SPARC Installation Guide

This guide provides step-by-step instructions to set up SPARC and its core dependencies.

Core Software Dependencies

Python 3.10 (recommended)
DeepMD-kit v3+
ASE — Atomic Simulation Environment
At least one DFT engine (VASP, CP2K, ORCA, Quantum ESPRESSO, xTB, or Gaussian)
PLUMED — optional, for enhanced sampling

Python Package Dependencies

numpy, pandas, dpdata, scipy

Step-by-Step Installation

Create and activate a conda environment:

conda create -n sparc python=3.10
conda activate sparc

Install DeepMD-kit v3:
```
pip install deepmd-kit
```
Refer to the DeepMD-kit installation guide for GPU-enabled or backend-specific (PyTorch / TensorFlow) installation options.

Clone and install SPARC:

git clone https://github.com/rahulumrao/sparc.git
cd sparc
pip install .

Install PLUMED (optional):

For standard CVs, install directly from conda-forge:

conda install -c conda-forge py-plumed

Note

Some Collective Variables (e.g., SPRINT) require the additional module and are not included in the standard PLUMED build. To enable them, build from source:

./configure --enable-mpi=no --enable-modules=all \
    PYTHON_BIN=$(which python) --prefix=$CONDA_PREFIX

make -j$(nproc) && make install

To verify PLUMED installation:

>>> from ase.calculators import plumed
>>> from plumed import Plumed

Install xTB (optional):

conda install -c conda-forge xtb-python

DFT Engine Requirements

SPARC supports the following DFT engines. Only install what you need:

Engine	Use case	Notes
VASP	Periodic solids	License required; set `VASP_PP_PATH` env variable
CP2K	Periodic solids / molecules	Free; ASE CP2K interface included
ORCA	Molecules	Free for academia; set `exe_command` in `input.yaml`
Quantum ESPRESSO	Periodic solids	Free; set `exe_command`
xTB	Semi-empirical (fast)	`conda install -c conda-forge xtb-python`
Gaussian	Molecules	License required

Environment Setup

For VASP, set the POTCAR files path:

export VASP_PP_PATH=/path/to/vasp/potcar_files

If PLUMED was installed from source (not conda-forge):

export PLUMED_KERNEL="$CONDA_PREFIX/lib/libplumedKernel.so"
export PYTHONPATH="$CONDA_PREFIX/lib/plumed/python:$PYTHONPATH"

Verification

sparc -h

Expected output:

sparc [-h] [-i INPUT_FILE]

options:
  -h, --help            show this help message and exit
  -i INPUT_FILE, --input_file INPUT_FILE
                        Input YAML file

Important

The pip install tensorflow[and-cuda] installation may not always detect the GPU due to potential configuration issues. To verify whether TensorFlow has successfully recognized the GPU, run:

python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

If the output is an empty list, check:

Your NVIDIA driver and CUDA toolkit installation
The CUDA version compatibility with TensorFlow
That environment variables (e.g., LD_LIBRARY_PATH) are correctly set

Refer to the TensorFlow GPU troubleshooting guide for details.