Did anyone had saved tf tokenzied data on Mlflow and loaded it for prediction in production code? I know there is a method to write as json or pickle but I have no idea how
