{"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\/lv\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","title":{"rendered":"MLOps: DevOps ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>2018. gad\u0101 profesion\u0101los lokos un tematiskaj\u0101s konferenc\u0113s, kas velt\u012btas AI, par\u0101d\u012bj\u0101s j\u0113dziens MLOps, kas \u0101tri nostiprin\u0101j\u0101s nozar\u0113 un tagad att\u012bst\u0101s k\u0101 patst\u0101v\u012bga virz\u012bba. N\u0101kotn\u0113 MLOps var k\u013c\u016bt par vienu no piepras\u012bt\u0101kaj\u0101m jom\u0101m IT. Kas tad tas ir un k\u0101 to izmantot, noskaidrojam t\u0101l\u0101k.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113\" 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>Kas ir MLOps<\/h2>\n<p>\nMLOps (tehnolo\u0123iju un procesu apvienojums ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s un izstr\u0101d\u0101to mode\u013cu integr\u0101cijas biznesa procesos) \u2014 tas ir jauns sadarb\u012bbas veids starp biznesa p\u0101rst\u0101vjiem, zin\u0101tniekiem, matem\u0101ti\u0137iem, ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s speci\u0101listiem un IT in\u017eenieriem, izstr\u0101d\u0101jot m\u0101ksl\u012bg\u0101 intelekta sist\u0113mas.<\/p>\n<p>Citiem v\u0101rdiem sakot, tas ir veids, k\u0101 p\u0101rv\u0113rst ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s metodes un tehnolo\u0123ijas par noder\u012bgu instrumentu biznesa uzdevumu risin\u0101\u0161anai.\u00a0<\/p>\n<p>J\u0101\u0146em v\u0113r\u0101, ka produktiviz\u0101cijas \u0137\u0113de s\u0101kas krietni pirms mode\u013ca izstr\u0101des. Pirmais solis ir biznesa uzdevuma noteik\u0161ana, hipot\u0113zes par v\u0113rt\u012bbu, ko var g\u016bt no datiem, un biznesa idejas par t\u0101s izmanto\u0161anu izstr\u0101de.\u00a0<\/p>\n<p>Pats j\u0113dziens MLOps rad\u0101s k\u0101 analo\u0123ija DevOps j\u0113dzienam, kas attiecas uz ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013ciem un tehnolo\u0123ij\u0101m. DevOps ir pieeja programmat\u016bras izstr\u0101dei, kas \u013cauj palielin\u0101t atsevi\u0161\u0137u izmai\u0146u ievie\u0161anas \u0101trumu, saglab\u0101jot elast\u012bbu un uzticam\u012bbu, izmantojot vair\u0101kus pieejas veidus, tostarp nep\u0101rtrauktu izstr\u0101di, funkciju sadal\u012b\u0161anu atsevi\u0161\u0137os neatkar\u012bgos mikroservisos, automatiz\u0113tu test\u0113\u0161anu un atsevi\u0161\u0137u izmai\u0146u ievie\u0161anu, glob\u0101lo darbsp\u0113ju uzraudz\u012bbu, operat\u012bv\u0101s reakcijas sist\u0113mu uz identific\u0113tiem trauc\u0113jumiem utt.\u00a0<\/p>\n<p>DevOps defin\u0113ja programmat\u016bras dz\u012bves ciklu, un speci\u0101listu kopien\u0101 rad\u0101s ideja izmantot \u0161o pa\u0161u metodolo\u0123iju attiec\u012bb\u0101 uz lielajiem datiem. DataOps ir m\u0113\u0123in\u0101jums piel\u0101got un papla\u0161in\u0101t metodiku, \u0146emot v\u0113r\u0101 lielu datu apjomu glab\u0101\u0161anas, p\u0101rs\u016bt\u012b\u0161anas un apstr\u0101des \u012bpatn\u012bbas da\u017e\u0101d\u0101s savstarp\u0113ji saist\u012btaj\u0101s platform\u0101s.<br \/>\n\u00a0\u00a0<br \/>\nPar\u0101d\u012bjos noteiktai kritiskai masai ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013cu, kas tiek ieviesti uz\u0146\u0113mumu biznesa procesos, tika nov\u0113rota sp\u0113c\u012bga l\u012bdz\u012bba ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s matem\u0101tisko mode\u013cu dz\u012bves cikla un programmat\u016bras dz\u012bves cikla starp\u0101. At\u0161\u0137ir\u012bba ir tikai taj\u0101, ka mode\u013cu algoritmus izveido, izmantojot ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s r\u012bkus un metodes. T\u0101p\u0113c dabiski rad\u0101s ideja piem\u0113rot un piel\u0101got jau zin\u0101mos programmat\u016bras izstr\u0101des pieejas ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013ciem. T\u0101d\u0113j\u0101di ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013cu dz\u012bves cikl\u0101 var izcelt \u0161\u0101dus galvenos posmus:<\/p>\n<ul>\n<li>biznesa idejas noteik\u0161ana;\n<\/li>\n<li>mode\u013ca apm\u0101c\u012bba;\n<\/li>\n<li>mode\u013ca test\u0113\u0161ana un ievie\u0161ana biznesa proces\u0101;\n<\/li>\n<li>mode\u013ca ekspluat\u0101cija.\n<\/li>\n<\/ul>\n<p>\nKad ekspluat\u0101cijas proces\u0101 rodas nepiecie\u0161am\u012bba main\u012bt vai papildin\u0101t modeli ar jauniem datiem, cikls s\u0101kas no jauna \u2014 modelis tiek pilnveidots, test\u0113ts, un tiek izvietota jauna versija.<\/p>\n<blockquote><p>Novirze. K\u0101p\u0113c papildin\u0101t, nevis p\u0101r-apm\u0101c\u012bt? Termins \u00abmode\u013ca p\u0101r-apm\u0101c\u012bba\u00bb ir divdom\u012bgs: starp speci\u0101listiem tas apz\u012bm\u0113 mode\u013ca defektu, kad modelis labi prognoz\u0113, faktiski atk\u0101rtojot prognoz\u0113jamo parametrus apm\u0101c\u012bbas izlas\u0113, bet daudz slikt\u0101k darbojas \u0101r\u0113j\u0101 datu izlas\u0113. Dabiski, \u0161\u0101ds modelis ir defekts, jo \u0161is tr\u016bkums ne\u013cauj to izmantot.<\/p><\/blockquote>\n<p>\n\u0160aj\u0101 dz\u012bves cikl\u0101 \u0161\u0137iet lo\u0123iski izmantot DevOps r\u012bkus: automatiz\u0113tu test\u0113\u0161anu, izvieto\u0161anu un uzraudz\u012bbu, mode\u013cu apr\u0113\u0137inu noform\u0113\u0161anu k\u0101 atsevi\u0161\u0137us mikroservisus. Bet ir ar\u012b vair\u0101ki aspekti, kas trauc\u0113 \u0161o r\u012bku tie\u0161u pielietojumu bez papildu ML papildin\u0101jumiem.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113\" src=\"\/wp-content\/uploads\/2020\/06\/e1d01fd68c3f0071f925a82c08d29d25.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<\/p>\n<h2>K\u0101 likt mode\u013ciem str\u0101d\u0101t un nest pe\u013c\u0146u<\/h2>\n<p>\nK\u0101 piem\u0113ru, ar kuru m\u0113s demonstr\u0113sim MLOps pieeju, \u0146emsim klasisko uzdevumu banku (vai jebkura cita) produkta klientu atbalsta \u010data robotiz\u0101cijai. Parasti atbalsta \u010data biznesa process izskat\u0101s \u0161\u0101di: klients ievada \u010data zi\u0146ojumu ar jaut\u0101jumu un sa\u0146em atbildi no eksperta iepriek\u0161 noteikt\u0101 dialogu koka ietvaros. \u0160\u0101da \u010data automatiz\u0101cijas uzdevums parasti tiek risin\u0101ts, izmantojot ekspertu noteiktus noteikumu kopumus, kas ir \u013coti darbietilp\u012bgi izstr\u0101d\u0113 un uztur\u0113\u0161an\u0101. \u0160\u0101das automatiz\u0101cijas efektivit\u0101te, atkar\u012bb\u0101 no uzdevuma sare\u017e\u0123\u012bt\u012bbas, var sv\u0101rst\u012bties no 20\u201330%. Protams, rodas doma, ka izdev\u012bg\u0101k ir ieviest m\u0101ksl\u012bg\u0101 intelekta moduli \u2014 modeli, kas izstr\u0101d\u0101ts ar ma\u0161\u012bnm\u0101c\u012bbas pal\u012bdz\u012bbu, kas:<\/p>\n<ul>\n<li>sp\u0113j apstr\u0101d\u0101t bez operatora iek\u013cau\u0161anas liel\u0101ku piepras\u012bjumu skaitu (atkar\u012bb\u0101 no t\u0113mas, da\u017eos gad\u012bjumos efektivit\u0101te var sasniegt 70\u201380%);\n<\/li>\n<li>lab\u0101k piel\u0101gojas nestandarta formul\u0113jumiem dialog\u0101 \u2014 sp\u0113j noteikt nodomu, re\u0101lo lietot\u0101ja v\u0113lmi neskaidri formul\u0113t\u0101 piepras\u012bjum\u0101;\n<\/li>\n<li>sp\u0113j noteikt, kad mode\u013ca atbilde ir adekv\u0101ta, bet kad par to ir \u0161aubas un ir j\u0101uzdod papildu preciz\u0113jo\u0161s jaut\u0101jums vai j\u0101p\u0101rsl\u0113dzas uz operatoru;\n<\/li>\n<li>var tikt papildu apm\u0101c\u012bta automatiz\u0113ti (viet\u0101 grupai izstr\u0101d\u0101t\u0101ju, kuri nep\u0101rtraukti piel\u0101go un kori\u0123\u0113 atbil\u017eu scen\u0101rijus, modeli papildu apm\u0101ca datu zin\u0101tnes speci\u0101lists, izmantojot atbilsto\u0161\u0101s ma\u0161\u012bnm\u0101c\u012bbas bibliot\u0113kas).\u00a0\n<\/li>\n<\/ul>\n<p>\n<img decoding=\"async\" alt=\"MLOps: DevOps ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113\" src=\"\/wp-content\/uploads\/2020\/06\/9ef6356090a250df6cb212ab8bedb5f3.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nK\u0101 pan\u0101kt, ka \u0161\u0101ds progres\u012bvs modelis darbojas?\u00a0<\/p>\n<p>K\u0101 ar\u012b, risinot jebkuru citu uzdevumu, pirms \u0161\u0101da modu\u013ca izstr\u0101des ir j\u0101nosaka biznesa process un form\u0101li j\u0101piesaka konkr\u0113tais uzdevums, ko m\u0113s risin\u0101sim, izmantojot ma\u0161\u012bnm\u0101c\u012bbu metodi. \u0160aj\u0101 punkt\u0101 s\u0101kas operacioniz\u0101cijas process, kas apz\u012bm\u0113ts ar akron\u012bmu Ops.\u00a0<\/p>\n<p>N\u0101kamais solis datu zin\u0101tnieka speci\u0101lista sadarb\u012bb\u0101 ar datu in\u017eenieri ir p\u0101rbaud\u012bt datu pieejam\u012bbu un atbilst\u012bbu, k\u0101 ar\u012b biznesa hipot\u0113zi par biznesa idejas dz\u012bvotsp\u0113ju, izstr\u0101d\u0101jot mode\u013ca prototipu un p\u0101rbaudot t\u0101 faktisko efektivit\u0101ti. Tikai p\u0113c tam, kad bizness to apstiprina, var uzs\u0101kt p\u0101reju no mode\u013ca izstr\u0101des uz t\u0101 integr\u0101ciju sist\u0113m\u0101s, kas veic konkr\u0113tu biznesa procesu. Svar\u012bgs moments ir r\u016bp\u012bga pl\u0101no\u0161ana integr\u0113\u0161an\u0101, k\u0101 ar\u012b dzi\u013ca izpratne par to, k\u0101 katr\u0101 posm\u0101 tiks izmantots modelis un k\u0101du ekonomisko efektu tas rad\u012bs, jo tas ir b\u016btisks jaut\u0101jums MLOps pieeju ievie\u0161an\u0101 uz\u0146\u0113muma tehnolo\u0123iskaj\u0101 ain\u0101.<\/p>\n<p>Ar m\u0101ksl\u012bg\u0101 intelekta tehnolo\u0123iju att\u012bst\u012bbu strauji pieaug uzdevumu skaits un da\u017e\u0101d\u012bba, kurus var risin\u0101t, izmantojot ma\u0161\u012bnm\u0101c\u012b\u0161anos. Katrs \u0161\u0101ds biznesa process ir uz\u0146\u0113muma ietaup\u012bjums, automatiz\u0113jot darbinieku darbu masveida poz\u012bcij\u0101s (izsaukumu centri, dokumentu p\u0101rbaude un \u0161\u0137iro\u0161ana uc), tas ir klientu b\u0101zes papla\u0161in\u0101\u0161ana, pievienojot jaunas pievilc\u012bgas un \u0113rtas funkcijas, k\u0101 ar\u012b ekonomija resursu optim\u0101l\u0101s izmanto\u0161anas un p\u0101rdales d\u0113\u013c. Galu gal\u0101 katrs process ir v\u0113rsts uz v\u0113rt\u012bbas rad\u012b\u0161anu un, k\u0101 rezult\u0101t\u0101, tam j\u0101nes noteikts ekonomiskais efekts. \u0160eit \u012bpa\u0161i svar\u012bgi ir skaidri formul\u0113t biznesa ideju un apr\u0113\u0137in\u0101t paredzamo pe\u013c\u0146u no mode\u013ca ievie\u0161anas kop\u0113j\u0101 uz\u0146\u0113muma v\u0113rt\u012bbas rad\u012b\u0161anas strukt\u016br\u0101. Pien\u0101k br\u012b\u017ei, kad mode\u013ca ievie\u0161ana neattaisno sevi, un laiks, ko iegulda ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s speci\u0101listi, var izr\u0101d\u012bties d\u0101rg\u0101ks nek\u0101 operators, kas veic \u0161o uzdevumu. Tie\u0161i t\u0101p\u0113c \u0161\u0101dus gad\u012bjumus ir j\u0101cen\u0161as atkl\u0101t jau agr\u012bnajos m\u0101ksl\u012bg\u0101 intelekta sist\u0113mu izstr\u0101des posmos.<\/p>\n<p>T\u0101d\u0113j\u0101di mode\u013ca pe\u013c\u0146a s\u0101kas tikai tad, kad MLOps proces\u0101 ir pareizi formul\u0113ts biznesa uzdevums, noteikti priorit\u0101tes un agr\u012bnajos izstr\u0101des posmos ir formul\u0113ts mode\u013ca integr\u0101cijas process sist\u0113m\u0101.<\/p>\n<h2>Jauns process \u2013 jauni izaicin\u0101jumi<\/h2>\n<p>\nVisaptvero\u0161a atbilde uz princip\u0101lo uz\u0146\u0113m\u0113jdarb\u012bbas jaut\u0101jumu par to, cik piem\u0113rotas ir ML mode\u013ci uzdevumu risin\u0101\u0161anai, un visp\u0101r\u0113jo uztic\u0113\u0161an\u0101s jaut\u0101jumu m\u0101ksl\u012bgajam intelektam \u2013 tas ir viens no galvenajiem izaicin\u0101jumiem MLOps pieejas att\u012bst\u012bbas un \u012bsteno\u0161anas proces\u0101. S\u0101kotn\u0113ji uz\u0146\u0113m\u0113ji skeptiski raug\u0101s uz ma\u0161\u012bnm\u0101c\u012bbas ievie\u0161anu procesos \u2013 ir gr\u016bti pa\u013cauties uz mode\u013ciem viet\u0101s, kur iepriek\u0161 parasti str\u0101d\u0101ja cilv\u0113ki. Uz\u0146\u0113mumiem programmas \u0161\u0137iet k\u0101 \u201emeln\u0101 kaste\u201d, kuras atbilsto\u0161\u012bba v\u0113l ir j\u0101pier\u0101da. Turkl\u0101t banku sektor\u0101, telekomunik\u0101ciju uz\u0146\u0113mumos un citur ir stingras valsts regul\u0113t\u0101ju pras\u012bbas. Visas sist\u0113mas un algoritmi, kas ieviesti banku procesos, tiek pak\u013cauti auditam. Lai atrisin\u0101tu \u0161o uzdevumu un pier\u0101d\u012btu uz\u0146\u0113m\u0113jdarb\u012bbai un regul\u0113t\u0101jiem m\u0101ksl\u012bg\u0101 intelekta atbilsto\u0161\u012bbu un precizit\u0101ti, kop\u0101 ar modeli tiek ieviesti uzraudz\u012bbas r\u012bki. Turkl\u0101t past\u0101v neatkar\u012bgas valid\u0101cijas proced\u016bra, kas ir oblig\u0101ta regul\u0113juma mode\u013ciem, un t\u0101 atbilst Centr\u0101l\u0101s bankas pras\u012bb\u0101m. Neatkar\u012bga ekspertu grupa veic mode\u013ca rezult\u0101tu auditu, \u0146emot v\u0113r\u0101 ievades datus.<\/p>\n<p>Otrs izaicin\u0101jums ir mode\u013cu risku nov\u0113rt\u0113\u0161ana un uzskaite, ievie\u0161ot ma\u0161\u012bnm\u0101c\u012bbas modeli. Pat ja cilv\u0113ks nevar ar simtprocent\u012bgu p\u0101rliec\u012bbu atbild\u0113t uz jaut\u0101jumu, vai tas ir balts vai zils kleita, t\u0101pat ar\u012b m\u0101ksl\u012bgajam intelektam ir ties\u012bbas uz k\u013c\u016bd\u0101m. V\u0113rts \u0146emt v\u0113r\u0101, ka laika gait\u0101 dati var main\u012bties, un mode\u013ciem nepiecie\u0161ams t\u0101l\u0101k m\u0101c\u012bties, lai sniegtu pietiekami prec\u012bzus rezult\u0101tus. Lai uz\u0146\u0113mumu process necietu, ir j\u0101vada mode\u013cu riski un regul\u0101ri j\u0101uzrauga mode\u013ca darb\u012bba, to apm\u0101cot uz jauniem datiem.<\/p>\n<p><img decoding=\"async\" alt=\"MLOps: DevOps ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113\" src=\"\/wp-content\/uploads\/2020\/06\/cb97ff0eab95d81e17bf76aad3c3994e.jpeg\" style=\"display:block;margin: 0 auto;\" \/><br \/>\n<br \/>\nTa\u010du p\u0113c pirm\u0101 neuztic\u0113\u0161an\u0101s posma s\u0101k par\u0101d\u012bties pret\u0113js efekts. Jo vair\u0101k mode\u013cu veiksm\u012bgi tiek ieviesti procesos, jo liel\u0101ka ir uz\u0146\u0113mumu v\u0113lme izmantot m\u0101ksl\u012bgo intelektu \u2013 rodas jauni un jauni uzdevumi, kurus var risin\u0101t ar ma\u0161\u012bnm\u0101c\u012bbas metod\u0113m. Katrs uzdevums uzs\u0101k veselu procesu, kam nepiecie\u0161amas noteiktas kompetences:<\/p>\n<ul>\n<li>datu in\u017eenieri sagatavo un apstr\u0101d\u0101 datus;\n<\/li>\n<li>datu zin\u0101tnieki izmanto ma\u0161\u012bnm\u0101c\u012bbas r\u012bkus un izstr\u0101d\u0101 modeli;\n<\/li>\n<li>IT ievie\u0161 modeli sist\u0113m\u0101;\n<\/li>\n<li>ML in\u017eenieris nosaka, k\u0101 \u0161o modeli pareizi integr\u0113t proces\u0101, k\u0101dus IT r\u012bkus izmantot atkar\u012bb\u0101 no mode\u013ca lieto\u0161anas pras\u012bb\u0101m, \u0146emot v\u0113r\u0101 piepras\u012bjuma pl\u016bsmu, atbildes laiku utt.\u00a0\n<\/li>\n<li>ML arhitekts projekt\u0113, k\u0101 fiziski var \u012bstenot programmat\u016bras produktu industrijas sist\u0113m\u0101.\n<\/li>\n<\/ul>\n<p>\nVisam ciklam ir nepiecie\u0161ams liels augsti kvalific\u0113tu speci\u0101listu skaits. Noteikt\u0101 att\u012bst\u012bbas posm\u0101 un ML mode\u013cu integr\u0113\u0161anas pak\u0101p\u0113 biznesa procesos izr\u0101d\u0101s, ka line\u0101ri palielin\u0101t speci\u0101listu skaitu proporcion\u0101li uzdevumu skaita pieaugumam k\u013c\u016bst d\u0101rgi un neefekt\u012bvi. T\u0101d\u0113\u013c rodas jaut\u0101jums par MLOps procesa automatiz\u0101ciju \u2014 noteikt vair\u0101ku standarta ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s uzdevumu klases, izstr\u0101d\u0101t tipiskus datu apstr\u0101des un mode\u013cu t\u0101lm\u0101c\u012bbas pl\u016bsmas. Ide\u0101l\u0101 situ\u0101cij\u0101 \u0161\u0101du uzdevumu risin\u0101\u0161anai ir nepiecie\u0161ami profesion\u0101\u013ci, kas vienl\u012bdz labi p\u0101rvalda kompetences BigData, Data Science, DevOps un IT krustpunkt\u0101. T\u0101d\u0113j\u0101di liel\u0101k\u0101 probl\u0113ma Data Science nozar\u0113 un liel\u0101kais izaicin\u0101jums MLOps procesu organiz\u0113\u0161an\u0101 ir \u0161\u0101das kompetences tr\u016bkums eso\u0161aj\u0101 darba tirg\u016b. Speci\u0101listi, kuri atbilst \u0161\u012bm pras\u012bb\u0101m, pa\u0161laik ir vien\u012bg\u0101 skait\u0101 darba tirg\u016b un tiek v\u0113rt\u0113ti zelta v\u0113rt\u0113.<\/p>\n<h2>Par kompetenc\u0113m<\/h2>\n<p>\nTeorij\u0101 visas MLOps uzdevumus var risin\u0101t ar klasiskajiem DevOps r\u012bkiem un neizmantot specializ\u0113tu lomu mode\u013ca papla\u0161in\u0101\u0161anu. K\u0101 jau min\u0113ts iepriek\u0161, datu zin\u0101tniekam j\u0101b\u016bt ne tikai matem\u0101ti\u0137im un datu anal\u012bzes speci\u0101listam, bet ar\u012b visa pl\u016bsmas guru \u2014 vi\u0146am j\u0101dodas arhitekt\u016bras izstr\u0101de, mode\u013cu programm\u0113\u0161ana vair\u0101k\u0101s valod\u0101s atkar\u012bb\u0101 no arhitekt\u016bras, datu vitrine un pa\u0161a lietojumprogrammas izvietojums. Tom\u0113r tehnisk\u0101s apak\u0161strukt\u016bras izstr\u0101de, ko \u012bsteno MLOps procesos, aiz\u0146em l\u012bdz 80% no darba resursiem, kas noz\u012bm\u0113, ka kvalific\u0113ts matem\u0101ti\u0137is, kur\u0161 ir kvalitat\u012bvs Data Scientist, velt\u012bs tikai 20% sava laika savai specialit\u0101tei. T\u0101d\u0113\u013c speci\u0101listu lomu no\u0161\u0137ir\u0161ana, kuri \u012bsteno ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s mode\u013cu ievie\u0161anas procesos, k\u013c\u016bst vit\u0101li nepiecie\u0161ama.\u00a0<\/p>\n<p>Cik detaliz\u0113ti ir j\u0101no\u0161\u0137ir lomas, atkar\u012bgs no uz\u0146\u0113muma apm\u0113riem. Ir viena lieta, kad start-up uz\u0146\u0113mum\u0101 ir viens speci\u0101lists, str\u0101dnieks uz rezerves ener\u0123ijas, kur\u0161 pats ir gan in\u017eenieris, gan arhitekts, gan DevOps. Piln\u012bgi cita lieta ir tad, kad liel\u0101 uz\u0146\u0113mum\u0101 visi mode\u013cu izstr\u0101des procesi ir koncentr\u0113ti vair\u0101kiem augsta l\u012bme\u0146a Data Science speci\u0101listiem, kam\u0113r programm\u0113t\u0101js vai datub\u0101zu speci\u0101lists \u2014 bie\u017e\u0101k sastopama un l\u0113t\u0101ka kompetence darba tirg\u016b \u2014 var uz\u0146emties liel\u0101ko da\u013cu ikdienas uzdevumu.<\/p>\n<p>T\u0101d\u0113j\u0101di robe\u017ea, kur\u0101 ir j\u0101sasniedz specialistu izv\u0113le, lai nodro\u0161in\u0101tu MLOps procesu, un k\u0101 tiek organiz\u0113ts izstr\u0101d\u0101to mode\u013cu operacionaliz\u0101cijas process, tie\u0161i ietekm\u0113 izstr\u0101d\u0101to mode\u013cu \u0101trumu un kvalit\u0101ti, komandas produktivit\u0101ti un t\u0101s mikroklimatu.<\/p>\n<h2>Kas jau tagad ir izdar\u012bts m\u016bsu komand\u0101<\/h2>\n<p>\nM\u0113s nesen s\u0101k\u0101m veidot kompeten\u010du strukt\u016bru un MLOps procesus. Bet jau tagad MVP test\u0113\u0161an\u0101 ir m\u016bsu projekti par mode\u013cu dz\u012bves cikla p\u0101rvald\u012bbu un mode\u013cu pielieto\u0161anu k\u0101 servisu.<\/p>\n<p>T\u0101pat m\u0113s esam noteiku\u0161i optim\u0101lu kompeten\u010du strukt\u016bru lieliem uz\u0146\u0113mumiem un organizatorisko strukt\u016bru sadarb\u012bbai starp visiem procesa dal\u012bbniekiem. Tika izveidotas Agile komandas, kas risina uzdevumus visam biznesa pas\u016bt\u012bt\u0101ju spektram, k\u0101 ar\u012b ir izveidots sadarb\u012bbas process ar projekta komand\u0101m platformu un infrastrukt\u016bras izstr\u0101dei, kas ir MLOps veidojo\u0161\u0101 pamata pamats.<\/p>\n<h2>Jaut\u0101jumi n\u0101kotnei<\/h2>\n<p>\nMLOps ir att\u012bsto\u0161a joma, kur\u0101 tr\u016bkst kompetences, un n\u0101kotn\u0113 t\u0101 ieg\u016bs jaudu. Bet pagaid\u0101m vislab\u0101k ir pamatus veikt no DevOps pieredzes un praks\u0113m. Galvenais MLOps m\u0113r\u0137is ir efekt\u012bv\u0101k izmantot ML mode\u013cus biznesa uzdevumu risin\u0101\u0161anai. Bet pie t\u0101 rodas daudzi jaut\u0101jumi:<\/p>\n<ul>\n<li>K\u0101 samazin\u0101t laiku mode\u013cu nodo\u0161anai ra\u017eo\u0161anai?\n<\/li>\n<li>K\u0101 samazin\u0101t birokr\u0101tiskos \u0161\u0137\u0113r\u0161\u013cus starp da\u017e\u0101du kompeten\u010du komand\u0101m un paaugstin\u0101t sadarb\u012bbas virzienu?\n<\/li>\n<li>K\u0101 izsekot mode\u013cus, p\u0101rvald\u012bt versijas un organiz\u0113t efekt\u012bvu uzraudz\u012bbu?\n<\/li>\n<li>K\u0101 izveidot patie\u0161\u0101m ciklisku dz\u012bves ciklu m\u016bsdienu ML modelim?\n<\/li>\n<li>K\u0101 standartiz\u0113t ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s procesu?\n<\/li>\n<\/ul>\n<p>\nAtbil\u017eu uz \u0161iem jaut\u0101jumiem liel\u0101 m\u0113r\u0101 b\u016bs atkar\u012bgs, cik \u0101tri MLOps atkl\u0101s savu pilno potenci\u0101lu.<br \/>\n<br \/>Avots: <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 5.0.4 - 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\/lv\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.4\" \/>\n\t\t<meta property=\"og:locale\" content=\"lv_LV\" \/>\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\/lv\/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 ma\u0161\u012bnm\u0101c\u012b\u0161an\u0101s pasaul\u0113 | ProHoster","description":"2018. gad\u0101 profesion\u0101laj\u0101s aprind\u0101s un tematiskaj\u0101s konferenc\u0113s, kas velt\u012btas AI, par\u0101d\u012bj\u0101s j\u0113dziens MLOps, kas \u0101tri nostiprin\u0101j\u0101s nozar\u0113 un \u0161obr\u012bd att\u012bst\u0101s k\u0101 neatkar\u012bga virziena.","canonical_url":"https:\/\/prohoster.info\/lv\/blog\/administrirovanie\/mlops-devops-v-mire-machine-learning","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"lv_LV","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\/lv\/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\/lv\/wp-json\/wp\/v2\/posts\/86523","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/comments?post=86523"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/posts\/86523\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/media\/86524"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/media?parent=86523"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/categories?post=86523"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/lv\/wp-json\/wp\/v2\/tags?post=86523"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}