Coppite yu am solo ngir modeli kode
Coppite yu xalaat yi dañuy tax soppi kodu IA di yëg ci saasi ndax dafay wax luy waaja am ci fichier bi du soppeeku, te dafay xool pàcc yu ndaw yi wuute.
Résumé
It matters because it can cut latency for large rewrites by an order of magnitude in coding tools.
Plongeur bu xóot
Su IA soppee benn fichier, token yu bari yi muy génne dañuy nuru kod bi njëkk; yenn rëdd yu néew rek ñooy soppiku. Generation naïf dafay gennewaat fichier bi yépp token par token, te loolu lu yeex la ci fichier yu mag yi. Coppite yu xalaat yi dañuy jëfandikoo jumtukaay bi soppikuwul: balluwaay bi fi nekk dafay nekk 'projet' bu baax ci li model bi di génne. Sistem bi dafay joxe ay pàccu kodu njëkk bi ni ay xalaat yu ñuy xalaat ba noppi may model bi mu xool lu bari ci ñoom ci benn yoon kese. Fépp fu model bi nangu, dañuy nangu jeton yooyu ci saasi; fuñu ko déggoowul, dafay defar span buñ jubbanti ci anam wu jaar yoon. Lii cousin code-specialized la ci decodage speculatif, waaye ludul benn modelu projet bu ndaw bu wuute, projet bi dafay ñëw ci fichier bi ñuy soppali, di joxe gaawaay yu bari ci liggéey yu diis.
Gis-gis xarala
Dekodage autorégressif buñ miin dafay defar benn token ci paas bu nekk. Pexem speculatif yi dañuy proposer ay token yu bari benn yoon, ba noppi ñu xool leen ci paralel: benn model mën na xool, ci benn yoon, ndax benn run of tokens proposé méngoo ak limu nara génne. Coppite yiñ xalaat dañuy joxe proposition yooyu ci kodu source bi soppikuwul, du ci model buñ tëral. Dawal yiñ nangu ñu ngi koy njëg lu tollu ci benn paas ngir jeton yu bari; divergence yi kese ñooy indi generation bu bees, kon njëg yi dañuy aju ci dayo soppali, du dayo fichier bi.
njeextalu pexe
Njëgg ak budget
Dogal yi architecture di jël dañuy indi njariñ ak njëgu liggéey bi ay at ci ginaaw.
dogal yu gëna leer
Njàngalem xarala yi dafay jàppale ekip yi ñu tànn li gën, te baña yam ci li gëna bees daal.
Xool kalite
Tanneef yu gëna baax ci wàllu ingeñër dina wàññi jafe-jafe yi ci wàllu wóor ci liggéey bi.
Ëlëgu coppite yuñ xalaat ngir modeli kode
Agent yu diis yi ak assistant IDE yi dina ñu wéeru ci lii ngir wéy di jëfandikoo différence yu mag ci saasi, ba ci fichier yu am junni ligne. Xaarandi lëkkaloo bu gëna dëgër ak formaa diff yuñ yamale, digle yu xam garab yuy sargal ay pexe syntax, ak boole ak seet suko defee projet speculatif bi amaale ay refactor yu mëna am. Ginaaw agent codage autonome yi dañuy def coppite yu bari ci liggéey bu nekk, coppite yu speculatif yi dañuy nekk levier bu am solo ngir tëye workflow yu bari yi ñuy tontu ak gëna yomb ci doxal.
Doxal ci àdduna dëgg
Jàppalekatu IDE bi dafay binndaat fichier bu am 500 ligne ngir soppi tur wi ci fonction bi, nangu ligne yi soppikuwul ci ay paas yu néew, ba noppi génne ay span yuñ soppi tur yi kese.
Komànd 'fix this lint error' mooy defar fichier biñ saafara ci saasi ndax 99% ci kod bi dañu koy jëfandikoowaat ni projet speculatif.
Agent codage autonome buy jëfandikoo ay fukki-fukki diff yu ndaw ci repo ak latency bu néew ci coppite bu nekk, di gaawloo liggéey bi yépp.
Jumtukaayu refactoring buy formatewaat ak yokk xeetu tegtal ci benn modle bu mag, di xool li ëpp ci logique bi soppikuwul ci paralel moo gën ñu koy defaraat.
Risk yi ak balustrade yi
Optimize benn benchmark mën na nëbb ñakk kattan yu gëna yaatu ci sistem bi.
Njëg li ñuy fay ci infrastructure yi ak ci toppatoo dañuy faral di suufeel.
Bu sistem yi di gëna xawa jafee xam, jafe-jafe yi am ci wàllu kaaraange ak seetlu mën nañu gëna bari.
Roadmap ngir samp gi
Mandargal latency, kalite, ak njëg yi laata ngay jëfandikoo.
Benchmark ci biir sargal ak done yu dëggu.
Jumtukaay bi di saytu njuumte yi, derive bi ak njeextalu jëfandikukat bi.
Waajal rollback ak yooni tontu ci jafe-jafe yi laata ngay eskale.
Weyal di banneexu
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Speculative Edits for Code Models quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Gis bi ci topp
Modèlu projet de decodage speculatif
Laaj yi ñuy faral di laaj
What is Speculative Edits for Code Models?
Coppite yu xalaat yi dañuy tax soppi kodu IA di yëg ci saasi ndax dafay wax luy waaja am ci fichier bi du soppeeku, te dafay xool pàcc yu ndaw yi wuute. Dafa am solo ndax mën na wàññi latency ngir binndaat yu bari ci xeetu magnitude ci jumtukaayi kodage.
Ban seetlu bu am solo mooy tax coppite yi am njariñ?
Soppi yi dañuy faral di soppi yenn liiñ yu néew, kon li ëpp ci token yi ñuy génne dañu méngoo ak source bi njëkk, te mën nañu leen jëfandikoowaat.
Fan la 'projet' bi bawoo ci soppali spekulatif, buñu ko méngale ak dekodaas spekulatif buñ miin?
Lu moy ab modelu projet bu wuute, dosiye bi ci boppam mooy joxe proposition speculatif yi ci anam wu amul fayda.
naka la dekodage autorégressif buñ miin di defaree ay jeton?
Decodage classique dafay toppalante, di defar benn token ci pass forward bu nekk, loolu mooy tax génne yu mag yi di yeex.
Ci anam yi ñuy xalaat, lan mooy dall ci ay token yuñ proposé te model bi dafay firndeel ni jub nañu?
proposition yuñ firndeel dañu leen di nangu ci paralel, moo tax token yu bari yu méngoo dañuy njëg lu tollu ci benn pass ci barabu bari.
Ak coppite yu speculatif, njëgu defar bi gëna jege ak lan?
span yiñ soppiwul dañuy nangu ci njëg yu yomb, kon liggéeyu generation dëgg mingi aju ci ni coppite yi di doxee, du guddaayi fichier bi.