machine_learning_fit

machine_learning_fit#

autoplex.fitting.common.jobs.machine_learning_fit(database_dir, species_list, isolated_atoms_energies=None, num_processes_fit=32, auto_delta=True, glue_xml=False, mlip_type=None, ref_energy_name='REF_energy', ref_force_name='REF_forces', ref_virial_name='REF_virial', device='cuda', hyper_param_optimization=False, **fit_kwargs)[source]#

Job for fitting potential(s).

Parameters:

  • database_dir (str | Path) – the database directory.

  • isolated_atoms_energies (dict | None) – Dict of isolated atoms energies.

  • num_processes_fit (int) – number of processes for fitting.

  • auto_delta (bool) – automatically determine delta for 2b, 3b and SOAP terms.

  • glue_xml (bool) – use the glue.xml core potential instead of fitting 2b terms.

  • mlip_type (str) – Choose one specific MLIP type: ‘GAP’ | ‘J-ACE’ | ‘P-ACE’ | ‘NEQUIP’ | ‘M3GNET’ | ‘MACE’

  • species_list (list.) – List of element names (str)

  • ref_energy_name (str, optional) – Reference energy name.

  • ref_force_name (str, optional) – Reference force name.

  • ref_virial_name (str, optional) – Reference virial name.

  • device (str) – specify device to use cuda or cpu.

  • hyper_param_optimization (bool) – call hyperparameter optimization (HPO) or not

  • fit_kwargs (dict.) – dict including more fit keyword args.