Kediska Goynta Model

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Maxey ka saartaa qaabka manjooyinku shabakad tababaran?
   • Ka saarida miisaanka ama qaababka wax yar ka geysanaya wax soo saarka
   • Ku darida xaddi badan si loo hagaajiyo saxnaanta
   • Ku tababarida moodal wakhtiyo badan
   • Kordhinta xallinta gelinta
2. Maxaa ka sooca manjyada habaysan iyo manjrada aan habaysan?
   • Goynta habaysan waxay ku shaqaysaa sawirada oo kaliya
   • Manjooyin habaysan ayaa meesha ka saaraya unugyo dhan sida neurons ama madax; aan qaabaysanayn waxay ka saartaa miisaanka shakhsi ahaaneed
   • Goynta aan qaabaysan waxay u baahan tahay dib u tabobar xaga eber ah
   • Ma jiro farqi la taaban karo oo u dhexeeya
3. Waa maxay sababta manjooyin aan habaysanayn ay inta badan ugu guuldareystaan inay dedejiyaan qaabka qalabka caadiga ah?
   • Waxay tirtirtaa lakabyo badan
   • Waxay had iyo jeer hoos u dhigtaa saxnaanta eber
   • Matrisyada miisaanka yar ayaa u baahan qalab gaar ah ama maktabado si looga faa'iidaysto eber
   • Waxay kordhisaa tirinta cabbirka
4. Waa maxay habka caadiga ah ee manjooyinku?
   • Wax walba hal mar goo goo, dabadeed geyn
   • Marka hore qiyaas, ka dibna dib ha u tababarin
   • Si aan kala sooc lahayn u tirtir lakabyada ilaa saxnaanta hoos u dhacdo
   • Tabobar, gooyo cabbirada ugu muhimsan, si fiican u hagaaji si aad u soo kabsato, oo ku celiso
5. Muxuu sheeganayaa mala-awaalka tigidhka bakhtiyaanasiibka?
   • Shabakadaha cufan waxay ka kooban yihiin shabag-hoosaadyo aad u yar oo la jaan qaadi kara qaabka buuxa marka laga soo bilaabo bilawga saxda ah
   • Goynta had iyo jeer waxay hagaajisaa saxnaanta
   • Manjooyin aan toos ahayn ayaa had iyo jeer ugu fiican
   • Moodooyinka waaweyn lama cadaadin karo
6. Halbeegkee fudud ayaa caadi ahaan loo isticmaalaa in lagu go'aamiyo miisaanka la jarayo?
   • Habka alifbeetada ee magacyada lakabka
   • Baaxadda buuxda ee miisaanka
   • Midabka khariidadaha hawlgelinta
   • Tirada tusaalooyinka tababarka
7. When is pruning most likely to reduce inference latency on ordinary dense hardware?
   • When individual zeros are added but the dense kernel still computes them
   • When the model is renamed without changing its graph
   • When whole channels or other structures are removed to produce a smaller dense model
   • When a sparse checkpoint is stored in a larger file
8. What should a team measure after pruning for a deployment target?
   • Only the number of zeros in a checkpoint
   • Only training accuracy before pruning
   • Whether the file extension changed
   • Task quality and latency or memory on the intended hardware