KServe na Model Gukorera Kubernetes
KServe ni urwego rusanzwe, Kubernetes-kavukire yo gukora imashini yiga imashini murwego.
Incamake
It gives teams a single, declarative way to deploy models with autoscaling, canary rollouts, and scale-to-zero, abstracting away most of the Kubernetes plumbing.
Kwibira cyane
Kera izwi nka KFServation kandi yavutse kumushinga wa Kubeflow, KServe isobanura ibikoresho byihariye bya InferenceService. Wanditse dosiye ngufi ya YAML yerekana icyitegererezo kibitswe mububiko (S3, GCS, Azure Blob), na KServe ikora ibisigaye. Ifasha byombi guhanura kandi, kurushaho, kubyara LLM itanga. Amato ya KServe yabanje kubakwa 'gukorera igihe' kumurongo rusange (TensorFlow Serving, TorchServe, Triton, scikit-yiga, XGBoost, Guhobera Isura) kandi ishyigikira ibikoresho byabigenewe. Yubatswe hejuru ya Knative Serving hamwe nurusobekerane (Istio cyangwa ibisa nayo), itanga ibyifuzo-biterwa na autoscaling harimo igipimo nyacyo-kugeza kuri zeru, moderi zidafite akamaro ntizikoresha compte. Irateganya kandi guhanura API hafi ya Protokole ifunguye, abakiriya rero bavugana na buri cyitegererezo kimwe batitaye kumurongo.
Ubushishozi
KServe ya autoscaling yishingikiriza kuri Knative, igapima umubare wikigereranyo gishingiye ku guhuza cyangwa gusaba-isegonda kandi irashobora kugabanuka kuri zeru iyo umuhanda uhagaze, hanyuma ubukonje-butangire kubisabwa. InferenceService ikuramo umuyoboro wuzuye muburyo bwo guhanura, guhinduranya (mbere / nyuma yo gutunganya), hamwe nibisobanuro. Abanyamideli bapakurura mububiko bwibintu binyuze mububiko bwa 'ububiko bwambere' bukurura ibihangano muri pod mugitangira, gukuramo ububiko bwikitegererezo uhereye kumashusho yabigenewe.
Ingaruka z'Ingamba
Igiciro na bije
Ibyemezo byubwubatsi bitwara imikorere nigiciro cyimikorere kumyaka.
Ibyemezo bisobanutse
Ubuhanga bwa tekinike bufasha amakipe guhitamo umurongo ukwiye, ntabwo ari shyashya gusa.
Kugenzura ubuziranenge
Guhitamo neza bya injeniyeri bigabanya ibintu byizewe mubikorwa.
Kazoza ka KServe na Model Gukorera Kubernetes
KServe irihuta cyane yerekeza kuri AI ibyara umusaruro, wongeyeho LLM yibanda kumurongo hamwe nibintu nka KV-cache-ibizi neza, guhuza icyitegererezo, hamwe na prefill / decode itandukanijwe ikorera kumurongo munini wururimi. Tegereza kwishyira hamwe byimbitse hamwe na moteri yerekana nka vLLM, ibyiza byinshi-node ikorera kuri moderi nini cyane kuri GPU imwe, hamwe ninzira-marembo yo kurwego rwo kwerekana ibimenyetso bishingiye kumitwaro iringaniye. Nkumushinga wa CNCF, urimo kuba igipimo gifunguye cyo gushyira moderi inyuma ya Kubernetes, kugabanya itandukaniro riri hagati yubushakashatsi nubushakashatsi bwihuse.
Gushyira mu bikorwa Isi
Banki ikoresha icyitegererezo cyo gutanga inguzanyo yandika imirongo 10 InferenceService YAML yerekana icyitegererezo muri S3, hamwe na KServe ikora autoscaling no kwinjira.
Itsinda rya e-ubucuruzi rikoresha KServe canary rollouts kugirango yohereze 10 ku ijana byimodoka muburyo bushya bwo gusaba, hanyuma igere ku 100 ku ijana iyo ibipimo bisa neza.
Laboratoire yubushakashatsi ikora moderi zidasanzwe zikoreshwa hamwe nubunini-kuri zeru, buri moderi rero izunguruka gusa mugihe icyifuzo kigeze kandi ntigikoresha GPU mugihe idafite akazi.
Itsinda rya MLOps rikoresha ibikoresho bya transformateur ya KServe kugirango ikore ishusho ihindagurika kandi isanzwe mbere yuko uwahanuye akora moderi ya Triton yatanzwe.
Ingaruka & Kurinda
Gutezimbere igipimo kimwe gishobora guhisha intege nke za sisitemu.
Ibikorwa Remezo no kubungabunga akenshi usanga bidahabwa agaciro.
Icyuho cyumutekano no kwitegereza birashobora kwiyongera uko sisitemu igenda igorana.
Igishushanyo mbonera
Sobanura ubukererwe, ubuziranenge, nigiciro cyibiciro mbere yo kubishyira mubikorwa.
Ibipimo byerekana umutwaro ufatika hamwe namakuru yimiterere.
Gukurikirana ibikoresho kubikosa, drift, ningaruka zabakoresha.
Tegura inzira yo gusubiza ibyabaye mbere yo gupima.
Komeza Ubushakashatsi
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 KServe and Model Serving on Kubernetes 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
Ubuyobozi bukurikira
Guhindura Ibishushanyo Kuri Kode Model
Ibibazo bikunze kubazwa
What is KServe and Model Serving on Kubernetes?
KServe ni urwego rusanzwe, Kubernetes-kavukire yo gukora imashini yiga imashini murwego. Iha amakipe inzira imwe, yo gutangaza uburyo bwo gukoresha moderi hamwe na autoscaling, canary rotout, hamwe nubunini-kuri-zeru, ikuramo ibyinshi mumazi ya Kubernetes.
Ni irihe zina rya KServe mbere yuko riba umushinga wihariye?
KServe yatangiriye nka KFServation mumushinga Kubeflow mbere yo kuba urubuga rwigenga.
Nibihe byambere Kubernetes umutungo wihariye usobanura kugirango ukoreshe icyitegererezo hamwe na KServe?
Uratangaza InferenceService ibikoresho byihariye, mubisanzwe muri YAML, kugirango ushireho kandi ugene icyitegererezo cyatanzwe.
Ni ubuhe bushobozi butuma moderi ya KServe idafite akamaro ikoresha compte zeru kugeza icyifuzo kigeze?
Yubatswe kuri Knative, KServe ishyigikira igipimo-kuri-zeru, ikamanura kopi kuri zeru iyo nta traffic nubukonje-butangiye kubisabwa.
Nuwuhe mushinga wibanze utanga ibyifuzo bya KServe?
KServe yubaka kuri Knative Serving, igapima kopi ishingiye ku guhuza cyangwa igipimo cyo gusaba kandi igafasha igipimo-kuri-zeru.
Nigute KServe ibona ibihangano byintangarugero mugutangira gutangira?
Ububiko bwambere bubika gukuramo ibihangano byintangarugero mububiko bwibintu (nka S3 cyangwa GCS) muri pod, gukuramo moderi kuva kumashusho.