A PyTorch library for Graph Convolutional Networks.
Torchgraphs is developed in Python 3.7 and depende on PyTorch 1.1. It is suggested but not required to install PyTorch beforehand, to correctly match the hardware capabilities, see the official installation instruction.
All requirements listed in requirements.txt will be installed automatically when running:
pip install https://github.com/baldassarreFe/torchgraphs
To develop the library itself, a conda environment with additional dependencies is provided, as well as a test suite for pytest:
git clone https://github.com/baldassarreFe/torchgraphs
cd torchgraphs
ENV_NAME=torchgraphs_env
conda env create -n "${ENV_NAME}" -f conda.yaml
conda activate "${ENV_NAME}"
pip install --editable .
pytest