You want to get the loss as low as possible withou...
# tacotron-2-support
p
You want to get the loss as low as possible without overfitting the model