quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz ngir defar nataal yuy dellu ginaaw

Jàngaleel ni modelu nataal autoregressive yi di kodee nataal yi ñuy toppalante ay token ba noppi ñu defar leen benn jéego benn yoon ci laaj mbind.

Yooni njiit yu jëm ciDefar nataal yuy dellu ginaaw
Laaj 1 de 8

Luy 'autoregressive' ci wàllu defar nataal?

Xoolaatal laaj bu nekk8 laaj

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

  1. Luy 'autoregressive' ci wàllu defar nataal?
    • Bépp élément bu bees ñu ngi koy seentu ci élément yiñ njëkka defar
    • Bépp mbir bu bees ñu ngi koy seetlu ci jëfandikoo bind bi kese
    • Nataal bi yépp dafay denoise ci saasi ci bruit yu bari
    • Pixel yi dañu leen di defar te kenn ci ñoom amul benn boor
  2. Ak naka la xeetu nataali autoregressif yu bees yu melni DALL-E 1 di wane nataal balaa ñu koy wax luy am?
    • Benn vecteur bu wéy
    • Ni griy bu takku bu ay jeton yu wuute yu bawoo ci benn téere kode buñ jàng
    • Pixel RGB yu 24 bit yu ñor
    • Ni limu jëmmu grafik vecteur
  3. Yan model yu njëkk ñoo defar nataal benn pixel bu ñor benn yoon?
    • GANs ak VAEs
    • Diffusion ak nataal bu dëgër
    • PixelRNN ak PixelCNN
    • CLIP ak BLIP
  4. Ban mekaniism mooy tax Transformer bi di toppatoo jeton yu njëkk yi ci jamonoy defar autorégresif?
    • Dàddu
    • Normalisasioŋ ci lots
    • Teg lëkkaloo yi
    • Sabab (maske) bàyyi xel ci sa bopp
  5. Lan mooy njariñ li gëna mag ci sampling nataal autorégressif?
    • Sampling dafay toppalante, moo tax dafay yeex
    • Mënuñu ko tënk ci mbind
    • Mënul wane ay detay yu rafet
    • Soxlawolen ay done tàggat
  6. Ban pexe mooy gaawlu defar nataal bu sukkandiko ci token ci di wax luy waaja am ci token yu bari ci paralel?
    • Jàngalekat bi di forse
    • Dekodage paralel buñ maskeer (lu melni, MaskGIT/Muse)
    • Dagg degrade
    • Kodage bu tàng
  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