Kedis ku xidhan hagaha · Dhexdhexaad Heerka

Lakabiyaha Adapter ee Kediska Wareejinta

Hubi sida adaabiyeyaasha qoor-gooyaha ay u ilaashadaan qaabka la qaboojiyey, u suurtageliyaan hawsha modulka ah ama wareejinta goobta, oo barbar dhig LoRA.

Wadooyinka haga ee la xidhiidhaLakabyada Adabtarka ee Wareejinta
Su'aal 1 ee 8

Maxaa ku dhacaya miisaanka asalka ah ee hore loo tababaray marka la tababarayo adapters?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Maxaa ku dhacaya miisaanka asalka ah ee hore loo tababaray marka la tababarayo adapters?
    • Waa la barafeeyay halka kaliya cabbirada adabtarada la tababaray
    • Waa la tirtiray
    • Si aan kala sooc lahayn ayaa dib loo bilaabay
    • Way labanlaabmeen cabbirkooda
  2. Waa maxay qaabka caadiga ah ee gudaha ee adabtarada qoor-gooyaha?
    • Hal lakab oo cufan oo weyn
    • Mashruuc hoose, aan toos ahayn, ka dibna mashruuc-sare, oo leh xidhiidh hadhaaga ah
    • Kaliya feejignaan madax
    • Shaandheeyaha sawirka convolutional
  3. Qiyaastii waa maxay jajabka cabbiraadaha ee adaabayaashu sida caadiga ah u tababaraan?
    • 100%
    • Dhab ahaan 50%
    • Inta badan wax ka yar 5%
    • Midna haba yaraatee
  4. Waa maxay sababta adapters-ku ugu habboon yihiin u adeegidda hawlo badan?
    • Waxay u baahan yihiin koobi qaabaysan oo gaar ah hawl kasta
    • Lama isku dari karo
    • Waxay tirtiraan qaabka saldhigga hawl kasta ka dib
    • Waxaad haysataa hal nooc oo saldhig ah oo waxaad beddeshaa adabiyeyaasha yaryar ee hawl kasta
  5. Qoyskee farsamada ayaa ka tirsan adapters?
    • Hagaajinta hagaajinta wax-ku-oolnimada leh (PEFT)
    • Kordhinta xogta
    • Isku-buufinta sawirka
    • Nidaamyada calaamadaynta
  6. Faa'iido intee le'eg ayay LoRA ku siinaysaa adabtarada caadiga ah ee la geliyo waqtiga qiimaynta?
    • Lama tababari karo
    • Miisaankeeda la bartay waxaa lagu dhex milmi karaa matries-ka aasaasiga ah, iyada oo laga fogaanayo daahitaanka dheeraadka ah ee lakab kasta
    • Waxay u baahan tahay dib u dhiska moodelka oo dhan
    • Waxay si joogto ah meesha uga saartaa isku xirka haraaga
  7. What does a residual connection contribute to a classic bottleneck adapter?
    • A second full pretrained model copy for each task
    • A mandatory external retrieval database
    • A path that adds the adapter’s transformed representation back to the original hidden state
    • A guarantee that no added computation occurs
  8. What is one benefit of stacking a domain adapter with a task adapter?
    • Merge all user data into the frozen base weights automatically
    • Avoid using any adapter modules
    • Convert a language model into a vision encoder
    • Reuse a domain specialization together with a separate task specialization