fix convert token to id

sentencepiece_ja.py

@@ -41,6 +41,9 @@ class SentencePieceJA(PreTrainedTokenizer):
         return self._tokenizer.encode(text).tokens
 
     def _convert_token_to_id(self, token):
+        ids = self._tokenizer.encode(token).ids
+        if len(ids) == 0:
+            return self.unk_token_id
         return self._tokenizer.encode(token).ids[0]
         
     def _convert_id_to_token(self, index: int) -> str: