readme and cleanup

README.md

@@ -32,7 +32,7 @@ duration_predictor = DurationPredictor(
     )
 )
 
-mel = torch.randn(2, 1024, 80)
+mel = torch.randn(2, 1024, 100)
 text = ['Hello', 'Goodbye']
 
 loss = duration_predictor(mel, text = text)

e2_tts_pytorch/e2_tts.py

@@ -168,10 +168,8 @@ def forward(
             return x
 
         max_seq_len = x.shape[1]
-        text_mask = text == -1
 
-        text = text + 1 # use 0 as filler token
-        text = text.masked_fill(text_mask, 0)
+        text = text + 1 # shift all other token ids up by 1 and use 0 as filler token
 
         text = text[:, :max_seq_len] # just curtail if character tokens are more than the mel spec tokens, one of the edge cases the paper did not address
         text = F.pad(text, (0, max_seq_len - text.shape[1]), value = 0)

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "e2-tts-pytorch"
-version = "0.1.0"
+version = "0.1.1"
 description = "E2-TTS in Pytorch"
 authors = [
     { name = "Phil Wang", email = "[email protected]" }