quiz bu lëkkale ak tegtal · Diggante Niveau

Baamtu Penalty ak Dekodaas Kontrol Quiz

Teg leen tàngoor, top-k, top-p, ak baamtu ay daan, boole ci ni logits di soppikoo balaa ñuy jël misaal ci LLM defar liggéey.

Yooni njiit yu jëm ciBaamtu Penalty ak Kontrolu Dekodaas
Laaj 1 de 8

Lan mooy yokk parametru 'temperatiir' bi ci génnug model bi?

Xoolaatal laaj bu nekk8 laaj

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

  1. Lan mooy yokk parametru 'temperatiir' bi ci génnug model bi?
    • Daf koy gëna bari ay anam
    • Daf koy gëna deterministe ak focus
    • Dafay soppi diisaayu model bi ba fàww
    • Gaawaay defar
  2. naka la top-p (nucleus) di jëlee dogal ci ban jeton lañu wara bàyyi xel?
    • Dafay tëye benn limu jetons ak lu mëna xew
    • Dafay denc ensemble bi gëna ndaw ci jetons yi nga xamni probabilite yi dañuy boole ci benn threshold
    • Dafay denc benn token bi gëna am wërsëg
    • Dafay denc jeton yu am probabilite bu ëpp 50%
  3. Ban etap la penalti buy baamtu di dox?
    • Ci kaw mbind mu ñor ginaaw generation
    • Ci done yiñ tàggat
    • ci logit yi balaa softmax
    • Ci kaw vokabileer tokenizer bi
  4. In OpenAI’s text-generation API, how do presence and frequency penalties differ?
    • Balance penalti ak lim token; Daan frequence bi dafa plat
    • Dañuy nuru
    • Ñoom ñaar ñépp ñu ngi jëm ci token bu njëkk bi
    • Frequency penalty scales with how often a token appeared; presence penalty applies once if it appeared at all
  5. For code completion where deterministic sampling is desired, which setting best matches the guide?
    • Tamperatiir bi jege amul dara
    • Tàngoor wu metti lool
    • Penalti bu bari ci baamtu bu ëpp 2.0
    • Top-p bu 1.0 ak tàngoor wu yéeg
  6. Luy top-k échantillonnage?
    • Dafay tëye jeton yi ci probabilite k
    • Dafay tëye k token yi gëna am wërsëg nekk kàndidaa
    • Yokk logit yépp ak k
    • Dindi k jetons yi gëna néew wërsëg ci tàggat yaram
  7. With a repetition penalty above 1, how does a standard logits processor adjust a previously seen token?
    • It sets every repeated-token logit to zero
    • It adds the penalty to every token equally
    • It divides a positive logit by the penalty and multiplies a negative logit by it
    • It removes the token from the vocabulary permanently
  8. How does top-k differ from top-p sampling?
    • Top-k uses cumulative probability while top-p fixes a count
    • Top-k changes model weights while top-p changes the tokenizer
    • They are both deterministic decoding methods
    • Top-k keeps a fixed number of candidates; top-p keeps a probability-mass-based set