MACEReplayFinetuningSettings

Included in QATK.MLFF

class MACEReplayFinetuningSettings(replay_data_filepath, number_of_samples=None, replay_subselect_method=None, replay_filtering_type=None)

Initialize replay fine-tuning settings.

Parameters:
  • replay_data_filepath (str) – Path to the replay data file (.xyz format) - required.

  • number_of_samples (int) – Number of samples to select from replay data.

  • replay_subselect_method (str) – Method for subselecting samples. Possible values are MLParameterOptions.REPLAY_SUBSELECT options (MLParameterOptions.REPLAY_SUBSELECT.FPS, MLParameterOptions.REPLAY_SUBSELECT.RANDOM) or strings.
    Default: MLParameterOptions.REPLAY_SUBSELECT.RANDOM

  • replay_filtering_type (str) – Type of filtering to apply on the configurations based on elements. Possible values are MLParameterOptions.REPLAY_FILTERING options (MLParameterOptions.REPLAY_FILTERING.NONE, MLParameterOptions.REPLAY_FILTERING.COMBINATIONS, MLParameterOptions.REPLAY_FILTERING.EXCLUSIVE, MLParameterOptions.REPLAY_FILTERING.INCLUSIVE) or strings.
    Default: MLParameterOptions.REPLAY_FILTERING.COMBINATIONS

nlinfo()
Returns:

The nlinfo.

Return type:

dict

numberOfSamples()
Returns:

Number of samples to select from replay data

Return type:

int

replayDataFilepath()
Returns:

Path to the replay data file

Return type:

str

replayFilteringType()
Returns:

Type of filtering to apply on the configurations based on elements.

Return type:

str

replaySubselectMethod()
Returns:

Subselection method

Return type:

str

requiresFoundationModel()

Check if configuration requires a foundation model.

Returns:

True if foundation model is required, False otherwise

Return type:

bool

uniqueString()

Return a unique string representing the state of the object.