Electrolyte Design Toolkit
Design toolkit for molecular electrolytes. Includes tools needed to launch high-throughput workflows for assessing properties of molecular using QCArchive, create machine learning models using a variety of methods, and using the machine learning models to steer simulation campaigns with Colmena.
Our goal is to create a software that is easy to deploy at different HPC centers and retool for new molecular applications. The code itself is very experimental, so expect the APIs to change frequently at this stage but please do complain if documentation is lacking.
Installation
The environment needed to run the tools is defined in environment.yml.
Create the environment with Miniconda by calling:
conda env create --file environment.yml --forceThe environment should be run on a Linux system. Detailed installation instructions will be provided after merging this project with Colmena.
Running Workflows
The colmena folder contains the run scripts for different design applications, each housed in different directories.
Each directory contains the necessary run files and a README.
Specific applications include:
ip-single-fidelity: The application studied in our SC21 submissionatomization-energy-no-retrain: The test application used to evaluate scaling of inference tasks