Installation

Install from PyPI

pele wheels support Python 3.11–3.14 on Linux x86_64 with glibc 2.28 or newer, and macOS 14 or newer on Intel and Apple Silicon. Windows is not supported. Install pele into your Python environment with:

pip install pele

Check that the compiled potential and optimization modules import successfully:

python -c "from pele.potentials import LJ; from pele.optimize import lbfgs_py; print('pele is ready')"

Then follow Getting started to minimize a configuration and store minima from a basinhopping run.

Build from source

Building from source requires Python 3.11 or newer, C/C++/Fortran compilers, OpenMP, SUNDIALS in double precision, Eigen, and LAPACK headers. A conda environment can supply these dependencies:

conda create -n pele -c conda-forge python=3.12 pip compilers sundials eigen blas-devel pkg-config
conda activate pele
pip install git+https://github.com/pele-python/pele

If you already have GCC, G++, and GFortran, you can omit compilers from the environment command. SUNDIALS, Eigen, and LAPACK are taken from the active conda environment, or from the system when no conda environment is active.

For editable installs and build options, see the development instructions in the README.

Optional features

The GUI uses PyQt5 and PyOpenGL. With a wheel installation, add them with:

pip install 'pele[gui]'

The OpenGL viewer also needs the system GLU and GLUT libraries. On Ubuntu:

sudo apt install libglu1-mesa libglut3.12

LAMMPS potentials require the LAMMPS Python package:

pip install 'pele[lammps]'

Wheels use the Python LAMMPS interface. To build the compiled interface or generate GUI forms from source, follow the source installation instructions in the README.

scikit-sparse optionally provides sparse Cholesky for rate calculations, and pymol-open-source provides structure viewing.