BOA

This repository contains the code to reproduce the results in the ICLR paper A Function-Centric Graph Neural Network Approach For Predicting Electron Densities.

Installation

For environment management, we use UV:

uv venv --python=3.11
source .venv/bin/activate
uv pip install -r requirements.txt --index-strategy unsafe-best-match
uv pip install -e scdp/ -e sciai-dft/ -e .

Environment Variables

Fill .env with following lines to set environment variables:

BOA_DATA=""
BOA_MODELS=""

Data setup

Detailed instructions for MD and QM9 data setup will follow soon.

Training

python boa/train.py experiment=<your_experiment>

Testing

python boa/test.py eval=<your_eval>