Ikibazo gihuza ibibazo · Hagati Urwego

Ikibazo gisanzwe

Kugenzura Imibare idasanzwe hamwe na affine ibipimo, imyitwarire yubunini bwayo, epsilon, RMSNorm hamwe na pre-norm yashyizwe. mu miyoboro yimbitse.

Inzira zijyanye no kuyoboraUbusanzwe
Ikibazo 1 Bya 8

Hirya no hino kurwego rusanzwe rugereranya iki kandi rutandukanye?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Hirya no hino kurwego rusanzwe rugereranya iki kandi rutandukanye?
    • Kurenga ibiranga urugero rumwe
    • Hafi yingero zose muri mini-batch
    • Kurenza ibipimo byose murusobe
    • Mubihe byinshi byamahugurwa
  2. Ni ukubera iki ibisanzwe bisanzwe bikundwa kuruta icyiciro gisanzwe muri transformateur?
    • Irasaba ingano nini cyane
    • Irigenga ingano yicyiciro kandi ikora impinduka-z'uburebure
    • Bikuraho ibikenewe byose byiga
    • Cyakora mugihe cyo gufata umwanzuro
  3. Niki ibipimo byiga gamma na beta reka LayerNorm ikora?
    • Hitamo igipimo cyo kwiga
    • Shiraho umubare
    • Kuraho no guhindura ibisohoka bisanzwe kugirango urusobe rushobore kugarura ibyingenzi
    • Kuraho ibikenewe mumikorere yo gukora
  4. Nigute RMSNorm itandukaniye na Layeri isanzwe?
    • Birasanzwe murwego rwose
    • Ntabwo ikoresha ibipimo kandi nta kugabana
    • Ikora kumashusho gusa
    • Gusimbuka bisobanura gukuramo kandi bigabanywa gusa numuzi-bivuze-kare
  5. Ni ikihe kibazo gisanzwe gisanzwe gifasha gukemura mumiyoboro yimbitse?
    • Ibikorwa bigenda kumunzani itandukanye cyane, bihagarika imyitozo
    • Kubura umwanya wa disiki
    • Guhitamo imibare yukuri
    • Shishoza uburemere bw'icyitegererezo
  6. Ni ukubera iki gushyira 'pre-norm' bikunze gukoreshwa muri transformateur yimbitse?
    • Bituma icyitegererezo ari gito
    • Ubusanzwe mbere ya buri sublayer ituma ibice byimbitse byoroha kandi bihamye guhugura
    • Ikuraho uburyo bwo kwitondera
    • Bikubye kabiri igipimo cyo kwiga mu buryo bwikora
  7. Why is a small epsilon added to LayerNorm’s variance before taking its square root?
    • To make the batch size larger
    • To replace the learnable scale
    • To avoid division by zero and improve numerical stability
    • To compute the sequence length
  8. If gamma is all ones and beta is all zeros, what does LayerNorm’s affine stage do?
    • It restores the original unnormalized inputs exactly
    • It applies a softmax over features
    • It changes the statistics to be batch-dependent
    • It leaves the normalized values unchanged after the mean-and-variance scaling