[Feature Request] Obtaining global and atomic dipole data from ab initio softwares

[shuaijiang-ustc]

Summary

Since in dpmd v2.0.0, tensor like dipole is supported, I am wondering if tensor data prep is under the way.

Detailed Description

I am conducting ab initio calculations based on Gaussian16. But I am not sure how to get the global and atomic dipole data including the units used in DPMD.

[amcadmus]

Hello, the dipole fitting method is to fit a vector defined on an atom. The "vector" can be atomic dipole, or any other vectorial property. The labels of dipole fitting can be of any unit.

[shuaijiang-ustc]

I see, thanks. But I found that in DataModifier.py, the trained dipole model is loaded to modify data. So does the dipole unit used in dataset affects this procedure? BTW, is the method in datamodifier.py the same as that used in https://doi.org/10.1063/5.0031215 except that the SPCE charges are replaced by WFCC? Many thanks.

[amcadmus]

No the DataModifier is for the long-range version of DP, which has not been published.

[shuaijiang-ustc]

Thanks, look foward to the long-range version!

[njzjz]

I recheck this issue and it seems to be a feature request of dpdata instead of DeePMD-kit.