{"id":86523,"date":"2020-06-26T07:42:14","date_gmt":"2020-06-26T05:42:14","guid":{"rendered":"https:\/\/prohoster.info\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning"},"modified":"2020-06-26T07:42:14","modified_gmt":"2020-06-26T05:42:14","slug":"mlops-devops-v-mire-machine-learning","status":"publish","type":"post","link":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","title":{"rendered":"MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>V roce 2018 se v profesion\u00e1ln\u00edch kruz\u00edch a na tematick\u00fdch konferenc\u00edch v\u011bnovan\u00fdch AI objevilo pojem MLOps, kter\u00fd se rychle usadil v oboru a nyn\u00ed se rozv\u00edj\u00ed jako samostatn\u00fd sm\u011br. Do budoucna by se MLOps mohl st\u00e1t jedn\u00edm z nej\u017e\u00e1dan\u011bj\u0161\u00edch oblast\u00ed v IT. Co to vlastn\u011b je a k \u010demu slou\u017e\u00ed, zjist\u00edme n\u00ed\u017ee.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed\" src=\"\/wp-content\/uploads\/2020\/06\/03bcaaaf5b9ddd864c7e8e0a607d81c4.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/p>\n<h2>Co je MLOps<\/h2>\n<p>\nMLOps (sjednocen\u00ed technologi\u00ed a proces\u016f strojov\u00e9ho u\u010den\u00ed a p\u0159\u00edstup\u016f k implementaci vyvinut\u00fdch model\u016f do podnikov\u00fdch proces\u016f) \u2014 to je nov\u00fd zp\u016fsob spolupr\u00e1ce mezi z\u00e1stupci podnik\u016f, v\u011bdci, matematiky, specialisty na strojov\u00e9 u\u010den\u00ed a IT in\u017een\u00fdry p\u0159i vytv\u00e1\u0159en\u00ed syst\u00e9m\u016f um\u011bl\u00e9 inteligence.<\/p>\n<p>Jin\u00fdmi slovy, to je zp\u016fsob, jak prom\u011bnit metody a technologie strojov\u00e9ho u\u010den\u00ed na u\u017eite\u010dn\u00fd n\u00e1stroj pro \u0159e\u0161en\u00ed podnikov\u00fdch \u00faloh.\u00a0<\/p>\n<p>Je t\u0159eba ch\u00e1pat, \u017ee \u0159et\u011bzec produktivizace za\u010d\u00edn\u00e1 d\u00e1vno p\u0159ed v\u00fdvojem modelu. Jeho prvn\u00edm krokem je definov\u00e1n\u00ed podnikov\u00e9ho \u00fakolu, hypot\u00e9zy o hodnot\u011b, kterou lze z dat z\u00edskat, a podnikatelsk\u00e9ho z\u00e1m\u011bru k jej\u00edmu pou\u017eit\u00ed.\u00a0<\/p>\n<p>Samo o sob\u011b pojet\u00ed MLOps vzniklo jako analogie konceptu DevOps aplikovan\u00e9ho na modely a technologie strojov\u00e9ho u\u010den\u00ed. DevOps \u2014 to je p\u0159\u00edstup k v\u00fdvoji softwaru, kter\u00fd umo\u017e\u0148uje zv\u00fd\u0161it rychlost zav\u00e1d\u011bn\u00ed jednotliv\u00fdch zm\u011bn p\u0159i zachov\u00e1n\u00ed flexibility a spolehlivosti pomoc\u00ed \u0159ady p\u0159\u00edstup\u016f, mezi kter\u00fdmi jsou nep\u0159etr\u017eit\u00fd v\u00fdvoj, rozd\u011blen\u00ed funkc\u00ed na \u0159adu nez\u00e1visl\u00fdch mikroslu\u017eeb, automatizovan\u00e9 testov\u00e1n\u00ed a deployment jednotliv\u00fdch zm\u011bn, glob\u00e1ln\u00ed monitoring funk\u010dnosti, syst\u00e9m okam\u017eit\u00e9 reakce na zji\u0161t\u011bn\u00e9 poruchy a dal\u0161\u00ed.\u00a0<\/p>\n<p>DevOps definoval \u017eivotn\u00ed cyklus softwaru a ve spole\u010denstv\u00ed odborn\u00edk\u016f vznikla my\u0161lenka pou\u017e\u00edt stejnou metodiku aplikovanou na velk\u00e1 data. DataOps \u2014 pokus o p\u0159izp\u016fsoben\u00ed a roz\u0161\u00ed\u0159en\u00ed metodiky s ohledem na zvl\u00e1\u0161tnosti ukl\u00e1d\u00e1n\u00ed, p\u0159enosu a zpracov\u00e1n\u00ed velk\u00fdch objem\u016f dat na r\u016fzn\u00fdch a vz\u00e1jemn\u011b propojen\u00fdch platform\u00e1ch.<br \/>\n\u00a0\u00a0<br \/>\nS p\u0159\u00edchodem ur\u010dit\u00e9 kritick\u00e9 masy model\u016f strojov\u00e9ho u\u010den\u00ed implementovan\u00fdch do obchodn\u00edch proces\u016f podnik\u016f bylo pozorov\u00e1no siln\u00e9 podobnosti mezi \u017eivotn\u00edm cyklem matematick\u00fdch model\u016f strojov\u00e9ho u\u010den\u00ed a \u017eivotn\u00edm cyklem softwaru. Rozd\u00edl spo\u010d\u00edv\u00e1 pouze v tom, \u017ee algoritmy model\u016f jsou vytv\u00e1\u0159eny za pomoci n\u00e1stroj\u016f a metod strojov\u00e9ho u\u010den\u00ed. Tak vznikla p\u0159irozen\u00e1 my\u0161lenka aplikovat a p\u0159izp\u016fsobit ji\u017e zn\u00e1m\u00e9 p\u0159\u00edstupy k v\u00fdvoji softwaru pro modely strojov\u00e9ho u\u010den\u00ed. V \u017eivotn\u00edm cyklu model\u016f strojov\u00e9ho u\u010den\u00ed lze proto vy\u010dlenit n\u00e1sleduj\u00edc\u00ed kl\u00ed\u010dov\u00e9 etapy:<\/p>\n<ul>\n<li>definice podnikatelsk\u00e9 my\u0161lenky;\n<\/li>\n<li>vytr\u00e9nov\u00e1n\u00ed modelu;\n<\/li>\n<li>testov\u00e1n\u00ed a nasazen\u00ed modelu do obchodn\u00edho procesu;\n<\/li>\n<li>prev\u00e1dzkov\u00e1n\u00ed modelu.\n<\/li>\n<\/ul>\n<p>\nKdy\u017e se v pr\u016fb\u011bhu provozu objev\u00ed pot\u0159eba zm\u011bnit nebo p\u0159e\u0161kolit model na nov\u00fdch datech, cyklus se spou\u0161t\u00ed znovu \u2014 model se vylep\u0161uje, testuje a nasazuje se nov\u00e1 verze.<\/p>\n<blockquote><p>Odbo\u010dka. Pro\u010d p\u0159e\u0161kolit, a ne p\u0159etr\u00e9novat? Term\u00edn \"p\u0159etr\u00e9nov\u00e1n\u00ed modelu\" m\u00e1 dvoj\u00ed v\u00fdklad: mezi specialisty znamen\u00e1 defekt modelu, kdy\u017e model dob\u0159e predikuje, vlastn\u011b opakuje predikovan\u00fd parametr na tr\u00e9ninkov\u00e9 sad\u011b, ale na extern\u00ed sad\u011b dat pracuje mnohem h\u016f\u0159e. P\u0159irozen\u011b je takov\u00fd model vadn\u00fd, proto\u017ee tento defekt neumo\u017e\u0148uje jeho aplikaci.<\/p><\/blockquote>\n<p>\nV tomto \u017eivotn\u00edm cyklu se jev\u00ed jako logick\u00e9 vyu\u017eit\u00ed DevOps n\u00e1stroj\u016f: automatizovan\u00e9 testov\u00e1n\u00ed, nasazen\u00ed a monitorov\u00e1n\u00ed, zpracov\u00e1n\u00ed v\u00fdpo\u010dt\u016f model\u016f ve form\u011b jednotliv\u00fdch mikro slu\u017eeb. Existuje v\u0161ak tak\u00e9 \u0159ada zvl\u00e1\u0161tnost\u00ed, kter\u00e9 br\u00e1n\u00ed p\u0159\u00edm\u00e9mu pou\u017eit\u00ed t\u011bchto n\u00e1stroj\u016f bez dal\u0161\u00ed ML vazby.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed\" src=\"\/wp-content\/uploads\/2020\/06\/e1d01fd68c3f0071f925a82c08d29d25.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<\/p>\n<h2>Jak p\u0159im\u011bt modely pracovat a p\u0159in\u00e1\u0161et zisk<\/h2>\n<p>\nJako p\u0159\u00edklad, na kter\u00e9m demonstrujeme pou\u017eit\u00ed p\u0159\u00edstupu MLOps, vezmeme si klasick\u00fd \u00fakol automatizace chatu pro podporu bankovn\u00edho (nebo jin\u00e9ho) produktu. Obvykle proces podpory prost\u0159ednictv\u00edm chatu vypad\u00e1 n\u00e1sledovn\u011b: klient zad\u00e1v\u00e1 do chatu zpr\u00e1vu s dotazem a dost\u00e1v\u00e1 odpov\u011b\u010f od specialisty v r\u00e1mci p\u0159edem definovan\u00e9ho stromu dialog\u016f. \u00dakol automatizace takov\u00e9ho chatu se obvykle \u0159e\u0161\u00ed pomoc\u00ed experty definovan\u00fdch sad pravidel, kter\u00e9 jsou velmi pracn\u00e9 na v\u00fdvoj a \u00fadr\u017ebu. Efektivita takov\u00e9 automatizace, v z\u00e1vislosti na \u00farovni slo\u017eitosti \u00fakolu, m\u016f\u017ee dos\u00e1hnout 20\u201330%. P\u0159irozen\u011b se objevuje my\u0161lenka, \u017ee v\u00fdhodn\u011bj\u0161\u00ed je zaveden\u00ed modulu um\u011bl\u00e9 inteligence \u2014 modelu vyvinut\u00e9ho pomoc\u00ed strojov\u00e9ho u\u010den\u00ed, kter\u00fd:<\/p>\n<ul>\n<li>je schopen zpracovat bez \u00fa\u010dasti oper\u00e1tora v\u011bt\u0161\u00ed po\u010det dotaz\u016f (v z\u00e1vislosti na t\u00e9matu, v n\u011bkter\u00fdch p\u0159\u00edpadech m\u016f\u017ee efektivita dos\u00e1hnout 70\u201380%);\n<\/li>\n<li>l\u00e9pe se p\u0159izp\u016fsobuje nestandardn\u00edm formulac\u00edm v dialogu \u2014 um\u00ed ur\u010dit z\u00e1m\u011br, skute\u010dn\u00e9 p\u0159\u00e1n\u00ed u\u017eivatele podle nejasn\u011b formulovan\u00e9ho dotazu;\n<\/li>\n<li>dok\u00e1\u017ee rozpoznat, kdy je odpov\u011b\u010f modelu adekv\u00e1tn\u00ed a kdy jsou pochybnosti o \"uv\u011bdom\u011bn\u00ed\" t\u00e9to odpov\u011bdi a je pot\u0159eba polo\u017eit dodate\u010dnou up\u0159es\u0148uj\u00edc\u00ed ot\u00e1zku nebo p\u0159epnout na oper\u00e1tora;\n<\/li>\n<li>m\u016f\u017ee b\u00fdt automatizovan\u011b dosv\u011bd\u010dena (m\u00edsto skupiny v\u00fdvoj\u00e1\u0159\u016f, kte\u0159\u00ed neust\u00e1le p\u0159izp\u016fsobuj\u00ed a upravuj\u00ed sc\u00e9n\u00e1\u0159e odpov\u011bd\u00ed, model dosv\u011bd\u010duje specialista na Data Science, kter\u00fd vyu\u017e\u00edv\u00e1 p\u0159\u00edslu\u0161n\u00e9 knihovny strojov\u00e9ho u\u010den\u00ed).\u00a0\n<\/li>\n<\/ul>\n<p>\n<img decoding=\"async\" alt=\"MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed\" src=\"\/wp-content\/uploads\/2020\/06\/9ef6356090a250df6cb212ab8bedb5f3.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nJak dostat takov\u00fd pokro\u010dil\u00fd model do chodu?\u00a0<\/p>\n<p>Stejn\u011b jako p\u0159i \u0159e\u0161en\u00ed jak\u00e9hokoli jin\u00e9ho \u00fakolu, ne\u017e za\u010dneme vyv\u00edjet takov\u00fd modul, je nezbytn\u00e9 ur\u010dit obchodn\u00ed proces a form\u00e1ln\u011b popsat konkr\u00e9tn\u00ed \u00fakol, kter\u00fd budeme \u0159e\u0161it pomoc\u00ed metody strojov\u00e9ho u\u010den\u00ed. V tomto bod\u011b za\u010d\u00edn\u00e1 proces operacionalizace, ozna\u010dovan\u00fd zkratkou Ops.\u00a0<\/p>\n<p>Dal\u0161\u00edm krokem je, \u017ee specialist\u00e9 datov\u00e9 v\u011bdy ve spolupr\u00e1ci s datov\u00fdmi in\u017een\u00fdry zkontroluj\u00ed dostupnost a dostate\u010dnost dat a obchodn\u00ed hypot\u00e9zu ohledn\u011b \u017eivotaschopnosti obchodn\u00ed my\u0161lenky, vyvinou prototyp modelu a ov\u011b\u0159\u00ed jeho skute\u010dnou \u00fa\u010dinnost. Teprve po potvrzen\u00ed ze strany podnik\u00e1n\u00ed m\u016f\u017ee za\u010d\u00edt p\u0159echod od v\u00fdvoje modelu k jeho integraci do syst\u00e9m\u016f, kter\u00e9 vykon\u00e1vaj\u00ed konkr\u00e9tn\u00ed obchodn\u00ed proces. D\u016fkladn\u00e9 pl\u00e1nov\u00e1n\u00ed implementace, hlubok\u00e9 porozum\u011bn\u00ed v ka\u017ed\u00e9 f\u00e1zi, jak bude model pou\u017e\u00edv\u00e1n a jak\u00fd ekonomick\u00fd efekt p\u0159inese, je kl\u00ed\u010dov\u00fdm momentem v procesech implementace p\u0159\u00edstup\u016f MLOps do technologick\u00e9ho prost\u0159ed\u00ed spole\u010dnosti.<\/p>\n<p>S rozvojem technologi\u00ed AI se dramaticky zvy\u0161uje mno\u017estv\u00ed a rozmanitost \u00faloh, kter\u00e9 mohou b\u00fdt \u0159e\u0161eny pomoc\u00ed strojov\u00e9ho u\u010den\u00ed. Ka\u017ed\u00fd takov\u00fd obchodn\u00ed proces znamen\u00e1 \u00fasporu pro spole\u010dnost prost\u0159ednictv\u00edm automatizace pr\u00e1ce zam\u011bstnanc\u016f ve frontov\u00fdch pozic\u00edch (call centrum, kontrola a t\u0159\u00edd\u011bn\u00ed dokument\u016f atd.), roz\u0161\u00ed\u0159en\u00ed z\u00e1kaznick\u00e9 z\u00e1kladny p\u0159id\u00e1n\u00edm nov\u00fdch atraktivn\u00edch a pohodln\u00fdch funkc\u00ed, \u00fasporu prost\u0159edk\u016f d\u00edky optim\u00e1ln\u00edmu vyu\u017eit\u00ed a p\u0159eorganizaci zdroj\u016f a mnoho dal\u0161\u00edho. V kone\u010dn\u00e9m d\u016fsledku je ka\u017ed\u00fd proces zam\u011b\u0159en na vytv\u00e1\u0159en\u00ed hodnoty a jako d\u016fsledek by m\u011bl p\u0159in\u00e1\u0161et ur\u010dit\u00fd ekonomick\u00fd efekt. Zde je velmi d\u016fle\u017eit\u00e9 jasn\u011b formulovat obchodn\u00ed my\u0161lenku a vypo\u010d\u00edtat o\u010dek\u00e1van\u00fd zisk z implementace modelu v celkov\u00e9 struktu\u0159e vytv\u00e1\u0159en\u00ed hodnoty spole\u010dnosti. St\u00e1v\u00e1 se, \u017ee implementace modelu se neosv\u011bd\u010d\u00ed a \u010das, kter\u00fd specialist\u00e9 na strojov\u00e9 u\u010den\u00ed v\u011bnuj\u00ed, vyjde mnohem dr\u00e1\u017ee ne\u017e pracovn\u00ed m\u00edsto oper\u00e1tora, kter\u00fd tuto \u00falohu vykon\u00e1v\u00e1. Proto je t\u0159eba se sna\u017eit odhalit takov\u00e9 p\u0159\u00edpady v rann\u00fdch f\u00e1z\u00edch tvorby syst\u00e9m\u016f AI.<\/p>\n<p>Z toho tedy vypl\u00fdv\u00e1, \u017ee model za\u010d\u00edn\u00e1 p\u0159in\u00e1\u0161et zisk teprve tehdy, kdy\u017e byla v procesu MLOps spr\u00e1vn\u011b formulov\u00e1na obchodn\u00ed \u00faloha, nastaveny priority a v ran\u00fdch f\u00e1z\u00edch v\u00fdvoje byl formulov\u00e1n proces implementace modelu do syst\u00e9mu.<\/p>\n<h2>Nov\u00fd proces \u2013 nov\u00e9 v\u00fdzvy<\/h2>\n<p>\nKomplexn\u00ed odpov\u011b\u010f na z\u00e1sadn\u00ed ot\u00e1zku podnik\u00e1n\u00ed ohledn\u011b toho, jak jsou modely strojov\u00e9ho u\u010den\u00ed aplikovateln\u00e9 na \u0159e\u0161en\u00ed \u00fakol\u016f, a obecn\u00e1 ot\u00e1zka d\u016fv\u011bry v AI \u2014 to je jeden z kl\u00ed\u010dov\u00fdch v\u00fdzev v procesu rozvoje a implementace p\u0159\u00edstup\u016f MLOps. Zpo\u010d\u00e1tku podniky skepticky vn\u00edmaj\u00ed zav\u00e1d\u011bn\u00ed strojov\u00e9ho u\u010den\u00ed do proces\u016f \u2014 je obt\u00ed\u017en\u00e9 spol\u00e9hat se na modely na m\u00edstech, kde d\u0159\u00edve obvykle pracovali lid\u00e9. Pro podniky se programy jev\u00ed jako \u201e\u010dern\u00e1 sk\u0159\u00ed\u0148ka\u201c, jej\u00ed\u017e relevantnost odpov\u011bd\u00ed mus\u00ed b\u00fdt je\u0161t\u011b prok\u00e1z\u00e1na. Krom\u011b toho existuj\u00ed v bankovn\u00ed \u010dinnosti, v podnik\u00e1n\u00ed telekomunika\u010dn\u00edch oper\u00e1tor\u016f a jinde p\u0159\u00edsn\u00e9 po\u017eadavky st\u00e1tn\u00edch regul\u00e1tor\u016f. V\u0161echny syst\u00e9my a algoritmy, kter\u00e9 jsou implementov\u00e1ny v bankovn\u00edch procesech, podl\u00e9haj\u00ed auditu. Abychom tuto \u00falohu vy\u0159e\u0161ili a prok\u00e1zali podnik\u016fm a regul\u00e1tor\u016fm opr\u00e1vn\u011bnost a spr\u00e1vnost odpov\u011bd\u00ed um\u011bl\u00e9 inteligence, spolu s modelem jsou implementov\u00e1ny monitoringov\u00e9 n\u00e1stroje. D\u00e1le existuje postup nez\u00e1visl\u00e9 validace, kter\u00fd je povinn\u00fd pro regulovan\u00e9 modely a kter\u00fd odpov\u00edd\u00e1 po\u017eadavk\u016fm centr\u00e1ln\u00ed banky. Nez\u00e1visl\u00e1 expertn\u00ed skupina prov\u00e1d\u00ed audit v\u00fdsledk\u016f z\u00edskan\u00fdch modelem s ohledem na vstupn\u00ed data.<\/p>\n<p>Druh\u00e1 v\u00fdzva \u2014 hodnocen\u00ed a zohledn\u011bn\u00ed modelov\u00fdch rizik p\u0159i zav\u00e1d\u011bn\u00ed modelu strojov\u00e9ho u\u010den\u00ed. I kdy\u017e ani \u010dlov\u011bk nem\u016f\u017ee stoprocentn\u011b jist\u011b odpov\u011bd\u011bt na ot\u00e1zku, zda bylo to konkr\u00e9tn\u00ed \u0161aty b\u00edl\u00e9 nebo modr\u00e9, m\u00e1 um\u011bl\u00e1 inteligence tak\u00e9 pr\u00e1vo na chybu. Je tak\u00e9 t\u0159eba vz\u00edt v \u00favahu, \u017ee s \u010dasem se data mohou m\u011bnit a modely se mus\u00ed d\u00e1le \u0161kolit, aby vyd\u00e1valy dostate\u010dn\u011b p\u0159esn\u00fd v\u00fdsledek. Aby podnikov\u00e9 procesy neutrp\u011bly, je nutn\u00e9 \u0159\u00eddit modelov\u00e1 rizika a sledovat pr\u00e1ci modelu, pravideln\u011b jej d\u00e1le \u0161kolit na nov\u00fdch datech.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed\" src=\"\/wp-content\/uploads\/2020\/06\/cb97ff0eab95d81e17bf76aad3c3994e.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nAle po prvn\u00ed f\u00e1zi ned\u016fv\u011bry za\u010d\u00edn\u00e1 p\u0159ich\u00e1zet opa\u010dn\u00fd efekt. \u010c\u00edm v\u00edce model\u016f je \u00fasp\u011b\u0161n\u011b implementov\u00e1no do proces\u016f, t\u00edm v\u011bt\u0161\u00ed m\u00e1 podnik chu\u0165 vyu\u017e\u00edvat um\u011blou inteligenci \u2014 objevuj\u00ed se nov\u00e9 \u00fakoly, kter\u00e9 lze vy\u0159e\u0161it metodami strojov\u00e9ho u\u010den\u00ed. Ka\u017ed\u00fd \u00fakol spou\u0161t\u00ed cel\u00fd proces, kter\u00fd vy\u017eaduje r\u016fzn\u00e9 kompetence:<\/p>\n<ul>\n<li>datov\u00ed in\u017een\u00fd\u0159i p\u0159ipravuj\u00ed a zpracov\u00e1vaj\u00ed data;\n<\/li>\n<li>datov\u00ed v\u011bdci aplikuj\u00ed n\u00e1stroje strojov\u00e9ho u\u010den\u00ed a vyv\u00edjej\u00ed model;\n<\/li>\n<li>IT zav\u00e1d\u011bj\u00ed model do syst\u00e9mu;\n<\/li>\n<li>ML in\u017een\u00fdr ur\u010duje, jak spr\u00e1vn\u011b integrovat tento model do procesu, jak\u00e9 IT n\u00e1stroje pou\u017e\u00edt v z\u00e1vislosti na po\u017eadavc\u00edch na re\u017eim aplikace modelu s ohledem na tok po\u017eadavk\u016f, dobu odezvy atd.\u00a0\n<\/li>\n<li>ML architekt navrhuje, jak fyzicky realizovat softwarov\u00fd produkt v pr\u016fmyslov\u00e9m syst\u00e9mu.\n<\/li>\n<\/ul>\n<p>\nCel\u00fd cyklus vy\u017eaduje velk\u00e9 mno\u017estv\u00ed vysoce kvalifikovan\u00fdch specialist\u016f. V ur\u010dit\u00e9m bod\u011b v\u00fdvoje a stupn\u011b pronik\u00e1n\u00ed ML model\u016f do obchodn\u00edch proces\u016f se ukazuje, \u017ee line\u00e1rn\u00ed \u0161k\u00e1lov\u00e1n\u00ed po\u010dtu specialist\u016f \u00fam\u011brn\u011b r\u016fstu po\u010dtu \u00faloh se st\u00e1v\u00e1 drah\u00fdm a neefektivn\u00edm. Proto vyvst\u00e1v\u00e1 ot\u00e1zka automatizace procesu MLOps - definice n\u011bkolika standardn\u00edch t\u0159\u00edd \u00faloh strojov\u00e9ho u\u010den\u00ed, v\u00fdvoj typov\u00fdch pipeline pro zpracov\u00e1n\u00ed dat a dotrenov\u00e1n\u00ed model\u016f. V ide\u00e1ln\u00edm p\u0159\u00edpad\u011b pro \u0159e\u0161en\u00ed takov\u00fdch \u00faloh pot\u0159ebujeme profesion\u00e1ly, kte\u0159\u00ed ovl\u00e1daj\u00ed kompetence na pomez\u00ed Big Data, Data Science, DevOps a IT. Proto je nejv\u011bt\u0161\u00edm probl\u00e9mem v oboru Data Science a nejv\u011bt\u0161\u00ed v\u00fdzvou p\u0159i organizaci proces\u016f MLOps nedostatek takov\u00fdch kompetenc\u00ed na st\u00e1vaj\u00edc\u00edm trhu pracovn\u00edch sil. Specialist\u00e9 spl\u0148uj\u00edc\u00ed takov\u00e9 po\u017eadavky jsou v sou\u010dasnosti na trhu pr\u00e1ce ojedin\u011bl\u00ed a maj\u00ed cenu zlata.<\/p>\n<h2>K ot\u00e1zce kompetenc\u00ed<\/h2>\n<p>\nV teorii lze v\u0161echny \u00falohy MLOps \u0159e\u0161it klasick\u00fdmi n\u00e1stroji DevOps a neuchylovat se k specializovan\u00e9mu roz\u0161\u00ed\u0159en\u00ed rolov\u00e9 modelu. Jak jsme ji\u017e uvedli v\u00fd\u0161e, data scientist mus\u00ed b\u00fdt nejen matematikem a specialistou na anal\u00fdzu dat, ale i guru cel\u00e9ho pipeline \u2013 na jeho bedrech le\u017e\u00ed v\u00fdvoj architektury, programov\u00e1n\u00ed model\u016f v n\u011bkolika jazyc\u00edch v z\u00e1vislosti na architektu\u0159e, p\u0159\u00edprava datov\u00e9 vitr\u00edny a nasazen\u00ed samotn\u00e9 aplikace. Nicm\u00e9n\u011b vytv\u00e1\u0159en\u00ed technologick\u00e9ho r\u00e1mce, realizovan\u00e9ho v kondenza\u010dn\u00edm procesu MLOps, zab\u00edr\u00e1 a\u017e 80% pracovn\u00edch zdroj\u016f, co\u017e znamen\u00e1, \u017ee kvalifikovan\u00fd matematik, kter\u00fdm je kvalitn\u00ed Data Scientist, bude pouze 20% \u010dasu v\u011bnovat sv\u00e9 specializaci. Proto se odd\u011blen\u00ed rol\u00ed specialist\u016f, kte\u0159\u00ed realizuj\u00ed proces zav\u00e1d\u011bn\u00ed model\u016f strojov\u00e9ho u\u010den\u00ed, st\u00e1v\u00e1 \u017eivotn\u011b d\u016fle\u017eit\u00fdm.\u00a0<\/p>\n<p>Jak podrobn\u011b by m\u011bly b\u00fdt role rozd\u011bleny, z\u00e1vis\u00ed na velikosti podniku. Je jin\u00e1 situace, kdy\u017e ve startupu pracuje jeden odborn\u00edk, kter\u00fd je s\u00e1m sob\u011b in\u017een\u00fdrem, architektem a DevOps specialistou. Naproti tomu, v r\u00e1mci velk\u00e9ho podniku jsou v\u0161echny procesy v\u00fdvoje model\u016f soust\u0159ed\u011bny na n\u011bkolik vysoce kvalifikovan\u00fdch odborn\u00edk\u016f na Data Science, zat\u00edmco program\u00e1tor \u010di odborn\u00edk na datab\u00e1ze \u2014 o n\u011bco b\u011b\u017en\u011bj\u0161\u00ed a levn\u011bj\u0161\u00ed kompetence na trhu pr\u00e1ce \u2014 si m\u016f\u017ee vz\u00edt na starost v\u011bt\u0161inu rutinn\u00edch \u00fakol\u016f.<\/p>\n<p>Od toho, kde le\u017e\u00ed hranice p\u0159i v\u00fdb\u011bru specialist\u016f pro zaji\u0161t\u011bn\u00ed procesu MLOps a jak je organizov\u00e1n proces operacionalizace vyv\u00edjen\u00fdch model\u016f, p\u0159\u00edmo z\u00e1vis\u00ed rychlost a kvalita vyvinut\u00fdch model\u016f, produktivita t\u00fdmu a mikroklima v n\u011bm.<\/p>\n<h2>Co u\u017e nyn\u00ed ud\u011blala na\u0161e t\u00fdm<\/h2>\n<p>\nNed\u00e1vno jsme za\u010dali budovat strukturu kompetenc\u00ed a procesy MLOps. P\u0159esto u\u017e nyn\u00ed jsou na\u0161e projekty pro spr\u00e1vu \u017eivotn\u00edho cyklu model\u016f a aplikaci model\u016f jako slu\u017eby ve f\u00e1zi testov\u00e1n\u00ed MVP.<\/p>\n<p>Tak\u00e9 jsme ur\u010dili optim\u00e1ln\u00ed strukturu kompetenc\u00ed a organiza\u010dn\u00ed strukturu interakce mezi v\u0161emi \u00fa\u010dastn\u00edky procesu pro velk\u00fd podnik. Byly organizov\u00e1ny t\u00fdmy Agile, kter\u00e9 \u0159e\u0161\u00ed \u00fakoly pro cel\u00fd spektrum obchodn\u00edch z\u00e1kazn\u00edk\u016f, a byl zaji\u0161t\u011bn proces interakce s projektov\u00fdmi t\u00fdmy pro vytv\u00e1\u0159en\u00ed platforem a infrastruktury, kter\u00e1 je z\u00e1kladem budovan\u00e9 MLOps struktury.<\/p>\n<h2>Ot\u00e1zky pro budoucnost<\/h2>\n<p>\nMLOps je rozv\u00edjej\u00edc\u00ed se oblast, kter\u00e1 trp\u00ed nedostatkem kompetenc\u00ed a v budoucnu bude nab\u00edrat na obr\u00e1tk\u00e1ch. V tuto chv\u00edli je nejlep\u0161\u00ed vych\u00e1zet z poznatk\u016f a praktik DevOps. Hlavn\u00edm c\u00edlem MLOps je efektivn\u011bj\u0161\u00ed vyu\u017eit\u00ed ML model\u016f pro \u0159e\u0161en\u00ed obchodn\u00edch \u00fakol\u016f. P\u0159i tom v\u0161ak vyvst\u00e1v\u00e1 \u0159ada ot\u00e1zek:<\/p>\n<ul>\n<li>Jak zkr\u00e1tit \u010das pot\u0159ebn\u00fd k uveden\u00ed model\u016f do produkce?\n<\/li>\n<li>Jak sn\u00ed\u017eit byrokratick\u00e9 t\u0159en\u00ed mezi t\u00fdmy r\u016fzn\u00fdch kompetenc\u00ed a zv\u00fd\u0161it zam\u011b\u0159en\u00ed na spolupr\u00e1ci?\n<\/li>\n<li>Jak sledovat modely, spravovat verze a organizovat efektivn\u00ed monitoring?\n<\/li>\n<li>Jak vytvo\u0159it skute\u010dn\u011b cyklick\u00fd \u017eivotn\u00ed cyklus pro modern\u00ed ML model?\n<\/li>\n<li>Jak standardizovat proces strojov\u00e9ho u\u010den\u00ed?\n<\/li>\n<\/ul>\n<p>\nOdpov\u011bdi na tyto ot\u00e1zky do zna\u010dn\u00e9 m\u00edry ur\u010d\u00ed, jak rychle MLOps napln\u00ed sv\u016fj potenci\u00e1l.<br \/>\n<br \/>Zdroj: <a content=\"nofollow\" rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/vtb\/blog\/508012\/\">habr.com<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>\u0412 2018 \u0433\u043e\u0434\u0443 \u0432 \u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\u043e\u043d\u0430\u043b\u044c\u043d\u044b\u0445 \u043a\u0440\u0443\u0433\u0430\u0445 \u0438 \u043d\u0430 \u0442\u0435\u043c\u0430\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043a\u043e\u043d\u0444\u0435\u0440\u0435\u043d\u0446\u0438\u044f\u0445, \u043f\u043e\u0441\u0432\u044f\u0449\u0435\u043d\u043d\u044b\u0445 AI, \u043f\u043e\u044f\u0432\u0438\u043b\u043e\u0441\u044c \u043f\u043e\u043d\u044f\u0442\u0438\u0435 MLOps, \u043a\u043e\u0442\u043e\u0440\u043e\u0435 \u0431\u044b\u0441\u0442\u0440\u043e \u0437\u0430\u043a\u0440\u0435\u043f\u0438\u043b\u043e\u0441\u044c \u0432 \u043e\u0442\u0440\u0430\u0441\u043b\u0438 \u0438 \u0441\u0435\u0439\u0447\u0430\u0441 \u0440\u0430\u0437\u0432\u0438\u0432\u0430\u0435\u0442\u0441\u044f \u043a\u0430\u043a \u0441\u0430\u043c\u043e\u0441\u0442\u043e\u044f\u0442\u0435\u043b\u044c\u043d\u043e\u0435 \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435. \u0412 \u043f\u0435\u0440\u0441\u043f\u0435\u043a\u0442\u0438\u0432\u0435 MLOps \u043c\u043e\u0436\u0435\u0442 \u0441\u0442\u0430\u0442\u044c \u043e\u0434\u043d\u043e\u0439 \u0438\u0437 \u043d\u0430\u0438\u0431\u043e\u043b\u0435\u0435 \u0432\u043e\u0441\u0442\u0440\u0435\u0431\u043e\u0432\u0430\u043d\u043d\u044b\u0445 \u0441\u0444\u0435\u0440 \u0432 IT. \u0427\u0442\u043e \u0436\u0435 \u044d\u0442\u043e \u0442\u0430\u043a\u043e\u0435 \u0438 \u0441 \u0447\u0435\u043c \u0435\u0433\u043e \u0435\u0434\u044f\u0442, \u0440\u0430\u0437\u0431\u0438\u0440\u0430\u0435\u043c\u0441\u044f \u043f\u043e\u0434 \u043a\u0430\u0442\u043e\u043c. \u0427\u0442\u043e \u0442\u0430\u043a\u043e\u0435 MLOps MLOps (\u043e\u0431\u044a\u0435\u0434\u0438\u043d\u0435\u043d\u0438\u0435 [&hellip;]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":86524,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[688],"tags":[],"class_list":["post-86523","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-administrirovanie"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 5.0.3 - aioseo.com -->\n\t<meta name=\"description\" content=\"\u0412 2018 \u0433\u043e\u0434\u0443 \u0432 \u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\u043e\u043d\u0430\u043b\u044c\u043d\u044b\u0445 \u043a\u0440\u0443\u0433\u0430\u0445 \u0438 \u043d\u0430 \u0442\u0435\u043c\u0430\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043a\u043e\u043d\u0444\u0435\u0440\u0435\u043d\u0446\u0438\u044f\u0445, \u043f\u043e\u0441\u0432\u044f\u0449\u0435\u043d\u043d\u044b\u0445 AI, \u043f\u043e\u044f\u0432\u0438\u043b\u043e\u0441\u044c \u043f\u043e\u043d\u044f\u0442\u0438\u0435 MLOps, \u043a\u043e\u0442\u043e\u0440\u043e\u0435 \u0431\u044b\u0441\u0442\u0440\u043e \u0437\u0430\u043a\u0440\u0435\u043f\u0438\u043b\u043e\u0441\u044c \u0432 \u043e\u0442\u0440\u0430\u0441\u043b\u0438 \u0438 \u0441\u0435\u0439\u0447\u0430\u0441 \u0440\u0430\u0437\u0432\u0438\u0432\u0430\u0435\u0442\u0441\u044f \u043a\u0430\u043a \u0441\u0430\u043c\u043e\u0441\u0442\u043e\u044f\u0442\u0435\u043b\u044c\u043d\u043e\u0435 \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435.\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Yuri Gagarin\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.3\" \/>\n\t\t<meta property=\"og:locale\" content=\"cs_CZ\" \/>\n\t\t<meta property=\"og:site_name\" content=\"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"\ud83e\udd47MLOps: DevOps \u0432 \u043c\u0438\u0440\u0435 Machine Learning | ProHoster\" \/>\n\t\t<meta property=\"og:description\" content=\"\u0412 2018 \u0433\u043e\u0434\u0443 \u0432 \u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\u043e\u043d\u0430\u043b\u044c\u043d\u044b\u0445 \u043a\u0440\u0443\u0433\u0430\u0445 \u0438 \u043d\u0430 \u0442\u0435\u043c\u0430\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043a\u043e\u043d\u0444\u0435\u0440\u0435\u043d\u0446\u0438\u044f\u0445, \u043f\u043e\u0441\u0432\u044f\u0449\u0435\u043d\u043d\u044b\u0445 AI, \u043f\u043e\u044f\u0432\u0438\u043b\u043e\u0441\u044c \u043f\u043e\u043d\u044f\u0442\u0438\u0435 MLOps, \u043a\u043e\u0442\u043e\u0440\u043e\u0435 \u0431\u044b\u0441\u0442\u0440\u043e \u0437\u0430\u043a\u0440\u0435\u043f\u0438\u043b\u043e\u0441\u044c \u0432 \u043e\u0442\u0440\u0430\u0441\u043b\u0438 \u0438 \u0441\u0435\u0439\u0447\u0430\u0441 \u0440\u0430\u0437\u0432\u0438\u0432\u0430\u0435\u0442\u0441\u044f \u043a\u0430\u043a \u0441\u0430\u043c\u043e\u0441\u0442\u043e\u044f\u0442\u0435\u043b\u044c\u043d\u043e\u0435 \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435.\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning\" \/>\n\t\t<meta property=\"og:image\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:secure_url\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:width\" content=\"350\" \/>\n\t\t<meta property=\"og:image:height\" content=\"350\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2020-06-26T05:42:14+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2020-06-26T05:42:14+00:00\" \/>\n\t\t<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"\ud83e\udd47MLOps: DevOps ve sv\u011bt\u011b strojov\u00e9ho u\u010den\u00ed | ProHoster","description":"V roce 2018 se v profesion\u00e1ln\u00edch kruz\u00edch a na tematick\u00fdch konferenc\u00edch v\u011bnovan\u00fdch AI objevila my\u0161lenka MLOps, kter\u00e1 se rychle etablovala v oboru a nyn\u00ed se vyv\u00edj\u00ed jako samostatn\u00fd sm\u011br.","canonical_url":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"cs_CZ","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47MLOps: DevOps \u0432 \u043c\u0438\u0440\u0435 Machine Learning | ProHoster","og:description":"\u0412 2018 \u0433\u043e\u0434\u0443 \u0432 \u043f\u0440\u043e\u0444\u0435\u0441\u0441\u0438\u043e\u043d\u0430\u043b\u044c\u043d\u044b\u0445 \u043a\u0440\u0443\u0433\u0430\u0445 \u0438 \u043d\u0430 \u0442\u0435\u043c\u0430\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u043a\u043e\u043d\u0444\u0435\u0440\u0435\u043d\u0446\u0438\u044f\u0445, \u043f\u043e\u0441\u0432\u044f\u0449\u0435\u043d\u043d\u044b\u0445 AI, \u043f\u043e\u044f\u0432\u0438\u043b\u043e\u0441\u044c \u043f\u043e\u043d\u044f\u0442\u0438\u0435 MLOps, \u043a\u043e\u0442\u043e\u0440\u043e\u0435 \u0431\u044b\u0441\u0442\u0440\u043e \u0437\u0430\u043a\u0440\u0435\u043f\u0438\u043b\u043e\u0441\u044c \u0432 \u043e\u0442\u0440\u0430\u0441\u043b\u0438 \u0438 \u0441\u0435\u0439\u0447\u0430\u0441 \u0440\u0430\u0437\u0432\u0438\u0432\u0430\u0435\u0442\u0441\u044f \u043a\u0430\u043a \u0441\u0430\u043c\u043e\u0441\u0442\u043e\u044f\u0442\u0435\u043b\u044c\u043d\u043e\u0435 \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435.","og:url":"https:\/\/prohoster.info\/cs\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2020-06-26T05:42:14+00:00","article:modified_time":"2020-06-26T05:42:14+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"86523","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":null,"breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-02-28 14:13:23","updated":"2022-09-28 09:22:54","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/86523","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/comments?post=86523"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/posts\/86523\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media\/86524"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/media?parent=86523"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/categories?post=86523"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/cs\/wp-json\/wp\/v2\/tags?post=86523"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}