Installation#

This guide covers installing Enerzyme from source on the main branch.

Clone the repository#

git clone https://github.com/Benzoin96485/Enerzyme.git
cd Enerzyme

Create a conda environment#

We recommend creating a conda environment from requirements.yaml:

conda env create -f requirements.yaml
conda activate enerzyme

The file installs core dependencies. For a pinned release for reproducibility of paper results, install from the requirements.yaml in the corresponding subdirectory of examples/. The environment may have a different name from enerzyme.

Install torch-scatter#

torch-scatter must match your PyTorch, CUDA, Python, and platform. Go to https://data.pyg.org/whl/ and pick the wheel that matches your stack. For example, with PyTorch 2.5.1, CUDA 12.4, Python 3.12, and Linux x86_64:

pip install https://data.pyg.org/whl/torch-2.5.0%2Bcu124/torch_scatter-2.1.2%2Bpt25cu124-cp312-cp312-linux_x86_64.whl

If the wheel for your stack is not found, you can build it from source.

Install Enerzyme#

From the repository root:

pip install -e .

Optional dependencies#

Some workflows need extra packages or external programs that are not installed by default:

  • NequIP models — nequip

  • XPaiNN models — XequiNet, SciPy, PySCF, pydantic

  • PLUMED enhanced sampling — py-plumed and a PLUMED-enabled build

  • QM annotation with TeraChem — TeraChem (licensed)

  • Bond assignment — QuantumPDB hjkgrp/quantumPDB (optional but recommended)

  • Enerzymette launchers — install from the Enerzymette repository Benzoin96485/Enerzymette with pip install -e .

  • fairchem — install from the fairchem repository Benzoin96485/fairchem

Verify the installation#

python -c "import enerzyme"
enerzyme -h
enerzyme predict -h
enerzyme simulate -h

If all commands print help text without import errors, the core install is ready.