Ray don Rarraba AI
Ray wani tsari ne mai buɗe ido wanda ke sauƙaƙa auna aikin Python da AI daga kwamfutar tafi-da-gidanka zuwa tarin dubban inji.
Dubawa
It matters because it gives a simple, unified way to distribute training, tuning, data processing, and serving without rewriting your code for each.
Zurfafa nutsewa
Babban ra'ayin Ray shine juya ayyukan Python na yau da kullun da azuzuwan zuwa raka'a da aka rarraba tare da ƙaramin canji. Ayyukan da aka yiwa alama azaman 'aiki' mai nisa yana gudana ba tare da jituwa ba akan kowane ma'aikaci a cikin tari; ajin da aka yiwa alama a matsayin 'dan wasan kwaikwayo' mai nisa ya zama sabis na gaskiya da ke zaune akan ma'aikaci. Ray yana dawo da makoma mara nauyi (nassoshi na abu) kuma yana sarrafa tsara lokaci, motsin bayanai ta wurin ajiyar abu da aka raba, da haƙurin kuskure. A saman wannan babban ɗakin karatu da aka gina manufa: Ray Train don horar da ƙirar ƙira, Ray Tune don binciken hyperparameter, Ray Data don kwararar bututun bayanai, RLlib don ƙarfafa koyo, da Ray Serve don sabis na ƙira mai ƙima. Wannan yana bawa gungu ɗaya damar sarrafa gabaɗayan aikin ML ƙarshe zuwa ƙarshe.
Fahimtar Fasaha
Maɓalli na farko sune ayyuka (marasa ƙasa, kiran aikin layi ɗaya) da ƴan wasan kwaikwayo (ma'aikata na ƙasa waɗanda ke riƙe abubuwa kamar samfurin da aka ɗora ko ƙira). Lokacin da kuka kira aiki mai nisa, nan da nan Ray ya dawo gaba kuma ya tsara aikin a cikin CPUs/GPUs da ke akwai; ka kira ray.get() don samun sakamako. Shagon abubuwan da aka rarraba cikin ƙwaƙwalwar ajiya tare da kwafin sifili mai raba ƙwaƙwalwar ajiya yana motsa manyan abubuwa kamar tsararru tsakanin ma'aikata yadda ya kamata, guje wa maimaita maimaitawa da yin bututun AI masu nauyi da bayanai cikin sauri.
Dabarun Tasiri
Kudin da kasafin kuɗi
Hukunce-hukuncen gine-gine suna haifar da aiki da tsadar aiki na shekaru.
Shawarwari masu haske
Ilimin fasaha yana taimaka wa ƙungiyoyi su zaɓi tari mai kyau, ba kawai sabon abu ba.
Kula da inganci
Zaɓuɓɓukan injiniya mafi kyau suna rage abin dogaro a cikin samarwa.
Makomar Ray don Rarraba AI
Ray ya zama kashin baya ga babban sikelin AI, wanda aka yi amfani da shi musamman wajen horarwa da kuma bautar manyan samfuran harshe. Yi tsammanin haɓakawa a takamaiman sabis na LLM (Ray Serve tare da vLLM), tsarin GPU iri-iri, haɗin kai tare da tafkunan bayanai da Kubernetes ta hanyar KubeRay, da ingantacciyar ƙira don ɗaukar nauyi mai ƙima. Yayin da samfura ke girma, rawar Ray a cikin ƙirƙira horarwar kuɗaɗe da yawa, bututun RLHF, da ba da ƙima a cikin dubunnan na'urori na iya haɓaka.
Aiwatar da Gaskiyar Duniya
Gudun Ray Tune don bincika ɗaruruwan haɗin haɗin hyperparameter a layi daya a cikin gungu na GPU don nemo mafi kyawun ƙirar ƙira.
Amfani da Ray Train don rarraba horon ƙirar koyo mai zurfi a cikin yawancin GPUs da nodes tare da ƙananan canje-canje na lamba
Gina bututun batch-inference tare da Ray Data don cin miliyoyin bayanai ta hanyar yawo su ta hanyar ƙira a cikin tari.
Ƙaddamar da ƙira da yawa a bayan madaidaicin atomatik guda ɗaya tare da Ray Serve don sarrafa zirga-zirgar samarwa mai canzawa
Hatsari & Tsare-tsare
Haɓaka ma'auni ɗaya na iya ɓoye manyan raunin tsarin.
Sau da yawa ana raina kayan more rayuwa da kuma kuɗin kulawa.
Tsaro da gibin lura na iya girma yayin da tsarin ke ƙara haɓaka.
Taswirar Hanya
Ƙayyade latency, inganci, da maƙasudin farashi kafin aiwatarwa.
Alamar ma'auni a ƙarƙashin ainihin kaya da yanayin bayanai.
Kula da kayan aiki don kurakurai, ɗigo, da tasirin mai amfani.
Shirya bijirowa da hanyoyin mayar da martani kafin sikeli.
Ci gaba da Bincike
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Ray for Distributed AI quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Jagora na gaba
Duban hankali na Gradient
Tambayoyin da ake yawan yi
What is Ray for Distributed AI?
Ray wani tsari ne mai buɗe ido wanda ke sauƙaƙa auna aikin Python da AI daga kwamfutar tafi-da-gidanka zuwa tarin dubban inji. Yana da mahimmanci saboda yana ba da hanya mai sauƙi, haɗin kai don rarraba horo, daidaitawa, sarrafa bayanai, da hidima ba tare da sake rubuta lambar ku ga kowane ba.
Wace matsala Ray ya warware da farko?
Ray yana ba da haɗin kai, hanya mai sauƙi don rarraba ƙididdiga a cikin injina da yawa ba tare da sake rubuta lambar ku ba don kowane nau'in nauyin aiki.
A cikin Ray, menene bambanci tsakanin 'aiki' da 'dan wasan kwaikwayo'?
Ayyuka ayyuka ne masu nisa marasa jiha waɗanda ke gudana a layi daya, yayin da 'yan wasan kwaikwayo abubuwa ne na dogon lokaci waɗanda ke riƙe da yanayi, kamar ƙirar ƙira.
Menene Ray zai dawo nan da nan lokacin da kuka ƙaddamar da aiki mai nisa?
Ray yana dawo da makoma mara nauyi nan da nan don haka aikin yana gudana ba tare da an daidaita shi ba; kuna kiran ray.get() don dawo da ainihin sakamakon lokacin da ake buƙata.
Wane ɗakin karatu na Ray aka ƙera don binciken hyperparameter da aka rarraba?
Ray Tune ya ƙware wajen gudanar da gwaje-gwajen hyperparameter da yawa a layi daya a cikin tari.
Ta yaya ma'ajiyar kayan Ray ke sa manyan bututun AI masu nauyi da bayanai su yi inganci?
Shagon abubuwan da aka raba a cikin ƙwaƙwalwar ajiya yana amfani da karatun sifili don haka manyan jeri-jeru na iya samun dama ga ma'aikata ba tare da sake tsarawa mai tsada ba.