Ndozi nkọwa maka ụdị koodu
Ndozi nkọwa na-eme ka ndezi koodu AI na-enwe mmetụta ozugbo site n'ịkọ na ọtụtụ faịlụ agaghị agbanwe agbanwe ma na-enyocha obere akụkụ ndị dị iche.
Nchịkọta
It matters because it can cut latency for large rewrites by an order of magnitude in coding tools.
Ime miri emi
Mgbe AI na-edezi faịlụ, ọtụtụ akara ngosi ọ na-ewepụta na-adịkarị ka koodu mbụ; naanị ahịrị ole na ole na-agbanwe n'ezie. Ọgbọ na-enweghị isi na-ewepụtaghachi faịlụ niile site na token, nke na-adị ngwa maka nnukwu faịlụ. Ndezi nkọwa na-erigbu usoro agbanweghị agbanwe: isi mmalite dị adị na-eme dị ka 'draft' dị elu nke ihe nlereanya ahụ ga-ewepụta. Sistemu na-enye chunks nke koodu mbụ dị ka ntule ma na-ahapụ ka ihe nlereanya ahụ nyochaa ọtụtụ n'ime ha n'otu ngafe na-aga n'ihu. Ebe ihe nlereanya ahụ kwenyere, a na-anabata akara ngosi ndị ahụ ozugbo; ebe ọ na-ekwenyeghị, ọ na-ewepụta oge edoziziri nke ọma. Nke a bụ nwa nwanne nne nwere koodu pụrụ iche nke nhọpụta ntule, mana kama ịbụ obere mpempe akwụkwọ dị iche iche, usoro a na-abịa n'efu na faịlụ a na-edezi, na-enye nnukwu ọsọ ọsọ na ọrụ ndozi dị arọ.
Nghọta nka nka
Ndozi autoregressive ọkọlọtọ na-ewepụta otu akara maka ngafe ọ bụla. Ụzọ ntule na-atụ aro ọtụtụ token n'otu oge ma nyochaa ha n'otu oge: ihe nlereanya nwere ike ịlele, n'otu ngafe, ma ọsọ nke akara ngosi dabara na ihe ọ gaara eme. Ndozi nkọwa na-enye atụmatụ ndị ahụ site na koodu isi mmalite na-agbanweghi agbanwe kama ịdepụta ụkpụrụ. Ọsọ anabatara na-eri ihe dị ka otu ngafe maka ọtụtụ akara; naanị divergences na-akpalite ọhụrụ ọgbọ, otú ọnụ ọnụ akpịrịkpa na edit size, ọ bụghị faịlụ size.
Mmetụta atụmatụ
Ọnụ ego na mmefu ego
Mkpebi ihe owuwu ụlọ na-akwalite arụmọrụ yana ọnụ ahịa ọrụ ruo ọtụtụ afọ.
Mkpebi doro anya
Nkà mmụta nka na-enyere ndị otu egwuregwu aka ịhọrọ nchịkọta ziri ezi, ọ bụghị naanị nke kachasị ọhụrụ.
Quality akara
Nhọrọ injinia ka mma na-ebelata ihe omume ntụkwasị obi na mmepụta.
Ọdịnihu nke ndezi nkọwa maka ụdị koodu
Ndị ọrụ na-edezi-dị arọ na ndị enyemaka IDE ga-adabere na nke a iji nọgide na-etinye nnukwu ọdịiche dị nso ozugbo, ọbụlagodi na faịlụ ahịrị puku. Na-atụ anya mwekota siri ike na usoro dị iche iche ahaziri ahazi, atụmatụ ndị maara osisi na-asọpụrụ ókèala syntax, yana nchikota na iweghachite ya mere atụmatụ ntule na-agụnye ndị nrụpụta. Dị ka ndị ọrụ nzuzo kwụụrụ onwe na-eme ọtụtụ ndezi n'otu ọrụ, ndezi ntule na-aghọ isi ihe na-eme ka usoro ọrụ ọtụtụ nzọụkwụ na-anabata ma dị ọnụ ala iji na-agba ọsọ.
Mmejuputa n'ezie n'ụwa
Onye enyemaka IDE na-edegharị faịlụ ahịrị 500 iji nyegharịa ọrụ, na-anabata ahịrị niile na-agbanweghị na ngafe ole na ole na naanị na-ewepụta oge akpọgharịrị aha.
Iwu 'dozie njehie lint a' nke na-emepụta faịlụ emeziri ihe fọrọ nke nta ka ọ bụrụ ozugbo n'ihi na ejiri 99% nke koodu ahụ dị ka ihe nlere anya.
Onye na-ahụ maka nzuzo kwụụrụ onwe ya na-etinye ọtụtụ obere iche iche n'ofe ebe nchekwa nwere obere oge dezie, na-edobe ọrụ niile ngwa ngwa.
Ngwá ọrụ na-emegharị ihe nke na-emezigharị ma na-agbakwụnye ụdị ihe ngosi na nnukwu modul, na-achọpụta ọtụtụ echiche na-agbanweghị agbanwe n'otu n'otu kama ịmaliteghachi ya.
Ihe ize ndụ & okporo ụzọ nche
Ịkwalite otu akara ngosi nwere ike zoo adịghị ike sistemụ sara mbara.
A na-eledakarị ihe akụrụngwa na ụgwọ ọrụ anya.
Ọdịiche nchekwa na nleba anya nwere ike itolite ka sistemu na-adịwanye mgbagwoju anya.
Map mmejuputa
Kọwaa latency, ịdịmma na ebumnuche ọnụ ahịa tupu mmejuputa ya.
Benchmark n'okpuru ibu dị adị na ọnọdụ data.
Nleba anya akụrụngwa maka mperi, ịkpafu na mmetụta onye ọrụ.
Kwadebe ụzọ nzaghachi azụghachi azụ na ihe omume tupu ịchachaa.
Nọgide na-eme nchọpụta
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
Ntuziaka na-esote
Modeldị Draft Ndozi Apụrụiche
Ajụjụ a na-ajụkarị
What is Speculative Edits for Code Models?
Ndozi nkọwa na-eme ka ndezi koodu AI na-enwe mmetụta ozugbo site n'ịkọ na ọtụtụ faịlụ agaghị agbanwe agbanwe ma na-enyocha obere akụkụ ndị dị iche. Ọ dị mkpa n'ihi na ọ nwere ike belata nkwụsị maka nnukwu idegharị site n'usoro ịdị ukwuu na ngwa koodu.
Kedu ihe nlebanya na-eme ka ndezi ntule dị irè?
Ndezigharị na-agbanwekarị naanị ahịrị ole na ole, yabụ imirikiti akara ngosi mmepụta dabara na isi mmalite ma enwere ike ijikwa ya ọzọ.
Ebee ka 'draft' ahụ si abịa na ndezi ntule, ma e jiri ya tụnyere nhazi ntule nkịtị?
Kama ụdị akwụkwọ dị iche, faịlụ izizi n'onwe ya na-enye atụmatụ ntule n'efu.
Kedu otu ọkọlọtọ autoregressive decoding si ewepụta akara?
Ndozi omuma bụ usoro n'usoro, na-ewepụta otu akara n'otu ngafe na-aga n'ihu, nke na-eme ka nnukwu mpụta ngwa ngwa.
N'ime usoro ntule, kedu ihe na-eme ọsọ nke akara ngosi nke ihe nlereanya ahụ gosipụtara na ọ bụ eziokwu?
A na-anabata atụmatụ akwadoro n'otu aka ahụ, ọtụtụ akara ngosi dabara na-efu ihe dị ka otu ngafe karịa ọtụtụ.
Site na ndezi nkọ, ọgbọ na-eri ọnụ ọgụgụ kacha nso na gịnị?
A na-anabata oge na-agbanweghị agbanwe n'ụzọ dị ọnụ ala, yabụ ezigbo ọgbọ na-arụ ọrụ na-agbanwe n'ezie, ọ bụghị ogologo faịlụ.