quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz ci wàllu jëfandikoo mémoire GPU ak fragmentation

Saytu GPU biñ jox ak mémoire biñ denc, jëfandikoowaat allocateur bi, ak saytu fragmentation, boole ci àppu empty_cache ak segment yuñ mëna yokk.

Yooni njiit yu jëm ciGPU jëfandikoo mémoire ak xaaj
Laaj 1 de 8

Luy xaajalug mémoire GPU?

Xoolaatal laaj bu nekk8 laaj

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

  1. Luy xaajalug mémoire GPU?
    • Su GPU bi xaajaloo
    • Su model bi amee ay diisaay yu bari
    • Su ñaari GPU déggoowul ci benn resultaa
    • Sudee memory bi amul xaalis dañu ko xaaj ay pàcc yu tasaaroo yu tuuti lool ngir mëna satisfaat laaj bu yaatu bu toppalante
  2. Lan moo waral kaadar yu melni PyTorch di jëfandikoo ab tànneefu mémoire cache?
    • Ngir defar model bi mu gëna dëggu
    • Ngir moytu joxe dawalkat bu yeex / woote yu amul fayda ci jëfandikoowaat bloku mémoire yuñ boole
    • ngir encrypter poid yi
    • Wàññi limu GPU yiñ soxla ba benn
  3. A CUDA allocation fails despite apparently unused reserved memory. What should a developer investigate?
    • Whether fragmentation is proven by the error alone
    • Whether all live tensors have already been freed
    • Whether inactive split blocks or other allocations explain the failure
    • Whether changing the learning rate releases GPU memory
  4. Which PyTorch allocator option can be tested for changing-size workloads that leave unused memory slivers?
    • Expandable segments in the allocator configuration
    • Increasing every tensor dimension
    • Disabling the GPU cooling system
    • Increasing the optimizer learning rate
  5. Lan la PagedAttention bu vLLM mëna def ngir wàññi yàqu-yàqu ci mémoire bi ñuy jël dogal?
    • Done yiñ tàggat ci model bi
    • cache valeur bu am solo (KV), di jëfandikoo xët yu am dayo bu takku
    • Gaawaayu montu GPU bi
    • Dayo font log yi
  6. Ban pexe mooy wàññi mémoire biy aktive ngir moytu njuumti yi ci bitti mémoire bi?
    • Yokk yeneen diisaay
    • Jëfandikoo tàngoor bu gëna gudd
    • Saytu gradient (jëfandikoowaat)
    • Denc yeneen poñ de saytu ci disk bi
  7. A live CUDA tensor still has Python references. Will empty_cache release that tensor’s allocation?
    • No; it releases unused cached memory rather than live tensor storage
    • Yes; it deletes every referenced tensor
    • Yes; it moves the tensor automatically to NVMe
    • Only if the tensor contains zero values
  8. PyTorch reports reserved memory greater than allocated memory. What distinction explains the gap?
    • Reserved memory measures only CPU RAM
    • Allocated memory includes every process on the GPU
    • The difference directly proves a model accuracy error
    • The allocator can retain cached blocks beyond memory currently used by tensors