Colmena: Scalable Machine-Learning-Based Steering of Ensemble Simulations for High Performance Computing

TL;DR

Colmena is an open-source Python framework that enables scalable, machine-learning-guided steering of ensemble simulations on HPC systems, significantly accelerating scientific discovery processes.

Contribution

It introduces a flexible, scalable framework that integrates ML models with ensemble simulations, simplifying deployment and coordination on high-performance computing resources.

Findings

Scales to 65,536 CPUs on HPC systems.

Accelerates electrolyte discovery by 100x.

Demonstrates effective ML-guided ensemble steering.

Abstract

Scientific applications that involve simulation ensembles can be accelerated greatly by using experiment design methods to select the best simulations to perform. Methods that use machine learning (ML) to create proxy models of simulations show particular promise for guiding ensembles but are challenging to deploy because of the need to coordinate dynamic mixes of simulation and learning tasks. We present Colmena, an open-source Python framework that allows users to steer campaigns by providing just the implementations of individual tasks plus the logic used to choose which tasks to execute when. Colmena handles task dispatch, results collation, ML model invocation, and ML model (re)training, using Parsl to execute tasks on HPC systems. We describe the design of Colmena and illustrate its capabilities by applying it to electrolyte design, where it both scales to 65536 CPUs and accelerates the discovery rate for high-performance molecules by a factor of 100 over unguided searches.