Ikibazo gihuza ibibazo · Hagati Urwego

Ibikurikira-Byabajijwe Ikibazo

Ongera usubiremo ururimi-rwerekana amahugurwa, imiterere yabitera, guhatira abarimu no gutandukanya amahugurwa nibisekuru.

Inzira zijyanye no kuyoboraIbikurikira
Ikibazo 1 Bya 8

Niki gikurikira-ikimenyetso cyerekana ubuhanuzi gisohoka kuri buri ntambwe?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Niki gikurikira-ikimenyetso cyerekana ubuhanuzi gisohoka kuri buri ntambwe?
    • Ikwirakwizwa rishoboka hejuru yamagambo kubikurikira
    • Ijambo rimwe rihamye
    • Ubusobanuro bw'interuro
    • Ikibonezamvugo
  2. Ni ubuhe bwoko bwo kwitondera GPT-yuburyo bwa decoder ikoresha?
    • Ibyerekezo byombi
    • Impamvu (mask) kwitondera ihisha ibimenyetso bizaza
    • Nta kwitabwaho na gato
    • Kwitonda bisanzwe
  3. Igisekuru 'autoregressive' gisobanura iki hano?
    • Icyitegererezo gitanga ibimenyetso byose icyarimwe
    • Icyitegererezo gisubira inyuma
    • Buri kimenyetso gishya cyongeweho kandi kigaburirwa inyuma kugirango uhanure ibizakurikiraho
    • Tokens ikorwa muburyo butunguranye
  4. Ni ubuhe buryo bwo gutakaza busanzwe bugabanywa mugihe cy'amahugurwa?
    • Ikigereranyo cya kare
    • Gutakaza Hinge
    • Intera
    • Kwambukiranya-entropiya hejuru yikimenyetso gikurikira
  5. Kuzamura ubushyuhe mugihe cyo gutoranya bikora iki?
    • Bituma umusaruro usohoka kandi udasanzwe
    • Ikora ibisohoka rwose
    • Yongera ingano yamagambo
    • Kuraho igikoresho cya softmax
  6. Mbere yo guhanura, ni gute inyandiko mbisi isanzwe itegurwa?
    • Byahinduwe mu Cyongereza
    • Igabanijwemo ibimenyetso na tokenizer
    • Yahinduwe amajwi
    • Hasigaye nkinyuguti mbisi nta gutunganya
  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