Dear developers,
For ABACUS, CP2K, and some other DFT codes, their output energies are larger than -100 eV/atom.
When training the NEP model, the warning "There is energy < -100 eV/atom in the data set. Because we use single precision in NEP training, it means that the reference and calculated energies might only be accurate up to 1 meV/atom" will be raised. To avoid this, the user needs to use some scripts (e.g., tools/Analysis_and_Processing/shift_energy_to_zero/shift_energy_to_zero.py) to calculate per-element energy bias and subtract it from the total energy. , which may introduce additional burden.
Other MLIP packages, e.g., DeePMD-kit, MACE, can calculate per-element energy bias internally directly without the need for human intervention. It would be great if NEP could introduce a similar feature.
Dear developers,
For ABACUS, CP2K, and some other DFT codes, their output energies are larger than -100 eV/atom.
When training the NEP model, the warning "There is energy < -100 eV/atom in the data set. Because we use single precision in NEP training, it means that the reference and calculated energies might only be accurate up to 1 meV/atom" will be raised. To avoid this, the user needs to use some scripts (e.g., tools/Analysis_and_Processing/shift_energy_to_zero/shift_energy_to_zero.py) to calculate per-element energy bias and subtract it from the total energy. , which may introduce additional burden.
Other MLIP packages, e.g., DeePMD-kit, MACE, can calculate per-element energy bias internally directly without the need for human intervention. It would be great if NEP could introduce a similar feature.