How to decay your learning rate
This work addresses the need for simpler and more robust learning rate schedules for deep learning practitioners, though it appears incremental as it builds on existing scheduling methods.
The paper tackled the problem of complex learning rate schedules in deep learning by proposing ABEL, an automatic scheduler that decays the learning rate based on weight norm bounces, matching tuned schedules and showing robustness across vision, NLP, and RL tasks.
Complex learning rate schedules have become an integral part of deep learning. We find empirically that common fine-tuned schedules decay the learning rate after the weight norm bounces. This leads to the proposal of ABEL: an automatic scheduler which decays the learning rate by keeping track of the weight norm. ABEL's performance matches that of tuned schedules and is more robust with respect to its parameters. Through extensive experiments in vision, NLP, and RL, we show that if the weight norm does not bounce, we can simplify schedules even further with no loss in performance. In such cases, a complex schedule has similar performance to a constant learning rate with a decay at the end of training.