tfr.keras.model.InputCreator

Interface for input creator.

The InputCreator class is an abstract class to implement create_inputs in ModelBuilder in tfr.keras.

To be implemented by subclasses:

  • __call__(): Contains the logic to create tf.keras.Input for context and example inputs.

Example subclass implementation:

class SimpleInputCreator(InputCreator):

  def __call__(self):
    return {}, {"example_feature_1": tf.keras.Input((1,), dtype=tf.float32)}

Methods

__call__

View source

@abc.abstractmethod
__call__() -> Tuple[tfr.keras.model.TensorDict, tfr.keras.model.TensorDict]

Invokes the InputCreator instance.

Returns
A tuple of two dicts which map the context and example feature keys to the corresponding tf.keras.Input.