[{"raw_affiliation_string":"EECS, MIT, USA and NVIDIA, USA","institution_ids":["https://openalex.org/I4210127875"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5047194942"],"corresponding_institution_ids":["https://openalex.org/I1310837484"],"apc_list":null,"apc_paid":null,"fwci":1.5307,"has_fulltext":true,"cited_by_count":13,"citation_normalized_percentile":{"value":0.85354076,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":96,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1381","last_page":"1394"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9976000189781189,"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":0.9976000189781189,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.9890999794006348,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10273","display_name":"IoT and Edge/Fog Computing","score":0.9882000088691711,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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":0.8658803701400757},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.595318615436554},{"id":"https://openalex.org/keywords/graph","display_name":"Graph","score":0.4989960193634033},{"id":"https://openalex.org/keywords/server","display_name":"Server","score":0.47868281602859497},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.46859943866729736},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.44343894720077515},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.42224496603012085},{"id":"https://openalex.org/keywords/compiler","display_name":"Compiler","score":0.4137239456176758},{"id":"https://openalex.org/keywords/cloud-computing","display_name":"Cloud computing","score":0.3378257751464844},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.21789300441741943},{"id":"https://openalex.org/keywords/operating-system","display_name":"Operating system","score":0.21141907572746277}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8658803701400757},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.595318615436554},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.4989960193634033},{"id":"https://openalex.org/C93996380","wikidata":"https://www.wikidata.org/wiki/Q44127","display_name":"Server","level":2,"score":0.47868281602859497},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.46859943866729736},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.44343894720077515},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.42224496603012085},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.4137239456176758},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.3378257751464844},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.21789300441741943},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.21141907572746277}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.1145/3613424.3614307","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614307","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614307","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2310.17752","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.17752","pdf_url":"https://arxiv.org/pdf/2310.17752","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"pmh:oai:dspace.mit.edu:1721.1/153267","is_oa":true,"landing_page_url":"https://hdl.handle.net/1721.1/153267","pdf_url":"https://dspace.mit.edu/bitstream/1721.1/153267/1/3613424.3614307.pdf","source":{"id":"https://openalex.org/S4306400425","display_name":"DSpace@MIT (Massachusetts Institute of Technology)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I63966007","host_organization_name":"Massachusetts Institute of Technology","host_organization_lineage":["https://openalex.org/I63966007"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"Article"}],"best_oa_location":{"id":"doi:10.1145/3613424.3614307","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3613424.3614307","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3613424.3614307","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"56th Annual IEEE/ACM International Symposium on Microarchitecture","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4388032243.pdf","grobid_xml":"https://content.openalex.org/works/W4388032243.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W12634471","https://openalex.org/W95608104","https://openalex.org/W1566289585","https://openalex.org/W1977295328","https://openalex.org/W2108598243","https://openalex.org/W2138011018","https://openalex.org/W2533598788","https://openalex.org/W2963163009","https://openalex.org/W2963173190","https://openalex.org/W2968103728","https://openalex.org/W2981758446","https://openalex.org/W3037749908","https://openalex.org/W4385572634","https://openalex.org/W6713134421"],"related_works":["https://openalex.org/W2058965144","https://openalex.org/W2164382479","https://openalex.org/W2146343568","https://openalex.org/W98480971","https://openalex.org/W2150291671","https://openalex.org/W2013643406","https://openalex.org/W2027972911","https://openalex.org/W2157978810","https://openalex.org/W2778498407","https://openalex.org/W1966837078"],"abstract_inverted_index":{"On-device":[0],"learning":[1,39],"and":[2,7,34,49,59,78,87,107,119,145,153,159,165,178],"efficient":[3,60],"fine-tuning":[4,12,64,205],"enable":[5,63],"continuous":[6],"privacy-preserving":[8],"customization":[9],"(e.g.,":[10,31],"locally":[11],"large":[13,179],"language":[14,180],"models":[15,161,177],"on":[16,40,65,174,207],"personalized":[17],"data).":[18],"However,":[19],"existing":[20],"training":[21,102,132,140],"frameworks":[22],"are":[23],"designed":[24],"for":[25,38,122],"cloud":[26],"servers":[27],"with":[28,83],"powerful":[29],"accelerators":[30],"GPUs,":[32],"TPUs)":[33],"lack":[35],"the":[36,41,75,81,92,100,116,139,219],"optimizations":[37],"edge,":[42],"which":[43,114],"faces":[44],"challenges":[45],"of":[46,131],"resource":[47],"limitations":[48],"edge":[50,67],"hardware":[51,154],"diversity.":[52],"We":[53,171],"introduce":[54],"PockEngine:":[55],"a":[56,128],"tiny,":[57],"sparse":[58,71],"engine":[61],"to":[62,168,185],"various":[66],"devices.":[68],"PockEngine":[69,96,125,148,173,182,203],"supports":[70,149],"backpropagation:":[72],"it":[73,156],"prunes":[74],"backward":[76,106],"graph":[77,103,123,133],"sparsely":[79],"updates":[80],"model":[82,93],"measured":[84],"memory":[85,196],"saving":[86,197],"latency":[88],"reduction":[89],"while":[90],"maintaining":[91],"quality.":[94],"Secondly,":[95],"is":[97,110],"compilation":[98],"first:":[99],"entire":[101],"(including":[104],"forward,":[105],"optimization":[108],"steps)":[109],"derived":[111],"at":[112,212],"compile-time,":[113],"reduces":[115],"runtime":[117],"overhead":[118],"brings":[120],"opportunities":[121],"transformations.":[124],"also":[126],"integrates":[127],"rich":[129],"set":[130],"optimizations,":[134],"thus":[135],"can":[136],"further":[137],"accelerate":[138],"cost,":[141],"including":[142],"operator":[143],"reordering":[144],"backend":[146],"switching.":[147],"diverse":[150],"applications,":[151],"frontends":[152],"backends:":[155],"flexibly":[157],"compiles":[158],"tunes":[160],"defined":[162],"in":[163],"PyTorch/TensorFlow/Jax":[164],"deploys":[166],"binaries":[167],"mobile":[169],"CPU/GPU/DSPs.":[170],"evaluated":[172],"both":[175],"vision":[176],"models.":[181],"achieves":[183],"up":[184],"15":[186],"\u00d7":[187,195,216],"speedup":[188],"over":[189],"off-the-shelf":[190],"TensorFlow":[191],"(Raspberry":[192],"Pi),":[193],"5.6":[194],"back-propagation":[198],"(Jetson":[199],"AGX":[200,210],"Orin).":[201],"Remarkably,":[202],"enables":[204],"LLaMav2-7B":[206],"NVIDIA":[208],"Jetson":[209],"Orin":[211],"550":[213],"tokens/s,":[214],"7.9":[215],"faster":[217],"than":[218],"PyTorch.":[220]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":8}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}