tf.quantization.fake_quant_with_min_max_vars

Fake-quantize the 'inputs' tensor of type float via global float scalars

tf.quantization.fake_quant_with_min_max_vars(
    inputs: Annotated[Any, _atypes.Float32],
    min: Annotated[Any, _atypes.Float32],
    max: Annotated[Any, _atypes.Float32],
    num_bits: int = 8,
    narrow_range: bool = False,
    name=None
) -> Annotated[Any, _atypes.Float32]

Fake-quantize the inputs tensor of type float via global float scalars min and max to outputs tensor of same shape as inputs.

Attributes

  • [min; max] define the clamping range for the inputs data.
  • inputs values are quantized into the quantization range ( [0; 2^num_bits - 1] when narrow_range is false and [1; 2^num_bits - 1] when it is true) and then de-quantized and output as floats in [min; max] interval.
  • num_bits is the bitwidth of the quantization; between 2 and 16, inclusive.

Before quantization, min and max values are adjusted with the following logic. It is suggested to have min <= 0 <= max. If 0 is not in the range of values, the behavior can be unexpected:

  • If 0 < min < max: min_adj = 0 and max_adj = max - min.
  • If min < max < 0: min_adj = min - max and max_adj = 0.
  • If min <= 0 <= max: scale = (max - min) / (2^num_bits - 1), min_adj = scale * round(min / scale) and max_adj = max + min_adj - min.

This operation has a gradient and thus allows for training min and max values.

inputs A Tensor of type float32.
min A Tensor of type float32.
max A Tensor of type float32.
num_bits An optional int. Defaults to 8.
narrow_range An optional bool. Defaults to False.
name A name for the operation (optional).

A Tensor of type float32.