Ikibazo gihuza ibibazo · Biragoye Urwego

Gahunda yo Gutoranya no Kumurika Bias Ikibazo

Reba uburyo icyitegererezo cyateganijwe gihindura imiterere ya decoder, impanvu zahabu-token ishobora kuba igabanuka, hamwe nubucuruzi bugenda butangwa.

Ikibazo 1 Bya 8

Kubogama kubogama bivuka kuberako moderi yatojwe nabarimu bahatira gusa ibisabwa kuri:

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Kubogama kubogama bivuka kuberako moderi yatojwe nabarimu bahatira gusa ibisabwa kuri:
    • Ibimenyetso-byukuri, ariko bigomba gukoresha ibyabo
    • Ibimenyetso byabo byahanuwe
    • Ibimenyetso by'urusaku bisanzwe
    • Ibimenyetso bizaza
  2. Icyitegererezo cyateganijwe gikora iki mugihe cy'amahugurwa?
    • Buri gihe ugaburira ikimenyetso cyukuri
    • Bisanzwe uhitamo hagati yikimenyetso nyacyo nicyitegererezo ubwacyo
    • Guhagarika imikorere yigihombo buri gihe
    • Gukuba kabiri igipimo cyo kwiga mugihe
  3. Nigute bishoboka ko ukoresha ibimenyetso-byukuri byerekana ibimenyetso hejuru y'amahugurwa ateganijwe?
    • Igumaho
    • Itangira hasi kandi iriyongera
    • Itangira hejuru kandi ikangirika yerekeza kuri zeru
    • Iranyeganyega ku buryo butunguranye
  4. Ninde watangije icyitegererezo, kandi hafi ryari?
    • Vaswani n'abandi., 2017
    • Hinton n'abandi., 2006
    • LeCun n'abandi, 1998
    • Bengio n'abandi., 2015
  5. Ikimenyane cya tekinike kizwi cyo kugaburira ibimenyetso byerekanwe mugihe cy'amahugurwa ni:
    • Intambwe yintangarugero ntago itandukanye, igoye gradients
    • Bituma imyitozo igenwa neza
    • Ntabwo bisaba amakuru yemewe
    • Bikuraho gukenera gushyirwaho
  6. Ni ubuhe buryo bw'ingengabihe bukunze kuvugwa kubora umwarimu uhatira?
    • Guhoraho
    • Kubora-sigmoid kubora
    • Kwiyongera kumurongo
    • Kugenda bisanzwe
  7. At a decoding step where scheduled sampling chooses the model-output branch, what is fed as the next context token?
    • The next ground-truth token every time
    • A token from a separate validation sequence
    • A token sampled from the model’s predicted distribution
    • The average of all vocabulary embeddings
  8. Why does scheduled sampling usually begin with a high probability of using ground truth?
    • It guarantees unbiased gradients
    • It removes the need for a training schedule
    • It makes inference use ground-truth tokens
    • Early predictions are poor, so a correct prefix can stabilize initial learning