Efficiently Train ASR Models that Memorize Less and Perform Better with Per-core Clipping
This work addresses privacy and efficiency issues in ASR training for speech recognition applications, though it is incremental as it builds on existing gradient clipping methods.
The paper tackles the problem of unintended memorization in automatic speech recognition (ASR) models by investigating per-core clipping (PCC), showing that it reduces memorization and improves performance with faster convergence and lower word error rates. It also introduces adaptive per-core clipping (APCC) to avoid hyperparameter tuning.
Gradient clipping plays a vital role in training large-scale automatic speech recognition (ASR) models. It is typically applied to minibatch gradients to prevent gradient explosion, and to the individual sample gradients to mitigate unintended memorization. This work systematically investigates the impact of a specific granularity of gradient clipping, namely per-core clip-ping (PCC), across training a wide range of ASR models. We empirically demonstrate that PCC can effectively mitigate unintended memorization in ASR models. Surprisingly, we find that PCC positively influences ASR performance metrics, leading to improved convergence rates and reduced word error rates. To avoid tuning the additional hyperparameter introduced by PCC, we further propose a novel variant, adaptive per-core clipping (APCC), for streamlined optimization. Our findings highlight the multifaceted benefits of PCC as a strategy for robust, privacy-forward ASR model training.