Metadata-Version: 2.1
Name: dpdata-ani
Version: 0.1.1
Summary: The dpdata plugin for ANI.
Project-URL: Homepage, https://github.com/njzjz/dpdata_ani
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: dpdata (>=0.2.7)
Requires-Dist: torchani
Requires-Dist: torch
Requires-Dist: numpy
Requires-Dist: h5py
Provides-Extra: aiqm1
Requires-Dist: MLatom ; extra == 'aiqm1'
Requires-Dist: dftd4 ; extra == 'aiqm1'
Provides-Extra: test
Requires-Dist: pytest ; extra == 'test'
Requires-Dist: pytest-cov ; extra == 'test'

# dpdata_ani

The [dpdata](https://github.com/deepmodeling/dpdata) plugin for ANI.

## Features

- `ani/1x` format to load [ANI-1x dataset](https://doi.org/10.1038/s41597-020-0473-z)
- `ani/1x`, `ani/2x`, and `ani/1ccx` driver to predict using ANI models with automatic batch size
- `aiqm1` driver for [AIQM1](https://doi.org/10.1038/s41467-021-27340-2) which is based on ANI

## Installation

```sh
pip install dpdata-ani
```

To enable AIQM1,
```sh
pip install dpdata-ani[aiqm1]
```

## Usage

First, download ANI-1x dataset:
```sh
wget https://figshare.com/ndownloader/files/18112775 -O ani1x.hdf5
```

Load the ANI-1x dataset, and predict it with ANI-2x models:
```py
import dpdata
ani = dpdata.MultiSystems().from_ani_1x("ani1x.hdf5")
predict = ani.predict(driver="ani/2x")
predict.to_deepmd_hdf5("predict.hdf5")
```

The inferred data will be stored in `predict.hdf5`.
