Installation

Note

HyPER has been tested to be compatible with CUDA=11.8+ and Python=3.9+. For older versions, the stability is not guaranteed.

Quick Start

Get the latest HyPER release:

git clone https://github.com/tzuhanchang/HyPER.git

Note

We recommend setting up environemnt using Anaconda or Miniconda with our provided environment_linux.yml file.

After installing conda, simply run:

conda env create -f environment_linux.yaml

To setup environment manually, HyPER requires the following dependencies:

PyTorch:pytorch >= 2.0, torchvision, torchaudio and pytorch-cuda (for running on GPUs).
PyG: torch_geometric >= 2.0 and torch_scatter
PyTorch HEP:torch_hep == 0.0.4
PyTorch Lightning: lightning
TensorBoard: tensorboard
HDF5 for Python: h5py
Other: numpy, pandas, tqdm, rich

There are some on going issues with PyG (and additional PyG libraries) installation on MacOS, try install it from wheels.