模型开发接口
推理部署接口
API 相关说明
A module to do quant and dequant according to observer’s scale and zero_point.
dtype (Union[str, QuantDtypeMeta]) – a string or QuantDtypeMeta indicating the target quantization dtype of input.
Union
str
QuantDtypeMeta
enable (bool) – whether do normal_forward or fake_quant_forward.
bool
normal_forward
fake_quant_forward
previous
PassiveObserver
next
TQT