Deploy (these builds only succeed on tagged commits):
TorchANI is a pytorch implementation of ANI. It is currently under alpha release, which means, the API is not stable yet. If you find a bug of TorchANI, or have some feature request, feel free to open an issue on GitHub, or send us a pull request.
Install
TorchANI requires the latest preview version of PyTorch. You can install PyTorch by the following commands (assuming cuda10):
After installing the correct PyTorch, you can install TorchANI by:
pip install torchani
See also PyTorch's official site for instructions of installing latest preview version of PyTorch.
Please install nightly PyTorch through pip install instead of conda install. If your PyTorch is installed through conda install, then pip would mistakenly recognize the package name as torch instead of torch-nightly, which would cause dependency issue when installing TorchANI.
To run the tests and examples, you must manually download a data package
./download.sh
Paper
The original ANI-1 paper is:
Smith JS, Isayev O, Roitberg AE. ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chemical science. 2017;8(4):3192-203.
We are planning a seperate paper for TorchANI, it will be available when we are ready for beta release of TorchANI.