[BUG] exported pymatgen/structure cannot match exported ase/structure

[qchempku2017]

Bug summary

Dear dpdata developers,

Recently I'm having an issue exporting structures into pymatgen format, and I'm eagerly looking for your help. Thank you very much in advance!

My system is LixFePO4, comprising of Li, Fe, P and O elements. When exporting into ase/structure, everything worked out fine:

However, as I switched to exporting pymatgen/structure, weird behavior appeared. Some O atom changed position with P atom, and Fe exchanged with Li, giving unreasonable outcome:

I was using pymatgen == 2024.11.13.

dpdata Version

0.2.21

Input Files, Running Commands, Error Log, etc.

Original dpdata system:

rlx-1.zip

exported structure cif files:

rlx-1-exports.zip

Please uncompress.

Steps to Reproduce

Codes I used:

import dpdata as dd
sys = dd.LabeledSystem("./lfp-dpmd-train0/rlx_fin/rlx-1/", fmt="deepmd/npy")
atoms = sys.to("ase/structure")[0]
atoms.write("./lfp-dpmd-relax0/rlx-1-final-ase.cif")
struct = sys.to("pymatgen/structure")[0]
_ = struct.to("./lfp-dpmd-relax0/rlx-1-final-pmg.cif")

Further Information, Files, and Links

No response

[wanghan-iapcm]

will be fixed by #782

[qchempku2017]

will be fixed by #782

God speed! Appreciate that!