How to Set the Batch Size for Large-Scale Pre-training?
This work addresses a practical issue for researchers and engineers in large-scale AI training, offering an incremental improvement by updating batch size theory for modern schedulers.
The paper tackles the problem of setting batch size for large-scale pre-training under the Warmup-Stable-Decay scheduler, deriving a revised theoretical framework that identifies minimum and optimal batch sizes, and proposing a dynamic scheduler that improves training efficiency and model quality.
The concept of Critical Batch Size, as pioneered by OpenAI, has long served as a foundational principle for large-scale pre-training. However, with the paradigm shift towards the Warmup-Stable-Decay (WSD) learning rate scheduler, we observe that the original theoretical framework and its underlying mechanisms fail to align with new pre-training dynamics. To bridge this gap between theory and practice, this paper derives a revised E(S) relationship tailored for WSD scheduler, characterizing the trade-off between training data consumption E and steps S during pre-training. Our theoretical analysis reveals two fundamental properties of WSD-based pre-training: 1) B_min, the minimum batch size threshold required to achieve a target loss, and 2) B_opt, the optimal batch size that maximizes data efficiency by minimizing total tokens. Building upon these properties, we propose a dynamic Batch Size Scheduler. Extensive experiments demonstrate that our revised formula precisely captures the dynamics of large-scale pre-training, and the resulting scheduling strategy significantly enhances both training efficiency and final model quality.