1) Add ExponentialLR

TTS/vocoder/configs/modified_hifigan.json

@@ -112,11 +112,13 @@
     "disc_clip_grad": -1,     // Discriminator gradient clipping threshold.
     "lr_scheduler_gen": "ExponentialLR",   // one of the schedulers from https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
     "lr_scheduler_gen_params": {
-       "gamma": 0.999
+       "gamma": 0.999,
+        "last_epoch": -1
     },
     "lr_scheduler_disc": "ExponentialLR",   // one of the schedulers from https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
     "lr_scheduler_disc_params": {
-    	  "gamma": 0.999
+    	  "gamma": 0.999,
+        "last_epoch": -1
     },
     "lr_gen": 0.0002,                  // Initial learning rate. If Noam decay is active, maximum learning rate.
     "lr_disc": 0.0002,

TTS/vocoder/models/hifigan_generator.py

@@ -1,3 +1,4 @@
+import torch
 from torch import nn
 from TTS.vocoder.layers.hifigan import MRF
 
@@ -7,6 +8,9 @@ class Generator(nn.Module):
     def __init__(self, in_channels=80, out_channels=1, base_channels=512, upsample_kernel=[16, 16, 4, 4],
                  resblock_kernel_sizes=[3, 7, 11], resblock_dilation_sizes=[1, 3, 5]):
         super(Generator, self).__init__()
+
+        self.inference_padding = 2
+
         self.input = nn.Sequential(
             nn.ReflectionPad1d(3),
             nn.utils.weight_norm(nn.Conv1d(in_channels, base_channels, kernel_size=7))
@@ -39,6 +43,14 @@ class Generator(nn.Module):
         out = self.output(x2)
         return out
 
+    def inference(self, c):
+        c = c.to(self.layers[1].weight.device)
+        c = torch.nn.functional.pad(
+            c,
+            (self.inference_padding, self.inference_padding),
+            'replicate')
+        return self.forward(c)
+
     def remove_weight_norm(self):
         nn.utils.remove_weight_norm(self.input[1])
         nn.utils.remove_weight_norm(self.output[2])
@@ -48,4 +60,12 @@ class Generator(nn.Module):
                 try:
                     nn.utils.remove_weight_norm(layer)
                 except:
-                    layer.remove_weight_norm()
+                    layer.remove_weight_norm()
+
+    def load_checkpoint(self, config, checkpoint_path, eval=False):  # pylint: disable=unused-argument, redefined-builtin
+        state = torch.load(checkpoint_path, map_location=torch.device('cpu'))
+        self.load_state_dict(state['model'])
+        if eval:
+            self.eval()
+            assert not self.training
+            self.remove_weight_norm()