Seetug Beam
Beam search ab pexem decodage la buy tëye yenn pàcc yu gëna am njariñ ci jéego bu nekk ci barabu bëgg-bëgg ci benn.
Résumé
It matters because it produces higher-quality, more coherent text for tasks like translation and summarization than picking the single best word every time.
Plongeur bu xóot
Su benn xeetu làkk defaree mbind, dafay wax luy waaja xew ci token bi ci topp, ba noppi baamtu ko. Decodage Greedy dafay jël benn token bi gëna am wërsëg, waaye loolu mën nala dugal ci benn koñ - tànneef bu teel bu gëna baax ci dëkk bi mën na yóbbe ab frase bu gëna bon. Seetug traverse hedges ci mën topp-k ci topp-k ('yaatu traverse', lu bari 4-10). Jéego bu nekk dafay yaatal bepp limyéer ak ay token yu ci topp, di poñ kàndidaa yépp ci seen log-probabilite cumulatif, ba noppi tëye k bi gëna kawe. Resultaa bi mooy toppalante bi gëna am poñ. Nekk na standard ci tekki làkku masin te ba leegi mingi wéy di am solo, muy tekki làkk bu wóor te wóor, moo gëna am solo ci sos lu bees.
Gis-gis xarala
Seetug beam dafay am poñ ci toppalante yi ci boole log-probabilité yu token yi, loolu moo koy yóbbu ci toppalante yu gëna gàtt (token bu nekk dafay yokk benn terme bu baaxul). Ngir jàpp loolu, sistem yi dañuy jëfandikoo normalisasioŋ guddaay, xaaj poñ yi ak guddaayi toppalante (muy luñu koy yokk ba ci benn doole). Yaatu beam bu gëna mag dafay banneexu ci kàndidaa yu bari waaye dafay gëna njëg ci xayma, te yenn saa muy génne mbind mu gëna ñaaw wala mu yàqu - muy luñu bind bu baax ci tekki masin neural.
njeextalu pexe
Gaawaay ak yaatuwaay
Liggéeyukaay yi ci làkk yi mën nañu gëna gaaw te duñu yàq deggoo gi.
Dugg ak yegg
Dafay yaatal jëfandikoo gi ci làkk yi ak ci anam yi ñuy jokkoo.
dogal yu gëna leer
Ekip yi mën nañu gëna yàgg ci àtte ci jamono ji otomatisation di liggéey ci baamtu.
Ëlëgu Seetug Beam
Ci wàllu ubbeeku, gis-gis bu bees, seetlu beam dafay gëna wecci ak pexe sampling (top-k, nucleus) ndax beam yi dañuy faral di defar mbind yu bari te wuute. Waaye ci liggéey yu tënk - tekki, xàmmee kàddu, defar kode, génne buñ yamale - seetlu beam ak anam yi mu mëna doon (seetlu beam yu bari, seetlu beam bu tënk buy forse kàddu yiñ soxla) dañuy wéy di am solo. Xaarandil jegewaale hybrid yuy wéy di boole seetlu xeetu limyéer ak jël misaal, boole ci dekodaas bu xam liggéey biy méngale pexe ak njub wala wuute mooy li njëkk.
Doxal ci àdduna dëgg
Sistemu tekkikatu masin neuronal yi ñooy tànn tekkikatu frase bi gëna neexa jëfandikoo ci frase yu bari
Xamme kàddu yi ci saa si, di dekode transkripsioŋ bi gëna mëna am ci xeetu probabilite akustik
Modèlu sous-titre nataal yi defar benn sous-titre bu lëkkaloo te du benn sous-titre buñ mëna gëm
Defar lu tënk luy forse ay baatu-caabi wala ay terminoloji ñu feeñ ci génne gi jëfandikoo seetlu beam buñ tënk
Risk yi ak balustrade yi
Lépp lu jaarul yoon mën na dugg ci rapoor yi, jàppale ci liggéey bi, wala ci njariñu gëstu bi.
Sensibilite bu gaaw mën na jur njariñ yu wuute ci laajte yu noonu mel.
Done yu am solo mën nañu feeñ sudee seytu jëfandikoo gi néew doole.
Roadmap ngir samp gi
Mandargal formaa génne gi, melokaan bi, ak standard kalite yi laata ngay dugal ko.
Tontu yu am solo ak balluwaay yu wóor saa yu dëggu bi di am solo.
Fexeel am barabu xool nit ñi ngir am njariñ yu am solo.
Toppal anami gacce yi ak di faral di tàggataat ay laaj wala def-liggéey.
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 Beam Search 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
Seetug beam buñ teg loxo ak ay tënk
Laaj yi ñuy faral di laaj
What is Beam Search?
Beam search ab pexem decodage la buy tëye yenn pàcc yu gëna am njariñ ci jéego bu nekk ci barabu bëgg-bëgg ci benn. Dafa am solo ndax dafay defar mbind mu gëna baax, gëna méngoo ci liggéey yu melni tekki làkk ak tënk, moo gën tànn benn baat bi gëna baax saa yu nekk.
Lan la 'yaatu traverse' (k) di saytu ci seetug traverse?
Yaatu traverse bi mooy limu sequence partiel yu kaw yiñ tëye ak yaatal ci jéego bu nekk; yaatuwaayu gëna mag dafay banneexu ci kàndidaa yu bari.
Lan moo wuutale seetlu beam ak dekodaas bu bëgg xaalis?
Decodage bu bëgg lu bari du jël token bi gëna kawe ci jéego bu nekk; beam seet hedges ci topp ay xeetu dige yu bari benn yoon.
Lan moo waral ñuy faral di jëfandikoo normalisasioŋ ci guddaay ci seetug traverse?
Bépp jeton buñ yokk dafay yokk benn log-probabilite bu baaxul, kon su normalisasioŋ amul, toppalante yu gàtt yi dañuy gëna am poñ ci anam wu jaarul yoon.
Ban liggéey la seetlu beam bi gëna baax?
Seetug beam dafay gëna baax sudee mébet bi mooy benn tekki bu jaar yoon, te neex, moo tax mu nekk lu ñuy faral di def ci tekki làkk.
Lan mooy njariñu jëfandikoo yaatuwaayu traverse bu yaatu lool ngir defar lu ubbeeku?
Ci anam wu wuute ak xam-xam, limyéer yu gëna yaatu ci defar mbindu neuronal dañuy faral di taamu xeetu mbind yu bari, yu ñuy baamtu moo gën yu am solo.