Converting Con2d result to Linear

vision
1

I’m trying to implement a convolutional autoencoder and when I flatten the tensor resulting from Con2D layer to feed into the Linear layer, I realize that the tensor is not from one but many samples. What is the correct way to make this conversion?

2

Hi @uhotspot4,

I believe that’s to be expected if you feed the network in batch! So flatten the tensor resulting from Conv2D by batch too. You can do this using .view(-1, <the expected value of flattened conv2d>). The -1 means it will automatically adjust to your batch size.

Take a look at this sample code:

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))

        x = x.view(-1, 320) # this is the flatten part

        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x)

Please correct me if I’m wrong :smiley: