Ikibazo gihuza ibibazo · Hagati Urwego

Ikirangantego cya Lens hamwe na Lens Ikibazo

Kugenzura ibyasomwe hagati-ibice byasomwe, ibisigisigi-bitemba byubatswe, umusemuzi wa lens yahinduwe hamwe nimbibi za lensitike mbisi.

Inzira zijyanye no kuyoboraLens Lens na Lens Lens
Ikibazo 1 Bya 8

Lens ikora iki?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Lens ikora iki?
    • Koresha icyitegererezo cyanyuma kidashyizwe kumurongo hagati yihishe kugirango usome ibyahanuwe kuri buri cyiciro
    • Ongeraho ibice bishya kuri transformateur
    • Kuraho ibitekerezo kurugero
    • Guhindura indimi zabantu
  2. Ni ubuhe buryo bw'ingenzi butezimbere lens yongeweho yongeramo lens mbisi mbisi?
    • Ikoresha amagambo manini
    • Itoza ikarita ntoya kuri affine ikarita yo guhindura leta zihishe muburyo bwanyuma
    • Ikuraho ibikenewe byose bitarimo gushyirwaho
    • Ikora gusa kumashusho yerekana
  3. Ni ubuhe buryo bwo guhindura ibintu butuma izo lens zishoboka?
    • Akayunguruzo
    • Ikimenyetso
    • Inyongera isigara yimigezi idashizwemo nyuma imishinga yo kwinjira
    • Igipimo cyo kwiga neza
  4. Ninde watangije lens yumwimerere, kandi hafi ryari?
    • DeepMind muri 2015
    • OpenAI muri 2018
    • Google muri 2012
    • nostalgebraist muri 2020
  5. Ni hehe lens ya logit yerekana igisubizo cyanyuma 'korohereza'?
    • Hagati-yatinze
    • Gusa murwego rwa mbere
    • Nta na rimwe, igisubizo kigaragara icyarimwe
    • Hanze yicyitegererezo rwose
  6. Ni ukubera iki lens ya mbisi mbisi idashobora kwizerwa mubice byambere?
    • Ibice byambere ntabwo bifite neuron
    • Ibihugu byabo byihishe ntabwo bihujwe nibisohoka byanyuma gushingirwaho biteganijwe
    • Ibice byambere ntibikoreshwa muburyo bwo guhanura
    • Matrix idashyizwe hamwe ni ibintu bisanzwe
  7. What does the model’s final unembedding map a residual-stream vector into?
    • A new set of transformer layers
    • A similarity graph over training examples
    • Vocabulary logits that can be converted into token probabilities
    • A gradient clipping threshold
  8. Why can an early-layer raw logit-lens prediction be misleading?
    • The layer has already generated the full answer
    • The tokenizer has been removed
    • The unembedding only works on images
    • The intermediate state may not yet be aligned with the final output basis