tensorplay.ao.quantization.default_dynamic_qconfig

tensorplay.ao.quantization.default_dynamic_qconfig = (functools.partial(<class 'tensorplay.ao.quantization.observer.PlaceholderObserver'>, dtype=tensorplay.quint8, quant_min=0, quant_max=255, is_dynamic=True){}, functools.partial(<class 'tensorplay.ao.quantization.observer.MinMaxObserver'>, dtype=tensorplay.qint8, qscheme=tensorplay.per_tensor_symmetric){})

observer factories for activations and weights.

Both fields must be observer classes or callables returning observer instances. Use MyObserver.with_args(x=1) to override constructor arguments.

Type:

Describes how to quantize a layer