dpdata, vasp, NWRITE=0,2

[chtchelkatchev]

vasp_to_mace_xyz.py.txt

dpdata has a known issue with reading OUTCAR or vasprun.xml files when nwrite=0. However, ASE can successfully parse vasprun.xml in this case and convert it into the extxyz format, which contains all necessary data: lattice parameters, energies, forces, and stresses. It would be great to enhance dpdata to at least match ASE's capabilities in this regard.

It would also be convenient for an extxyz database (e.g., such as the one used for mace) to be freely readable by dpdata.

I typically convert VASP databases into the extxyz format, which can store not only structures, energies, forces, and stresses but also local magnetic moments, density of states, and other properties. The issue is that storing raw VASP output on disk is impractical—the files are enormous and quickly fill up storage. Most of the data generated by VASP isn’t needed for machine learning. Using the xyz format for storage saves a tremendous amount of space.