Open chigkim opened 1 year ago
Is there a recipe to fine tune with freezing beginning layers and only train the last few layers with smaller learning rate? If I just use train_yourtts.py with setting RESTORE_PATH, doesn't it train the all the layers?
Is there a recipe to fine tune with freezing beginning layers and only train the last few layers with smaller learning rate? If I just use train_yourtts.py with setting RESTORE_PATH, doesn't it train the all the layers?