Something I don't understand about the structure of alexnet

[ChenDRAG]

nn.Hardtanh(inplace=True),
BinarizeConv2d(int(192*self.ratioInfl), int(384*self.ratioInfl), kernel_size=3, padding=1),

this is a sample code from alexnet binary.py, what i don't understand is since you already binarize the input in

BinarizeConv2d function,

so what is the point of using hardtanh activation?

[masterdezign]

so what is the point of using hardtanh activation?

To approximate sign gradients in the backward pass