Rollout ak dagg bopp quiz

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

1. Luy mébetu fësal bàyyi xel ci nit ñi?
   • Topp leeral yiy dajale ci jeton yi ñuy dugal ba ci yi ñuy génne ci diisaay yi
   • Ngir yokk yeneen diisaay ci Transformatër
   • Yokk limu jàng
   • Soppi bàyyi xel ci convolusioŋ
2. Lan moo waral benn kàrtu bàyyi xel ci benn couche du doy ngir joxe leeral?
   • Duñu musa xayma bàyyi xel
   • Leeral yi dañuy jaxasoo ci diisaay yu bari ak bopp, moo tax benn kàrt mënul gis yoon wi yépp
   • Kart yi dañuy nuru saa yu nekk
   • Transformatër yi benn bopp kese lañu am
3. Lan la rollout attention di yokk ci matrix attention bu nekk ngir leeral lëkkaloo yi des?
   • Bias bu amul dara
   • Term bu bari bruit
   • ab komponent matrix dàntite balaa ñuy normalise
   • Softmax bu gëna mag
4. Lan la gëstub bàyyi xel ci bopp yu bari wane ci redundancy bopp?
   • Bopp bu nekk amna solo
   • Mënu dindi bopp yi
   • Dagg saa yu nekk dafay yokk njub ñaari yoon
   • Bopp yu bari mën nañu leen dagg te duñu yàq njub
5. naka lañuy xaymaa njariñu bopp ci dagg garab?
   • Jaaraleko ci sensitiwite bu sukkandiko ci gradient bu ñàkk ci masku bopp
   • ci barabu alfabet bi
   • Ni ñuy lim ay parametram kese
   • Ci waxtu bis bi
6. What different jobs do attention rollout and head pruning each do?
   • Pexe yu nuróo lañu
   • Rollout dafay jàppale ci tekki bi dagg di jëfandikoo redondance ngir kompresse model yi
   • Ñoom ñaar ñépp dañuy yokk ay paramet
   • Kenn ci ñoom du laal xel
7. In an attention-rollout calculation, why are residual paths included with layer attention?
   • To make every head have the same weight
   • To remove the value vectors from the Transformer
   • To approximate information that can pass along the identity path when layer transitions are composed
   • To prevent the attention matrix from being normalized
8. How should an attention-rollout score be interpreted?
   • As proof that a token caused the output
   • As the exact value vector used at every layer
   • As a measure of how many heads the model contains
   • As a post-hoc approximation of relative token influence, not a causal explanation