I-Keras, i-API yokufunda ejulile yomthombo ovulekile

UKeras

Ilogo ye-Keras

Nge ukuthandwa okukhulu okuzuziwe ukusetshenziswa kobuhlakani bokwenziwa ezindaweni ezahlukene, “Ukufunda Okujulile” (ukufunda okujulile), ikwazile futhi ukuthola ukuhlobana okukhulu, njengoba isetshenziselwa ukwenza izinqumo, ukutholwa kwento, ukubonwa kwenkulumo, ukuhumusha ulimi neminye imisebenzi eminingi, ngoba isho eminye yaleyo esetshenzisiwe.

Solo ukusho isibonelo, ukufunda okujulile Isetshenziswa kumakhamera wokugada namuhla futhi kulokhu sikhuluma ngokusetshenziswa kwezentengiselwano, okumelela imakethe enkulu futhi ngaphezu kwakho konke ngoba ukubhekwa kwevidiyo akuseyona into ewubukhazikhazi, kodwa sekuqalile ukuba yisidingo.

Ngale ndlela, kunezinhlobonhlobo zamaphrojekthi wezohwebo nemithombo evulekile, yomibili le ndawo yokubhekwa kwevidiyo kanye nezinye izimo zokusetshenziswa zokufunda okujulile.

Mayelana neKeras

Ukushiya eceleni udaba lwezohwebo nokugxila kancane esihlokweni se-athikili, Ngingathanda ukukhuluma kancane ngeKeras, okuyi- I-High-level neural network API ebhalwe ngePython. Le labhulali yenethiwekhi ye-neural umthombo ovulekile yakhelwe ukuhlinzeka ngokuhlolwa okusheshayo ngamanethiwekhi e-neural ajulile futhi ingasebenza phezu kwe-CNTK, i-TensorFlow, ne-Theano.

Ini Phuma kuya kuKeras kusuka kwamanye amaphrojekthi afanayo, yilokho Iklanywe ngendlela yokuthi abantu bayifunde kalula, njengoba idalwe ukuze ibe lula, ngama-API angaguquki futhi alula, inciphisa izenzo ezidingekayo ukuze kusetshenziswe ikhodi evamile futhi ichaze ngokucacile amaphutha omsebenzisi.

UKeras inikeza isikhombimsebenzisi se-Python ngezinga eliphezulu lokukhipha futhi, ngesikhathi esifanayo, unenketho yamaseva amaningi ngezinjongo zokubala. Lokhu kwenza i-Keras ihambe kancane kunezinye izinhlaka zokufunda ezijulile, kodwa ilungele kakhulu abaqalayo. njengoba igxile ekubeni yi-modular, kulula ukuyisebenzisa futhi inwetshwe. Ayizibambi izibalo ezisezingeni eliphansi; esikhundleni salokho, iwadlulisela komunye umtapo wolwazi okuthiwa i-Backend.

Elinye iphuzu elivuna uKeras ukuthi ivumela abasebenzisi ukuthi bakhiqize amamodeli ajulile kokubili ku iOS, Android, kanye nakuwebhu naku-JVM, futhi inokusekelwa okuqinile kwe-multi-GPU nokusekelwa kokuqeqeshwa okusabalalisiwe.

Isiqephu 3

Kuhle ukusho ukuthi uKeras, Njengamanje isegatsheni layo le-3.x, eyakhululwa emasontweni ambalwa edlule futhi leli gatsha elisha selivele lithola ukuthuthukiswa nokulungiswa okuthile, esingaqaphela ngakho ukuthi iphrojekthi ithuthukiswa njalo futhi inomphakathi omkhulu osebenzayo.

I-Keras ibilokhu isekuhlolweni okukhulu kwe-beta yomphakathi izinyanga ezimbalwa, futhi ukukhishwa kwe-Keras 3 kumele ukubhalwa kabusha okuphelele, ukuthuthukisa amakhono okuqeqesha nokuthumela amamodeli esikalini.

Ngaphakathi izici eziyinhloko zaleli gatsha elisha kwe-Keras 3, okulandelayo kuyagqama:

Ukusekelwa kwe-Multi-backend

Ngaphandle kokungabaza enye yezinto ezintsha ezinhle ze-Keras 3.0 ukusekelwa kwayo okungakaze kubonwe kwama-backends amaningi, njengoba isebenza njengesixhumi esikhulu esinekhono lokukhetha ngokuguquguqukayo i-backend ezohlinzeka ngokusebenza okungcono kakhulu ngaphandle kokushintsha noma yini kukhodi.

Ukuthuthukiswa kokusebenza

Okunye okugqamile okubalulekile kwe-Keras 3.0 ukuthuthukiswa kokusebenza, ngoba isebenzisa ukuhlanganiswa kwe-XLA (Accelerated Linear Algebra) ukuze kuthuthukiswe izibalo zezibalo, ngaphezu kokuphindaphindeka kabili ekuthuthukiseni ukusebenza kahle, amasu okuhlanganisa afana nokuqeqeshwa kokunemba okuxubile nokuqeqeshwa okusabalalisiwe.

I-ecosystem enwetshiwe

Ngalesi sibuyekezo esisha, i-Keras ithole ukuthuthukiswa kokusekelwa futhi ingaqiniswa njenge-PyTorch, ithunyelwe njengemodeli ye-TensorFlow, noma iqiniswe njengomsebenzi we-JAX ongenamthetho. Lokhu kusho ukuthi ungakwazi ukusebenzisa amandla ohlaka ngalunye olunwetshiwe lwe-ecosystem ye-Keras ngaphandle kokuvaleleka ku-ecosystem eyodwa.

Kuhle ukusho ukuthi i-Keras 3 isebenzisana kakhulu ne-Keras 2, njengoba isebenzisa i-Keras 2 API, ngenani elilinganiselwe lokuhlukile, ngakho-ke abasebenzisi abaningi ngeke kudingeke benze izinguquko zekhodi ukuze baqale ukusebenzisa imibhalo yabo. I-Keras kulokhu okusha inguqulo.

Ekugcineni, Uma unesifiso sokwazi okwengeziwe ngakho, ungabheka imininingwane yaleli gatsha elisha Kulesi sixhumanisi esilandelayo. Uma ufuna ukwazi ifayela le- kanjani ukusebenzisa Keras? kusistimu yakho, ungabheka izindlela zokufaka kulesi sixhumanisi, ngenkathi enjalo nje imibhalo kanye namacala okusetshenziswa Ukuze ufunde ngakho, ungakwenza ku- lesi sixhumanisi


Shiya umbono wakho

Ikheli lakho le ngeke ishicilelwe. Ezidingekayo ibhalwe nge *

*

*

  1. Unomthwalo wemfanelo ngedatha: AB Internet Networks 2008 SL
  2. Inhloso yedatha: Lawula Ugaxekile, ukuphathwa kwamazwana.
  3. Ukusemthethweni: Imvume yakho
  4. Ukuxhumana kwemininingwane: Imininingwane ngeke idluliselwe kubantu besithathu ngaphandle kwesibopho esisemthethweni.
  5. Isitoreji sedatha: Idatabase ebanjwe yi-Occentus Networks (EU)
  6. Amalungelo: Nganoma yisiphi isikhathi ungakhawulela, uthole futhi ususe imininingwane yakho.