tf.raw_ops.QuantizedInstanceNorm

Quantized Instance normalization.

tf.raw_ops.QuantizedInstanceNorm(
    x, x_min, x_max, output_range_given=False, given_y_min=0, given_y_max=0,
    variance_epsilon=1e-05, min_separation=0.001, name=None
)

Args:

  • x: A Tensor. Must be one of the following types: qint8, quint8, qint32, qint16, quint16. A 4D input Tensor.
  • x_min: A Tensor of type float32. The value represented by the lowest quantized input.
  • x_max: A Tensor of type float32. The value represented by the highest quantized input.
  • output_range_given: An optional bool. Defaults to False. If True, given_y_min and given_y_min and given_y_max are used as the output range. Otherwise, the implementation computes the output range.
  • given_y_min: An optional float. Defaults to 0. Output in y_min if output_range_given is True.
  • given_y_max: An optional float. Defaults to 0. Output in y_max if output_range_given is True.
  • variance_epsilon: An optional float. Defaults to 1e-05. A small float number to avoid dividing by 0.
  • min_separation: An optional float. Defaults to 0.001. Minimum value of y_max - y_min
  • name: A name for the operation (optional).

Returns:

A tuple of Tensor objects (y, y_min, y_max).

  • y: A Tensor. Has the same type as x.
  • y_min: A Tensor of type float32.
  • y_max: A Tensor of type float32.