Done yuñ xaaj lépp

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

1. Lan la FSDP xaaj ci GPU yi nga xamni parallelism done buñ miin deful?
   • Parametre yi, degrade yi ak tolluwaayu optimisatër bi
   • du lenn ludul ay lots yuñ dugal
   • Xeetu njàng mi kese
   • Dosie yiñ dajale ci disk bi
2. Ban operation collective la FSDP di jëfandikoo ngir defaraat poid yu mat yi ci benn couche balaa ñu koy calculer?
   • Diffusion kese
   • Ñépp dajaloo
   • Tasaaroo
   • Barrière
3. Lan moo waral FSDP di dindi poids yiñ dajale ci saasi ginaaw biñu xaymaa benn couche?
   • Ngir yàq gradient yiñ jëfandikoowul
   • ngir forse ab bouclage de formation bu yeex
   • Ngir tëye emprent memory budul jeex ci benn shard ci aparey bu nekk
   • Ndax PyTorch daf leen di efaase ci anam wu mujjee
4. Ban xeetu mémoire-optimisation moo gënoon inspiré FSDP?
   • Google
   • RDD yu Apache
   • CUDA yu NVIDIA
   • Zero bu Microsoft (Optimisatëru Redondance Zéro)
5. naka la FSDP di nëbbe lu bari ci latency reso bi ci dajale ay poid?
   • Sooy jaxasoo jokkoo ak xayma, nga njëkka jël parametru layer bi ci topp
   • Soo desee paas biy dellu ginaaw
   • Ni ñuy kompressee poid yi ci benn bit
   • Taggat ci CPU kese
6. Ban liggéey bu mbooloo mooy séddale daggitu gradient yuñ boole mu delloo ci seeni GPU ci paas bi ci ginaaw?
   • Ñaari yoon
   • Wàññi tasaaroo
   • Kopi benn-GPU
   • Dajale rek
7. Lan moo waral staadu sharding bu mat sëkk mën wàññi mémoire bu GPU bu nekk suñu ko méngale ak parallelism done yuñ toppandoo?
   • Dafay dindi bépp parametru model ci diiru tàggat yaram
   • Dafay soppi staadu optimisatër bi ci laaj reso bi
   • Rank bu nekk dafay denc benn xeetu etaa ci barabu benn kopi bu mat
   • Dafay def bépp aktivasioŋ ci CPU bi ba fàww
8. Luy kompromis sistem bu mag ci FSDP?
   • Memoire bu woyof te amul jokkoo ci benn poñ
   • Parametre yu bari yuñ toppandoo ngir am misaali done yu néew
   • Wecci tàggat yaram ak gis-gis
   • Jëfandikoo mémoire bu néew ci aparey bu nekk ngir weccoo jokkoo bu mbooloo