fixed jianfch#454

-fixed repetition of earliest failed word when alignment failure occurs at end of the audio

Author: jianfch

stable_whisper/non_whisper/alignment.py

@@ -847,6 +847,7 @@ def _redo_words(self, index: int = None, ):
             self._remaining_len += sum(len(w.word) for w in self._seg_word_tokens)
             self._all_word_tokens = self._seg_word_tokens + self._all_word_tokens
             self._curr_words = []
+            self._temp_data.word = None
         elif index != len(self._seg_word_tokens):  # redo from _idx
             self._remaining_len += sum(len(w.word) for w in self._seg_word_tokens[index:])
             self._all_word_tokens = self._seg_word_tokens[index:] + self._all_word_tokens