multipAL is an active learning package that optimizes material compositions to maximize one or more properties. MultipAL works closely in tandem with the NIST JARVIS DFT database and VASP DFT software. It implements a random forest regressor to make predictions and quantifies the uncertainties using the ForestCI package.
First, make sure the dependencies are installed with the correct versions. You can use the environment.yml file in the extras/ directory to recreate a working conda environment. The Unix command is conda env create -f environment.yml.
Second, overwrite the vasp.py in your environment's .../lib/python3.8/site-packages/jarvis/tasks/vasp/ directory with the vasp.py file provided in the extras/ directory. This fixes a few issues with the default VASP settings.
You're now ready to use the multipal package! Please follow the tutorials in examples/ to learn more about what the package can do, and how to use it.
This code is made available under the MIT License.