Training (orange) vs test (blue) accuracy on our trained model