Is it possible to fine-tune one of your pretrained models?

I've only seen the training example with which we can create a new model from scratch with our training data. What if we wanted to make one of the pretrained models stronger to our domain data?