Nhungamiro yehunyanzvi

Kufungidzira Kutenderera uye Multi-Token Prediction

Kutepfenyura uye kufanotaura kwezviratidzo zvakawanda kumhanyisa chizvarwa chemhando yemutauro nekufungidzira akati wandei emangwana tokeni kamwechete uye nekuaona nekupfuura kumwe chete, pane kugadzira chiratidzo chimwe panguva.

2 min verengaLast update

Pfupiso

They cut latency without changing the text the model would have written.

Kudzika Kwakadzika

Yakajairika autoregressive decoding inononoka nekuti chiratidzo chega chega chinoda kupfuudza kumberi uye ma tokeni anogadzirwa akanyatso mumwe mushure memumwe, achisiya GPU isingashandiswe. Kudhirodha kwekufungidzira kunogadzirisa izvi nedhizaini rakachipa iro rinokurudzira chunk yematokeni evamiriri, iyo iyo hombe yaivavarirwa modhi yobva yaona zvakafanana; chero prefix inofanana neicho chinangwa chingave chaburitsa chinogamuchirwa mahara, uye yekutanga kusawirirana inogadziriswa. Kutepfenyura kutenderera uye Medusa-maitiro akawanda-tokeni kufanotaura peta dhizaini mumodhi pachayo: yakawedzera kureruka misoro yekufungidzira (kana rwizi rwekufungidzira tokens) rega modhi imwe inyore uye simbisa, ichinzvenga yakaparadzana dhizaini modhi. Nekuti verification ndeyechokwadi, kugovera kwakabuda kwakafanana neyakajairwa decoding, iwe unongowana 2 kusvika ku3 kakapetwa mashoma anotevedzana matanho.

Technical Insight

Chinokosha ndechekuti transformer inogona kukwevera zvinzvimbo mune imwe yekumberi kupfuura zvakachipa seimwe, sezvo iri ndangariro-bandwidth yakasungwa, kwete compute yakasungwa, panguva yekunyora. Misoro yakawanda yekufembera inoburitsa tokeni dzevamiriri venzvimbo dzinoverengeka dzinotevera; muti kana kutevedzana kwevanoda kukwikwidza kunosimbiswa pamwe chete, uye kugamuchirwa kunoshandisa sampling yekuramba (kana makaro kuenzanisa) kuitira kuti zviratidzo zvinogamuchirwa zviteedzere kugovera kwakanangana. Kureba kunogamuchirwa panhanho kunosarudza kukurumidza.

Strategic Impact

Mutengo uye bhajeti

Zvisarudzo zvezvivakwa zvinotyaira kuita uye mutengo wekushandisa kwemakore.

Sarudzo dzakajeka

Dzidzo yehunyanzvi inobatsira zvikwata kusarudza murwi wakakodzera, kwete iwo mutsva chete.

Kudzora kwemhando yepamusoro

Sarudzo dzeinjiniya dziri nani dzinoderedza zviitiko zvekuvimbika mukugadzira.

Ramangwana reKufungidzira Kutenderera uye Multi-Token Prediction

Nzira dzekungofungidzira dzisingade dzakapatsanurwa dhizaini modhi dzave kuita dzisiri muinjini dzekufungidzira, uye tsvakiridzo iri kusundidzira mitengo yekugamuchirwa yakakwira nemisoro yedhirowa iri nani, vagadziri vemiti, uye kudzidzisa iyo base modhi pamwe chete kune akawanda-token kufanotaura (ayo anogona zvakare kuvandudza mhando). Tarisira aya matekiniki kuti abatanidze nequantization uye batching saka vanopindirana vabatsiri vanonzwa nekukasira kunyangwe mamodheru achikura.

Real-World Implementation

Kucheka mhinduro latency yemubatsiri wekutaura ne2 kusvika 3x uchishandisa Medusa-maitiro ekuwedzera ekufanotaura misoro.

Kuwedzera-yekufungidzira decoding kune inference server saka hapana yakaparadzana dhizaini modhi inoda kutambirwa

Kumhanyisa kupedzisa kodhi apo refu, inofanotaurwa tokeni inotambirwa muzvikamu zvakakura

Kuderedza mutengo weGPU pachikumbiro nekuburitsa mamwe ma tokens kubva kune yega yega-inosungwa yekupfuura pass

Njodzi & Guardrails

Kugadzirisa imwe bhenji kunogona kuvanza yakafara system kushaya simba.

Infrastructure uye mari yekugadzirisa inowanzotarisirwa pasi.

Chengetedzo uye kucherechedzwa mapundu anogona kukura sezvo masisitimu anowedzera kuoma.

Implementation Roadmap

1

Tsanangura latency, mhando, uye mutengo zvinangwa usati waitwa.

2

Benchmark pasi pechokwadi mutoro uye data mamiriro.

3

Chishandiso chekutarisa zvikanganiso, kudonha, uye mushandisi maitiro.

4

Gadzirira nzira dzekudzosera kumashure uye dzezviitiko usati wawedzera.

Ramba Uchiongorora

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 Streaming and Multi-Token Prediction quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Tanga mibvunzo

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Gaidhi rinotevera

Multi-Token Prediction Training

Mibvunzo inowanzo bvunzwa

What is Speculative Streaming and Multi-Token Prediction?

Kutepfenyura uye kufanotaura kwezviratidzo zvakawanda kumhanyisa chizvarwa chemhando yemutauro nekufungidzira akati wandei emangwana tokeni kamwechete uye nekuaona nekupfuura kumwe chete, pane kugadzira chiratidzo chimwe panguva. Ivo vanocheka latency vasina kushandura zvinyorwa izvo modhi ingadai yakanyora.

Nei yakajairwa autoregressive decoding ichiwanzononoka paGPU?

Chiratidzo chega chega chinoda chayo chekupfuura uye chakanyatso kutevedzana, saka iyo GPU ndeyeyeuko-bandwidth inosungwa uye isingashandiswe panguva yekudhirodha.

Ko 'mudhirowa' anoita sei mukufungidzira decoding?

Iyo yakachipa dhizaini inokurudzira chunk yevamiriri tokeni iyo hombe yainotarirwa modhi yobva yaona zvakafanana.

Hunhu hwekubuda hunochengetedzwa sei mukufungidzira decoding?

Verification (kukara kuenzanisa kana kurambwa sampling) inova nechokwadi chekuti tokeni dzinogamuchirwa dzinotevera kugovera chaiko kwaizoburitswa modhi inotangwa, saka zvinobuda hazvichinji.

Chii chinosiyanisa Medusa-maitiro akawanda-token kufanotaura kubva kukirasi yekufungidzira decoding?

Medusa-maitiro nzira dzinopeta dhizaini mune modhi ine mamwe ekufanotaura misoro, kudzivirira kukosha kwekugamuchira yakaparadzana dhizaini modhi.

Sei transformer ichigonesa zvigaro zvakawanda zvevamiriri zvakachipa seimwe panguva yekunyora?

Panguva yekudhirodha iyo bhodhoro iri kufambisa uremu kubva mundangariro, saka kuhwina zvimwe zvinzvimbo mune imwechete pass kunowedzera mari shoma yekombuta.