Ikibazo gihuza ibibazo · Hagati Urwego

Ikibazo Cyibishusho Byibisekuruza

Wige uburyo moderi yerekana amashusho yerekana amashusho nkibimenyetso bikurikirana hanyuma ubyare intambwe imwe icyarimwe uhereye kumyandiko.

Inzira zijyanye no kuyoboraIgisekuru cya Autoregressive Igisekuru
Ikibazo 1 Bya 8

'Autoregressive' bisobanura iki murwego rwo kubyara amashusho?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. 'Autoregressive' bisobanura iki murwego rwo kubyara amashusho?
    • Buri kintu gishya giteganijwe gutegurwa kubintu byakozwe mbere
    • Buri kintu gishya cyahanuwe hakoreshejwe gusa inyandiko
    • Ishusho yose yerekanwe icyarimwe biturutse ku rusaku rudasanzwe
    • Pixels ikorwa rwose itigenga
  2. Nigute bigezweho bigezweho byerekana amashusho nka DALL-E 1 mubisanzwe byerekana ishusho mbere yo kubihanura?
    • Nka vector imwe ikomeza
    • Nka gride ihamye yikimenyetso cya discret kuva kode yize
    • Nka mbisi 24-bit RGB pigiseli gusa
    • Nkurutonde rwibishushanyo mbonera
  3. Ni ubuhe bwoko bwambere bwatanze amashusho pigiseli mbisi icyarimwe?
    • GANs na VAEs
    • Gutandukana gushikamye hamwe na Imagen
    • PixelRNN na PixelCNN
    • KANDA na BLIP
  4. Nubuhe buryo butuma Transformer yitabira gusa ibimenyetso byambere mugihe cya autoregressive generation?
    • Kureka
    • Ibisanzwe
    • Siba amahuza
    • Impamvu (yitwikiriye) kwiyitaho
  5. Ni izihe ngaruka nyamukuru zifatika zo gutoranya amashusho?
    • Gutoranya birakurikiranye bityo biratinda
    • Ntishobora gutondekwa ku nyandiko
    • Ntishobora kwerekana urugero rwiza
    • Ntabwo bisaba amakuru yamahugurwa
  6. Nubuhe buryo bwihuta bwerekana ibimenyetso bishingiye kumashusho muguhishurira ibimenyetso byinshi murwego rumwe?
    • Guhatira abarimu
    • Masked parallel decoding (urugero, MaskGIT / Muse)
    • Gukata buhoro buhoro
    • Kode imwe
  7. How does the original DALL·E model use text and image tokens?
    • Separate unconnected classifiers process text and pixels
    • A diffusion decoder predicts all pixels simultaneously
    • A single autoregressive Transformer models the sequence conditioned on text and image tokens
    • The model uses only a text encoder and no image-token sequence
  8. What scale of image-text pairs did the original DALL·E paper report collecting?
    • About 250 pairs
    • About 25 thousand pairs
    • About 25 billion pairs
    • About 250 million image-text pairs