Neural network model quantification method
The embodiment of the invention provides a neural network model quantification method, and the method comprises the steps: obtaining a quantification data set, and determining the initial parameter sensitivity feature information of each neural network layer according to the quantification data set;...
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Patent |
| Sprache: | chi ; eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | |
|---|---|
| container_issue | |
| container_start_page | |
| container_title | |
| container_volume | |
| creator | XIA JINPENG ZHANG YUEWEI OU LIN |
| description | The embodiment of the invention provides a neural network model quantification method, and the method comprises the steps: obtaining a quantification data set, and determining the initial parameter sensitivity feature information of each neural network layer according to the quantification data set; generating reference parameter information according to the initial parameter sensitivity feature information corresponding to each neural network layer; quantizing each reference parameter information based on at least one quantization strategy, and generating at least one neural network layer quantization result corresponding to each neural network layer; and determining a neural network model quantization result corresponding to the neural network model according to a target storage threshold value of a target storage device, the initial volume of the neural network model and each neural network layer quantization result corresponding to each neural network layer. The parameter information is generated accordin |
| format | Patent |
| fullrecord | <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_CN118171698A</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>CN118171698A</sourcerecordid><originalsourceid>FETCH-epo_espacenet_CN118171698A3</originalsourceid><addsrcrecordid>eNrjZNDySy0tSsxRyEstKc8vylbIzU9JzVEoLE3MK8lMy0xOLMnMz1PITS3JyE_hYWBNS8wpTuWF0twMim6uIc4euqkF-fGpxQWJyalAU-Kd_QwNLQzNDc0sLRyNiVEDAF7eKYc</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Neural network model quantification method</title><source>esp@cenet</source><creator>XIA JINPENG ; ZHANG YUEWEI ; OU LIN</creator><creatorcontrib>XIA JINPENG ; ZHANG YUEWEI ; OU LIN</creatorcontrib><description>The embodiment of the invention provides a neural network model quantification method, and the method comprises the steps: obtaining a quantification data set, and determining the initial parameter sensitivity feature information of each neural network layer according to the quantification data set; generating reference parameter information according to the initial parameter sensitivity feature information corresponding to each neural network layer; quantizing each reference parameter information based on at least one quantization strategy, and generating at least one neural network layer quantization result corresponding to each neural network layer; and determining a neural network model quantization result corresponding to the neural network model according to a target storage threshold value of a target storage device, the initial volume of the neural network model and each neural network layer quantization result corresponding to each neural network layer. The parameter information is generated accordin</description><language>chi ; eng</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; COUNTING ; ELECTRIC DIGITAL DATA PROCESSING ; PHYSICS</subject><creationdate>2024</creationdate><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240611&DB=EPODOC&CC=CN&NR=118171698A$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,780,885,25564,76547</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240611&DB=EPODOC&CC=CN&NR=118171698A$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>XIA JINPENG</creatorcontrib><creatorcontrib>ZHANG YUEWEI</creatorcontrib><creatorcontrib>OU LIN</creatorcontrib><title>Neural network model quantification method</title><description>The embodiment of the invention provides a neural network model quantification method, and the method comprises the steps: obtaining a quantification data set, and determining the initial parameter sensitivity feature information of each neural network layer according to the quantification data set; generating reference parameter information according to the initial parameter sensitivity feature information corresponding to each neural network layer; quantizing each reference parameter information based on at least one quantization strategy, and generating at least one neural network layer quantization result corresponding to each neural network layer; and determining a neural network model quantization result corresponding to the neural network model according to a target storage threshold value of a target storage device, the initial volume of the neural network model and each neural network layer quantization result corresponding to each neural network layer. The parameter information is generated accordin</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>ELECTRIC DIGITAL DATA PROCESSING</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2024</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZNDySy0tSsxRyEstKc8vylbIzU9JzVEoLE3MK8lMy0xOLMnMz1PITS3JyE_hYWBNS8wpTuWF0twMim6uIc4euqkF-fGpxQWJyalAU-Kd_QwNLQzNDc0sLRyNiVEDAF7eKYc</recordid><startdate>20240611</startdate><enddate>20240611</enddate><creator>XIA JINPENG</creator><creator>ZHANG YUEWEI</creator><creator>OU LIN</creator><scope>EVB</scope></search><sort><creationdate>20240611</creationdate><title>Neural network model quantification method</title><author>XIA JINPENG ; ZHANG YUEWEI ; OU LIN</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_CN118171698A3</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>chi ; eng</language><creationdate>2024</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>ELECTRIC DIGITAL DATA PROCESSING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>XIA JINPENG</creatorcontrib><creatorcontrib>ZHANG YUEWEI</creatorcontrib><creatorcontrib>OU LIN</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>XIA JINPENG</au><au>ZHANG YUEWEI</au><au>OU LIN</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Neural network model quantification method</title><date>2024-06-11</date><risdate>2024</risdate><abstract>The embodiment of the invention provides a neural network model quantification method, and the method comprises the steps: obtaining a quantification data set, and determining the initial parameter sensitivity feature information of each neural network layer according to the quantification data set; generating reference parameter information according to the initial parameter sensitivity feature information corresponding to each neural network layer; quantizing each reference parameter information based on at least one quantization strategy, and generating at least one neural network layer quantization result corresponding to each neural network layer; and determining a neural network model quantization result corresponding to the neural network model according to a target storage threshold value of a target storage device, the initial volume of the neural network model and each neural network layer quantization result corresponding to each neural network layer. The parameter information is generated accordin</abstract><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | |
| ispartof | |
| issn | |
| language | chi ; eng |
| recordid | cdi_epo_espacenet_CN118171698A |
| source | esp@cenet |
| subjects | CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING COUNTING ELECTRIC DIGITAL DATA PROCESSING PHYSICS |
| title | Neural network model quantification method |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-07T10%3A00%3A03IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-epo_EVB&rft_val_fmt=info:ofi/fmt:kev:mtx:patent&rft.genre=patent&rft.au=XIA%20JINPENG&rft.date=2024-06-11&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3ECN118171698A%3C/epo_EVB%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true |