Isbaarada Sharing iyo Kediska Tababarka ee Dib loo Bilaaban karo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Waa maxay sababta isbaarooyinkii waaweynaa ay u kala qaybsameen jajab?
   • Si aad moodelka uga dhigto in uu si degdeg ah ugu tabobaro agabyada yaryar
   • Si aad u qariso miisaanka isticmaalayaasha
   • Sababtoo ah jeexjeexintu waxay hagaajinaysaa saxnaanta
   • Sababtoo ah hal isbaaro ayaa aad u weyn karta in hal fayl ama mishiin uu si hufan u maareeyo
2. Which extra state is typically needed for training resumption beyond model weights?
   • Kaliya saxnaanta ugu dambeysa
   • Optimizer and scheduler state, step progress, RNG state, and relevant data-loader progress
   • Only a model-weight shard index
   • A screenshot of the training dashboard
3. Waa maxay faa'iidada ugu weyn ee tababarka dib loo bilaabayo ee shaqooyinka dheer?
   • Waxay meesha ka saaraysaa baahida loo qabo go'an ansixinta
   • Waxay dammaanad qaadaysaa in moodelku aanu waligii dhaafin
   • Shilka ama ka-hor-istaaggu kuma qasbana in dib looga bilaabo eber
   • Waxa ay meesha ka saaraysaa wax-ka-beddelka
4. In a sharded distributed checkpoint, how can ranks avoid gathering the entire state on one worker?
   • By saving their assigned local tensor shards with checkpoint metadata
   • By each emailing a full model copy
   • By deleting optimizer state before every save
   • By disabling all state serialization
5. Doorkee ayuu ka ciyaaraa faylka tusaha ee isbaarooyinka jeexan?
   • Waxay kaydisaa erayga sirta ah ee isticmaalaha
   • Waxay ku khariidadaysaa magaca halbeeg kasta faylka shard ee ka kooban yahay
   • Waxay cadaadisaa sawirada ku jira kaydka xogta
   • Waxay jadwaleysaa heerka waxbarashada
6. Qaabkee ayaa inta badan loo isticmaalaa si badbaado leh, kaydinta khariidad xusuusta ku samaysay ee jeexjeexyada moodeelka?
   • MP3
   • CSV
   • badbaadiyeyaal
   • Kaydka ZIP-kaliya
7. A package contains safetensors weight shards and an index, but no optimizer state. What can it support by itself?
   • Loading model weights, without fully restoring optimizer training history
   • Guaranteed exact resumption of every interrupted run
   • Recovering the next batch position automatically
   • Reconstructing every random-number-generator state
8. A sharded checkpoint loader supports a new rank layout. What should the team verify before trusting recovery?
   • That fewer ranks guarantee identical floating-point results
   • That the weight index alone includes every application state
   • That no restart test is necessary
   • That the necessary saved state loads correctly and the resumed training step behaves as expected