SS refactoring

TTS/tts/layers/speedy_speech/decoder.py

@@ -1,5 +1,6 @@
 from torch import nn
 from TTS.tts.layers.generic.res_conv_bn import ConvBNBlock, ResidualConvBNBlock
+from TTS.tts.layers.generic.wavenet import WNBlocks
 from TTS.tts.layers.glow_tts.transformer import Transformer
 
 
@@ -19,7 +20,7 @@ class Decoder(nn.Module):
             Default decoder_params...
 
             for 'transformer'
-                encoder_params={
+                decoder_params={
                     'hidden_channels_ffn': 128,
                     'num_heads': 2,
                     "kernel_size": 3,
@@ -30,12 +31,22 @@ class Decoder(nn.Module):
                 },
 
             for 'residual_conv_bn'
-                encoder_params = {
+                decoder_params = {
                     "kernel_size": 4,
                     "dilations": 4 * [1, 2, 4, 8] + [1],
                     "num_conv_blocks": 2,
                     "num_res_blocks": 17
                 }
+
+            for 'wavenet'
+                decoder_params = {
+                    "num_blocks": 12,
+                    "hidden_channels":192,
+                    "kernel_size": 5,
+                    "dilation_rate": 1,
+                    "num_layers": 4,
+                    "dropout_p": 0.05
+                }
     """
     # pylint: disable=dangerous-default-value
     def __init__(
@@ -60,6 +71,8 @@ class Decoder(nn.Module):
         elif decoder_type == 'residual_conv_bn':
             self.decoder = ResidualConvBNBlock(self.hidden_channels,
                                                **decoder_params)
+        elif decoder_type == 'wavenet':
+            self.decoder = WNBlocks(in_channels=self.in_channels, hidden_channels=self.hidden_channels, **decoder_params)
         else:
             raise ValueError(f'[!] Unknown decoder type - {decoder_type}')
 

TTS/tts/models/speedy_speech.py

@@ -24,7 +24,7 @@ class SpeedySpeech(nn.Module):
             "num_res_blocks": 13
         },
         decoder_type='residual_conv_bn',
-        decoder_residual_conv_bn_params={
+        decoder_params={
             "kernel_size": 4,
             "dilations": 4 * [1, 2, 4, 8] + [1],
             "num_conv_blocks": 2,
@@ -41,7 +41,7 @@ class SpeedySpeech(nn.Module):
         if positional_encoding:
             self.pos_encoder = PositionalEncoding(hidden_channels)
         self.decoder = Decoder(out_channels, hidden_channels,
-                               decoder_type, decoder_residual_conv_bn_params)
+                               decoder_type, decoder_params)
         self.duration_predictor = DurationPredictor(hidden_channels + c_in_channels)
 
         if num_speakers > 1 and not external_c:

TTS/tts/utils/generic_utils.py

@@ -117,22 +117,20 @@ def setup_model(num_chars, num_speakers, c, speaker_embedding_dim=None):
                         num_speakers=num_speakers,
                         c_in_channels=0,
                         num_splits=4,
-                        num_sqz=2,
+                        num_squeeze=2,
                         sigmoid_scale=False,
                         mean_only=True,
-                        hidden_channels_enc=192,
-                        hidden_channels_dec=192,
-                        use_encoder_prenet=True,
+                        use_encoder_prenet=c["use_encoder_prenet"],
                         external_speaker_embedding_dim=speaker_embedding_dim)
     elif c.model.lower() == "speedy_speech":
         model = MyModel(num_chars=num_chars + getattr(c, "add_blank", False),
                         out_channels=c.audio['num_mels'],
-                        hidden_channels=128,
+                        hidden_channels=c['hidden_channels'],
                         positional_encoding=c['positional_encoding'],
                         encoder_type=c['encoder_type'],
                         encoder_params=c['encoder_params'],
                         decoder_type=c['decoder_type'],
-                        decoder_residual_conv_bn_params=c['decoder_residual_conv_bn_params'],
+                        decoder_params=c['decoder_params'],
                         c_in_channels=0)
     return model
 

TTS/utils/io.py

@@ -57,7 +57,7 @@ def copy_model_files(c, config_file, out_path, new_fields):
             in the config file.
     """
     # copy config.json
-    copy_config_path = OUT_PATH
+    copy_config_path = os.path.join(out_path, 'config.json')
     config_lines = open(config_file, "r").readlines()
     # add extra information fields
     for key, value in new_fields.items():
@@ -70,5 +70,6 @@ def copy_model_files(c, config_file, out_path, new_fields):
     config_out_file.writelines(config_lines)
     config_out_file.close()
     # copy model stats file if available
-    copy_stats_path = os.path.join(out_path, 'scale_stats.npy')
-    copyfile(c.audio['stats_path'], copy_stats_path)
+    if c.audio['stats_path'] is not None:
+        copy_stats_path = os.path.join(out_path, 'scale_stats.npy')
+        copyfile(c.audio['stats_path'], copy_stats_path)