Known Issues

• Discrepancy of the description on the hyperparameter k in the manuscript and the implementation.
  • Fixed. The parameter neighbor_num performs the same as the hyperparameter k in the manuscript
• Under pytorch 1.7.0, there will be a "backward error" (required FloatTensor but received DoubleTensor.) We are inspecting the cause of the error. The library will be upgraded to support the latest pytorch.
  • Fixed. Now support up to pytorch 1.8.1

---

If you find an error reported by the program that is not listed here. Feel free to contact me via [email protected] or raise an issue in this repo.