PyTorch: Adding hooks to model for saving the intermediate layers output returns the features two times

Question

I want to implement the Style Transfer paper using Pytorch and the VGG19 network. For this, I need the intermediate output features for some layers. I named the convolutional mo