AdaLRS: Loss-Guided Adaptive Learning Rate Search for Efficient Foundation Model Pretraining
This work addresses the need for efficient hyperparameter tuning in large-scale AI training, offering a plug-in solution that generalizes across models and scenarios, though it is incremental as it builds on existing learning rate optimization methods.
The authors tackled the problem of inefficient learning rate tuning in foundation model pretraining by proposing AdaLRS, an adaptive algorithm that optimizes loss descent velocities online, which improved model performance with minimal extra computation.
Learning rate is widely regarded as crucial for effective foundation model pretraining. Recent research explores and demonstrates the transferability of learning rate configurations across varying model and dataset sizes, etc. Nevertheless, these approaches are constrained to specific training scenarios and typically necessitate extensive hyperparameter tuning on proxy models. In this work, we propose \textbf{AdaLRS}, a plug-in-and-play adaptive learning rate search algorithm that conducts online optimal learning rate search via optimizing loss descent velocities. We provide experiment results to show that the optimization of training loss and loss descent velocity in foundation model pretraining are both convex and share the same optimal learning rate. Relying solely on training loss dynamics, AdaLRS involves few extra computations to guide the search process, and its convergence is guaranteed via theoretical analysis. Experiments on both LLM and VLM pretraining show that AdaLRS adjusts suboptimal learning rates to the neighborhood of optimum with marked efficiency and effectiveness, with model performance improved accordingly. We also show the robust generalizability of AdaLRS across varying training scenarios, such as different model sizes, training paradigms, and base learning rate scheduler choices.