diff --git a/research/nlp/senta/src/data/field_reader/ernie_text_field_reader.py b/research/nlp/senta/src/data/field_reader/ernie_text_field_reader.py
index 91c20a0114b5f34f25a7029c567604909ab3c890..faa24672b7baf99a7bbc0a961c3220cfe2932ef7 100644
--- a/research/nlp/senta/src/data/field_reader/ernie_text_field_reader.py
+++ b/research/nlp/senta/src/data/field_reader/ernie_text_field_reader.py
@@ -89,7 +89,7 @@ class ErnieTextFieldReader():
         padded_ids, input_mask = pad_batch_data(src_ids,
                                                 pad_idx=self.field_config.padding_id,
                                                 return_input_mask=True,
-                                                return_seq_lens=True)
+                                                return_seq_lens=False)
         sent_ids_batch = pad_batch_data(sentence_ids, pad_idx=self.field_config.padding_id)
         pos_ids_batch = pad_batch_data(position_ids, pad_idx=self.field_config.padding_id)
 
diff --git a/research/nlp/senta/src/data/tokenizer/tokenization_utils.py b/research/nlp/senta/src/data/tokenizer/tokenization_utils.py
index 3c4ca7a859a878bfbe1d15f6f49636f684030c95..e07c91610a9d54aefa63ec1579b4ae70d6b84dac 100644
--- a/research/nlp/senta/src/data/tokenizer/tokenization_utils.py
+++ b/research/nlp/senta/src/data/tokenizer/tokenization_utils.py
@@ -136,7 +136,7 @@ class BpeEncoder():
                     j = word.index(first, i)
                     new_word.extend(word[i:j])
                     i = j
-                except IOError:
+                except ValueError:
                     new_word.extend(word[i:])
                     break