Search papers, labs, and topics across Lattice.
The paper introduces Olmix, a framework designed to address challenges in data mixing for language model training, specifically focusing on understanding the configuration space of mixing methods and efficiently adapting to evolving domain sets. Through an empirical study, the authors identify key design choices for effective mixing methods and propose "mixture reuse," a technique that leverages past mixture ratios to efficiently recompute mixtures after domain set updates. Experiments show that mixture reuse achieves comparable performance to full recomputation with significantly reduced compute (74% less) and outperforms training without mixing by 11.6% on downstream tasks.
By reusing existing data mixture ratios and only recomputing for affected domains, Olmix slashes compute costs by 74% without sacrificing downstream task performance during iterative LM development.
Data mixing -- determining the ratios of data from different domains -- is a first-order concern for training language models (LMs). While existing mixing methods show promise, they fall short when applied during real-world LM development. We present Olmix, a framework that addresses two such challenges. First, the configuration space for developing a mixing method is not well understood -- design choices across existing methods lack justification or consensus and overlook practical issues like data constraints. We conduct a comprehensive empirical study of this space, identifying which design choices lead to a strong mixing method. Second, in practice, the domain set evolves throughout LM development as datasets are added, removed, partitioned, and revised -- a problem setting largely unaddressed by existing works, which assume fixed domains. We study how to efficiently recompute the mixture after the domain set is updated, leveraging information from past mixtures. We introduce mixture reuse, a mechanism that reuses existing ratios and recomputes ratios only for domains affected by the update. Over a sequence of five domain-set updates mirroring real-world LM development, mixture reuse matches the performance of fully recomputing the mix after each update with 74% less compute and improves over training without mixing by 11.6% on downstream tasks.
AI2
Stanford HAI