Installation

MintFlow is available for Python 3.10 and 3.11.

We do not recommend installation on your system Python. Please set up a virtual environment, e.g. via venv or conda through the Mambaforge distribution, or create a Docker image.

To set up and activate a virtual environment with venv, run:

python3 -m venv ~/.venvs/mintflow
source ~/.venvs/mintflow/bin/activate

To create and activate a conda environment instead, run:

conda create -n mintflow python=3.11
conda activate mintflow

Step 1: Install PyTorch

Visit the PyTorch website and install its appropriate version based on your OS and compute platform.

Step 2: Install PyTorch Geometric Dependencies

Step 2.1: Figure out your PyTorch and CUDA versions

To learn your PyTorch version, if you installed PyTorch with conda, you can run

conda list | grep torch

or if you installed PyTorch via pip, you can run

pip list | grep torch

If you want to use GPU acceleration, please run the following command to know your CUDA version:

nvidia-smi

Step 2.2: Install additional libraries related to PyTorch Geometric

Before installing PyTorch Geometric, you need to install some additional external libraries. These include:

• PyTorch Scatter

• PyTorch Sparse

To install these libraries, run

pip install torch_scatter torch_sparse -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html

where ${TORCH} and ${CUDA} should be replaced by the specific PyTorch and CUDA versions (see previous section how to obtain them).

For example, for PyTorch 2.6.0 and CUDA 12.4, type:

pip install torch_scatter torch_sparse -f https://data.pyg.org/whl/torch-2.6.0+cu124.html

If you have chosen not to use GPU acceleration, ${CUDA} should be replaced by “cpu”.

Step 3: Install MintFlow

Install MintFlow via pip:

pip install mintflow

Or install including optional dependencies required for running tutorials with:

pip install mintflow[all]