High Throughput Training of Deep Surrogates from Large Ensemble Runs

Abstract

Recent years have seen a surge in deep learning approaches to accelerate numerical solvers, which provide faithful but computationally intensive simulations of the physical world. These deep surrogates are generally trained in a supervised manner from limited amounts of data slowly generated by the same solver they intend to accelerate. We propose an open-source framework that enables the online training of these models from a large ensemble run of simulations. It leverages multiple levels of parallelism to generate rich datasets. The framework avoids I/O bottlenecks and storage issues by directly streaming the generated data. A training reservoir mitigates the inherent bias of streaming while maximizing GPU throughput. Experiment on training a fully connected network as a surrogate for the heat equation shows the proposed approach enables training on 8TB of data in 2 hours with an accuracy improved by 47% and a batch throughput multiplied by 13 compared to a traditional offline procedure.

The International Conference for High Performance Computing, Networking, Storage, and Analysis, Nov 2023, Denver, CO, United States

References in corpus (7)

• Physics-Constrained Deep Learning for High-dimensional Surrogate Modeling and Uncertainty Quantification without Labeled Data
• Massively Parallel Methods for Deep Reinforcement Learning
• Parsl: Pervasive Parallel Programming in Python
• PyTorch Distributed: Experiences on Accelerating Data Parallel Training
• Revisiting Fundamentals of Experience Replay
• RADICAL-Cybertools: Middleware Building Blocks for Scalable Science
• Collect & Infer -- a fresh look at data-efficient Reinforcement Learning