Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/njzjz/dpdata_qdpi
dpdata plugin for QDπ
https://github.com/njzjz/dpdata_qdpi
Last synced: 2 months ago
JSON representation
dpdata plugin for QDπ
- Host: GitHub
- URL: https://github.com/njzjz/dpdata_qdpi
- Owner: njzjz
- License: lgpl-3.0
- Created: 2022-09-26T07:48:27.000Z (about 2 years ago)
- Default Branch: master
- Last Pushed: 2024-05-28T20:16:10.000Z (7 months ago)
- Last Synced: 2024-05-29T14:11:19.287Z (7 months ago)
- Language: Python
- Homepage:
- Size: 62.5 KB
- Stars: 0
- Watchers: 2
- Forks: 2
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# dpdata_qdpi
[dpdata](https://github.com/deepmodeling/dpdata) plugin for [QDπ](https://pubs.acs.org/doi/10.1021/acs.jctc.2c01172).
## Features
* `qdpi` driver
## Installation
If you have installed [DeePMD-kit](https://github.com/deepmodeling/deepmd-kit),
```sh
pip install dpdata_qdpi
```
Otherwise, add `gpu` or `cpu` to install DeePMD-kit:
```sh
pip install dpdata_qdpi[gpu]
# or for the CPU version of TensorFlow
pip install dpdata_qdpi[cpu]
```
At this time, you need to install either [AMBERTools](ambermd.org/) SQM (`sqm`) or [DFTB+](https://github.com/dftbplus/dftbplus/) (`dftb+`).
```sh
conda install ambertools -c conda-forge
# OR
conda install dftbplus dftbplus-python -c conda-forge
```
## Usage
Download the QDπ model from [RutgersLBSR/qdpi](https://gitlab.com/RutgersLBSR/qdpi).
```py
from dpdata_qdpi import QDPiDriver
qdpi = QDPiDriver(
model="qdpi-1.0.pb",
charge=0,
backend="sqm",
)
```
`backend` can be either `sqm`, `dftb+`, or `dftb+api`.
Assume you have an XYZ file `ch4.xyz`
```xyz
5
C 0.92334 0.06202 0.01660
H 2.01554 0.06202 0.01660
H 0.55927 1.09164 0.03247
H 0.55927 -0.46653 0.90033
H 0.55927 -0.43903 -0.88301
```
Load the structure:
```py
from dpdata import System
ch4 = System("ch4.xyz")
```
Perform single point calculation using the QDπ model:
```py
p = ch4.predict(driver=qdpi)
print("Energies:", p["energies"][0])
print("Forces:", p["forces"][0])
```
```
Energies: -1102.0472189112793
Forces: [[-4.92853860e-05 3.71129259e-04 -1.00154387e-04]
[ 2.07637527e-02 -1.98691092e-06 -7.85158242e-07]
[-6.81949398e-03 1.93688568e-02 3.32209598e-04]
[-6.96972976e-03 -1.01335356e-02 1.69318148e-02]
[-6.92524354e-03 -9.60446352e-03 -1.71630849e-02]]
```
Or do an optimization:
```py
from dpdata.plugins.ase import ASEMinimizer
lbfgs = ASEMinimizer(
driver=qdpi,
)
p = ch4.minimize(minimizer=lbfgs)
print("Coordinates:", p["coords"][0])
print("Energies:", p["energies"][0])
print("Forces:", p["forces"][0])
```
```
Coordinates: [[ 0.92333966 0.06202338 0.01659862]
[ 2.0161223 0.06202041 0.0165999 ]
[ 0.55907714 1.0921887 0.03247964]
[ 0.559075 -0.46681303 0.9008036 ]
[ 0.5590758 -0.43929946 -0.88349175]]
Energies: -1102.0472
Forces: [[-1.0836746e-04 9.8321143e-05 -2.5905898e-05]
[ 6.7257555e-05 8.3519126e-06 -2.5666191e-06]
[ 4.4526685e-05 -5.3852447e-05 1.7676884e-05]
[-9.4877823e-06 -1.9283394e-05 2.8249622e-05]
[ 6.0710017e-06 -3.3537217e-05 -1.7453991e-05]]
```
Read [dpdata's documentation](https://docs.deepmodeling.com/projects/dpdata) for more usage of dpdata.