convinception.py

class conv_block(nn.Module):
    def __init__(self, in_channels, out_channels, **kwargs):
        super(conv_block, self).__init__()
        self.relu = nn.ReLU()
        self.conv = nn.Conv2d(in_channels, out_channels, **kwargs)
        self.batchnorm = nn.BatchNorm2d(out_channels)
        
    def forward(self, x):
        return self.relu(self.batchnorm(self.conv(x)))