Bridging the machine learning deployment gap in gravitational wave physics

Deep learning algorithms have achieved state-of-the-art performance across a wide array of settings in recent years, and with this success has come an abundance of papers applying these methods to gravitational wave physics problems. Despite this, there are still comparatively few deep learning models planned for online deployment during the O4 data collection run of the LIGO-Virgo-KAGRA collaboration. One reason for this disparity is a lack of standardized software tooling adequate to quickly implement and iterate upon novel ideas and validate them on sufficiently large volumes of data to achieve confidence in production performance. In this work, we describe a suite of libraries intended to bridge this gap, ml4gw and hermes, and outline how their use in two specific applications has led to increases in efficiency and model robustness.