/pytorch-dimenet

Primary LanguagePythonMIT LicenseMIT

DimeNet PyTorch

This repository is DimeNet PyTorch version which is ported from the original TensorFlow repo.

Getting Started

# Download processed QM9 data.
./download-data.sh

# Train model to predict mu.
cd src
python run_train.py params/001.yaml

Results

Epochs 800 is used.

Target Unit MAE
mu Debye 0.0285
U0 meV TODO

Differences from original

  • Use RAdam as an optimizer.
  • Use Mish as an activation.
  • The number of layers and n_hidden in OutputBlock might be different.
  • The loss func might be different.
  • Data splitting might be different.

Cite

@inproceedings{klicpera_dimenet_2020,
  title = {Directional Message Passing for Molecular Graphs},
  author = {Klicpera, Johannes and Gro{\ss}, Janek and G{\"u}nnemann, Stephan},
  booktitle={International Conference on Learning Representations (ICLR)},
  year = {2020}
}