ZeroPointDomain

class torch.ao.quantization.observer.ZeroPointDomain(value)

枚举，指示 zero_point 是在整数域还是浮点域

整数域: quantized_val = (float_val / scale) (整数) + zero_point (整数) 浮点域: quantized_val = (float_val - (zero_point (浮点) - scale * mid_point)) / scale 无域: quantized_val = (float_val / scale)