Quiz paralelismu done

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

1. Ci parallelism done buñ miin, lan la GPU bu nekk di tëye?
   • Kopi bu mat bu model bi, di liggéey ci beneen wàll ci lote bi
   • Benn wàll bu wuute ci diisaayi model bi
   • Optimiser bi kese, ak poid yiñ denc ci digg bi
   • Benn xeetu model buñ komprime ak paramet yu néew
2. Ban jëfu jokkoo mooy tëye replica yu model yi ci jéego bu nekk?
   • Broadcast, mooy kopie benn done GPU ci yeneen yi
   • Lépp-wàññi, mooy xayma gradient yi ci GPU yépp
   • Scatter, mooy xaaj done yi ci diggante GPU yi
   • Amul benn jokkoo buñu soxla ci digganté GPU yi
3. What memory limit remains in plain data parallelism?
   • The full model is sharded automatically across the GPUs
   • Only the input batch uses device memory
   • Each GPU keeps model parameters and optimizer state replicated, limiting the model size that fits
   • Each worker stores different model layers by default
4. Which factors does NCCL use when selecting collective algorithms?
   • Only the random seed and learning rate
   • Only the total number of model parameters
   • The optimizer type and validation score
   • Message size and topology
5. naka la ko PyTorch di nëbbee ay done yuñ séddale parallel?
   • Sooy jaxasoo gradient bi nga wàññi lépp ak paas biy wéy
   • Soo sànnee senkronizaasioŋ gradient ci jéego yu wuute
   • Sooy boole gradient yépp ñu nekk benn nimero
   • Soo defee paas bi jëm kanam ci CPU bi
6. With equal local batch size B on each of W data-parallel workers, what is the global batch for one synchronized step?
   • B divided by W
   • B multiplied by W, before any gradient-accumulation factor
   • Always exactly W samples
   • The model parameter count divided by B
7. Why may adding more data-parallel GPUs fail to increase throughput proportionally?
   • All-reduce makes model parameters disappear
   • The model is copied only to one worker
   • Gradient communication, data loading, and synchronization add overhead
   • More workers always reduce the global batch to zero
8. What does gradient accumulation let a data-parallel training loop do?
   • Shard parameters automatically across all workers
   • Skip forward and backward computation
   • Replace the model with an all-reduce operation
   • Combine gradients from multiple microbatches before an optimizer update