To help you get started, we've collected the most common ways that schnetpack is being used within popular public projects.

atomistic-machine-learning / schnetpack / tests / fixtures / qm9.py View on Github

def qm9_splits(qm9_dataset, qm9_split):
    return spk.data.train_test_split(qm9_dataset, *qm9_split)

View more ways to use schnetpack

SchNetPack - Deep Neural Networks for Atomistic Systems

SchNetPack is a toolbox for the development and application of deep neural networks to the prediction of potential energy surfaces and other quantum-chemical properties of molecules and materials. It contains basic building blocks of atomistic neural networks, manages their training and provides simple access to common benchmark datasets. This allows for an easy implementation and evaluation of new models.

The documentation can be found here.

Features

• SchNet - an end-to-end continuous-filter CNN for molecules and materials [1-3]
• PaiNN - equivariant message-passing for molecules and materials [4]
• Output modules for dipole moments, polarizability, stress, and general response properties
• Modules for electrostatics, Ewald summation, ZBL repulsion
• GPU-accelerated molecular dynamics code incl. path-integral MD, thermostats, barostats

Installation

Install with pip

The simplest way to install SchNetPack is through pip which will automatically get the source code from PyPI:

pip install schnetpack

Install from source

You can also install the most recent code from our repository:

git clone https://github.com/atomistic-machine-learning/schnetpack.git
cd schnetpack
pip install .

Visualization with Tensorboard

SchNetPack supports multiple logging backends via PyTorch Lightning. The default logger is Tensorboard. SchNetPack also supports TensorboardX.

Getting started

The best place to get started is training a SchNetPack model on a common benchmark dataset via the command line interface (CLI). When installing SchNetPack, the training script spktrain is added to your PATH. The CLI uses Hydra and is based on the PyTorch Lightning/Hydra template that can be found here. This enables a flexible configuration of the model, data and training process. To fully take advantage of these features, it might be helpful to have a look at the Hydra and PyTorch Lightning docs.

Example 1: QM9

In the following, we focus on using the CLI to train on the QM9 dataset, but the same procedure applies for the other benchmark datasets as well. First, create a working directory, where all data and runs will be stored:

mkdir spk_workdir
cd spk_workdir

Then, the training of a SchNet model with default settings for QM9 can be started by:

spktrain experiment=qm9_atomwise

The script prints the defaults for the experiment config qm9_atomwise. The dataset will be downloaded automatically to spk_workdir/data, if it does not exist yet. Then, the training will be started.

All values of the config can be changed from the command line, including the directories for run and data. By default, the model is stored in a directory with a unique run id hash as a subdirectory of spk_workdir/runs. This can be changed as follows:

spktrain experiment=qm9_atomwise run.data_dir=/my/data/dir run.path=~/all_my_runs run.id=this_run

If you call spktrain experiment=qm9_atomwise --help, you can see the full config with all the parameters that can be changed. Nested parameters can be changed as follows:

spktrain experiment=qm9_atomwise run.data_dir=