Quiz ngir yamale couche

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

1. Ci lan la normalisasioŋ couche di xayma moyenne bi ak variance bi?
   • Ci màndarga yi ci benn misaal
   • Ci misaal yépp ci mini-batch
   • Ci diisaay yépp ci reso bi
   • Ci jamonoy tàggat yaram yu bari
2. Lan moo tax ñu taamu normalisasioŋ couche moo gën normalisasioŋ lote ci transformatër yi?
   • Dafay laaj nga am lote bu rëy
   • Dafay moom boppam ci dayo lote bi te dafay jëfandikoo ay toppalante yu am guddaay bu mën soppiku
   • Dafay dindi soxla bépp parametre buñ mëna jàng
   • Du dox ludul ci diiru inference
3. Lan la parametre yiñ mëna jàng gamma ak beta may LayerNorm mu def?
   • Tannal tolluwaayu jàng
   • Defal limu diisaay yi
   • Rescale ak toxal génne gi normalisé suko defee reso bi mëna am représentation yu am njariñ
   • Dindil soxla ab fonction aktive
4. Lan moo wuutale RMSNorm ak LayerNorm biñ miin?
   • Dafay normalise ci lote bi yépp
   • Jëfandikuwul benn paramet wala xaaj
   • Ci nataal kese lay dox
   • Dafay sànni dindi moyenne bi ba noppi séddale ko ci rasine-moyenne-kare
5. Ban jafe-jafe la normalisation couche di njëkka jàppale ci saafara ci reso yu xóot yi?
   • Liggéeyukaay yi dañu dem ci yeneen anam yu wuute lool, te loolu dafay yàq tàggat yaram
   • Espace disk bi jeexna
   • Tann done yu baax
   • Encrypter poids model
6. Lan moo waral ñuy faral di jëfandikoo 'pre-norm' ci transformatër yu xóot lool?
   • Dafay gëna ndaw model bi
   • Normalise balaa sublayer bu nekk dafay yombal stack yu xóot yi gëna dëgër ngir tàggat
   • Dafay dindi jumtukaayu bàyyi xel
   • Dafay yokk limu jàng ci saasi ñaari yoon
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