How can I train a part of the network?

vision
1 
class Network(nn.Module):
    self.network1()
    self.network2()

    def forward(self, x):
        x = self.network1(x)
        y = self.network2(x)
        return x, y

class Classifier(nn.Module):
   self.classifier()
   ...

What I want to do is

step1) 
when Epoch < 10:
         train classifier with whole Network()
step2)
when Epoch >= 10:
        train classifier with only network1() in Network()

How can I freeze(?) the network2 in step2 ?

2

You can set the requires_grad flag of network2 to False after 10 epochs.