quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz buy wax luy am ci Token bi ci topp

Xoolaat tàggat modelu làkk autoregressive, contexte causal, forse jàngalekat ak wuute gi nekk ci digganté tàggat ak jikkoom.

Yooni njiit yu jëm ciToken bi ci topp
Laaj 1 de 8

Lan la modelu wax luy am ci token bi ci topp di génne ci jéego bu nekk?

Xoolaatal laaj bu nekk8 laaj

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

  1. Lan la modelu wax luy am ci token bi ci topp di génne ci jéego bu nekk?
    • Séddale probabilite ci vokabileer bi ngir token bi ci topp
    • Benn baat bu takku
    • Tekki làkku frase bi
    • ab poñ ci grammaire
  2. Ban xeetu bàyyi xel la dekodeer bu nuroo ak GPT di jëfandikoo?
    • Fexe ci ñaari yoon
    • Causal (maske) bàyyi xel buy nëbb jetons ëlëg
    • Amul benn ndànk
    • Fexe ci anam wu bari
  3. Luy tekki ci jamonoy 'autorégressif' fii?
    • Modèle bi dafay defar token yépp benn yoon
    • Modèle bi dafay dellu ci valeur moyenne
    • Bépp jeton bu bees dañu koy yokk ba noppi delloo ko ngir xam li ci topp
    • Token yi ñu ngi leen di defar ci anam wu ñu ko bëggee
  4. Ban fonction perte lañuy gëna wàññi ci diiru tàggat yaram?
    • Jumtukaay kaare moyen
    • Perte de charnière
    • Distance cosinus
    • Entropie cross ci token bi ci topp
  5. Luy yokk tàngoor wi ci diiru jël misaal?
    • Dafay tax génne gi gëna bari ay anam yu wuute ak xarañteef
    • Dafay tax génne gi nekk deterministik
    • Yokk dayo vokabileer bi
    • Dindi couche softmax bi
  6. Laata ñuy wax luy am, naka lañuy faral di waajal mbind mu ñor mi?
    • Dañu ko tekki ci Àngle
    • Tokenizer bi moo koy xaaj ay token
    • Dañu koy soppi def ko audio
    • Dañu koy bàyyi muy ay araf yu ñor te kenn duko def dara
  7. During next-token training, what is teacher forcing?
    • The model receives future target tokens before the input
    • The model randomly changes the tokenizer at every step
    • The model receives the ground-truth preceding tokens when predicting the next token
    • The model always samples its own previous prediction during training
  8. What mismatch can arise because generation feeds back model-produced tokens?
    • The model can no longer use attention
    • The vocabulary disappears after the first token
    • The model’s loss is computed only on images
    • Errors in generated prefixes can affect later predictions, unlike training on the ground-truth prefix