quiz bu lëkkale ak tegtal · Diggante Niveau

Quiz de formation de précision mixte

Dafay muur tolluwaayu FP16 ak eskaalu perte, bfloat16, poid master, autocast, dajale Tensor Core, ak natt benefiis yi aju ci aparey bi.

Yooni njiit yu jëm ciTaggat yaram bu jaxaso
Laaj 1 de 8

In an FP16 training recipe that keeps an FP32 master weight, why retain it?

Xoolaatal laaj bu nekk8 laaj

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

  1. In an FP16 training recipe that keeps an FP32 master weight, why retain it?
    • Ngir mëna am njub ci yeesali diisaay yu ndaw yi 16-bit mëna rond
    • Ngir wàññi fichier model bi ci disk bi
    • Ngir may tàggat yaram te du am GPU
    • ngir enkripte parametru model bi
  2. Ban jafe-jafe la perte scaling di saafara ci FP16 formation?
    • Modèle yu gëna méngoo ak done yi
    • Degrade yiy wàcci ba zero ndax FP16 dafa tuuti
    • GPU yi amul bayaal ci disk bi
    • Done yuy yab ci disk bi yeex
  3. Lan mooy njariñu bfloat16 ci kaw FP16?
    • Dafay jëfandikoo bit yu néew
    • Dafay denc mbind ci barabu nimero
    • Amna xeetu exponent bu yaatu ni FP32, wàññi soxla scaling perte
    • Ci CPU yi kese lay dox
  4. For many supported Tensor Core matrix operations, how are partial sums commonly accumulated?
    • In integer arithmetic with no rounding
    • In text before converting to FP16
    • In FP16 only, which prevents every numerical error
    • In FP32, which reduces rounding error compared with low-precision accumulation
  5. What is a common benefit of using 16-bit values for supported training operations?
    • They can reduce tensor memory and improve throughput, with gains depending on hardware and workload
    • They guarantee higher model accuracy
    • They remove the need to compute gradients
    • They always double end-to-end training speed
  6. Which current PyTorch APIs are commonly used for automatic mixed-precision training?
    • torch.cuda.amp only, which is the recommended new namespace
    • torch.amp.autocast and, for FP16 gradient scaling, torch.amp.GradScaler
    • torch.load and torch.save
    • torch.nn.DataParallel and torch.manual_seed
  7. Why does bfloat16 often need less loss scaling than FP16?
    • It has more mantissa bits than FP32
    • It stores gradients as integers
    • It has an exponent range like FP32, so very small gradients are less likely to underflow
    • It never rounds values
  8. What does autocast do in automatic mixed precision?
    • Converts every tensor and operation to FP16 unconditionally
    • Replaces the optimizer with a different algorithm
    • Guarantees identical bitwise results to FP32
    • Selects operation dtypes so supported operations can use lower precision while others retain safer precision