Xiga xigaHagaha xiga
Ciqaabta Ku-noqoshada iyo Koontaroolada Go'aaminta
Luuqadda AI
Luqadda AI HAGAHA
Madaxa Goynta Medusa
Medusa waa hab male-awaal-dekode ah kaas oo dhawr saadaasha dheeraadka ah ku xidhaya qaabka luqadda si ay hal mar u qiyaasto calaamado badan oo mustaqbalka ah.
Dhageyso hagahanAI-waa laga sheekeeyay · ~2 min
Boggaan2 daqiiqo akhri
Dulmar
Marka la xaqiijiyo qiyaasahan hal baas oo hore ah, waxa ay dedejisaa jiilka qoraalka qiyaas ahaan 2-3x iyada oo aan la beddelin qaybinta wax soo saarka moodeelka.
quusid qoto dheer
Normal language models generate one token per forward pass, which is slow because each step must wait for the previous one. Medusa adds lightweight feed-forward heads on top of the frozen base model; each head predicts a token a few positions ahead (head 1 predicts the next token, head 2 the token after, and so on). These predictions form a tree of candidate continuations. The full model then verifies the whole tree in one pass using a 'tree attention' mask, accepting the longest prefix that matches what the model would have produced anyway. For Medusa-1, the backbone stays frozen and original-model rejection sampling can preserve the base model's output distribution, enabling lossless inference acceleration. Medusa-2 jointly trains the heads and backbone with a special recipe that preserves the backbone's next-token capability while improving head prediction. The paper also studies typical acceptance, which can increase acceptance and maintain similar generation quality but does not make the same exact distribution-matching guarantee.
Dhaqangelinta Adduunka-dhabta ah
Goynta daahitaanka jawaabta chatbot adigoo aqbala calaamado badan oo la xaqiijiyay midkiiba gudbin
Dardar gelinta kaaliyeyaasha dhamaystirka koodka halkaasoo taxanaha calaamaduhu ay fududahay in la qiyaaso
Hoos u dhigida qiimaha goynta ee LLM API-yada taraafikada sare iyada oo aan la dirin qaab qabyo ah oo gooni ah
Dardar gelinta qaab-dhismeedka qoraalka qaab-dheer sida koobitaanka iyadoo la ilaalinayo wax-soo-saarka mid la mid ah dejinta koodka caadiga ah
Mustaqbalka Madaxa Goynta Medusa
Go'aaminta male-awaalka ah waxay noqonaysaa halbeegyada isku xidhka soo-saarka wax-soo-saarka, iyo habab is-xakameysan sida Medusa, oo iska ilaalinaya u baahan qaab qabyo ah, ayaa soo jiidasho leh sababtoo ah way ka fudud yihiin in la geeyo. Shaqada mustaqbalka waxay isku daraysaa madaxda qaabka Medusa oo leh qaabka EAGLE saadaasha, dhismaha geedka oo ka wanaagsan, iyo xaqiijinta qalabka garanaya. Filo is dhexgalka adag ee qaab-dhismeedka u adeegida, hagaajinta tooska ah ee qaabka geedka shaqo kasta, iyo isku darka isku dhafka KV-cache si daahitaanku u dhaco iyada oo aan la helin GPU-yo dheeraad ah ama lumin tayo leh.
Ku dhaqangelinta
Maxay muhiim u tahay
- Xawaaraha iyo miisaanka. Socodka shaqada luqaddu si dhakhso leh ayay u socon kartaa iyada oo aan la hurayn joogteynta.
- Helitaanka iyo gaarsiinta. Waxay balaadhisaa gelitaanka luqadaha iyo qaababka isgaarsiinta.
- Go'aamo cad. Kooxuhu waxay waqti badan ku qaadan karaan xukunka halka otomaatiggu uu qabanayo ku celcelinta.
Iska ilaali
- Xaqiiqooyinka dhalanteed waxay si deggan u geli karaan warbixinnada, taageerada socodka, ama natiijooyinka cilmi-baarista.
- Dareenka degdega ahi wuxuu abuuri karaa natiijooyin aan iswaafaqayn codsiyada la midka ah.
- Xogta qoraalka xasaasiga ah ayaa laga yaabaa in la kashifo haddii kontaroolada gelitaanka ay daciif yihiin.
Tallaabooyinka xiga
- Qeex qaabka wax soo saarka, codka, iyo heerarka tayada ka hor inta aan la baahin.
- Jawaabaha salka ku haya ilo lagu kalsoon yahay mar kasta oo saxnidu ay muhiim tahay.
- Hayso isbaarada dib u eegista bini aadamka ee wax soo saarka sare.
- Lasoco qaababka guuldarada oo dib u leyli dardargelinta ama socodka shaqada si joogto ah.
Sii wad Sahaminta
Warqad bilaash ah
Hel warbixinta maalinlaha ah ee AI
Saddex sheeko oo la xaqiijiyay oo AI ah maalin kasta oo shaqo ah, oo lagu qoray Ingiriis fudud. Bilaash had iyo jeer, aan xayaysiis lahayn.
Hal iimayl maalin kasta oo shaqo ah. Ka bax hal gujin. Marna ma iibinno ama ma wadaagno ciwaankaaga.
Is tijaabi
Take the Medusa Decoding Heads quiz
Fikrad degdeg ah oo ku saabsan jawaab kasta, iyo shahaado la wadaajin karo oo leh aqoonsi la xaqiijin karo markaad ka gudubto koorso.
Taageer waxbarashada AI bilaashka ah. AI Understanding waa 501(c)(3) aan faa'iido doon ahayn — xayaysiis la'aan, albaab lacag ah ma jiro, had iyo jeer. Samee deeq
Su'aalaha soo noqnoqda
Waa maxay Medusa Decoding Heads?
Medusa waa hab male-awaal-dekode ah kaas oo dhawr saadaasha dheeraadka ah ku xidhaya qaabka luqadda si ay hal mar u qiyaasto calaamado badan oo mustaqbalka ah. Marka la xaqiijiyo qiyaasahan hal baas hore ah, waxa ay dedejisaa jiilka qoraalka qiyaas ahaan 2-3x iyada oo aan la beddelin qaybinta wax soo saarka moodeelka.
Maxay saadaaliyeen madaxda Medusa ee dheeraadka ah?
Madax kasta oo Medusa ah wuxuu saadaaliyaa calaamad dhowr boos mustaqbalka, sidaas darteed dhowr calaamadood ayaa la soo jeedin karaa hal mar.
Under which Medusa configuration does the paper describe lossless inference acceleration?
The Medusa paper identifies Medusa-1, which freezes the backbone, as enabling lossless acceleration; exact rejection sampling verifies accepted candidates against the base model distribution.
Qaab noocee ah ayaa sii wadida musharaxiinta Medusa loo diyaariyey si loo xaqiijiyo?
Musharixiinta waxay sameeyaan geed, maaskarada dareenka geedku waxay u ogolaataa qaabka salku inuu xaqiijiyo dhammaan laamaha hal hore u dhaafo.
Waa maxay faa'iidada muhiimka ah ee Medusa marka loo eego qaab-qabyo-qodoynta mala-awaalka ah?
Medusa waxay ku dhejisaa madaxyo fudud qaabka jira, markaa looma baahna in la tababaro oo loo adeego shabakad qabyo ah oo gaar ah.
Qiyaastii xawaare noocee ah ayay Medusa caadi ahaan ka warbixisaa?
Medusa guud ahaan waxay ku guulaysataa qiyaas ahaan 2-3x hoos u dhac ku yimaada jiilka iyadoo ku xidhan culayska shaqada.
Sii wad waxbarashada
Tilmaamaha la xidhiidha
Tilmaamayaal badan ayaa loo doortay mawduucan