ZeroPointDomain#

class torch.ao.quantization.observer.ZeroPointDomain(value)[source]#

Enum that indicate whether zero_point is in integer ___domain or floating point ___domain

integer ___domain: quantized_val = (float_val / scale) (integer) + zero_point (integer) float ___domain: quantized_val = (float_val - (zero_point (float) - scale * mid_point)) / scale none ___domain: quantized_val = (float_val / scale)