Change dilation size, channel numbers in a trained model?

Hi, I have a question.
If I have a trained model, say VGG, is there a way for me to change the output channel numbers, dilation sizes in a certain convolution layer?