Questionário sobre Leis de Dimensionamento para Redes Neurais

As respostas e explicações aparecem conforme você responde ao teste.

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. Quais são as três quantidades que as leis de escala originais relacionam com a perda?
   • Tamanho do lote, abandono e épocas
   • Parâmetros do modelo, tamanho do conjunto de dados e computação
   • Taxa de aprendizagem, redução de peso e profundidade
   • Contagem de GPU, memória e latência
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. O que representa o termo constante (irredutível) em uma fórmula de perda de lei de escala?
   • Um bug de software
   • A taxa de aprendizagem
   • O número de GPUs
   • A entropia intrínseca dos dados que nenhum modelo pode superar
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. Qual tendência está levando o campo além da simples escala de “quanto maior, melhor”?
   • GPUs se tornando gratuitas
   • Escassez de dados de alta qualidade, além de limites de energia e custos
   • Aparecendo texto ilimitado de alta qualidade
   • Curvas de perda se transformando em linhas verticais retas
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