tensorplay.ao.quantization.default_weight_only_qconfig

tensorplay.ao.quantization.default_weight_only_qconfig = (<class 'tensorplay.nn.modules.linear.Identity'>, functools.partial(<class 'tensorplay.ao.quantization.fake_quantize.FakeQuantize'>, dtype=tensorplay.qint8, qscheme=tensorplay.per_tensor_symmetric, quant_min=-128, quant_max=127){})

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