remove attention mask

config.json

@@ -12,20 +12,20 @@
   "text_cleaner": "english_cleaners",
 
   "epochs": 2000,
-  "lr": 0.00001875,
+  "lr": 0.001,
   "warmup_steps": 4000,
-  "batch_size": 2,
+  "batch_size": 32,
   "eval_batch_size": 32,
   "r": 5,
 
   "griffin_lim_iters": 60,
   "power": 1.5,
 
-  "num_loader_workers": 16,
+  "num_loader_workers": 12,
 
   "checkpoint": false,
   "save_step": 69,
   "data_path": "/run/shm/erogol/LJSpeech-1.0",
-  "min_seq_len": 90, 
+  "min_seq_len": 0, 
   "output_path": "result"
 }

layers/tacotron.py

@@ -231,8 +231,8 @@ class Decoder(nn.Module):
         # RNN_state -> |Linear| -> mel_spec
         self.proj_to_mel = nn.Linear(256, memory_dim * r)
 
-    def forward(self, inputs, memory=None, input_lengths=None):
+    def forward(self, inputs, memory=None):
-        r"""
+        """
         Decoder forward step.
 
         If decoder inputs are not given (e.g., at testing time), as noted in
@@ -242,8 +242,6 @@ class Decoder(nn.Module):
             inputs: Encoder outputs.
             memory (None): Decoder memory (autoregression. If None (at eval-time),
               decoder outputs are used as decoder inputs.
-            input_lengths (None): input lengths, used for
-              attention masking.
 
         Shapes:
             - inputs: batch x time x encoder_out_dim
@@ -251,12 +249,6 @@ class Decoder(nn.Module):
         """
         B = inputs.size(0)
 
-        
-        # if input_lengths is not None:
-            # mask = get_mask_from_lengths(processed_inputs, input_lengths)
-        # else:
-            # mask = None
-
         # Run greedy decoding if memory is None
         greedy = memory is None
 

models/tacotron.py

@@ -8,12 +8,11 @@ from TTS.layers.tacotron import Prenet, Encoder, Decoder, CBHG
 
 class Tacotron(nn.Module):
     def __init__(self, embedding_dim=256, linear_dim=1025, mel_dim=80,
-                 freq_dim=1025, r=5, padding_idx=None,
+                 freq_dim=1025, r=5, padding_idx=None):
-                 use_atten_mask=False):
+                 
         super(Tacotron, self).__init__()
         self.mel_dim = mel_dim
         self.linear_dim = linear_dim
-        self.use_atten_mask = use_atten_mask
         self.embedding = nn.Embedding(len(symbols), embedding_dim,
                                       padding_idx=padding_idx)
         print(" | > Embedding dim : {}".format(len(symbols)))
@@ -26,16 +25,13 @@ class Tacotron(nn.Module):
         self.postnet = CBHG(mel_dim, K=8, projections=[256, mel_dim])
         self.last_linear = nn.Linear(mel_dim * 2, freq_dim)
 
-    def forward(self, characters, mel_specs=None, input_lengths=None):
+    def forward(self, characters, mel_specs=None):
         B = characters.size(0)
 
         inputs = self.embedding(characters)
         # (B, T', in_dim)
         encoder_outputs = self.encoder(inputs)
 
-        if not self.use_atten_mask:
-            input_lengths = None
-
         # (B, T', mel_dim*r)
         mel_outputs, alignments = self.decoder(
             encoder_outputs, mel_specs, input_lengths=input_lengths)

train.py

@@ -112,8 +112,7 @@ def train(model, criterion, data_loader, optimizer, epoch):
 
         # forward pass
         mel_output, linear_output, alignments =\
-            model.forward(text_input_var, mel_spec_var,
+            model.forward(text_input_var, mel_spec_var)
-                          input_lengths= torch.autograd.Variable(torch.cuda.LongTensor(sorted_lengths)))
         
         # loss computation
         mel_loss = criterion(mel_output, mel_spec_var)
@@ -337,8 +336,7 @@ def main(args):
                      c.hidden_size,
                      c.num_mels,
                      c.num_freq,
-                     c.r,
+                     c.r)
-                     use_atten_mask=True)
                      
     optimizer = optim.Adam(model.parameters(), lr=c.lr)