Where to Begin: Efficient Pretraining via Subnetwork Selection and Distillation
This work addresses the challenge of resource-efficient SLM development for practitioners, though it is incremental as it builds on existing methods like distillation and evolutionary search.
The paper tackles the problem of inefficient pretraining for small language models (SLMs) by introducing a framework that combines subnetwork selection and distillation, resulting in a model that matches the validation perplexity of a comparable SLM while requiring 9.2x fewer pretraining tokens.
Small Language models (SLMs) offer an efficient and accessible alternative to Large Language Models (LLMs), delivering strong performance while using far fewer resources. We introduce a simple and effective framework for pretraining SLMs that brings together three complementary ideas. First, we identify structurally sparse sub-network initializations that consistently outperform randomly initialized models of similar size under the same compute budget. Second, we use evolutionary search to automatically discover high-quality sub-network initializations, providing better starting points for pretraining. Third, we apply knowledge distillation from larger teacher models to speed up training and improve generalization. Together, these components make SLM pretraining substantially more efficient: our best model, discovered using evolutionary search and initialized with LLM weights, matches the validation perplexity of a comparable Pythia SLM while requiring 9.2x fewer pretraining tokens. We release all code and models at https://github.com/whittle-org/whittle/, offering a practical and reproducible path toward cost-efficient small language model development at scale.