Merge branch 'larger-attn-larger-model-sigmoid' into attn-smoothing-bgs-sigmoid-wd

config.json

@@ -1,6 +1,6 @@
 {
-    "model_name": "TTS-weight-decay",
-    "model_description": "Weight decay as in FastAI",
+    "model_name": "TTS-sigmoid",
+    "model_description": "Net outputting Sigmoid unit",
     "audio_processor": "audio",
     "num_mels": 80,
     "num_freq": 1025,

layers/tacotron.py

@@ -374,6 +374,7 @@ class Decoder(nn.Module):
             decoder_output = decoder_input
             # predict mel vectors from decoder vectors
             output = self.proj_to_mel(decoder_output)
+            output = torch.sigmoid(output)
             stop_input = output
             # predict stop token
             stop_token, stopnet_rnn_hidden = self.stopnet(

models/tacotron.py

@@ -23,7 +23,9 @@ class Tacotron(nn.Module):
         self.encoder = Encoder(embedding_dim)
         self.decoder = Decoder(256, mel_dim, r)
         self.postnet = PostCBHG(mel_dim)
-        self.last_linear = nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim)
+        self.last_linear = nn.Sequential(
+            nn.Linear(self.postnet.cbhg.gru_features * 2, linear_dim),
+            nn.Sigmoid())
 
     def forward(self, characters, mel_specs=None, mask=None):
         B = characters.size(0)