VITS avg_loss_0 and avg_loss_disc are identical

[MedAymenF]

I was training a VITS model when I noticed that the avg_loss_0 and avg_loss_disc loss curves are identical. Is this normal? and if it is, is it not redundant to have both?
Here's another training run showing the same thing.

[eginhard]

Yeah, they're always equal:

coqui-ai-TTS/TTS/tts/layers/losses.py

Lines 768 to 775 in 9035e36

	loss = 0.0
	return_dict = {}
	loss_disc, loss_disc_real, _ = self.discriminator_loss(
	scores_real=scores_disc_real, scores_fake=scores_disc_fake
	)
	return_dict["loss_disc"] = loss_disc * self.disc_loss_alpha
	loss = loss + return_dict["loss_disc"]
	return_dict["loss"] = loss

Having it named loss_disc helps to tell what the loss means. I'm not sure if the other one could be safely removed or if that's a standard name that Coqui expects to be there.

[MedAymenF]

Thanks.