0% found this document useful (0 votes)
13 views7 pages

ERNIE

Official implementations for various pre-training models of ERNIE-family, covering topics of Language Understanding & Generation, Multimodal Understanding & Generation, and beyond

Uploaded by

Marcos Luis
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
13 views7 pages

ERNIE

Official implementations for various pre-training models of ERNIE-family, covering topics of Language Understanding & Generation, Multimodal Understanding & Generation, and beyond

Uploaded by

Marcos Luis
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 7
9115/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-faly covering tops of Langu Bi PaddlePaddle / ERNIE Pui: Coicial implementations for various pre-traning models of ERNIE-family, covering topics of Language Understanding & Generation, Multimodal Understanding & Generation, and beyond fr GSksuis Y 1Skfons PF Branches © Tays-Ar Acoty sur © Code © tssves Th Pullrequests © Discussions © Actions PF emieckiopen-vid ~ YG 8rancher OB ta PO @ Vocorine Updte README situ Be mete Research 1 spokeations ert D sitgnore L. pre-commit-conigyant D ReAoMen README ENV 1D eaoMe sockemd README: htps:igthub.com/PadélePadale/ERNIE Update issue tempts (#753) peste README ma pate dta.20997 Dyaraph e436) update to pasele 20 Update README md Upaate README ENV.na sc ing 2 notations © Security LX Insights Goto fie entes-beryer ® 2 years 390 years 390 4 yeas 290 3 years 290 yeas 290 yeas 290 yeas 290 wn 9115/24, 10:27AM GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-Iamly covering tops of Langu © ERNIE Milestones a) O22. O a=. ee | ee hae Spe Uaeretntee! Won the CAAT Award LOAN ESRI, TART NPLATE ARM, 201983, FRET RAE ARR O ERNIE10, RTE SSAA RAMSAR, HRVGNTIRSTIOT ROUEN, ARMA SULA, 2M: RNAS CTE Eepro DL, MUMIA TROT AMSA MUTRERNERATTR, FP, twH LcasyDL. BML sam, [7 MES) FFifRoadmap + 2022818: © BISLEBSRSHAVIERERE, cevrevit 2.0 (are) JESCHEIE + 2022520: © SBSFFFIRERNIE 3 OF FURAN? 1 TOMASHRRREERNIE 3.0 Base = 200M HS ESEFRIAIERNIE 3.0 XBase = TAMPERRSLIEFRIREZERNIE 3.0 Mecium ‘Hie ESR MEOMMIERNIE SAT It SPSBERNIE-Gen (4252) RVI, SEAR TRCEARE: PRMOREAREE, TRNCEOR, NE, (oH EMEA OERNICTTAREYE: AF CRADORUNR, SISTLOERNICRRIRT SARI, CORASELEMRUIUG TENA, (SB — NRT A, AUSTRIAN, © SeCRAILE, RRUNGMEE, RISRHNER, HARIIPSS, HeRURSENLeTTARRABIMTE, © SRIYNLPMFESS: SCARAB, SCARCE, ARUN, (BIRR, SER, RTE, © SAORI, RUBIA, ial SUI, JUNE RURILA, # 2021-123: © SiEVISIR! concen TESLA + 2021520: © ERNIE SR FFIADUA SU IE © SPERAIOREE core-oron ETE = HERSCARRARERIERL ewwre-ooc TESTI = AS AMRIT ASRS ARID eance-van TESST = BSS AOORUIERE ence-uno JECT + 2020824 © eMe-viL SORAES! GGEAD htps:igthub.com/PadélePadale/ERNIE ar 9115/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-Iaily covering tops of Langu + HORROR TIME, EXER RS TRUINNS) orem, + ARGRECOM, ORT OK RENE, Mee LNT, + DRS WEARER, UNMIS, + 2020520 © ERVTE-cEN BRESCIA = BEARERS, THEA vear-2070 BOR, = ARE ERNIE HUGO REAGAN, STR FRE, = COMTEROET PRIEIRAONTATIAN (ELA baseslarge/anne=4206) , 1 PVRTEIRUIESI ARMA span-by-span SESRES5, LAULSHRRR SEN —NELRBH, 1 LUCE AL EIR RRL SANTOR EROS, 1 RISHY Mult-Flow Attention SVEN, + 2020420 ARAHCRNIESo9e, —APRTHRRATSRIEHENY, SRARERNIEHEagoreagtor. HPCLSA, + 2020327 EESemtvalP0205 NFA be, + 2019.12.26 cLUEW— ee, + 2019.116 #AFERNEE Te + 2019.77 MAsEANIE 20, + 2019.3.16 ARABLRNIE 10, RRR 1 eR: Tae 2. SeREMieRHE S ait clone htps://gtthub.con/Padderaddle/ERM. pit PRRLE: GASTOERNIEAR RTI + SRARERNIES OfFPNSRUSGAREY, ERS LYFE: o Regine fe © RRRLUIEBjson © unenae © ReREEaIjsons © abe + BUNA REIS ND), SRL FERNIEA BRAN S Fea + HeRERNIES OTMUIMRIRMR ASCARI HEESS 1 ERNMES ORAS PSE ernie 3.0 BEF e 8 Be Aroseis noes applications nodes rub 2 ere eR + OS MESA RTOS AINE, ET, TERETE. + SORE HONE htps:igthub.com/PadélePadale/ERNIE an 9115/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-Iaily covering tops of Langu DBASE RH 6 «/applseations/ tases PRELASREEARRER 1 fase o + TRS, ERIN, REBURison se + FAPAMIsonSRH47E /oxamples/ERRER, {RFEIERNIES OFLU RUE FT IMBEOMR ESL! ISH) /examples/clserie f.chjson, TERJsonSCHAEPRS 28, OD, TSRRS TER, RT CRIES. oI. UMA AAT RESON e at Jexanples/els_ernte_fe_eh-3son aap + ORRWSRETERE, FHRRMRITKls_erie fe chjson, SADAFILLNRRMUUURASIES, + ich, BLRIRS I pytron run trainer.py , SITRAER, SEARS Femieieh ek ARNE LIA aHoH a, # emie seascape @ 1 EF /sonsCRAMMLGU, RUAITIERLL. /exanples/cis_ernie_fc_ch-3son python run_srainer.py ~paran_path «/evanples/cls_ernie_fe_ch- 301 + Sein ‘Alectoun —-poussnsy run_trainer. py. exanples/els_eonde_fe_ch. on o + UIMRBEFRO RES RNR /logitestlogt SCH, + IME a RRS SOMES RRNRAPTE /outputlEINTR, Hisave_inference.model/ SCHR it AF AMIR, save_checkpoint/ XHREARE ET INSEE, RBRMjsonxcee + RAB sonSRHHZE /eramples/ERRER, {RFRERNIE2 ORTUESAAMY IS COHRNWIY 5 RUMOR INE /examples/clsernie_fchnferjson + EREHVAL /examples/ls_emie fe ch inferjson FHORUREIOMA RE, HUAI ARE. FARUGIROOARIRNS, ISR PRAN ‘ o er" igh: ("data_path":"-/6ata/precict sata"), Snfererce” ("inference rodel path”: /outpit/els_ernie_fe_ch/save_snference roel /Snfarence_ step 751", output_path"? * Jou “y > aa run inferpy , RRENAASAMRRM HRT]. SRA + BRISA ARE /outpul/predict_resultcxHep, FURR * SHRUG RURE MRNA htps:igthub.com/PadélePadale/ERNIE ar 9115/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-Iaily covering tops of Langu + SRA F&: 2 /applicaions/models hubEIRTR, FER, BAINBMETRER e 4 /applications/nodels, nub PRES. base + BSF, MResearcn ot od CLUES ule ont sia NERA BERRA + SHOOTERURAACLUERLEEA, CLUE FRU: i a CLUEWSC2020IFLYTEK TNEWS §AFQMC =CMNU CSL) OCNLI SSG 24.1024 ROBERTo-exlarge 9079 e202 $9.33 760063888367 7081 71636, 20L1024H ERNIE 30-xBase gnae 6222 6037595 84988427 SOT TTA 12LT6BH | ROBERTa-wowm-ext-base 855 e122 seus 74751617728 12U762H ERNIE 30-Base 818 sore 58737553365 8330 DSI 75.63 6L768H — RAT6, Chinese 7500 5968 $662 73S 7926004 73S 7099 ©L766H —-ERNIE30-Medium 7993, cork 5716 7456 60878123. 7702 72.9, 1 CLEARS ESSISF Grid Search 75: SSN, SIAEITIGATIIN 100 steps FHEMEMAUR, MUIR RAMRLURYE PoaethE 2. SYKCESF Grd Search REEPEEE batch siz’ 16, 32,64 learning rates: Te-5, 26-5, 30-5, Se-5; AY9 CLUEWSC2020 RARAEAR/), AUBLRTE RAL LAORHERT batch size ARBUAR, FELURY CLUEWSC2020 SEJURTERSMEEMIT batch size = 8 AORRESHER; PRISCLUEWSC2020 A TFLYTEK BIGRSERY dropout MARIEARUGRUNL, FFLLRY CLUEWSC2020 X] IFLYTEX BRIGSEFAUindropout prob = 0.0 ANARSHDR. TiS hheesre ‘SOROS: TASK. |AFOMC NEWS IFLYTEK CMINLI OCNLI CLUEWSC2020 CSL epoch 3 3 3 2 5 so 5 maxseatensth 128 28 82818 1B 256 vwarmup_proportion 0:1 04 or ont 01 htps:igthub.com/PadélePadale/ERNIE 37 9115/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersining models of ERNIE-Iaily covering tops of Langu ERNIES@@Grid Search MHS Model FMC ‘TNEWS ryTEK cM cn cLuEWsc2020 cst ERNE Aa tee Giese. | BRIE! 30. oe Ce BROS osjosz 64 Ih2e Medium os Lire sz 32 Ir bapinzeos Be sz. 8.20 05idrop_out oy ERNIE ber 64 Ide bez 16 hse 52 16.in2e 30-Base 05 0s os betehzeos 9 bez 16 In 0s NIE 30 XBase RGR Sep (LAD) SRR (ATE) bez 16Ie2e- be 16 h3e- 5 Bae 32 Ine e e betehteos 9 bse. 6 20-05 eexoene (oe) saat (CURT) TA (Lae) SCBKS FA ERNIE 1.0 areicte{sunzeisernte, o ‘itles(éenie: Emhanced representation through knowledge integration), SJournal=(ankiv preprint apxiv3904.09225), year=(2039) ) ERNIE 2.0, inproceedings (sunzezderaie, e titie=(crnle 2.0: A continual pre-tratning ¢ranewore for language understanding}, suthor=(Son, Yu and kong, Shuohuan ane (4, Yin and Feng, Shikon and Tian, #90 ané hu, Hus and Kang, nfene), booktitles(Poceecings of the ANAT Conference on Artificial Intelligence), ERNIE-GEN farcicie(xtaozeneernie, o ‘Uities{Ennle-gens An emrarced multi-taw pre-tratning and fine-tuntng framework for natural language generation}, htps:igthub.com/PadélePadale/ERNIE er 9118/24, 10:27AM —_ GitHub -PadelePadle/ERNIE: Official implementations for various prersning models of ERNIE-Ialy, covering tops of Langu author=(Xia0, Oongling and Zhang, Han and Li, YuKun and Sun, Yu ané Tian, Hao and wy Ha and Wangs Haifene), Sournale(anciv preprint sexsys290s.1314), vyear=(202e) > ERNIE-ViL eseticie(yureavernse, e ‘ities(tenie-vil: knowledge enhanced viston-2anguage representations through scene 8r3ph), futhor=(Vu, Fel and Tangy 24ft and Vin, Wetchong and Sin, Yu and Tian, #20 and Way Hun and Mang, Halong), ERNIE-Gram areiere(xtsoz022ernie, e ‘titlee{RNIE-Gran: Pre-Training with GxLicithy N-cran Masked Language Modeling for Natural Language Understanding, suthor=(Xiao, Dangling and Li, Westin and Zang, Non and Sun, Yu and Than, Mao and Wu, Ha and Wang, HoLferg), $Journale(arxiv preprint apkivi2ei9, 124M yeara{202e) Releases © ERNIE aygrpah Packages Contributors 34 SOS eC ONELOR TOs 220 contrtor Languages htps:igthub.com/PadélePadale/ERNIE 7

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy