FakeQuantMovingAverageAbsMax

class paddle.nn.quant.quant_layers. FakeQuantMovingAverageAbsMax ( name: str | None = None, moving_rate: float = 0.9, quant_bits: int = 8, dtype: DTypeLike = 'float32', reduce_type: Literal['max'] | None = None ) [source]

FakeQuantMovingAverageAbsMax layer does the moving_average_abs_max quant and then dequant. Its computational formula is described as below:

$scale = (moving\_rate*accum+max(abs(x)))/(moving\_rate*state+1)$ $range = 2^{bit\_length - 1} - 1$ $Out = round(X / scale * range) * scale / range$

forward ( input: Tensor ) → Tensor

forward¶

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

• *inputs (tuple) – unpacked tuple arguments

• **kwargs (dict) – unpacked dict arguments