wf_psf.training.training_data_adapter

Training Data Adapter.

A module containing training data adapter methods.

Author(s): Jennifer Pollack <jennifer.pollack@cea.fr>

Classes

TrainingDataAdapter(base_adapter[, loss_type])

TrainingDataAdapter.

class wf_psf.training.training_data_adapter.TrainingDataAdapter(base_adapter: DataAdapter, loss_type: str = 'mse')

Bases: object

TrainingDataAdapter.

Wraps a generic DataAdapter to prepare training-specific inputs and targets for TensorFlow models.

Responsibilities: - Stack sources and masks if loss requires it. - Return train / validation inputs and targets separately. - Keep loss-specific logic localized.

Attributes:

train_inputs

Return train inputs.

train_targets

Return train targets.

validation_inputs

Return validation inputs.

validation_targets

Return Validation targets.

property train_inputs: list[Tensor]

Return train inputs.

Train inputs as a list [positions, seds].

property train_targets: Tensor

Return train targets.

Train targets for the model

property validation_inputs: list[Tensor]

Return validation inputs.

Validation inputs as a list [positions, seds].

property validation_targets: Tensor

Return Validation targets.

Validation targets