Flatten before Sigmoid()?

vision
#1

I want to build classification model with global average pooling.
Should I flatten before using Sigmoid() ?

        self.main = nn.Sequential(
            nn.Conv2d(in_channels=out_channel, out_channels=1, kernel_size=1),
            nn.AvgPool2d(28, 1)
            # flatten -> sigmoid
            # sigmoid
        )
#2

Flatten is an elemebt-wise operation so it doesn’t really matter if you just apply spatial transformations afterwards

#3

As Juan mentioned, you don’t need to flatten before sigmoid. You do need to flatten it so it will fit in the last fully connected layer since you are doing classification