When applying Transfer Learning to a pre-trained network in PyTorch, how do you "freeze" the feature extraction layers so their weights remain untouched during optimization?

Question

Answer

By setting the learning rate of the optimizer to zero for the entire model.

Answer

By calling  requires_grad_(False)  on the feature extractor module or its parameters.

Answer

By moving the feature extraction layers to a different device than the classifier.

Answer

By re-initializing the weights of the feature extractor using  nn.init.kaiming_uniform_ .

Answer

By calling  network.eval()  before passing data through the feature extractor.

Crowdly

When applying Transfer Learning to a pre-trained network in PyTorch, how do you ...

Want instant access to all verified answers on moodle.jku.at?