pytorch自定义二值化网络层

任务要求:

自定义一个层主要是定义该层的实现函数,只需要重载Function的forward和backward函数即可,如下:

import torch
from torch.autograd import Function
from torch.autograd import Variable

定义二值化函数

class BinarizedF(Function):
    def forward(self, input):
        self.save_for_backward(input)
        a = torch.ones_like(input)
        b = -torch.ones_like(input)
        output = torch.where(input>=0,a,b)
        return output
    def backward(self, output_grad):
        input, = self.saved_tensors
        input_abs = torch.abs(input)
        ones = torch.ones_like(input)
        zeros = torch.zeros_like(input)
        input_grad = torch.where(input_abs<=1,ones, zeros)
        return input_grad

定义一个module

class BinarizedModule(nn.Module):
    def __init__(self):
        super(BinarizedModule, self).__init__()
        self.BF = BinarizedF()
    def forward(self,input):
        print(input.shape)
        output =self.BF(input)
        return output

进行测试

a = Variable(torch.randn(4,480,640), requires_grad=True)
output = BinarizedModule()(a)
output.backward(torch.ones(a.size()))
print(a)
print(a.grad)

其中,二值化函数部分也可以按照方式写,但是速度慢了0.05s

class BinarizedF(Function):
    def forward(self, input):
        self.save_for_backward(input)
        output = torch.ones_like(input)
        output[input<0] = -1
        return output
    def backward(self, output_grad):
        input, = self.saved_tensors
        input_grad = output_grad.clone()
        input_abs = torch.abs(input)
        input_grad[input_abs>1] = 0
        return input_grad