some questions about vocoder

[LqNoob]

Hi, @MoonInTheRiver, thank you for your great work. I have some questions about the training of vocoder as follows:

1. In the training phase, did you use the same loss function as the HiFi-GAN vocoder model's loss function? That is, did you use other loss functions besides feature_loss, discriminator_loss and generator_loss?
2. During the forward computation of the Discriminator, are there mel features input in addition to the real and predicted audio?