torch.ao.quantization.observer.default_float_qparams_observer_4bit

torch.ao.quantization.observer.default_float_qparams_observer_4bit = functools.partial(<class 'torch.ao.quantization.observer.PerChannelMinMaxObserver'>, dtype=torch.quint4x2, qscheme=torch.per_channel_affine_float_qparams, ch_axis=0){}

Default observer for a floating point zero-point and 4 bit activations.

上一页

torch.ao.quantization.observer.default_float_qparams_observer

下一页

torch.ao.quantization.observer.default_observer