Upload bert_tokenizer.py

bert_tokenizer.py

@@ -135,9 +135,9 @@ class ChineseBertTokenizer(BertTokenizerFast):
         pinyin_ids = torch.LongTensor(pinyin_tokens).view(-1)
         return input_ids, pinyin_ids
 
-    def convert_ids_to_pinyin_ids(self, ids: List[str]):
+    def convert_ids_to_pinyin_ids(self, ids: List[int]):
         pinyin_ids = []
-        tokens = self.convert_tokens_to_ids(ids)
+        tokens = self.convert_ids_to_tokens(ids)
         for token in tokens:
             if len(token) > 1:
                 pinyin_ids.append([0] * 8)