Campus Units

Electrical and Computer Engineering, Mechanical Engineering, Plant Sciences Institute

Document Type

Article

Publication Version

Submitted Manuscript

Publication Date

2021

Journal or Book Title

arXiv

Abstract

We consider the distributed training of large-scale neural networks that serve as PDE solvers producing full field outputs. We specifically consider neural solvers for the generalized 3D Poisson equation over megavoxel domains. A scalable framework is presented that integrates two distinct advances. First, we accelerate training a large model via a method analogous to the multigrid technique used in numerical linear algebra. Here, the network is trained using a hierarchy of increasing resolution inputs in sequence, analogous to the 'V', 'W', 'F', and 'Half-V' cycles used in multigrid approaches. In conjunction with the multi-grid approach, we implement a distributed deep learning framework which significantly reduces the time to solve. We show the scalability of this approach on both GPU (Azure VMs on Cloud) and CPU clusters (PSC Bridges2). This approach is deployed to train a generalized 3D Poisson solver that scales well to predict output full-field solutions up to the resolution of 512x512x512 for a high dimensional family of inputs.

Comments

This is a pre-print of the article Balu, Aditya, Sergio Botelho, Biswajit Khara, Vinay Rao, Chinmay Hegde, Soumik Sarkar, Santi Adavani, Adarsh Krishnamurthy, and Baskar Ganapathysubramanian. "Distributed Multigrid Neural Solvers on Megavoxel Domains." arXiv preprint arXiv:2104.14538 (2021). Posted with permission.

Copyright Owner

The Author(s)

Language

en

File Format

application/pdf

Published Version

Share

COinS