Installation

MXtalTools has two classes of dependency:

• PyTorch + PyG — must be installed manually because the correct wheels depend on your CUDA version.

• Everything else — handled automatically by pip install mxtaltools.

Step 1 — Install PyTorch and PyTorch Geometric

Choose the commands for your CUDA version from the official guides:

• PyTorch installation guide

• PyG installation guide

For example, with CUDA 11.8 and PyTorch 2.4:

pip install torch==2.4.1 --index-url https://download.pytorch.org/whl/cu118

pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv \
    -f https://data.pyg.org/whl/torch-2.4.0+cu118.html

pip install torch-geometric==2.7.0

Step 2 — Install MXtalTools

Users:

pip install mxtaltools

Developers (editable install from a clone):

git clone git@github.com:InfluenceFunctional/MXtalTools.git
cd MXtalTools
pip install -e .

Optional Dependencies

UMA model support (fairchem-core) — only needed if you want to use the UMA machine-learning interatomic potential:

pip install mxtaltools[uma]
# or, separately:
pip install fairchem-core

CSD Python API — only needed for constructing crystal datasets from .cif files. Requires a valid licence from CCDC. Install separately following the CCDC instructions.

Weights & Biases — required for model training. Already included in the base install; activate with:

wandb login

User config — for model training, create configs/users/YOUR_USERNAME.yaml with your paths and W&B settings, then pass --user YOUR_USERNAME on the command line.