# Two lines to see the model's biggest mistakes
interp = Interpretation.from_learner(learn)
interp.plot_top_losses(9)