Kubeflow: Kayan Aikin Koyo Na'ura don Kubernetes

Kubeflow: Kayan Aikin Koyo Na'ura don Kubernetes

Kubeflow: Kayan Aikin Koyo Na'ura don Kubernetes

Matsayinmu a yau zaiyi ma'amala da filin na Ilmantarwa na atomatik (Ilmantarwa na na'ura / ML). Musamman game da aikace-aikacen buɗe tushen kira da ake kira "Kubeflow", wanda kuma, ke aiki Kubernetes. Wanne, kamar yadda da yawa daga cikinku suka riga kuka sani, shine tsarin buɗe tushen buɗe kai tsaye don tura kayan aiki, ƙerawa da sarrafa aikace-aikacen kwantena.

"Kubeflow" duk da kasancewar a halin yanzu akwai a karkashin barga version 1.2, kamar yadda yake a cikin shafin yanar gizon hukuma da GitHub, a cikin Blog ɗin hukuma, an riga an yi sharhi akan na gaba 1.3. Abin da ya sa a yau, za mu shiga cikin wannan aikace-aikacen.

Toolarin Kayan Aikin :warewa: Bude Tushen Ilimi Mai zurfi SW

Toolarin Kayan Aikin :warewa: Bude Tushen Ilimi Mai zurfi SW

Kuma kamar yadda aka saba, ga waɗanda ke da sha'awar shiga cikin batun da aka karanta, za mu bar waɗannan mahaɗan masu zuwa daga abubuwan da suka gabata don ku bincika sau ɗaya bayan an gama wannan rubutun:

"Kayan Aikin Fahimtar Microsoft (wanda a da ake kira CNTK) babban kayan aikin koyo ne (Machine Learning) de «Código Abierto» tare da babbar dama. Hakanan kyauta ne, mai sauƙin amfani, da ƙimar darajar kasuwanci wanda ke ba ku damar ƙirƙirar algorithms mai zurfin ilmantarwa wanda ke iya koyo a matakin kusa da na kwakwalwar ɗan adam." Toolarin Kayan Aikin :warewa: Bude Tushen Ilimi Mai zurfi SW

Labari mai dangantaka:
Toolarin Kayan Aikin :warewa: Bude Tushen Ilimi Mai zurfi SW

Labari mai dangantaka:
.NET da ML.NET: Manhajojin Buɗe Ido na Microsoft
Labari mai dangantaka:
TensorFlow da Pytorch: Bude Tushen Aiyukan dandamali

Kubeflow: Bude Burin Koyon Injin Inji

Kubeflow: Bude Burin Koyon Injin Inji

Menene Kubeflow?

A cewar ka shafin yanar gizo, wannan aikin bude an bayyana shi kamar haka:

"Aiki ne wanda aka keɓe don samar da aikin koyo na na'ura (ML) tura kayan aiki akan Kubernetes mai sauƙi, šaukuwa, kuma mai daidaitawa. Manufarta ba shine sake sake wasu ayyuka ba, amma don samar da hanya mai sauƙi don ƙaddamar da mafi kyawun tsarin buɗe tushen buɗewa don ML a duk faɗin abubuwan more rayuwa. Don haka duk inda Kubernetes ya gudana, Kubeflow na iya gudu."

Duk da yake, a shafinku a GitHub, a taƙaice ƙara da waɗannan:

"Kubeflow shine asalin dandamali a cikin girgije don ayyukan koyon na'ura: bututu, horo da turawa."

Daga wannan, ana iya sauƙaƙe hakan, babban maƙasudin "Kubeflow" es:

"Sanya tsarin koyon na'ura (ML) samfurin sikelin da turawa a matsayin mai sauki, ta hanyar barin Kubernetes suyi abinda yakeyi: Sauƙi, sake maimaitawa, da ɗaukewar ɗawainiya a cikin kayan aiki daban-daban, tura kayan aikin microservices da gudanarwa sassauƙa hade da sikelin buƙata."

Halaye?

Daga cikin halaye na kwarai na "Kubeflow" Zamu iya ambaci waɗannan abubuwa masu zuwa:

  • Ya haɗa da sabis don ƙirƙira da sarrafa litattafan rubutun Jupiter masu ma'amala. Ba da damar keɓance kayan aiki iri ɗaya da sauran kayan aikin komputa don daidaita su da bukatun kimiyyar bayanai. Don haka, sauƙaƙa gwaji tare da gudanawar aiki na cikin gida, sannan sanya su a cikin gajimare idan ya zama dole.
  • Yana bayar da mai ba da aikin horo na TensorFlow na al'ada. Wanne za a iya amfani dashi don horar da samfurin ML. Musamman, mai ba da aikin Kubeflow zai iya ɗaukar ayyukan horo na TensorFlow da aka rarraba. Bada iko don saita mai kula da horo don amfani da CPUs ko GPUs, kuma don haka don daidaitawa zuwa nau'ikan masu tarin yawa.
  • Tana tallafawa TensorFlow Mai bautar kwantena don aika ƙirar ƙirar TensorFlow zuwa Kubernetes. Bugu da ƙari, Kubeflow an haɗa ta da Seldon Core, wani dandamali ne na buɗe tushen tura samfuran ilmantarwa akan Kubernetes, da NVIDIA Triton Inference Server don ƙara yawan amfani da GPU yayin tura samfurin ML / DL a sikeli.
  • Ya hada da fasahar Bututun Kubeflow. Wanne shine cikakkiyar mafita don turawa da sarrafa ayyukan ML na ƙarshe zuwa ƙarshe. Bada izinin gwaji mai sauri da abin dogaro, don tsarawa da kuma kwatanta gudu, da yin bita dalla-dalla kan kowacce gudu.
  • Yana bayar da tushe mai yawa. Tunda, ban da aiki sosai tare da TensorFlow, da sannu zai sami tallafi ga PyTorch, Apache MXNet, MPI, XGBoost, Chainer, da ƙari.

Arin bayanai na yau da kullun akan "Kubeflow" za a iya samu kai tsaye a kan Shafin yanar gizo.

Menene Kubernetes?

Ba da, "Kubeflow" yana aiki akan "Kubernetes", ya cancanci tantancewa gwargwadon naka shafin yanar gizo cewa karshen shine mai zuwa:

"Kubernetes (K8s) wani dandamali ne na buɗaɗɗiyar hanyar samar da kayan aiki, ƙira, da kuma sarrafa aikace-aikacen kwantena."

Kuma idan akwai, ana so a zurfafa "Kubernetes" Kuna iya bincika abubuwan da suka gabata da sabbin abubuwan da suka dace a ƙasa:

Labari mai dangantaka:
Kubernetes 1.19 ya zo tare da tallafi na shekara guda, TLS 1.3, haɓakawa da ƙari
Labari mai dangantaka:
Docker vs Kubernetes: fa'idodi da rashin amfani

Hoton hoto don ƙarshen labarin

ƙarshe

Muna fatan wannan "amfani kadan post" game da «Kubeflow», wani shiri na budewa mai ban sha'awa da kuma na zamani a fagen zurfin ilmantarwa, wanda aka sanya don kara isar da tsarin bude tushen «Kubernetes »; yana da matukar amfani da amfani, ga duka «Comunidad de Software Libre y Código Abierto» kuma yana da babbar gudummawa wajan yada kyawawan al'adu, manyan halittu da girma na aikace-aikacen «GNU/Linux».

A yanzu, idan kuna son wannan publicación, Kar ka tsaya raba shi tare da wasu, akan rukunin yanar gizon da kuka fi so, tashoshi, ƙungiyoyi ko al'ummomin hanyoyin sadarwar jama'a ko tsarin aika saƙon, zai fi dacewa kyauta, buɗewa da / ko amintacce kamar yadda sakon wayaSignalMastodon ko wani na Mai rarrabewa, zai fi dacewa.

Kuma ku tuna ziyarci gidanmu na farko a «DagaLinux» don bincika ƙarin labarai, da shiga tashar tashar mu ta hukuma Sakon waya daga FromLinuxDuk da yake, don ƙarin bayani, zaku iya ziyarci kowane Laburaren kan layi kamar yadda OpenLibra y JITIT, don samun dama da karanta littattafan dijital (PDFs) akan wannan batun ko wasu.


Abubuwan da ke cikin labarin suna bin ka'idodinmu na ka'idojin edita. Don yin rahoton kuskure danna a nan.

Kasance na farko don yin sharhi

Bar tsokaci

Your email address ba za a buga.

*

*

  1. Mai alhakin bayanan: Miguel Ángel Gatón
  2. Dalilin bayanan: Gudanar da SPAM, gudanar da sharhi.
  3. Halacci: Yarda da yarda
  4. Sadarwar bayanan: Ba za a sanar da wasu bayanan ga wasu kamfanoni ba sai ta hanyar wajibcin doka.
  5. Ajiye bayanai: Bayanin yanar gizo wanda Occentus Networks (EU) suka dauki nauyi
  6. Hakkoki: A kowane lokaci zaka iyakance, dawo da share bayanan ka.