Loss-to-Loss Prediction: Scaling Laws for All Datasets
This work addresses the challenge of scaling law predictions for varied data distributions, which is incremental as it extends existing scaling law methodologies to cross-dataset scenarios.
The paper tackles the problem of predicting loss across different data distributions, deriving a strategy to predict one loss from another for pre-training datasets and downstream tasks, with predictions extrapolating well even at 20x the largest FLOP budget used to fit the curves.
While scaling laws provide a reliable methodology for predicting train loss across compute scales for a single data distribution, less is known about how these predictions should change as we change the distribution. In this paper, we derive a strategy for predicting one loss from another and apply it to predict across different pre-training datasets and from pre-training data to downstream task data. Our predictions extrapolate well even at 20x the largest FLOP budget used to fit the curves. More precisely, we find that there are simple shifted power law relationships between (1) the train losses of two models trained on two separate datasets when the models are paired by training compute (train-to-train), (2) the train loss and the test loss on any downstream distribution for a single model (train-to-test), and (3) the test losses of two models trained on two separate train datasets (test-to-test). The results hold up for pre-training datasets that differ substantially (some are entirely code and others have no code at all) and across a variety of downstream tasks. Finally, we find that in some settings these shifted power law relationships can yield more accurate predictions than extrapolating single-dataset scaling laws.