Module delta.extensions.callbacks

Custom callbacks that come with DELTA.

Functions

def ExponentialLRScheduler(start_epoch: int = 10, multiplier: float = 0.95)

Schedule the learning rate exponentially.

Parameters

start_epoch : int

The epoch to begin.

multiplier : float

After start_epoch, multiply the learning rate by this amount each epoch.

Classes

class SetTrainable (layer_name: str, epoch: int, trainable: bool = True, learning_rate: float = None)

Changes whether a given layer is trainable during training.

This is useful for transfer learning, to do an initial training and then allow fine-tuning.

Parameters

layer_name : str

The layer to modify.

epoch : int

The change will take place at the start of this epoch (the first epoch is 1).

trainable : bool

Whether the layer will be made trainable or not trainable.

learning_rate : float

Optionally change the learning rate as well.

Expand source code

class SetTrainable(tensorflow.keras.callbacks.Callback):
    """
    Changes whether a given layer is trainable during training.

    This is useful for transfer learning, to do an initial training and then allow fine-tuning.
    """
    def __init__(self, layer_name: str, epoch: int, trainable: bool=True, learning_rate: float=None):
        """
        Parameters
        ----------
        layer_name: str
            The layer to modify.
        epoch: int
            The change will take place at the start of this epoch (the first epoch is 1).
        trainable: bool
            Whether the layer will be made trainable or not trainable.
        learning_rate: float
            Optionally change the learning rate as well.
        """
        super().__init__()
        self._layer_name = layer_name
        self._epoch = epoch - 1
        self._make_trainable = trainable
        self._lr = learning_rate
        self._triggered = False

    def on_epoch_begin(self, epoch, logs=None): # pylint: disable=unused-argument
        if epoch == self._epoch:
            if self._triggered:
                return
            self._triggered = True # don't repeat twice
            l = self.model.get_layer(self._layer_name)
            l.trainable = True
            # have to abort, recompile changed model, and continue training
            raise ContinueTrainingException(completed_epochs=epoch, recompile_model=True, learning_rate=self._lr)

Ancestors

• keras.src.callbacks.Callback

Methods

def on_epoch_begin(self, epoch, logs=None)

Called at the start of an epoch.

Subclasses should override for any actions to run. This function should only be called during TRAIN mode.

Args

epoch

Integer, index of epoch.

logs

Dict. Currently no data is passed to this argument for this method but that may change in the future.