openvino.runtime.opset14.fake_quantize¶

openvino.runtime.opset14.fake_quantize(data: Union[openvino._pyopenvino.Node, int, float, numpy.ndarray], input_low: Union[openvino._pyopenvino.Node, int, float, numpy.ndarray], input_high: Union[openvino._pyopenvino.Node, int, float, numpy.ndarray], output_low: Union[openvino._pyopenvino.Node, int, float, numpy.ndarray], output_high: Union[openvino._pyopenvino.Node, int, float, numpy.ndarray], levels: int, auto_broadcast: str = 'NUMPY', name: Optional[str] = None, *, prefix: Optional[str] = None, suffix: Optional[str] = None) → openvino._pyopenvino.Node¶

Perform an element-wise linear quantization on input data.

Parameters

data – The node with data tensor.
input_low – The node with the minimum for input values.
input_high – The node with the maximum for input values.
output_low – The node with the minimum quantized value.
output_high – The node with the maximum quantized value.
levels – The number of quantization levels. Integer value.
auto_broadcast – The type of broadcasting specifies rules used for auto-broadcasting of input tensors.
name – Optional name of the new node.
prefix – Optional keyword-only string to apply before original names of all generated input nodes (for example: passed as numpy arrays).
suffix – Optional keyword-only string to apply after original names of all generated input nodes (for example: passed as numpy arrays).

Returns

New node with quantized value.

Input floating point values are quantized into a discrete set of floating point values.

if x <= input_low:
    output = output_low
if x > input_high:
    output = output_high
else:
    output = fake_quantize(output)

Fake quantize uses the following logic:

f[ output =: dfrac{round( dfrac{data - input_low}{(input_high - input_low)cdot (levels-1)})} {(levels-1)cdot (output_high - output_low)} + output_low f]