courses
courses copied to clipboard
move model.flatten above so that model compiles
Currently, the model wouldn't compile for sizes other than (224,224) when fine tuning as under the branch if not include_top, the model is not yet flattened but later in the create function, a dense layer is added (without flattening), hence, raising an error. Changing the position of the model.flatten fixes this.
Can you please confirm this works correctly in keras 1?