Three changes:
- Fixed a bug related to triclinic box. In GPUGA-v1.0, the code and the manual are not consistent regarding the box matrix: they differ by a matrix transpose. This bug does not affect the results in the paper.
- Fixed a bug related to the S parameter in the minimal-Tersoff potential. When S=2, there is no problem, but when S !=2, the derivatives of the repulsive and attractive functions are not correctly calculated. This bug does not affect the results in the paper.
- Uploaded a C++ code to transform VASP data (should be in given forms) to
train.in. This is only for me, but the users can also check if it is helpful.