Fix typo, move check for set equality

Signed-off-by: Jocelyn Huang <[email protected]>

nemo/collections/common/tokenizers/text_to_speech/tts_tokenizers.py

@@ -562,6 +562,13 @@ def __init__(
         if fixed_vocab:
             tokens = {self.text_preprocessing_func(c) for c in fixed_vocab}
             self.set_fixed_vocab = True  # Used to check whether dataset entries need filtering
+
+            if g2p.symbols == tokens:
+                logging.info(
+                    "Did not replace G2P valid symbol set since the given set is equivalent to the existing one."
+                )
+                self.set_fixed_vocab = False
+                break
             g2p.replace_symbols(tokens)
         else:
             tokens = set(g2p.symbols)

nemo/collections/tts/g2p/modules.py

@@ -549,14 +549,11 @@ def replace_symbols(self, symbols, keep_alternate=True):
         Args:
             symbols (List, Set): User-provided set of valid symbols, both graphemes and phonemes
             keep_alternate (bool): Whether to keep the other pronunciation(s) of a word if not all contain
-                illegal phonemes (and the word doesn't containi illegal graphemes).
+                illegal phonemes (and the word doesn't contain illegal graphemes).
                 Warning: this may change a word from being ambiguous to having only one valid pronunciation.
                 Defaults to True.
         """
         new_symbols = set(symbols)
-        if self.symbols == new_symbols:
-            logging.info("Did not replace G2P valid symbol set since the given set is equivalent to the existing one.")
-            return
 
         # Keep track of what will need to be deleted or (if keep_alternate=True) replaced
         deletion_words = []