tfl.pwl_calibration_lib.verify_hyperparameters

View source on GitHub

Verifies that all given hyperparameters are consistent.

tfl.pwl_calibration_lib.verify_hyperparameters(
    input_keypoints=None,
    output_min=None,
    output_max=None,
    monotonicity=None,
    convexity=None,
    is_cyclic=False,
    lengths=None,
    weights_shape=None
)

See PWLCalibration class level comment for detailed description of arguments.

Args:

  • input_keypoints: input_keypoints of PWLCalibration layer.
  • output_min: Smallest output of PWLCalibration layer.
  • output_max: Largest output of PWLCalibration layer.
  • monotonicity: monotonicity hyperparameter of PWLCalibration layer.
  • convexity: convexity hyperparameter of PWLCalibration layer.
  • is_cyclic: is_cyclic hyperparameter of PWLCalibration layer.
  • lengths: Lengths of pieces of piecewise linear function.
  • weights_shape: Shape of weights of PWLCalibration layer.

Raises:

  • ValueError: If something is inconsistent.