We recommend creating a virtual Conda environment to run HGraphormer. Following is the instruction:

conda create --name HGraphormer python=3.7 conda activate HGraphormer

conda install pytorch==1.11.0 torchvision==0.12.0 torchaudio==0.11.0 cudatoolkit=11.3 -c pytorch

pip install path pip install pynvml pip install scipy

import torch print(torch.version) print(torch.cuda.is_available()) print(torch.cuda.current_device()) print(torch.cuda.device(0)) print(torch.cuda.get_device_name(0))