SchNetPack - Deep Neural Networks for Atomistic Systems





GitHub Stars



Last Commit

12d ago










SchNetPack - Deep Neural Networks for Atomistic Systems

Build Status codecov Code style: black

SchNetPack aims to provide accessible atomistic neural networks that can be trained and applied out-of-the-box, while still being extensible to custom atomistic architectures.

Currently provided models:
  • SchNet - an end-to-end continuous-filter CNN for molecules and materials [1-3]
  • wACSF - weighted atom-centered symmetry functions [4,5]

Note: This is the last version before a major update. In the next version, we plan to adopt Hydra and PyTorch Lightning, switch to indexing instead of masking and make the networks compatible with TorchScript. Therefore, there will be breaking changes.

  • python 3
  • ASE
  • numpy
  • PyTorch (>=0.4.1)
  • h5py
  • Optional: tensorboardX

Note: We recommend using a GPU for training the neural networks.


Install with pip

pip install schnetpack

Install from source

Clone the repository

git clone
cd schnetpack

Install requirements

pip install -r requirements.txt

Install SchNetPack

pip install .

You're ready to go!

Getting started

The best place to start is training a SchNetPack model on a common benchmark dataset. The example scripts provided by SchNetPack are inserted into your PATH during installation.

QM9 example

The QM9 example scripts allows to train and evaluate both SchNet and wACSF neural networks. The training can be started using: train <schnet/wacsf> qm9 <dbpath> <modeldir> --split num_train num_val [--cuda]

where num_train and num_val need to be replaced by the number of training and validation datapoints respectively.

You can choose between SchNet and wACSF networks and have to provide a path to the database file and a path to a directory which will be used to store the model. If the database path does not exist, the data is downloaded and stored there. Please note that the database path must include the file extension .db. With the --cuda flag, you can activate GPU training. The default hyper-parameters should work fine, however, you can change them through command-line arguments. Please refer to the help at train <schnet/wacsf> --help.

The training progress will be logged in <modeldir>/log, either as CSV (default) or as TensorBoard event files. For the latter, TensorBoard needs to be installed to view the event files. This can be done by installing the version included in TensorFlow

pip install tensorflow

or the standalone version.

To evaluate the trained model with the best validation error, call eval <datapath> <modeldir> --split test [--cuda]

which will run on the specified --split and write a result file evaluation.txt into the model directory.


For the full API reference, visit our documentation.

If you are using SchNetPack in you research, please cite:

K.T. Schütt, P. Kessel, M. Gastegger, K. Nicoli, A. Tkatchenko, K.-R. Müller. SchNetPack: A Deep Learning Toolbox For Atomistic Systems. J. Chem. Theory Comput. 10.1021/acs.jctc.8b00908 arXiv:1809.01072. (2018)


  • [1] K.T. Schütt. F. Arbabzadah. S. Chmiela, K.-R. Müller, A. Tkatchenko.
    Quantum-chemical insights from deep tensor neural networks. Nature Communications 8. 13890 (2017)

  • [2] K.T. Schütt. P.-J. Kindermans, H. E. Sauceda, S. Chmiela, A. Tkatchenko, K.-R. Müller.
    SchNet: A continuous-filter convolutional neural network for modeling quantum interactions. Advances in Neural Information Processing Systems 30, pp. 992-1002 (2017) link

  • [3] K.T. Schütt. P.-J. Kindermans, H. E. Sauceda, S. Chmiela, A. Tkatchenko, K.-R. Müller.
    SchNet - a deep learning architecture for molecules and materials. The Journal of Chemical Physics 148(24), 241722 (2018) 10.1063/1.5019779

  • [4] M. Gastegger, L. Schwiedrzik, M. Bittermann, F. Berzsenyi, P. Marquetand. wACSF—Weighted atom-centered symmetry functions as descriptors in machine learning potentials. The Journal of Chemical Physics, 148(24), 241709. (2018) 10.1063/1.5019667

  • [5] J. Behler, M. Parrinello. Generalized neural-network representation of high-dimensional potential-energy surfaces. Physical Review Letters, 98(14), 146401. (2007) 10.1103/PhysRevLett.98.146401

Rate & Review

Great Documentation0
Easy to Use0
Highly Customizable0
Bleeding Edge0
Responsive Maintainers0
Poor Documentation0
Hard to Use0
Unwelcoming Community0