How to use the saliency.xrai.XRAIParameters function in saliency

additional parameters for the XRAI saliency
                          method. If it is None, an XRAIParameters object
                          will be created with default parameters. See
                          XRAIParameters for more details.

    Raises:
        ValueError: If algorithm type is unknown (not full or fast).
                    If the shape of `base_attribution` dosn't match the shape of `x_value`.

    Returns:
        XRAIOutput: an object that contains the output of the XRAI algorithm.

    TODO(tolgab) Add output_selector functionality from XRAI API doc
    """
    if extra_parameters is None:
      extra_parameters = XRAIParameters()

    # Check the shape of base_attribution.
    if base_attribution is not None:
      if not isinstance(base_attribution, np.ndarray):
        base_attribution = np.array(base_attribution)
      if base_attribution.shape != x_value.shape:
        raise ValueError(
          'The base attribution shape should be the same as the shape of '
          '`x_value`. Expected {}, got {}'.format(
            x_value.shape, base_attribution.shape))

    # Calculate IG attribution if not provided by the caller.
    if base_attribution is None:
      _logger.info("Computing IG...")
      x_baselines = self._make_baselines(x_value, baselines)