Lookahead ak gaynde gi

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

1. Lan mooy jéego bi gëna am solo ci yeesali diisaayu Lion optimizer?
   • jël màndarga ab term bu sukkandiko ci momentum
   • Séddale degrade bi ak moyenne bimu am
   • Xayma ap xayma Hessian bu ñaareelu rang
   • Moyenne poids ci checkpoint yu bari
2. naka la Lookahead di defaree njariñam?
   • Dafay jëfandikoo benn xeetu poid buñ yeesal ñaari yoon ci jéego bu nekk
   • Dafay tëye poid yu 'gaaw' yu ñuy yeesal ci optimisatëru base ak poid yu 'yeex' yuy faral di interpoler ci ñoom
   • Dafay dindi momentum bi yépp ngir sakkanal mémoire bi
   • Dafay tàggat ñaari reso ba noppi di xayma seeni génne
3. naka lañu njëkka gisee Lion optimizer?
   • Ni ñu ko jëlee ci theorie optimisation convexe
   • Ni ñuy xaymaa yeesali SGD ak Adam
   • Ci seetug prograamu màndarga otomatik ci kaw algorithm yiy gëna xéewale
   • Ci simulation physique buy yëngu ci particule yi
4. Buñu ko méngale ak Adam, lan mooy njariñu fàttaliku bu mag bi Lion am?
   • Du denc benn réew
   • Dafay jëfandikoo poid yu xaaju njub ci saasi
   • Dafay séddoo etaa ci bépp diisaay
   • Dafay tëye benn tampon momentum ci palaasu ñaari tensor yu Adam
5. Lan moo waral Lion di soxla njàng mu gëna ndaw Adam?
   • Ndax jëfandikoo màndarga dafay tax magnitude yeesal bu nekk nekk benn, muy gëna yokk jéego bi
   • Ndax jëfandikoowul benn momentum
   • Ndax dafay liggéey ci lote yu gëna mag
   • Ndax dafay tere yaram wi wàññeeku
6. Benn njariñu Lookahead mooy dafay tax ñu tàggat:
   • Dafay gëna am solo ci tànneefi hiperparametre yi
   • Dafa gëna dëgër te dëgër, ak oscillation bu néew
   • Gëna yeex ci jéego bu nekk lu tollu ci fukki yoon
   • mënul ànd ak Adam wala SGD
7. Lookahead slow weights are 0 and the fast weights end an inner loop at 8. If alpha is 0.25, where do the slow weights move?
   • To 8, regardless of alpha
   • To −2, away from the fast weights
   • To 2, one quarter of the distance toward the fast weights
   • They remain at 0 because alpha only changes Lion
8. After Lookahead updates its slow weights at synchronization, where do the fast weights restart?
   • At the original random initialization
   • At the final validation predictions
   • At the previous slow weights before interpolation
   • At the updated slow-weight values for the next inner loop