Analytic nuclear gradients for machine-learned orbital-free DFT

This repository provides data-generation, training, density-optimization, and geometry-optimization workflows for developing and evaluating machine-learned orbital-free density functional theory (ML-OFDFT).

Quickstart

Clone the repository:

git clone git@github.com:sciai-lab/structures25-nuclear-gradients.git
cd structures25-nuclear-gradients

Install the GPU-enabled environment with uv:

uv sync --extra cuda --extra dev

For a CPU-only environment, replace cuda with cpu. Add the docs extra to build the documentation.

Set the data and model directories before running the workflows:

export DFT_DATA="/path/to/data"
export DFT_MODELS="/path/to/models"

Trained model checkpoints are available on Hugging Face.

Usage

The main command-line entry points are:

• mldft_ks: generate Kohn-Sham DFT reference data
• mldft_labelgen: generate training labels
• mldft_train: train ML-OFDFT models
• mldft_denop: optimize electron densities
• mldft_geomopt: optimize molecular geometries

Activate the environment and run an entry point, for example:

source .venv/bin/activate
mldft_geomopt

The workflows are configured with Hydra through the configuration files in configs.

Benchmark inputs, numerical results, and reproduction commands are provided in evaluation.

Documentation

Install the docs extra and build the documentation locally:

uv sync --extra cuda --extra docs
make docs

The generated documentation is available at docs/build/html/index.html.

Third-party licenses

This code adapts code from the following third party libraries:

• pytorch_geometric
• equiformer_v2

These are distributed under the

MIT License

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.