Questions about training data

[yesheng22]

Thanks for generously sharing the model and training dataset. Unfortunately, the dataset file does not have an associated description file. Combined with the published arxiv literature, there are the following points that need to be confirmed:
For example, the comment line in xyz is the following:

Lattice="-2.88422865 -4.99562956 0.0 -2.88422865 4.99562956 -0.0 -0.0 -0.0 -11.84729818" Properties=species:S:1:pos:R:3:forces:R:3:magmoms. R:1 energy=-154.52400753 stress="-0.00263071343401293 -0.0 -0.0 -0.0 -0.0 -0.0026307123729563788 0.0 -0.0 0.0 -0.0031734529601744146" bandgap=0.0 e _per_atom_relaxed=-8.023012054 mp_id=mp-1208409 energy_per_atom=-8.0280003765 ef_per_atom_relaxed=-8.023012054 corrected_total_energy= -160.56000753 ef_per_atom=-1.2779654928750004 task_id=mp-1208409 calc_id=0 ionic_step=6 pbc="T T T T"

1 Is the unit of energy eV?
2 Is the unit of stress eV/(A^3)?
3 Is the energy mean total energy which applied U?
4 Is the corrected_total_energy equal to energy plus DFT-D3 corrected energy ?
-154.52400753[energy]-6.036[DFTD3 correction]=-160.56000753[corrected_total_energy]

[janosh]

1. yes
2. yes
3. yes
4. where is your example coming from? training_data.zip doesn't contain the DFTD3 correction nor does 2023-11-22-mp-trj-extxyz-by-yuan.zip. those two should be identical anyway. maybe @chiang-yuan can comment about other versions of this file floating around. this is what my mp-1208409 comment line looks like

mptrj-gga-ggapu/mp-1208409.extxyz

Lattice="-2.88422865 -4.99562956 0.0 -2.88422865 4.99562956 -0.0 -0.0 -0.0 -11.84729818" Properties=species:S:1:pos:R:3:forces:R:3:magmoms:R:1 energy=-154.52400753 stress="-0.00263071343401293 -0.0 -0.0 -0.0 -0.0026307123729563788 -0.0 -0.0 0.0 -0.0031734529601744146" bandgap=0.0 e_per_atom_relaxed=-8.023012054 mp_id=mp-1208409 energy_per_atom=-8.0280003765 ef_per_atom_relaxed=-8.023012054 corrected_total_energy=-160.56000753 ef_per_atom=-1.2779654928750004 task_id=mp-1208409 calc_id=0 ionic_step=6 pbc="T T T"

[chiang-yuan]

corrected_total_energy was MP2020Compatibility composition-dependent correction https://github.com/ACEsuit/mace?tab=readme-ov-file#mace-mp-materials-project-force-fields