Workshop:PyHPC 2020: 9th Workshop on Python for High-Performance and Scientific Computing
Authors: Nicholson T. Collier, Jonathan Ozik, and Eric R. Tatara (Argonne National Laboratory (ANL))
Abstract: Distributed agent-based modeling (ABM) on high-performance computing resources provides the promise of capturing unprecedented details of large-scale complex systems. The specialized knowledge required for developing such ABMs, however, creates barriers to wider adoption and utilization. Here we present our experiences in developing an initial implementation of Repast4Py, a Python-based distributed ABM toolkit. We build on our experiences in developing ABM toolkits, including Repast for High-Performance Computing (Repast HPC), to identify the key elements of a useful distributed ABM toolkit. We leverage the Numba, NumPy, and PyTorch packages and the Python C-API to create a scalable modeling system that can exploit the largest HPC resources and emerging computing architectures.