#qconfig

#horizon_plugin_pytorch.quantization.get_qconfig(observer: ~typing.Type[~horizon_plugin_pytorch.quantization.observer_v2.ObserverBase] = <class 'horizon_plugin_pytorch.quantization.observer_v2.MinMaxObserver'>, in_dtype: ~torch.dtype | ~horizon_plugin_pytorch.dtype.QuantDType | None = None, weight_dtype: ~torch.dtype | ~horizon_plugin_pytorch.dtype.QuantDType | None = 'qint8', out_dtype: ~torch.dtype | ~horizon_plugin_pytorch.dtype.QuantDType | None = 'qint8', fix_scale: bool = False)

Get qconfig.

• Parameters:
  • observer – observer type for input and output. Support MinMaxObserver and MSEObserver
  • in_dtype – input dtype.
  • weight_dtype – weight dtype.
  • out_dtype – output dtype.
  • fix_scale – Whether fix input/output scale.