Laaj ak tontu ci architecture

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

1. Ci biir benn bloku ResNet, lan mooy wareefu convolution 1x1 bu njëkk bi?
   • ngir wàññi limu chaine yi balaa convolution 3x3
   • Yokk dayo barab
   • Jëfandikoo xaaj bu mujj bi
   • Yokk bruit ngir yamale
2. Lan moo waral bottleneck bi tax reso yu xóot yi gëna yomb ci ordinatër?
   • Dafay dindi bépp aktivasioŋ
   • Dafay def ay operaasioŋ yu seer ci chaine yu néew
   • Dafay jëfandikoo kernel yu gëna mag fépp
   • Dafay teggi tasaare bi ci ginaaw
3. Ci biir autoencodeur, couche bi mooy njëkka wax:
   • Tanneef bi gëna xéewale
   • Format fichier nataal bi
   • Ñaata lañuy kompresse
   • Dayo lote bi
4. MobileNetV2 ban anam lay jëfandikoo ci xalaatu buteel bi?
   • Boutel de pooling pur
   • bottleneck buy dellusi
   • Amul benn cou bottle
   • Bottleneck residuel buñ soppi (yaatu suko defee mu wàññeeku)
5. naka la LoRA di jëfandikoo njàngalem bottleneck ci misaali làkk yu mag yi?
   • Soo dugalee ab bottleneck buñu mëna tàggat (sew) ci yeesali diisaay bi
   • Ni ñuy efaase ay couche attention
   • Soo yokke poid yépp ñaari yoon
   • Ni ñuy xaymaa 1 bit
6. Ban motif chaine mooy topp benn bloku bottleneck bu ResNet?
   • Yaatu, yaatu, yaatu
   • Wàññi, defar, defaraat (lu melni, 256 ba 64 ba 256)
   • sew, yaatu, sew
   • Random ci couche bu nekk
7. A ResNet block reduces 256 channels to 64 before its 3×3 convolution. Ignoring the 1×1 layers, what is the 3×3 layer’s approximate multiply-add cost relative to 256 channels?
   • About one-quarter
   • About one-half
   • About one-sixteenth
   • About four times as much
8. An autoencoder uses a narrow latent code. What constraint does this impose on reconstruction?
   • The encoder must preserve every input pixel exactly
   • The network can no longer use learned weights
   • The decoder must expand the latent code before encoding
   • The decoder must reconstruct the input from a compressed representation