{"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\/ro\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","title":{"rendered":"MLOps: DevOps \u00een lumea Machine Learning","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>\u00cen 2018, \u00een cercurile profesionale \u0219i la conferin\u021bele tematice dedicate AI, a ap\u0103rut termenul MLOps, care s-a consolidat rapid \u00een industrie \u0219i acum se dezvolt\u0103 ca o direc\u021bie de sine st\u0103t\u0103toare. \u00cen perspectiva viitoare, MLOps ar putea deveni unul dintre cele mai c\u0103utate domenii din IT. Ce este acesta \u0219i cu ce se consum\u0103, afl\u0103m \u00een continuare.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps \u00een lumea Machine Learning\" 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>Ce este MLOps<\/h2>\n<p>\nMLOps (combinarea tehnologiilor \u0219i proceselor de \u00eenv\u0103\u021bare automat\u0103 \u0219i a abord\u0103rilor pentru implementarea modelelor dezvoltate \u00een procesele de afaceri) este o nou\u0103 modalitate de colaborare \u00eentre reprezentan\u021bii afacerilor, cercet\u0103tori, matematicieni, speciali\u0219ti \u00een \u00eenv\u0103\u021barea automat\u0103 \u0219i inginerii IT \u00een crearea sistemelor de inteligen\u021b\u0103 artificial\u0103.<\/p>\n<p>Cu alte cuvinte, este o modalitate de transformare a metodelor \u0219i tehnologiilor de \u00eenv\u0103\u021bare automat\u0103 \u00een unelte utile pentru solu\u021bionarea problemelor de afaceri.\u00a0<\/p>\n<p>Trebuie s\u0103 \u00een\u021belegem c\u0103 lan\u021bul de productivizare \u00eencepe mult \u00eenainte de dezvoltarea modelului. Primul s\u0103u pas este definirea obiectivului de afaceri, ipotezele despre valoarea care poate fi extras\u0103 din date \u0219i idea de afaceri pentru aplicarea acesteia.\u00a0<\/p>\n<p>Conceptul de MLOps a ap\u0103rut ca o analogie a conceptului DevOps aplicat modelelor \u0219i tehnologiilor de \u00eenv\u0103\u021bare automat\u0103. DevOps este o abordare a dezvolt\u0103rii software-ului ce permite cre\u0219terea vitezei de implementare a unor modific\u0103ri individuale, men\u021bin\u00e2nd \u00een acela\u0219i timp flexibilitatea \u0219i fiabilitatea printr-o serie de metode, dintre care se num\u0103r\u0103 dezvoltarea continu\u0103, divizarea func\u021biilor \u00een microservicii independente, testarea automatizat\u0103 \u0219i implementarea modific\u0103rilor individuale, monitorizarea global\u0103 a func\u021bion\u0103rii, sistemul de reac\u021bie rapid\u0103 la defec\u021biuni identificate etc.\u00a0<\/p>\n<p>DevOps a definit ciclul de via\u021b\u0103 al software-ului \u0219i \u00een comunitatea speciali\u0219tilor a ap\u0103rut ideea de a utiliza aceea\u0219i metodologie aplicat\u0103 datelor mari. DataOps este o \u00eencercare de a adapta \u0219i extinde metodologia \u021bin\u00e2nd cont de specificit\u0103\u021bile stoc\u0103rii, transiterii \u0219i proces\u0103rii unor seturi mari de date \u00een platforme diverse \u0219i interconectate.<br \/>\n\u00a0\u00a0<br \/>\nOdat\u0103 cu apari\u021bia unei mase critice de modele de \u00eenv\u0103\u021bare automat\u0103 integrate \u00een procesele de afaceri ale \u00eentreprinderilor, a fost observat\u0103 o asem\u0103nare puternic\u0103 \u00eentre ciclul de via\u021b\u0103 al modelului de \u00eenv\u0103\u021bare automat\u0103 \u0219i ciclul de via\u021b\u0103 al software-ului. Diferen\u021ba este c\u0103 algoritmii modelului sunt crea\u021bi cu ajutorul instrumentelor \u0219i metodelor de \u00eenv\u0103\u021bare automat\u0103. Prin urmare, a ap\u0103rut \u00een mod natural ideea de a aplica \u0219i adapta abord\u0103rile cunoscute de dezvoltare a software-ului pentru modelele de \u00eenv\u0103\u021bare automat\u0103. Astfel, \u00een ciclul de via\u021b\u0103 al modelelor de \u00eenv\u0103\u021bare automat\u0103 se pot identifica urm\u0103toarele etape cheie:<\/p>\n<ul>\n<li>definirea ideii de afaceri;\n<\/li>\n<li>antrenarea modelului;\n<\/li>\n<li>testarea \u0219i implementarea modelului \u00een procesul de afaceri;\n<\/li>\n<li>exploatarea modelului.\n<\/li>\n<\/ul>\n<p>\nAtunci c\u00e2nd \u00een timpul exploat\u0103rii apare necesitatea de a modifica sau de a reantrena modelul pe date noi, ciclul se reia - modelul este revizuit, testat \u0219i se lanseaz\u0103 o nou\u0103 versiune.<\/p>\n<blockquote><p>Dezvoltare. De ce reantrenare, nu reconstruc\u021bie? Termenul \u201ereconstruc\u021bie a modelului\u201d are o interpretare ambivalent\u0103: printre speciali\u0219ti, acesta semnific\u0103 un defect al modelului, atunci c\u00e2nd modelul face predic\u021bii bune, reproduc\u00e2nd efectiv parametrul previzibil pe setul de antrenament, dar func\u021bioneaz\u0103 mult mai slab pe un set extern de date. Evident, un astfel de model este considerat defect, deoarece acest defect nu \u00eei permite utilizarea.<\/p><\/blockquote>\n<p>\n\u00cen acest ciclu de via\u021b\u0103, utilizarea instrumentelor DevOps pare logic\u0103: testare automat\u0103, desf\u0103\u0219urare \u0219i monitorizare, prezentarea calculului modelului sub form\u0103 de microservicii separate. Totu\u0219i, exist\u0103 \u0219i o serie de particularit\u0103\u021bi care \u00eempiedic\u0103 aplicarea direct\u0103 a acestor instrumente f\u0103r\u0103 un adaos suplimentar de ML.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps \u00een lumea Machine Learning\" src=\"\/wp-content\/uploads\/2020\/06\/e1d01fd68c3f0071f925a82c08d29d25.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<\/p>\n<h2>Cum s\u0103 facem modelele s\u0103 func\u021bioneze \u0219i s\u0103 genereze profit<\/h2>\n<p>\nCa exemplu, \u00een care vom demonstra aplicarea abord\u0103rii MLOps, s\u0103 lu\u0103m o problem\u0103 clasic\u0103 de automatizare a chatului de suport pentru un produs bancar (sau orice alt produs). De obicei, procesul de afaceri de suport prin chat arat\u0103 \u00een felul urm\u0103tor: clientul introduce un mesaj cu o \u00eentrebare \u00een chat \u0219i prime\u0219te un r\u0103spuns de la specialist \u00een cadrul unui arbore de dialoguri prestabilit. Sarcina automatiz\u0103rii acestui chat este, de obicei, rezolvat\u0103 prin seturi de reguli definite de exper\u021bi, care sunt foarte laborioase de dezvoltat \u0219i \u00eentre\u021binut. Eficien\u021ba unei astfel de automatiz\u0103ri, \u00een func\u021bie de complexitatea sarcinii, poate atinge 20\u201330%. Evident, apare ideea c\u0103 este mai profitabil s\u0103 implement\u0103m un modul de inteligen\u021b\u0103 artificial\u0103 \u2014 un model dezvoltat prin \u00eenv\u0103\u021bare automat\u0103, care:<\/p>\n<ul>\n<li>este capabil s\u0103 proceseze f\u0103r\u0103 participarea operatorului un num\u0103r mai mare de solicit\u0103ri (\u00een func\u021bie de subiect, \u00een unele cazuri eficien\u021ba poate ajunge la 70\u201380%);\n<\/li>\n<li>se adapteaz\u0103 mai bine la formul\u0103ri neobi\u0219nuite \u00een dialog \u2014 \u0219tie s\u0103 determine intentia, dorin\u021ba real\u0103 a utilizatorului dintr-o solicitare formulat\u0103 neclar;\n<\/li>\n<li>\u0219tie s\u0103 determine c\u00e2nd r\u0103spunsul modelului este adecvat \u0219i c\u00e2nd, av\u00e2nd \u00een vedere 'con\u0219tien\u021ba' acestui r\u0103spuns, exist\u0103 \u00eendoieli \u0219i trebuie s\u0103 pun\u0103 o \u00eentrebare suplimentar\u0103 de clarificare sau s\u0103 se \u00eendrepte c\u0103tre operator;\n<\/li>\n<li>poate fi \u00eembun\u0103t\u0103\u021bit\u0103 automat (\u00een locul unui grup de dezvoltatori care adapteaz\u0103 \u0219i corecteaz\u0103 constant scenariile de r\u0103spuns, modelul este perfec\u021bionat de un specialist \u00een Data Science, folosind bibliotecile corespunz\u0103toare de \u00eenv\u0103\u021bare automat\u0103).\u00a0\n<\/li>\n<\/ul>\n<p>\n<img decoding=\"async\" alt=\"MLOps: DevOps \u00een lumea Machine Learning\" src=\"\/wp-content\/uploads\/2020\/06\/9ef6356090a250df6cb212ab8bedb5f3.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nCum putem face un astfel de model avansat s\u0103 func\u021bioneze?\u00a0<\/p>\n<p>A\u0219a cum se \u00eent\u00e2mpl\u0103 cu orice alt\u0103 problem\u0103, \u00eenainte de a dezvolta un astfel de modul, este necesar s\u0103 definim procesul de afaceri \u0219i s\u0103 descriem formal sarcina specific\u0103 pe care o vom rezolva cu ajutorul metodei de \u00eenv\u0103\u021bare automat\u0103. Acest punct marcheaz\u0103 \u00eenceputul procesului de operativizare, denumit \u00een prescurtare Ops.\u00a0<\/p>\n<p>Urm\u0103torul pas pentru specialistul \u00een Data Science, \u00een colaborare cu inginerul de date, este de a verifica disponibilitatea \u0219i suficien\u021ba datelor \u0219i a ipotezei de afaceri referitoare la viabilitatea ideii de afaceri, dezvolt\u00e2nd un prototip al modelului \u0219i test\u00e2nd efectiv eficien\u021ba acestuia. Numai dup\u0103 confirmarea de c\u0103tre afacere se poate \u00eencepe tranzi\u021bia de la dezvoltarea modelului la integrarea acestuia \u00een sistemele care execut\u0103 un proces de afaceri specific. Planificarea implement\u0103rii de la un cap\u0103t la altul, o \u00een\u021belegere profund\u0103 la fiecare etap\u0103 a modului \u00een care modelul va fi utilizat \u0219i a efectului economic pe care \u00eel va aduce, reprezint\u0103 un punct fundamental \u00een procesele de implementare a abord\u0103rilor MLOps \u00een peisajul tehnologic al companiei.<\/p>\n<p>Odat\u0103 cu dezvoltarea tehnologiilor AI, num\u0103rul \u0219i diversitatea sarcinilor care pot fi rezolvate prin \u00eenv\u0103\u021barea automat\u0103 cresc exponen\u021bial. Fiecare astfel de proces de afaceri reprezint\u0103 o economie pentru companie datorit\u0103 automatiz\u0103rii muncii angaja\u021bilor din func\u021bii de mas\u0103 (centru de apeluri, verificarea \u0219i clasificarea documentelor etc.), extinderea bazei de clien\u021bi prin ad\u0103ugarea de func\u021bii noi atr\u0103g\u0103toare \u0219i convenabile, economisirea de fonduri prin utilizarea optim\u0103 \u0219i redistribuirea resurselor \u0219i multe altele. \u00cen cele din urm\u0103, orice proces este orientat spre crearea de valoare \u0219i, ca urmare, ar trebui s\u0103 aduc\u0103 un anumit efect economic. Aici este foarte important s\u0103 formul\u0103m clar idea de afaceri \u0219i s\u0103 calcul\u0103m profitul estimat din implementarea modelului \u00een structura general\u0103 de creare a valorii companiei. Se \u00eent\u00e2mpl\u0103 situa\u021bii \u00een care implementarea modelului nu se justific\u0103, iar timpul petrecut de speciali\u0219tii \u00een \u00eenv\u0103\u021barea automat\u0103 cost\u0103 mult mai mult dec\u00e2t un loc de munc\u0103 pentru un operator care \u00eendepline\u0219te aceast\u0103 sarcin\u0103. De aceea, astfel de cazuri trebuie identificate la etapele timpurii de creare a sistemelor AI.<\/p>\n<p>Prin urmare, profitul modelului \u00eencepe s\u0103 apar\u0103 doar atunci c\u00e2nd \u00een procesul MLOps a fost formulat\u0103 corect sarcina de afaceri, priorit\u0103\u021bile au fost stabilite \u0219i, \u00een etapele timpurii de dezvoltare, a fost formulat procesul de integrare a modelului \u00een sistem.<\/p>\n<h2>Un nou proces \u2013 noi provoc\u0103ri<\/h2>\n<p>\nO r\u0103spuns cuprinz\u0103tor la \u00eentrebarea de baz\u0103 a afacerii despre c\u00e2t de aplicabile sunt modelele ML pentru rezolvarea problemelor, \u00eentrebarea general\u0103 a \u00eencrederii \u00een AI \u2014 este unul dintre provoc\u0103rile cheie \u00een procesul de dezvoltare \u0219i implementare a abord\u0103rilor MLOps. Ini\u021bial, afacerea percepe cu scepticism implementarea \u00eenv\u0103\u021b\u0103rii automate \u00een procese \u2014 este greu s\u0103 te bazezi pe modele \u00een locuri \u00een care anterior, de obicei, lucrau oamenii. Pentru afaceri, programele apar ca un \u201eblack box\u201d, relevan\u021ba r\u0103spunsurilor c\u0103ruia trebuie \u00eenc\u0103 dovedit\u0103. \u00cen plus, \u00een activitatea bancar\u0103, \u00een afacerile operatorilor de telecomunica\u021bii \u0219i \u00een altele exist\u0103 cerin\u021be stricte din partea autorit\u0103\u021bilor de reglementare. Toate sistemele \u0219i algoritmii integra\u021bi \u00een procesele bancare sunt supu\u0219i auditului. Pentru a rezolva aceast\u0103 problem\u0103, pentru a dovedi afacerii \u0219i reglementatorilor validitatea \u0219i corectitudinea r\u0103spunsurilor inteligen\u021bei artificiale, \u00eempreun\u0103 cu modelul sunt implementate instrumente de monitorizare. \u00cen plus, exist\u0103 o procedur\u0103 de validare independent\u0103, obligatorie pentru modelele de reglementare, care respect\u0103 cerin\u021bele BNR. Un grup de exper\u021bi independen\u021bi efectueaz\u0103 auditul rezultatelor ob\u021binute de model, av\u00e2nd \u00een vedere datele de intrare.<\/p>\n<p>A doua provocare \u2014 evaluarea \u0219i gestionarea riscurilor modelului la implementarea modelului de \u00eenv\u0103\u021bare automat\u0103. Chiar dac\u0103 o persoan\u0103 nu poate r\u0103spunde cu o sut\u0103 la sut\u0103 siguran\u021b\u0103 la \u00eentrebarea dac\u0103 rochia era alb\u0103 sau albastr\u0103, inteligen\u021ba artificial\u0103 are de asemenea dreptul la eroare. De asemenea, trebuie avut \u00een vedere c\u0103, de-a lungul timpului, datele se pot schimba, iar modelele trebuie s\u0103 fie actualizate pentru a oferi rezultate suficient de precise. Pentru a preveni afectarea procesului de afaceri, este necesar s\u0103 se gestioneze riscurile modelului \u0219i s\u0103 se urm\u0103reasc\u0103 performan\u021ba modelului, actualiz\u00e2ndu-l periodic cu date noi.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps \u00een lumea Machine Learning\" src=\"\/wp-content\/uploads\/2020\/06\/cb97ff0eab95d81e17bf76aad3c3994e.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nDar dup\u0103 prima etap\u0103 de ne\u00eencredere, apare efectul invers. Cu c\u00e2t mai multe modele sunt implementate cu succes \u00een procese, cu at\u00e2t apetitul afacerii pentru utilizarea inteligen\u021bei artificiale cre\u0219te \u2014 p\u00e2n\u0103 la a g\u0103si noi sarcini care pot fi rezolvate prin metode de \u00eenv\u0103\u021bare automat\u0103. Fiecare sarcin\u0103 ini\u021biaz\u0103 un \u00eentreg proces, necesitant diverse competen\u021be:<\/p>\n<ul>\n<li>inginerii de date preg\u0103tesc \u0219i prelucreaz\u0103 datele;\n<\/li>\n<li>speciali\u0219tii \u00een \u0219tiin\u021ba datelor aplic\u0103 instrumente de \u00eenv\u0103\u021bare automat\u0103 \u0219i dezvolt\u0103 modelul;\n<\/li>\n<li>IT implementeaz\u0103 modelul \u00een sistem;\n<\/li>\n<li>Inginerul ML determin\u0103 cum s\u0103 integreze corect acest model \u00een proces, ce instrumente IT s\u0103 foloseasc\u0103 \u00een func\u021bie de cerin\u021bele modului de aplicare a modelului, \u021bin\u00e2nd cont de fluxul de solicit\u0103ri, timpul de r\u0103spuns etc.\u00a0\n<\/li>\n<li>Arhitectul ML proiecteaz\u0103 cum poate fi implementat fizic produsul software \u00eentr-un sistem industrial.\n<\/li>\n<\/ul>\n<p>\n\u00centregul ciclu necesit\u0103 un num\u0103r mare de speciali\u0219ti foarte califica\u021bi. La un anumit punct de dezvoltare \u0219i grad de penetrare a modelelor ML \u00een procesele de afaceri, devine evident c\u0103 scalarea liniar\u0103 a num\u0103rului de speciali\u0219ti \u00een func\u021bie de cre\u0219terea num\u0103rului de sarcini devine costisitoare \u0219i ineficient\u0103. Prin urmare, apare \u00eentrebarea automatiz\u0103rii procesului MLOps - definirea unor clase standard de sarcini de \u00eenv\u0103\u021bare automat\u0103, dezvoltarea unor pipeline-uri tipice pentru procesarea datelor \u0219i recalibrarea modelului. \u00centr-o imagine ideal\u0103, pentru rezolvarea acestor sarcini sunt necesari profesioni\u0219ti care s\u0103 de\u021bin\u0103 competen\u021be la intersec\u021bia BigData, Data Science, DevOps \u0219i IT. De aceea, cea mai mare problem\u0103 din industria Data Science \u0219i cea mai mare provocare \u00een organizarea proceselor MLOps este lipsa acestei competen\u021be pe pia\u021ba actual\u0103 de preg\u0103tire a resurselor umane. Speciali\u0219tii care \u00eendeplinesc aceste cerin\u021be sunt, \u00een prezent, foarte rari pe pia\u021ba muncii \u0219i sunt aprecia\u021bi la pre\u021b de aur.<\/p>\n<h2>Despre competen\u021be<\/h2>\n<p>\n\u00cen teorie, toate sarcinile MLOps pot fi rezolvate cu instrumente clasice DevOps, f\u0103r\u0103 a recurge la o extindere specializat\u0103 a modelului de roluri. A\u0219a cum am men\u021bionat mai sus, data scientist-ul trebuie s\u0103 fie nu numai matematician \u0219i specialist \u00een analytics de date, ci \u0219i guru al \u00eentregului pipeline - responsabilitatea sa include dezvoltarea arhitecturii, programarea modelelor \u00een mai multe limbaje \u00een func\u021bie de arhitectur\u0103, preg\u0103tirea vitrinei de date \u0219i implementarea aplica\u021biei. Cu toate acestea, crearea unei structuri tehnologice, realizat\u0103 \u00een cadrul procesului MLOps, necesit\u0103 p\u00e2n\u0103 la 80% din eforturile de munc\u0103, ceea ce \u00eenseamn\u0103 c\u0103 un matematician calificat, a\u0219a cum este un Data Scientist de calitate, va dedica doar 20% din timp specialit\u0103\u021bii sale. Prin urmare, delimitarea rolurilor speciali\u0219tilor care implementeaz\u0103 modele de \u00eenv\u0103\u021bare automat\u0103 devine o necesitate vital\u0103.\u00a0<\/p>\n<p>C\u00e2t de bine trebuie delimitate rolurile depinde de dimensiunea \u00eentreprinderii. E o situa\u021bie c\u00e2nd \u00eentr-un startup exist\u0103 un specialist care se ocup\u0103 de tot, de inginerie \u0219i arhitectur\u0103, p\u00e2n\u0103 la DevOps. Este cu totul altceva c\u00e2nd \u00eentr-o mare \u00eentreprindere toate procesele de dezvoltare a modelelor sunt concentrate pe c\u00e2\u021biva speciali\u0219ti de Data Science de \u00eenalt\u0103 calitate, \u00een timp ce programatorii sau speciali\u0219tii \u00een baze de date \u2014 competen\u021be mai comune \u0219i mai pu\u021bin costisitoare pe pia\u021ba muncii \u2014 pot prelua o mare parte din sarcinile de rutin\u0103.<\/p>\n<p>Astfel, de modul \u00een care se traseaz\u0103 grani\u021ba \u00een alegerea speciali\u0219tilor pentru asigurarea procesului MLOps \u0219i de cum este organizat procesul de operativizare a modelelor dezvoltate, depind direct viteza \u0219i calitatea modelelor dezvoltate, productivitatea echipei \u0219i climatul de munc\u0103 din cadrul acesteia.<\/p>\n<h2>Ce a fost deja realizat de echipa noastr\u0103<\/h2>\n<p>\nNu cu mult timp \u00een urm\u0103 am \u00eenceput s\u0103 construim structura competen\u021belor \u0219i proceselor MLOps. Dar deja \u00een aceast\u0103 etap\u0103, proiectele noastre pentru gestionarea ciclului de via\u021b\u0103 al modelelor \u0219i pentru aplicarea modelelor ca serviciu sunt \u00een stadiul de testare MVP.<\/p>\n<p>De asemenea, am definit structura optim\u0103 de competen\u021be \u0219i structura organiza\u021bional\u0103 a interac\u021biunii \u00eentre to\u021bi participan\u021bii la proces pentru o mare \u00eentreprindere. Au fost organizate echipe Agile care rezolv\u0103 sarcini pentru \u00eentreaga gam\u0103 de clien\u021bi de afaceri, iar procesul de interac\u021biune cu echipele de proiect care dezvolt\u0103 platforme \u0219i infrastructura, care este funda\u021bia cl\u0103dirii MLOps, a fost \u00eembun\u0103t\u0103\u021bit.<\/p>\n<h2>\u00centreb\u0103ri pentru viitor<\/h2>\n<p>\nMLOps este un domeniu \u00een dezvoltare care se confrunt\u0103 cu o lips\u0103 de competen\u021be \u0219i care \u00een viitor va c\u00e2\u0219tiga av\u00e2nt. P\u00e2n\u0103 atunci, cel mai bine este s\u0103 ne baz\u0103m pe experien\u021bele \u0219i practicile DevOps. Scopul principal al MLOps este utilizarea mai eficient\u0103 a modelelor ML pentru a rezolva problemele de afaceri. Dar apar multe \u00eentreb\u0103ri:<\/p>\n<ul>\n<li>Cum s\u0103 reducem timpul necesar pentru a lansa modelele \u00een produc\u021bie?\n<\/li>\n<li>Cum s\u0103 reducem fric\u021biunile birocratice \u00eentre echipele cu competen\u021be diferite \u0219i s\u0103 sporim concentrarea pe colaborare?\n<\/li>\n<li>Cum s\u0103 monitoriz\u0103m modelele, s\u0103 gestion\u0103m versiunile \u0219i s\u0103 organiz\u0103m un monitorizare eficient\u0103?\n<\/li>\n<li>Cum s\u0103 cre\u0103m un ciclu de via\u021b\u0103 cu adev\u0103rat ciclic pentru un model ML modern?\n<\/li>\n<li>Cum s\u0103 standardiz\u0103m procesul de \u00eenv\u0103\u021bare automat\u0103?\n<\/li>\n<\/ul>\n<p>\nR\u0103spunsurile la aceste \u00eentreb\u0103ri vor determina \u00een mare m\u0103sur\u0103 c\u00e2t de repede MLOps \u00ee\u0219i va desf\u0103\u0219ura pe deplin poten\u021bialul.<br \/>\n<br \/>Sursa: <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 5.0.2.1 - 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\/ro\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning\" \/>\n\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.2.1\" \/>\n\t\t<meta property=\"og:locale\" content=\"ro_RO\" \/>\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\/ro\/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\udd47 MLOps: DevOps \u00een lumea Machine Learning | ProHoster","description":"\u00cen 2018, \u00een cercurile profesionale \u0219i la conferin\u021bele dedicate AI, a ap\u0103rut conceptul de MLOps, care s-a impus rapid \u00een industrie \u0219i acum se dezvolt\u0103 ca o direc\u021bie de sine st\u0103t\u0103toare.","canonical_url":"https:\/\/prohoster.info\/ro\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"ro_RO","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\/ro\/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\/ro\/wp-json\/wp\/v2\/posts\/86523","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/comments?post=86523"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/posts\/86523\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media\/86524"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/media?parent=86523"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/categories?post=86523"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/ro\/wp-json\/wp\/v2\/tags?post=86523"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}