Ikibazo gihuza ibibazo · Biragoye Urwego

Mirostat Perplexity Ikibazo cyo kugenzura

Gupfukirana intego ya Mirostat itunguranye / urujijo, igitekerezo cyo gutanga ibitekerezo, igipimo cyo kuvugurura, itandukaniro rya verisiyo, n'imbibi nka decoding heuristic.

Inzira zijyanye no kuyoboraMirostat Perplexity Igenzura
Ikibazo 1 Bya 8

What does Mirostat try to control during generation?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. What does Mirostat try to control during generation?
    • A target level of output perplexity or surprise
    • Umubare w'icyitegererezo
    • The training dataset size
    • Ubushyuhe bwa GPU
  2. Nigute Mirostat itandukanye cyane na top-k cyangwa top-p icyitegererezo?
    • Ntabwo yigeze ikoresha ibintu bisanzwe
    • Ihindura imikorere ihindagurika ikoresheje ibitekerezo aho guhagarara neza
    • Ikora kumashusho gusa
    • Bisaba kongera imyitozo
  3. Muri Mirostat, ni uruhe ruhare ibipimo eta bigira?
    • Ishiraho ingano yamagambo
    • Ikosora ibisohoka uburebure
    • Nibipimo byo kwiga bigenzura uburyo urwego rwihuta
    • Ihagarika icyitegererezo
  4. Ni ubuhe butumwa bukomeye bwa Mirostat verisiyo ya 2 kurenza verisiyo ya 1?
    • Itanga amashusho nayo
    • Birasaba kwibuka cyane
    • Ikuraho ibipimo bya tau
    • Ihagarika igitekerezo cyo gukwirakwiza Zipfian kugirango igenzure byoroshye, bikomeye
  5. Niba Mirostat ibonye ko ibisohoka bigenda bihanurwa, bizakora:
    • Kurekura kugabanuka kugirango wemere ibimenyetso bitandukanye
    • Komeza kugabanya kugirango wemerere ibimenyetso bike
    • Reka kubyara ako kanya
    • Ongera ingano yicyitegererezo
  6. Which Mirostat settings does current llama.cpp expose?
    • Tokenizer vocabulary and number of layers
    • Mode, target tau, and learning-rate eta
    • Only the prompt length
    • Training epochs and validation loss
  7. What does eta control in Mirostat’s feedback update?
    • The size of the model vocabulary
    • The target number of output tokens
    • How strongly the threshold changes in response to the observed error
    • The probability of a correct factual answer
  8. What does a Mirostat target not guarantee by itself?
    • That the sampler reads token probabilities
    • That the setting can influence surprise
    • That the output contains tokens
    • That the generated text is factually correct