Trainer

class motep.train.Trainer(mtp_data: MTPData, seed: int | None = None, rng: Generator | None = None, engine: str = 'cext', loss: dict | LossSetting | None = None, steps: list[dict] | None = None, *, update_mindist: bool = False, comm: DummyMPIComm = <motep.parallel.DummyMPIComm object>)

Bases: object

Trainer.

train(images: list[Atoms]) → LossFunctionBase

Train.

Parameters:

images (list[Atoms]) – List of ASE Atoms objects.

Returns:

loss – LossFunction object after training.

Return type:

LossFunctionBase

update_mindist(images: list[Atoms]) → None

Update min_dist of the MTP potential.