Improving Training Stability for Multitask Ranking Models in Recommender Systems
This addresses a critical problem for developers of large-scale recommender systems by providing practical solutions to prevent loss divergence and resource waste, though it is incremental in nature.
The paper tackles training instability issues in multitask ranking models for recommender systems, proposing a new algorithm that significantly improves stability without compromising convergence on YouTube production data.
Recommender systems play an important role in many content platforms. While most recommendation research is dedicated to designing better models to improve user experience, we found that research on stabilizing the training for such models is severely under-explored. As recommendation models become larger and more sophisticated, they are more susceptible to training instability issues, i.e., loss divergence, which can make the model unusable, waste significant resources and block model developments. In this paper, we share our findings and best practices we learned for improving the training stability of a real-world multitask ranking model for YouTube recommendations. We show some properties of the model that lead to unstable training and conjecture on the causes. Furthermore, based on our observations of training dynamics near the point of training instability, we hypothesize why existing solutions would fail, and propose a new algorithm to mitigate the limitations of existing solutions. Our experiments on YouTube production dataset show the proposed algorithm can significantly improve training stability while not compromising convergence, comparing with several commonly used baseline methods.