[MRG] voice generator refactor (#460)

* vg

* temp

* fixed

autokeras/__init__.py

@@ -1,6 +1,6 @@
 from autokeras.image.image_supervised import ImageClassifier, ImageRegressor, PortableImageSupervised
 from autokeras.text.text_supervised import TextClassifier, TextRegressor
-from autokeras.tabular.tabular_supervised import TabularClassifier, TabularRegressor
+# from autokeras.tabular.tabular_supervised import TabularClassifier, TabularRegressor
 
 from autokeras.net_module import CnnModule, MlpModule
 

autokeras/pretrained/voice_generator/deepvoice3_pytorch/builder.py

@@ -1,88 +1,48 @@
 from autokeras.pretrained.voice_generator.deepvoice3_pytorch.model import MultiSpeakerTTSModel, AttentionSeq2Seq
 
 
-def deepvoice3(n_vocab, embed_dim=256, mel_dim=80, linear_dim=513, r=4,
-               downsample_step=1,
-               n_speakers=1, speaker_embed_dim=16, padding_idx=0,
-               dropout=(1 - 0.95), kernel_size=5,
-               encoder_channels=128,
-               decoder_channels=256,
-               converter_channels=256,
-               query_position_rate=1.0,
-               key_position_rate=1.29,
-               use_memory_mask=False,
-               trainable_positional_encodings=False,
-               force_monotonic_attention=True,
-               use_decoder_state_for_postnet_input=True,
-               max_positions=512,
-               embedding_weight_std=0.1,
-               speaker_embedding_weight_std=0.01,
-               freeze_embedding=False,
-               window_ahead=3,
-               window_backward=1,
-               key_projection=False,
-               value_projection=False,
-               ):
+def deepvoice3(n_vocab, embed_dim=256, mel_dim=80, linear_dim=513, r=4, n_speakers=1, speaker_embed_dim=16,
+               padding_idx=0, dropout=(1 - 0.95), kernel_size=5, encoder_channels=128, decoder_channels=256,
+               converter_channels=256, query_position_rate=1.0, key_position_rate=1.29, use_memory_mask=False,
+               trainable_positional_encodings=False, force_monotonic_attention=True,
+               use_decoder_state_for_postnet_input=True, max_positions=512, embedding_weight_std=0.1,
+               freeze_embedding=False, window_ahead=3, window_backward=1):
     """Build deepvoice3
     """
     from autokeras.pretrained.voice_generator.deepvoice3_pytorch.deepvoice3 import Encoder, Decoder, Converter
 
-    time_upsampling = max(downsample_step // r, 1)
-
     # Seq2seq
     h = encoder_channels  # hidden dim (channels)
     k = kernel_size  # kernel size
-    encoder = Encoder(
-        n_vocab, embed_dim, padding_idx=padding_idx,
-        n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim,
-        dropout=dropout, max_positions=max_positions,
-        embedding_weight_std=embedding_weight_std,
-        # (channels, kernel_size, dilation)
-        convolutions=[(h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
-                      (h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
-                      (h, k, 1), (h, k, 3)],
-    )
+    encoder = Encoder(n_vocab, embed_dim, n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim,
+                      padding_idx=padding_idx, embedding_weight_std=embedding_weight_std,
+                      convolutions=[(h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
+                                    (h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
+                                    (h, k, 1), (h, k, 3)], dropout=dropout)
 
     h = decoder_channels
-    decoder = Decoder(
-        embed_dim, in_dim=mel_dim, r=r, padding_idx=padding_idx,
-        n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim,
-        dropout=dropout, max_positions=max_positions,
-        preattention=[(h, k, 1), (h, k, 3)],
-        convolutions=[(h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
-                      (h, k, 1)],
-        attention=[True, False, False, False, True],
-        force_monotonic_attention=force_monotonic_attention,
-        query_position_rate=query_position_rate,
-        key_position_rate=key_position_rate,
-        use_memory_mask=use_memory_mask,
-        window_ahead=window_ahead,
-        window_backward=window_backward,
-        key_projection=key_projection,
-        value_projection=value_projection,
-    )
+    decoder = Decoder(embed_dim, n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim, in_dim=mel_dim, r=r,
+                      max_positions=max_positions, preattention=[(h, k, 1), (h, k, 3)],
+                      convolutions=[(h, k, 1), (h, k, 3), (h, k, 9), (h, k, 27),
+                                    (h, k, 1)], attention=[True, False, False, False, True], dropout=dropout,
+                      use_memory_mask=use_memory_mask, force_monotonic_attention=force_monotonic_attention,
+                      query_position_rate=query_position_rate, key_position_rate=key_position_rate,
+                      window_ahead=window_ahead, window_backward=window_backward)
 
     seq2seq = AttentionSeq2Seq(encoder, decoder)
 
     # Post net
     in_dim = h // r
 
     h = converter_channels
-    converter = Converter(
-        n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim,
-        in_dim=in_dim, out_dim=linear_dim, dropout=dropout,
-        time_upsampling=time_upsampling,
-        convolutions=[(h, k, 1), (h, k, 3), (2 * h, k, 1), (2 * h, k, 3)],
-    )
+    converter = Converter(n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim, in_dim=in_dim, out_dim=linear_dim,
+                          convolutions=[(h, k, 1), (h, k, 3), (2 * h, k, 1), (2 * h, k, 3)], dropout=dropout)
 
     # Seq2seq + post net
-    model = MultiSpeakerTTSModel(
-        seq2seq, converter, padding_idx=padding_idx,
-        mel_dim=mel_dim, linear_dim=linear_dim,
-        n_speakers=n_speakers, speaker_embed_dim=speaker_embed_dim,
-        trainable_positional_encodings=trainable_positional_encodings,
-        use_decoder_state_for_postnet_input=use_decoder_state_for_postnet_input,
-        speaker_embedding_weight_std=speaker_embedding_weight_std,
-        freeze_embedding=freeze_embedding)
+    model = MultiSpeakerTTSModel(seq2seq, converter, mel_dim=mel_dim, linear_dim=linear_dim, n_speakers=n_speakers,
+                                 speaker_embed_dim=speaker_embed_dim,
+                                 trainable_positional_encodings=trainable_positional_encodings,
+                                 use_decoder_state_for_postnet_input=use_decoder_state_for_postnet_input,
+                                 freeze_embedding=freeze_embedding)
 
     return model