\(\renewcommand{\AA}{\text{Å}}\)

2.2. Installation

The LAMMPS Python module enables calling the LAMMPS C library API from Python by dynamically loading functions in the LAMMPS shared library through the Python ctypes module. Because of the dynamic loading, it is required that LAMMPS is compiled in “shared” mode.

Two components are necessary for Python to be able to invoke LAMMPS code:

  • The LAMMPS Python Package (lammps) from the python folder

  • The LAMMPS Shared Library (liblammps.so, liblammps.dylib or liblammps.dll) from the folder where you compiled LAMMPS.

2.2.1. Installing the LAMMPS Python Module and Shared Library

Making LAMMPS usable within Python and vice versa requires putting the LAMMPS Python package (lammps) into a location where the Python interpreter can find it and installing the LAMMPS shared library into a folder that the dynamic loader searches or inside of the installed lammps package folder. There are multiple ways to achieve this.

  1. Install both components into a Python site-packages folder, either system-wide or in the corresponding user-specific folder. This way no additional environment variables need to be set, but the shared library is otherwise not accessible.

  2. Do an installation into a virtual environment.

  3. Leave the files where they are in the source/development tree and adjust some environment variables.

Compile LAMMPS with either CMake or the traditional make procedure in shared mode. After compilation has finished, type (in the compilation folder):

make install-python

This will try to build a so-called (binary) wheel file, a compressed binary python package and then install it with the python package manager ‘pip’. Installation will be attempted into a system-wide site-packages folder and if that fails into the corresponding folder in the user’s home directory. For a system-wide installation you usually would have to gain superuser privilege first, e.g. though sudo

File

Location

Notes

LAMMPS Python package

  • $HOME/.local/lib/pythonX.Y/site-packages/lammps

X.Y depends on the installed Python version

LAMMPS shared library

  • $HOME/.local/lib/pythonX.Y/site-packages/lammps

X.Y depends on the installed Python version

For a system-wide installation those folders would then become.

File

Location

Notes

LAMMPS Python package

  • /usr/lib/pythonX.Y/site-packages/lammps

X.Y depends on the installed Python version

LAMMPS shared library

  • /usr/lib/pythonX.Y/site-packages/lammps

X.Y depends on the installed Python version

No environment variables need to be set for those, as those folders are searched by default by Python or the LAMMPS Python package.

Changed in version 24Mar2022.

Note

If there is an existing installation of the LAMMPS python module, make install-python will try to update it. However, that will fail if the older version of the module was installed by LAMMPS versions until 17Feb2022. Those were using the distutils package, which does not create a “manifest” that allows a clean uninstall. The make install-python command will always produce a lammps-<version>-<python>-<abi>-<os>-<arch>.whl file (the ‘wheel’). And this file can be later installed directly with python -m pip install <wheel file>.whl without having to type make install-python again and repeating the build step, too.

For the traditional make process you can override the python version to version x.y when calling make with PYTHON=pythonX.Y. For a CMake based compilation this choice has to be made during the CMake configuration step.

If the default settings of make install-python are not what you want, you can invoke install.py from the python directory manually as

python install.py -p <python package> -l <shared library> -v <version.h file> [-n]
  • The -p flag points to the lammps Python package folder to be installed,

  • the -l flag points to the LAMMPS shared library file to be installed,

  • the -v flag points to the LAMMPS version header file to extract the version date,

  • and the optional -n instructs the script to only build a wheel file but not attempt to install it.

To verify if LAMMPS can be successfully started from Python, start the Python interpreter, load the lammps Python module and create a LAMMPS instance. This should not generate an error message and produce output similar to the following:

$ python
Python 3.8.5 (default, Sep  5 2020, 10:50:12)
[GCC 10.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import lammps
>>> lmp = lammps.lammps()
LAMMPS (18 Sep 2020)
using 1 OpenMP thread(s) per MPI task
>>>

Note

Unless you opted for “In place use”, you will have to rerun the installation any time you recompile LAMMPS to ensure the latest Python package and shared library are installed and used.

Note

If you want Python to be able to load different versions of the LAMMPS shared library with different settings, you will need to manually copy the files under different names (e.g. liblammps_mpi.so or liblammps_gpu.so) into the appropriate folder as indicated above. You can then select the desired library through the name argument of the LAMMPS object constructor (see Creating or deleting a LAMMPS object).

2.2.2. Extending Python to run in parallel

If you wish to run LAMMPS in parallel from Python, you need to extend your Python with an interface to MPI. This also allows you to make MPI calls directly from Python in your script, if you desire.

We have tested this with MPI for Python (aka mpi4py) and you will find installation instruction for it below.

Installation of mpi4py (version 3.0.3 as of Sep 2020) can be done as follows:

  • Via pip into a local user folder with:

    pip install --user mpi4py
    
  • Via dnf into a system folder for RedHat/Fedora systems:

    # for use with OpenMPI
    sudo dnf install python3-mpi4py-openmpi
    # for use with MPICH
    sudo dnf install python3-mpi4py-openmpi
    
  • Via pip into a virtual environment (see above):

    $ source $HOME/myenv/activate
    (myenv)$ pip install mpi4py
    
  • Via pip into a system folder (not recommended):

    sudo pip install mpi4py
    

For more detailed installation instructions and additional options, please see the mpi4py installation page.

To use mpi4py and LAMMPS in parallel from Python, you must make certain that both are using the same implementation and version of MPI library. If you only have one MPI library installed on your system this is not an issue, but it can be if you have multiple MPI installations (e.g. on an HPC cluster to be selected through environment modules). Your LAMMPS build is explicit about which MPI it is using, since it is either detected during CMake configuration or in the traditional make build system you specify the details in your low-level src/MAKE/Makefile.foo file. The installation process of mpi4py uses the mpicc command to find information about the MPI it uses to build against. And it tries to load “libmpi.so” from the LD_LIBRARY_PATH. This may or may not find the MPI library that LAMMPS is using. If you have problems running both mpi4py and LAMMPS together, this is an issue you may need to address, e.g. by loading the module for different MPI installation so that mpi4py finds the right one.

If you have successfully installed mpi4py, you should be able to run Python and type

from mpi4py import MPI

without error. You should also be able to run Python in parallel on a simple test script

mpirun -np 4 python3 test.py

where test.py contains the lines

from mpi4py import MPI
comm = MPI.COMM_WORLD
print("Proc %d out of %d procs" % (comm.Get_rank(),comm.Get_size()))

and see one line of output for each processor you run on. Please note that the order of the lines is not deterministic

$ mpirun -np 4 python3 test.py
Proc 0 out of 4 procs
Proc 1 out of 4 procs
Proc 2 out of 4 procs
Proc 3 out of 4 procs