Automatic Memory Management for Compute Graphs
A method includes obtaining a compute graph for computing a first tensor, identifying in the graph a reduction operation in at least one dimension of the first tensor, locating, at the operation, a cut point that cuts the graph into first and second portions, and determining a plurality of slices of...
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Patent |
| Sprache: | 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 | Rizman, Matej Shmakov, Pavel Navarro Perez, Juan Antonio Naryanto, Vinsensius B. Vega S Chimisov, Cyril Saxena, Ashish |
| description | A method includes obtaining a compute graph for computing a first tensor, identifying in the graph a reduction operation in at least one dimension of the first tensor, locating, at the operation, a cut point that cuts the graph into first and second portions, and determining a plurality of slices of the first tensor. The method also includes backpropagating the cut point through the graph to define a plurality of first graph pieces for the first portion, each particular first graph piece representing a computation of a particular slice of the plurality of slices based on a particular portion of a plurality of portions of a second tensor. The method further includes defining one or more second graph pieces to combine outputs of the first graph pieces, and executing the first graph pieces and the second graph pieces to execute the first portion of the compute graph. |
| format | Patent |
| fullrecord | <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_US2024193421A1</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>US2024193421A1</sourcerecordid><originalsourceid>FETCH-epo_espacenet_US2024193421A13</originalsourceid><addsrcrecordid>eNrjZNBzLC3Jz00syUxW8E3NzS-qVPBNzEtMT81NzStRSMsvUnDOzy0oLUlVcC9KLMgo5mFgTUvMKU7lhdLcDMpuriHOHrqpBfnxqcUFicmpeakl8aHBRgZGJoaWxiZGho6GxsSpAgCRxytH</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Automatic Memory Management for Compute Graphs</title><source>esp@cenet</source><creator>Rizman, Matej ; Shmakov, Pavel ; Navarro Perez, Juan Antonio ; Naryanto, Vinsensius B. Vega S ; Chimisov, Cyril ; Saxena, Ashish</creator><creatorcontrib>Rizman, Matej ; Shmakov, Pavel ; Navarro Perez, Juan Antonio ; Naryanto, Vinsensius B. Vega S ; Chimisov, Cyril ; Saxena, Ashish</creatorcontrib><description>A method includes obtaining a compute graph for computing a first tensor, identifying in the graph a reduction operation in at least one dimension of the first tensor, locating, at the operation, a cut point that cuts the graph into first and second portions, and determining a plurality of slices of the first tensor. The method also includes backpropagating the cut point through the graph to define a plurality of first graph pieces for the first portion, each particular first graph piece representing a computation of a particular slice of the plurality of slices based on a particular portion of a plurality of portions of a second tensor. The method further includes defining one or more second graph pieces to combine outputs of the first graph pieces, and executing the first graph pieces and the second graph pieces to execute the first portion of the compute graph.</description><language>eng</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; COUNTING ; 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=20240613&DB=EPODOC&CC=US&NR=2024193421A1$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25542,76290</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240613&DB=EPODOC&CC=US&NR=2024193421A1$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>Rizman, Matej</creatorcontrib><creatorcontrib>Shmakov, Pavel</creatorcontrib><creatorcontrib>Navarro Perez, Juan Antonio</creatorcontrib><creatorcontrib>Naryanto, Vinsensius B. Vega S</creatorcontrib><creatorcontrib>Chimisov, Cyril</creatorcontrib><creatorcontrib>Saxena, Ashish</creatorcontrib><title>Automatic Memory Management for Compute Graphs</title><description>A method includes obtaining a compute graph for computing a first tensor, identifying in the graph a reduction operation in at least one dimension of the first tensor, locating, at the operation, a cut point that cuts the graph into first and second portions, and determining a plurality of slices of the first tensor. The method also includes backpropagating the cut point through the graph to define a plurality of first graph pieces for the first portion, each particular first graph piece representing a computation of a particular slice of the plurality of slices based on a particular portion of a plurality of portions of a second tensor. The method further includes defining one or more second graph pieces to combine outputs of the first graph pieces, and executing the first graph pieces and the second graph pieces to execute the first portion of the compute graph.</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2024</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZNBzLC3Jz00syUxW8E3NzS-qVPBNzEtMT81NzStRSMsvUnDOzy0oLUlVcC9KLMgo5mFgTUvMKU7lhdLcDMpuriHOHrqpBfnxqcUFicmpeakl8aHBRgZGJoaWxiZGho6GxsSpAgCRxytH</recordid><startdate>20240613</startdate><enddate>20240613</enddate><creator>Rizman, Matej</creator><creator>Shmakov, Pavel</creator><creator>Navarro Perez, Juan Antonio</creator><creator>Naryanto, Vinsensius B. Vega S</creator><creator>Chimisov, Cyril</creator><creator>Saxena, Ashish</creator><scope>EVB</scope></search><sort><creationdate>20240613</creationdate><title>Automatic Memory Management for Compute Graphs</title><author>Rizman, Matej ; Shmakov, Pavel ; Navarro Perez, Juan Antonio ; Naryanto, Vinsensius B. Vega S ; Chimisov, Cyril ; Saxena, Ashish</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_US2024193421A13</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng</language><creationdate>2024</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>Rizman, Matej</creatorcontrib><creatorcontrib>Shmakov, Pavel</creatorcontrib><creatorcontrib>Navarro Perez, Juan Antonio</creatorcontrib><creatorcontrib>Naryanto, Vinsensius B. Vega S</creatorcontrib><creatorcontrib>Chimisov, Cyril</creatorcontrib><creatorcontrib>Saxena, Ashish</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Rizman, Matej</au><au>Shmakov, Pavel</au><au>Navarro Perez, Juan Antonio</au><au>Naryanto, Vinsensius B. Vega S</au><au>Chimisov, Cyril</au><au>Saxena, Ashish</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Automatic Memory Management for Compute Graphs</title><date>2024-06-13</date><risdate>2024</risdate><abstract>A method includes obtaining a compute graph for computing a first tensor, identifying in the graph a reduction operation in at least one dimension of the first tensor, locating, at the operation, a cut point that cuts the graph into first and second portions, and determining a plurality of slices of the first tensor. The method also includes backpropagating the cut point through the graph to define a plurality of first graph pieces for the first portion, each particular first graph piece representing a computation of a particular slice of the plurality of slices based on a particular portion of a plurality of portions of a second tensor. The method further includes defining one or more second graph pieces to combine outputs of the first graph pieces, and executing the first graph pieces and the second graph pieces to execute the first portion of the compute graph.</abstract><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | |
| ispartof | |
| issn | |
| language | eng |
| recordid | cdi_epo_espacenet_US2024193421A1 |
| source | esp@cenet |
| subjects | CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING COUNTING PHYSICS |
| title | Automatic Memory Management for Compute Graphs |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-03T19%3A30%3A15IST&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=Rizman,%20Matej&rft.date=2024-06-13&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EUS2024193421A1%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 |