You are correct. If you create a WideResnet Pytorch model, then you should be able to input that into the Learner class like I did in the example in my last post. However, you will also (probably) have to create a custom dataset class which you can then instantiate with a DataBunch fastai class.
To say that all in a different way, if you have your pytorch WideResnet model in a variable named model and you create a DataBunch class instance called data with your custom dataset. Then you should be able to call
learner = Learner(data, model, loss_func=..., metrics=[...])
and have a working fastai learner which supports the .fit, .find_lr, etc. Someone can (please) correct me if I’m wrong about any of what I said, or chime in if there is an easier way to do this.