Install¶
Install SIMPLE-NN¶
SIMPLE-NN is tested and supported on the following versions of Python:
- Python
2.7
,3.4-3.6
Requirements¶
In SIMPLE-NN, various Python modules are used. Most of these modules are installed automatically during the install process of SIMPLE-NN. However, we recommend to install Tensorflow manually for better performance. In addition, you need to install mpi4py(optional) manually if you want to use MPI. (MPI is supported only in generate_features and preprocess part. See /simple_nn/Simple_nn section.) Detailed information for installing Tensorflow and mpi4py can be found from the links below.
Install Tensorflow: https://www.tensorflow.org/install/
Tensorflow r1.6
-r1.13
is supported.
Install mpi4py: https://mpi4py.readthedocs.io/en/stable/install.html
Install from source¶
You can download a current SIMPLE-NN source package from link below. Once you have a zip file, unzip it. This will create SIMPLE-NN directory. After unzipping the file, run the command below to install SIMPLE-NN.
Download SIMPLE-NN: https://github.com/MDIL-SNU/SIMPLE-NN
cd SIMPLE-NN
python setup.py install
# If root permission is not available, add --user command like below
# python setup.py install --user
Currently, pip install is not supported but will be addressed.
Install LAMMPS implementation¶
To utilize LAMMPS package for SIMPLE-NN generated NN potentials, you must copy the source codes to LAMMPS/src directory with the following command and compile LAMMPS package.
cp SIMPLE-NN/simple_nn/features/symmetry_function/symmetry_functions.h /path/to/lammps/src/
cp SIMPLE-NN/simple_nn/features/symmetry_function/pair_nn.* /path/to/lammps/src/