ConvLearner Layers for Pre-trained Models

It was mentioned that the ConvLearner adds a few layers to the pre-trained model and fine tunes. How to know how many layers and how to adjust the number of added layers? Also is there a way to adjust say stride or dilation in the new layers that are added?

Generally the defaults will be fine the vast majority of the time.

We’ll learn about more advanced customizations later. If you’d like to skip ahead, #fastai-users is the place to discuss library issues that we haven’t covered in the course yet.

Ok thanks