   Completely reuse the hyperparameter settings from the original text and we have Training data of equivalent magnitude for characters hope you can help us We can provide reasonable consultancy fees