DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining

Published: 21 Sept 2023, Last Modified: 20 Dec 2023NeurIPS 2023 spotlightEveryoneRevisionsBibTeX
Keywords: language models, pretraining, domain reweighting, data curation
TL;DR: We present an algorithm that reweights how much of each domain/data source is in a language modeling dataset (e.g. The Pile), resulting in faster LM training and improvements in perplexity and downstream accuracy.
Abstract: The mixture proportions of pretraining data domains (e.g., Wikipedia, books, web text) greatly affect language model (LM) performance. In this paper, we propose Domain Reweighting with Minimax Optimization (DoReMi), which first trains a small proxy model using group distributionally robust optimization (Group DRO) over domains to produce domain weights (mixture proportions) without knowledge of downstream tasks. We then resample a dataset with these domain weights and train a larger, full-sized model. In our experiments, we use DoReMi on a 280M-parameter proxy model to set the domain weights for training an 8B-parameter model (30x larger) more efficiently. On The Pile, DoReMi improves perplexity across all domains, even when it downweights a domain. DoReMi improves average few-shot downstream accuracy by 6.5% points over a baseline model trained using The Pile's default domain weights and reaches the baseline accuracy with 2.6x fewer training steps. On the GLaM dataset, DoReMi, which has no knowledge of downstream tasks, even matches the performance of using domain weights tuned on downstream tasks.
Submission Number: 2581