update vocoder loss implemenatations and fix MSEDLoss

vocoder/layers/losses.py

@@ -88,7 +88,7 @@ class MSEGLoss(nn.Module):
     """ Mean Squared Generator Loss """
     # pylint: disable=no-self-use
     def forward(self, score_fake):
-        loss_fake = torch.mean(torch.sum(torch.pow(score_fake, 2), dim=[1, 2]))
+        loss_fake = F.mse_loss(score_fake, score_fake.new_ones(score_fake.shape))
         return loss_fake
 
 
@@ -96,7 +96,8 @@ class HingeGLoss(nn.Module):
     """ Hinge Discriminator Loss """
     # pylint: disable=no-self-use
     def forward(self, score_fake):
-        loss_fake = torch.mean(F.relu(1. + score_fake))
+        # TODO: this might be wrong
+        loss_fake = torch.mean(F.relu(1. - score_fake))
         return loss_fake
 
 
@@ -107,10 +108,14 @@ class HingeGLoss(nn.Module):
 
 class MSEDLoss(nn.Module):
     """ Mean Squared Discriminator Loss """
+    def __init__(self,):
+        super(MSEDLoss, self).__init__()
+        self.loss_func = nn.MSELoss()
+
     # pylint: disable=no-self-use
     def forward(self, score_fake, score_real):
-        loss_real = torch.mean(torch.sum(torch.pow(score_real - 1.0, 2), dim=[1, 2]))
-        loss_fake = torch.mean(torch.sum(torch.pow(score_fake, 2), dim=[1, 2]))
+        loss_real = self.loss_func(score_real, score_real.new_ones(score_real.shape))
+        loss_fake = self.loss_func(score_fake, score_fake.new_zeros(score_fake.shape))
         loss_d = loss_real + loss_fake
         return loss_d, loss_real, loss_fake
 
@@ -126,11 +131,15 @@ class HingeDLoss(nn.Module):
 
 
 class MelganFeatureLoss(nn.Module):
+    def __init__(self,):
+        super(MelganFeatureLoss, self).__init__()
+        self.loss_func = nn.L1Loss()
+
     # pylint: disable=no-self-use
     def forward(self, fake_feats, real_feats):
         loss_feats = 0
         for fake_feat, real_feat in zip(fake_feats, real_feats):
-            loss_feats += torch.mean(torch.abs(fake_feat - real_feat))
+            loss_feats += self.loss_func(fake_feats, real_feats)
         loss_feats /= len(fake_feats) + len(real_feats)
         return loss_feats