Hyper-parameters for reproducing the reported results are provided in the train_QuatE_dataset.py.

How to run

Requirements: Pytorch 1.4+

STEP:

1. export LD_LIBRARY_PATH=/usr/local/lib:/usr/lib:/usr/local/lib64:/usr/lib64
2. CUDA_VISIBLE_DEVICES=0 python3 train_QuatE_dataset.py

Citation

@article{zhang2019quaternion,
  title={Quaternion Knowledge Graph Embedding},
  author={Zhang, Shuai and Tay, Yi and Yao, Lina and Liu, Qi},
  journal={arXiv preprint arXiv:1904.10281},
  year={2019}
}

This code is based on the OpenKE project.