Ikibazo gihuza ibibazo · Hagati Urwego

Gupima Amategeko Kubibazo bya Neural Network

Gerageza uburyo amategeko yo gupima afitanye isano no gutakaza ururimi-icyitegererezo kubipimo, amakuru, hamwe na compte, hamwe nibiteganijwe.

Inzira zijyanye no kuyoboraGupima Amategeko Kumurongo wa Neural
Ikibazo 1 Bya 8

What pattern did Kaplan et al. report for language-model loss as model size, data, or compute increased?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. What pattern did Kaplan et al. report for language-model loss as model size, data, or compute increased?
    • Approximate power-law relationships over the tested ranges
    • A sudden guaranteed drop to zero
    • A fixed increase in loss for every larger model
    • Random oscillation with no measurable trend
  2. Ni ubuhe buryo butatu amategeko yambere yo gupima afitanye isano nigihombo?
    • Ingano yubunini, abata ishuri, nibihe
    • Icyitegererezo cyibipimo, ingano yamakuru, na compte
    • Igipimo cyo kwiga, kubora ibiro, hamwe nubujyakuzimu
    • Kubara GPU, kwibuka, no gutinda
  3. How can a scaling-law fit help when planning a larger training run?
    • It guarantees the final benchmark score
    • It eliminates the need for validation
    • It can inform an extrapolation from smaller experiments, subject to the tested setup
    • It predicts results exactly for any dataset or architecture
  4. Ijambo rihoraho (ridashoboka) muburyo bwo gutakaza amategeko agereranya iki?
    • Ikosa rya software
    • Igipimo cyo kwiga
    • Umubare wa GPU
    • Entropic entropy yamakuru nta moderi ishobora gutsinda
  5. Why can absolute loss reductions become smaller as compute keeps increasing under a power-law fit?
    • The reducible loss term shrinks while the loss approaches its floor
    • The data is removed after each compute increase
    • The model stops using parameters
    • Power laws require loss to increase over time
  6. Ni ubuhe buryo busunika umurima kurenza 'binini ni byiza' gupima?
    • GPU iba ubuntu
    • Ubuke bwamakuru yo murwego rwohejuru hiyongereyeho ingufu nigipimo cyibiciro
    • Imipaka itagira imipaka yo mu rwego rwo hejuru igaragara
    • Gutakaza umurongo uhinduka umurongo uhagaze
  7. What did the original Kaplan study say about architecture details such as width and depth within its tested range?
    • They always dominated model size, data, and compute
    • They made loss independent of data
    • They had relatively small effects compared with the studied scaling variables
    • They guaranteed the same performance for all architectures
  8. What compute-allocation tendency did the original Kaplan paper report?
    • Every model should be trained to zero loss
    • Small models always use less compute at equal loss
    • Training data has no effect on compute allocation
    • Large models could be more sample-efficient and compute-efficient training could stop before convergence