peal.training.interfaces.PredictorConfig

class peal.training.interfaces.PredictorConfig

Bases: BaseModel

The config template for a model.

Methods

Attributes

architecture	The config of the architecture of the model.
base_model
base_path	The name of the class.
continue_training
data	The config of the data used for training the model.
distill_from	Where to distill from if used for distillation.
generator
is_loaded	A flag that indicates if the model is loaded from a checkpoint.
kwargs	A dict containing all variables that could not be given with the current config structure
model_path	The name of the model.
model_type	The name of the class.
only_last_layer
seed	The seed that is used to ensure reproducibility of results.
tracking_level
weights_path
training	The config of the training of the model.
task	The config of the task the model shall solve.

training: TrainingConfig

The config of the training of the model.

task: TaskConfig

The config of the task the model shall solve.

architecture: ArchitectureConfig | str | None = None

The config of the architecture of the model.

data: DataConfig | None = None

The config of the data used for training the model.

model_path: str = 'peal_runs/predictor1'

The name of the model.

kwargs: dict = {}

A dict containing all variables that could not be given with the current config structure

is_loaded: bool = False

A flag that indicates if the model is loaded from a checkpoint.

model_type: str = 'discriminator'

The name of the class.

base_path: str | None = None

The name of the class.

seed: int = 0

The seed that is used to ensure reproducibility of results.

distill_from: str = 'predictor'

Where to distill from if used for distillation. Could either be done from the dataset or from the model.

weights_path: str | None = None

continue_training: bool = False

tracking_level: int = 4

only_last_layer: bool = False

generator: None | GeneratorConfig = None

base_model: None | str = None