Scaling Laws for Neural Language Models

TL;DR

This paper investigates empirical scaling laws in neural language models, revealing power-law relationships between performance, model size, dataset size, and compute, guiding optimal training strategies for large models.

Contribution

It establishes simple, quantitative scaling laws for language model performance and compute efficiency, providing practical guidelines for training large models effectively.

Findings

Performance scales as a power-law with model size, dataset size, and compute.

Architectural details like width and depth have minimal impact within studied ranges.

Optimal training involves large models with modest data, stopping before convergence.

Abstract

We study empirical scaling laws for language model performance on the cross-entropy loss. The loss scales as a power-law with model size, dataset size, and the amount of compute used for training, with some trends spanning more than seven orders of magnitude. Other architectural details such as network width or depth have minimal effects within a wide range. Simple equations govern the dependence of overfitting on model/dataset size and the dependence of training speed on model size. These relationships allow us to determine the optimal allocation of a fixed compute budget. Larger models are significantly more sample-efficient, such that optimally compute-efficient training involves training very large models on a relatively modest amount of data and stopping significantly before convergence.