Environment Setup

Make sure you have Conda installed. Then clone the repository and set up the environment:

git clone https://github.com/YMa-lab/PHAROST.git
cd PHAROST

# Create the conda environment
conda env create -f environment.yml

# Activate the environment
conda activate PHAROST_env

Installation should take between less than 15 minutes on a standard desktop.

Install PyTorch Geometric

The PyG companion libraries are already declared in environment.yml and installed by conda env create. If you ever need to reinstall them manually:

pip install torch-scatter torch-sparse torch-cluster torch-spline-conv torch-geometric \
  -f https://data.pyg.org/whl/torch-2.8.0+cu128.html

Version Compatibility

torch_scatter, torch_sparse, torch_cluster, and torch_spline_conv ship as precompiled C extensions that must match the installed PyTorch ABI exactly. The pinned versions in environment.yml target PyTorch 2.8 + CUDA 12.8 (note the +pt28cu128 suffix on the wheel filenames).

If torch is later upgraded (e.g., to 2.10), import torch_geometric will segfault with no helpful error message. To verify versions are aligned:

python -c "import torch; print(torch.__version__, torch.version.cuda)"
pip list | grep -E "^torch_(scatter|sparse|cluster|spline)"
# Expect: torch 2.8.x / cuda 12.8 / companion libs all '+pt28cu128'

If they don’t match, either downgrade torch back to 2.8.0:

pip install torch==2.8.0 torchvision torchaudio \
  --index-url https://download.pytorch.org/whl/cu128

or reinstall the PyG companion libs against your current torch version using the matching index from https://data.pyg.org/whl/.

Optional: Jupyter Notebook Support

If you plan to use Jupyter notebooks:

conda install -c anaconda ipykernel
python -m ipykernel install --user --name PHAROST_env