Environment

Installation

We recommend environment management with Conda. The supports below are based on Conda.

To config a DGRL method and train the model on existing datasets, the following packages are required:

conda create -n dgrl python==3.10
conda activate dgrl
# torch
conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=12.1 -c pytorch -c nvidia
# PyG (>=2.4.0)
pip install torch_geometric==2.4.0
# PyG dependencies
pip install torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.1.0+cu121.html
# PyGSD
pip install torch-geometric-signed-directed
# RAY
pip install -U "ray[data,train,tune,serve]"
# RAY dependencies
pip install hyperopt
# other dependencies
pip install prettytable
pip install torchmetrics
pip install hyperopt
pip install easydict