Ikibazo gihuza ibibazo · Biragoye Urwego

Guhagarika-Ntibisanzwe na Kavukire Ntibisanzwe Ikibazo

Kugenzura guhagarika-gake kwitondera, NSA yo kwikuramo-guhitamo-idirishya amashami, guhuza ibyuma no gupima hamwe nurutonde rurerure.

Ikibazo 1 Bya 8

Ni ukubera iki kwiyitaho bisanzwe bihenze kumurongo muremure?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Ni ukubera iki kwiyitaho bisanzwe bihenze kumurongo muremure?
    • Igiciro cyacyo gikura inshuro enye n'uburebure bukurikirana
    • Ikoresha ibipimo bike cyane
    • Ntishobora gukora kuri GPUs
    • Yitabira gusa ikimenyetso cya mbere
  2. Ni ikihe gitekerezo cyibanze cyo guhagarika ibitekerezo?
    • Kuraho ibiryo-byimbere
    • Kubara kwitondera gusa hagati yatoranijwe yerekana ibimenyetso
    • Gukoresha umutwe umwe gusa
    • Kwikuba kabiri
  3. Niki gituma Kavukire idasanzwe (NSA) itandukana nuburyo bwinshi bwambere?
    • Ikuraho ibitekerezo rwose
    • Cyakora mugihe cyo gufata umwanzuro
    • Nibitozwa kurangiza-kurangira hamwe nibikoresho-byahujwe
    • Ntabwo bisaba GPU
  4. Ni ayahe mashami atatu NSA ihuza urufunguzo n'indangagaciro?
    • Encoder, decoder, hamwe no gushiramo
    • Kureka, gusanzwe, no guhuriza hamwe
    • Kwiyongera, gutema, no kurigata
    • Gucomeka, guhitamo, hamwe nidirishya
  5. Ni ukubera iki NSA ikoresha ibikorwa byo guhagarika urwego aho gukoresha ibimenyetso-buke buke?
    • Guhuza na GPU yibuka kwinjira hamwe na tensor-core yinjiza kubintu byihuse
    • Kugirango icyitegererezo kibe kinini
    • Kwirinda gukoresha ibitekerezo na gato
    • Kugabanya umubare wibice
  6. Mugihe cyo gutandukanya autoregressive, kwitonda gake bifasha cyane kuko iyo ntambwe akenshi:
    • Kubara-kubara kuri matrix iragwira
    • Kwibuka-kwibuka, kugabanya ibimenyetso byitabiriwe byihuta
    • Yigenga yuburebure bwikurikiranya
    • Koresha kuri CPU gusa
  7. What information does NSA’s compression branch provide to the attention computation?
    • A full dense score for every token pair
    • Only the optimizer’s first moment
    • Coarse-grained summaries of blocks of tokens
    • A random subset of output labels
  8. Why can learned sparsity patterns be more useful than a fixed arbitrary mask?
    • They remove the need to compute attention scores
    • They guarantee zero memory use
    • They force every token to attend globally
    • They can select blocks relevant to the input while remaining aligned with efficient hardware operations