Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Adhabu ya Kurudia na Vidhibiti vya Kusimbua

Linganisha halijoto, top-k, top-p, na adhabu za marudio, ikijumuisha jinsi kumbukumbu hubadilika kabla ya kuchukua sampuli katika mtiririko wa kazi wa LLM.

Njia za mwongozo zinazohusianaAdhabu ya Marudio na Vidhibiti vya Kusimbua
Swali 1 ya 8

Kuinua parameta ya 'joto' kunafanya nini kwa pato la mfano?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Kuinua parameta ya 'joto' kunafanya nini kwa pato la mfano?
    • Huifanya kuwa nasibu zaidi na tofauti
    • Huifanya iwe ya kuamua zaidi na yenye umakini
    • Inabadilisha kabisa uzito wa mfano
    • Huongeza kasi ya kizazi
  2. Je, sampuli za top-p (nucleus) huamua vipi ni tokeni za kuzingatia?
    • Huweka idadi maalum ya tokeni bila kujali uwezekano
    • Huweka seti ndogo zaidi ya tokeni ambazo uwezekano wake unafikia kizingiti
    • Huhifadhi ishara moja tu inayowezekana zaidi
    • Inaweka tokeni na uwezekano wa zaidi ya 50%
  3. Adhabu ya marudio huwa katika hatua gani?
    • Kwenye maandishi ghafi baada ya kizazi
    • Juu ya data ya mafunzo
    • Kwenye kumbukumbu kabla ya softmax
    • Kwenye msamiati wa ishara
  4. In OpenAI’s text-generation API, how do presence and frequency penalties differ?
    • Kuwepo kwa mizani ya adhabu na hesabu ya ishara; adhabu ya frequency ni bapa
    • Wanafanana
    • Zote mbili zinatumika tu kwa ishara ya kwanza
    • 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?
    • Joto karibu na sifuri
    • Joto la juu sana
    • Adhabu ya marudio ya juu zaidi ya 2.0
    • Juu-p ya 1.0 na joto la juu
  6. Sampuli za top-k hufanya nini?
    • Huweka muhtasari wa tokeni kwa uwezekano k
    • Huhifadhi tokeni k pekee zinazowezekana zaidi kama wagombeaji
    • Huzidisha kumbukumbu zote kwa k
    • Huondoa tokeni zenye uwezekano mdogo wa k kutoka kwa mafunzo
  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