TQT

class TQT(dtype, enable=True, **kwargs)[源代码]

TQT: https://arxiv.org/abs/1903.08066 Trained Quantization Thresholds for Accurate and Efficient Fixed-Point Inference of Deep Neural Networks

参数
  • dtype (Union[str, QuantDtypeMeta]) – 指明量化输入数据类型的字符串或 QuantDtypeMeta.

  • enable (bool) – 选择 normal_forward 或是 fake_quant_forward.