unknown token fix

speechbrain · Jul 3, 2023 · 2905a52 · 2905a52
1 parent 9937302
commit 2905a52
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 0 deletions.
diff --git a/recipes/LJSpeech/TTS/fastspeech2/train.py b/recipes/LJSpeech/TTS/fastspeech2/train.py
@@ -473,6 +473,7 @@ def dataio_prepare(hparams):
     # add a dummy symbol for idx 0 - used for padding.
     lexicon = ["@@"] + lexicon
     input_encoder.update_from_iterable(lexicon, sequence_input=False)
+    input_encoder.add_unk()
 
     # load audio, text and durations on the fly; encode audio and text.
     @sb.utils.data_pipeline.takes(
@@ -502,6 +503,7 @@ def audio_pipeline(
         label_phoneme = label_phoneme.strip()
         label_phoneme = label_phoneme.split()
         text_seq = input_encoder.encode_sequence_torch(label_phoneme).int()
+
         assert len(text_seq) == len(
             durs
         ), f"{len(text_seq)}, {len(durs), len(label_phoneme)}, ({label_phoneme})"  # ensure every token has a duration

diff --git a/speechbrain/pretrained/interfaces.py b/speechbrain/pretrained/interfaces.py
@@ -2858,6 +2858,8 @@ def __init__(self, *args, **kwargs):
         lexicon = ["@@"] + lexicon
         self.input_encoder = self.hparams.input_encoder
         self.input_encoder.update_from_iterable(lexicon, sequence_input=False)
+        self.input_encoder.add_unk()
+
         self.g2p = GraphemeToPhoneme.from_hparams("speechbrain/soundchoice-g2p")
 
         self.spn_token_encoded = (