[feature request] automatic lammps data and input header generation

[SanggyuChong]

Hello! Thanks for a great model! Been having fun using it for this and that lately.

After an offline discussion with @Luthaf, I would like to ask if it would be possible to provide a script that automatically generates the lammps data file from an arbitrary format (.xyz and friends) , then also provide the necessary header in which the atomic numbers are correctly passed to the model via the LAMMPS in file, properly accounting for the order in which they are defined in the LAMMPS atom types. This is currently a bit of an overhead left to the user, and I think it'd be nicer if this could be resolved from a PET-MAD standpoint.

Open for discussion, thanks a lot all!

[PicoCentauri]

I don't know if a script or just some lines of python codes with docs are better. For non-Python users as script might be useful though

Anyway the way to go is by writing LAMMPS datafile with ASE:

https://wiki.fysik.dtu.dk/ase/ase/io/formatoptions.html#ase.io.lammpsdata.write_lammps_data

The order of the LAMMPS types can be set using the specorder argument. If one wants them in the canonical order of elements, the lines below will do the job assuming the XYZ file is parsed with ase.io.read and called atoms:

atomic_numbers = np.unique(atoms.get_atomic_numbers())
specorder = [ase.symbols.chemical_symbols[n] for n in atomic_numbers]
write_lammps_data("lmp.data", atoms, specorder=specorder, masses=True)

atomic_numbers contains the list for the pair_coeff command, which can be easily formatted for the input file as

print(f"pair_coeff * * {' '.join(map(str, atomic_numbers))}")