dev-sungman / covid-challenge Goto Github PK

class Conv2d(nn.Conv2d):
    def __init__(self, in_channels, output_channels, kernel_size, stride=1,
                    padding=0, dilation=1, groups=1, bias=True):
        super(Conv2d, self).__init__(in_channels, output_channels, kernel_size, stride, padding,
                                    dilation, groups, bias)
    
    def forward(self, x):
        weight = self.weight
        weight_mean = weight.mean(dim=1, keepdim=True).mean(dim=2, keepdim=True).mean(dim=3, keepdim=True)
        weight = weight - weight_mean
        std = weight.view(weight.size(0), -1).std(dim=1).view(-1,1,1,1) + 1e-5
        weight = weight / std.expand_as(weight)
        return F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups) <--------

import torch.nn.functional