Zane-zane na Seldon Core da Inference
Seldon Core dandamali ne na bude-bude don tura samfuran koyo na inji akan Kubernetes, tare da fasalin fage: zane-zane.
Dubawa
Instead of serving one isolated model, it lets you chain models, routers, combiners, and transformers into a single directed graph that runs as one deployable service.
Zurfafa nutsewa
Yawancin shari'o'in amfani da samarwa na gaske sun ƙunshi fiye da kiran ƙira ɗaya. Kuna iya aiwatar da shigarwar, bi da buƙatun zuwa ɗaya daga cikin samfura da yawa, gudanar da taro, sannan bayan aiwatar da sakamakon. Seldon Core yana bayyana wannan azaman jadawali da aka siffanta a cikin SeldonDeployment (ko, a cikin gine-ginen v2, ta hanyar Seldon Core Operator da MLServer). An gina jadawali daga nau'ikan abubuwan da za a sake amfani da su: Model yana ba da tsinkaya, Mai Canjawa yana canza abubuwan shigarwa ko fitarwa, Mai ba da hanya tsakanin hanyoyin sadarwa yana yanke shawarar wane yaro zai kira (ba da damar gwajin A/B da 'yan fashi da yawa), kuma Mai Haɗawa yana tattara abubuwan samarwa daga ƙira da yawa don haɗawa. Seldon yana goyan bayan tsare-tsare da yawa ta hanyar sabar da aka riga aka shirya da kuma naɗaɗɗen Python na al'ada, kuma yana fallasa ma'auni masu wadatarwa, rarrabawa, da shigar da kaya daga cikin akwatin don lura da bayyanawa.
Fahimtar Fasaha
jadawali jadawali jadawali ne a cyclic jadawali inda kowane kumburi ya kasance microservice tare da madaidaicin fa'ida, kuma mawaƙan Seldon (Mawaƙin sabis ɗin / mai zartarwa) yana bibiyar buƙata ta cikin jadawali kuma ya haɗa amsa. Saboda masu ba da hanya tsakanin hanyoyin sadarwa na iya aiwatar da dabaru masu amfani da makamai masu yawa, zirga-zirgar zirga-zirga na iya canzawa daidai gwargwado zuwa ingantattun samfura dangane da siginar lada. Seldon Core v2 yana ƙaddamar da jadawali daga kowane sabar ƙirar ƙira ta amfani da MLServer da Buɗe Ka'idar Inference, yana ba da damar yin hidimar samfuri da yawa da wuce gona da iri akan kayan aikin da aka raba.
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 Seldon Core da Hotunan Inference
Seldon yana motsawa zuwa na zamani, MLOps na tsakiya na bayanai tare da bututun Core v2 da ƙirar bayanai, da ƙarin haɗakarwa tare da gano ɓarna (Alibi Gano) da iya bayyanawa (Alibi Bayyana). Kamar yadda LLMs da tsarin wakili suka zama zane-zane na maidowa, samfura, da kayan aikin, taswirar jadawali-jari bisa dabi'a akan waɗannan ayyukan aiki. Yi tsammanin ƙarin ba da fifiko kan ingantattun samfura masu yawa, watsa shirye-shirye, da daidaitaccen kiyayewa ta yadda hadaddun tsarin AI masu matakai da yawa su kasance masu ɓarna kuma ana iya sarrafa su a samarwa.
Aiwatar da Gaskiyar Duniya
Mai ba da lamuni yana ɗaure Transformer wanda zafi-zafi ɗaya ke ɓoye fasali zuwa ƙirar ƙira, sannan Mai Canjawa wanda ke tsara makin, duk a matsayin SeldonDeployment guda ɗaya.
Kamfanin watsa labarai yana amfani da kullin na'ura mai ba da hanya tsakanin hanyoyin sadarwa yana gudanar da gungun 'yan bindiga da yawa don aika ƙarin zirga-zirga zuwa kowane samfurin shawarwarin yana samun lada mai girma.
Ƙungiya ta haɗa nau'ikan zamba guda uku tare da kumburin Combiner wanda ke daidaita makin su kafin mayar da shawara ɗaya ga mai kira.
Mai inshorar da aka tsara yana haɗa rajistar kaya na Seldon da masu bayyana Alibi zuwa jadawali don haka za a iya gano kowane tsinkaya da kuma bayyana shi don tantancewa.
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 Seldon Core and Inference Graphs 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
TensorRT da Injin Inference
Tambayoyin da ake yawan yi
What is Seldon Core and Inference Graphs?
Seldon Core dandamali ne na bude-bude don tura samfuran koyo na inji akan Kubernetes, tare da fasalin fage: zane-zane. Maimakon yin hidimar keɓantaccen samfuri ɗaya, yana ba ku damar sarkar ƙira, masu amfani da hanyar sadarwa, masu haɗawa, da taswira zuwa jadawali guda ɗaya wanda ke aiki azaman sabis ɗin turawa ɗaya.
Menene ma'anar fasalin da ke bambanta Seldon Core daga uwar garken samfuri guda ɗaya?
Seldon Core yana ba ku damar tsara samfura, masu amfani da hanyar sadarwa, masu haɗawa, da masu canzawa zuwa jadawali ɗaya wanda aka tura azaman sabis ɗaya.
Wanne bangaren jadawali na Seldon ne ya yanke shawarar kodin yaro don aika buƙatu zuwa gare shi, yana ba da damar gwajin A/B?
Na'ura mai ba da hanya tsakanin hanyoyin sadarwa tana jagorantar buƙatun zuwa ɗaya daga cikin 'ya'yansa kuma yana iya aiwatar da gwajin A/B ko ɓangarori masu yawan gaske.
Wanne nau'in bangaren ne ke tattara abubuwan da aka samu daga samfura da yawa don haɗawa?
Mai haɗawa yana haɗa martanin ƙirar yara da yawa zuwa fitarwa guda ɗaya, wanda shine yadda ake gina ƙungiyoyi.
Wane irin tsarin jadawali ne Seldon jadawali ya samar?
Zane-zane na Seldon ana jagorantar jadawalin acyclic jadawalai inda kowane kulli microservice ne tare da madaidaicin dubawar hasashen hasashen.
A cikin Seldon Core v2, wanne uwar garken da yarjejeniya ke ba da damar yin aiki da samfura da yawa a cikin jadawali?
Core v2 yana ƙaddamar da jadawali daga sabobin ƙira ta amfani da MLServer da Buɗe Ƙa'idar Ƙimar don hidimar ƙirar ƙira.