Sharding ak quiz de formation buñu mëna delloosi

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

1. Lu tax ñuy xaaj checkpoint yu mag yi ay pàcc?
   • Ngir defar saxaaru model bi gëna gaaw ci ay dugal yu ndaw
   • Ngir nëbb jëfandikukat yi xeetu diisaay bi
   • Ndax sharding dafay gëna rafetal
   • Ndax benn checkpoint mën na nekk lu yaatu lool ba benn fichier wala masin mënu ko jëfandikoo bu baax
2. Which extra state is typically needed for training resumption beyond model weights?
   • Du lenn ludul poñ yi mujj
   • 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. Lan mooy njariñ li gëna mag ci tàggat yaram ngir liggéey yu yàgg?
   • Dina dindi soxla ensemble validation
   • Dafay garanti ni model bi du musa ëpp loxo
   • Crash wala preemption du forse tàmbaliwaat ci noonu
   • Dafay dindi optimisatër bi
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. Ban cër la fichier index di def ci checkpoint yuñ xaaj?
   • Dafay denc baatu-jàll jëfandikukat bi
   • Dafay méngale tur parametre bu nekk ak fichier shard bi ko def
   • Dafay kompresse nataal yi ci dataset bi
   • Dafay waajal ni ñuy jàngee
6. Ban format lañuy faral di jëfandikoo ngir yobbu modelu shards yi ci anam wu wóor?
   • MP3
   • CSV
   • tenseur de sécurité
   • Archive ZIP kese
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