IKubeflow: Ithuluzi Lokufunda Lomshini lamaKubernetes
Okuthunyelwe kwethu namuhla kuzobhekana nomkhakha we Ukufunda Okuzenzakalelayo (Ukufunda Komshini / ML). Ngokuqondile ngohlelo lokusebenza lomthombo ovulekile olubizwa nge- "Ukugeleza", yona esebenza ku Kubernetes. Okuyinto, njengoba iningi lenu selivele niyazi, uhlelo lomthombo ovulekile lokusebenzisa ngokuzenzakalela ukuthunyelwa, ukukala nokusingathwa kwezinhlelo zokusebenza eziqukethe izinto ezifakiwe.
"Ukugeleza" yize okwamanje itholakala ngaphansi kwe- inguqulo ezinzile 1.2, njengoba ibonakala kuwebhusayithi yayo esemthethweni naku-GitHub, kwi-Blog yayo esemthethweni, isivele iphawuliwe nge inguqulo elandelayo 1.3. Kungakho namuhla, sizongena kulolu hlelo lokusebenza.
Ithuluzi Lokuqonda: Umthombo Ovulekile Wokufunda Okujulile SW
Futhi njengenjwayelo, kulabo abalangazelela ukungena esihlokweni esifundwayo, sizoshiya izixhumanisi ezilandelayo kokuthunyelwe kwangaphambilini okuhlobene ukuze ukuhlole uma lokhu okuthunyelwe sekuqediwe:
"I-Microsoft Cognitive Toolkit (phambilini ebibizwa nge-CNTK) iyithuluzi lamathuluzi lokufunda elijulile
(Machine Learning)de«Código Abierto»ngamandla amakhulu. Futhi kumahhala, kulula ukuyisebenzisa, kanye nekhwalithi yebanga lokuhweba ekuvumela ukuthi udale ama-algorithms wokufunda ajulile akwazi ukufunda ezingeni eliseduze nelobuchopho bomuntu." Ithuluzi Lokuqonda: Umthombo Ovulekile Wokufunda Okujulile SW
IKubeflow: Iphrojekthi Yokufunda Yomshini Ovulekile
Kuyini iKubeflow?
Ngokusho kwe- Iwebhusayithi esemthethweni, le phrojekthi evulekile ichazwa ngale ndlela elandelayo:
"Yiprojekthi enikezelwe ekwenzeni ukuthunyelwa kokuhamba kokusebenza komshini (ML) kokusebenza kumaKubernetes okulula, okuphathekayo, nokwesabekayo. Akuhloselwe ukwenza ezinye izinsizakalo kabusha, kepha ukuhlinzeka ngendlela elula yokuhambisa amasistimu wemithombo evulekile angcono kakhulu we-ML kuzingqalasizinda ezahlukahlukene. Ngakho-ke noma yikuphi lapho iBubernetes isebenza khona, iKubeflow ingasebenza."
Ngenkathi, kusayithi lakho ku GitHub, engeza kafushane okulandelayo:
"IKubeflow iyipulatifomu yomdabu esefu yokusebenza kwemishini yokufunda: amapayipi, ukuqeqeshwa nokuthunyelwa."
Kulokhu, kungabonakala kalula ukuthi, inhloso enkulu ye- "Ukugeleza" es:
"Yenza imodeli yokufunda (ML) yokulinganisa imishini nokuthunyelwa kube lula ngangokunokwenzeka, ngokuvumela amaKubernetes enze lokho akwaziyo ukukwenza: Ukuthunyelwa okulula, okuphindaphindekayo nokuphathekayo kwingqalasizinda ehlukahlukene, ukuthunyelwa kwemicroservices kanye nokuphathwa kuhlanganiswe ngokukhululekile futhi kukalwe ngokufunwayo."
Izici?
Phakathi kwezici eziphawulekayo ze- "Ukugeleza" singasho okulandelayo:
- Kufaka nezinsizakalo zokwakha nokuphatha izincwadi zokubhalela ze-Jupiter. Ukuvumela ukwenza ngezifiso ukuthunyelwa kwezinsizakusebenza ezifanayo nezinye zekhompyutha ukuzivumelanisa nezidingo zesayensi yedatha. Ngakho-ke, okwenza kube lula ukuzama ukugeleza komsebenzi kwasendaweni, bese ukuhambisa efwini lapho kudingeka.
- Inikeza opharetha womsebenzi wokuqeqeshwa kweTensorFlow. Okungasetshenziswa ukuqeqesha imodeli ye-ML. Ikakhulu, opharetha womsebenzi weKubeflow angaphatha imisebenzi yokuhanjiswa kweTensorFlow. Ukuvumela amandla ukumisa isilawuli sokuqeqesha ukuthi sisebenzise ama-CPU noma ama-GPU, futhi ngaleyo ndlela sivumelane nosayizi ohlukahlukene weqoqo.
- Isekela isitsha seTensorFlow Serving sokuthekelisa amamodeli weTensorFlow aqeqeshiwe eKubernetes. Ngokwengeziwe, iKubeflow ibuye ihlanganiswe neSeldon Core, ipulatifomu yomthombo ovulekile wokuhambisa amamodeli wokufunda womshini kumaKubernetes, kanye ne-NVIDIA Triton Inference Server ukukhulisa ukusetshenziswa kwe-GPU lapho kusetshenziswa amamodeli we-ML / DL esikalini.
- Kufaka ubuchwepheshe beKubeflow Pipelines. Okuyisixazululo esibanzi sokuphaka nokuphatha ukugeleza kokusebenza kokuphela kokuphela kwe-ML. Ukuvumela ukuhlolwa okusheshayo nokunokwethenjelwa, okusetshenziselwa ukuhlela nokuqhathanisa ukugijima, nokubuyekeza imibiko enemininingwane ekugijimeni ngakunye
- Inikeza isisekelo sezinhlaka eziningi. Njengoba, ngaphezu kokusebenza kahle kakhulu neTensorFlow, kungekudala izokwesekwa yiPyTorch, i-Apache MXNet, i-MPI, i-XGBoost, i-Chainer, nokuningi.
Imininingwane eminingi esesikhathini ku- "Ukugeleza" ingatholakala ngqo ku- Ibhulogi esemthethweni.
Yini amaKubernetes?
Kunikezwe, "Ukugeleza" isebenza ku "Kubernetes", kufanelekile ukucacisa ngokusho kwakho Iwebhusayithi esemthethweni ukuthi lokhu okulandelayo kungokulandelayo:
"I-Kubernetes (i-K8s) iyipulatifomu yomthombo ovulekile wokusebenzisa ngokuzenzakalela ukuthunyelwa, ukukala, kanye nokuphathwa kwezinhlelo zokusebenza eziqukethe izinto ezifakiwe."
Futhi uma kwenzeka, ufisa ukujula "Kubernetes" Ungahlola ukushicilelwa kwethu kwangaphambilini nokwakamuva okuhlobene ngezansi:
Isiphetho
Siyethemba lokhu "okuthunyelwe okuwusizo okuncane" cishe «Kubeflow», iphrojekthi yomthombo ovulekile ethokozisayo futhi yesimanje emkhakheni wokufunda okujulile, okwenziwe ukwandisa ukufinyelela kwepulatifomu yomthombo ovulekile «Kubernetes »; inentshisekelo enkulu futhi iyasiza, kuyo yonke «Comunidad de Software Libre y Código Abierto» kanye negalelo elikhulu ekusabalalisweni kwemvelo emangalisayo, enkulu futhi ekhulayo yezicelo ze «GNU/Linux».
Okwamanje, uma ukuthandile lokhu publicación, Ungami yabelana ngayo nabanye, kumawebhusayithi wakho owathandayo, iziteshi, amaqembu noma imiphakathi yokuxhumana nabantu noma amasistimu wokuthumela imiyalezo, okungcono mahhala, okuvulekile kanye / noma okuphephe kakhulu njenge yocingo, Isignali, I-mastodon noma enye ye- I-Fediverse, okungcono.
Futhi khumbula ukuvakashela ikhasi lethu lasekhaya ku- «DesdeLinux» ukuhlola izindaba eziningi, kanye nokujoyina isiteshi sethu esisemthethweni se- I-Telegram ye DesdeLinux. Ngenkathi, ukuthola eminye imininingwane, ungavakashela noma yikuphi Umtapo wolwazi oku-inthanethi njengoba I-OpenLibra y I-JedIT, ukufinyelela nokufunda izincwadi zedijithali (ama-PDF) ngalesi sihloko noma ezinye.