diff --git a/PaddleNLP/neural_machine_translation/transformer/infer.py b/PaddleNLP/neural_machine_translation/transformer/infer.py
index 8a6804e6a496888e503449190ddc74846b1cff6e..9a3bfa2a78e475bd3ea996ae492bc7a08279d22a 100644
--- a/PaddleNLP/neural_machine_translation/transformer/infer.py
+++ b/PaddleNLP/neural_machine_translation/transformer/infer.py
@@ -4,6 +4,9 @@ import multiprocessing
 import numpy as np
 import os
 import sys
+if sys.version[0] == '2':
+    reload(sys)
+    sys.setdefaultencoding("utf-8")
 sys.path.append("../../")
 sys.path.append("../../models/neural_machine_translation/transformer/")
 from functools import partial
@@ -307,13 +310,13 @@ def fast_infer(args):
                     for j in range(end - start):  # for each candidate
                         sub_start = seq_ids.lod()[1][start + j]
                         sub_end = seq_ids.lod()[1][start + j + 1]
-                        hyps[i].append(" ".join([
+                        hyps[i].append(b" ".join([
                             trg_idx2word[idx]
                             for idx in post_process_seq(
                                 np.array(seq_ids)[sub_start:sub_end])
                         ]))
                         scores[i].append(np.array(seq_scores)[sub_end - 1])
-                        print(hyps[i][-1])
+                        print(hyps[i][-1].decode("utf8"))
                         if len(hyps[i]) >= InferTaskConfig.n_best:
                             break
         except (StopIteration, fluid.core.EOFException):
diff --git a/PaddleNLP/neural_machine_translation/transformer/train.py b/PaddleNLP/neural_machine_translation/transformer/train.py
index 09c50e453fd5b5bb7b25a69e65deb83c5e91b62f..0d30178ae98a0fc3cba1a71e7fba596aeb53201f 100644
--- a/PaddleNLP/neural_machine_translation/transformer/train.py
+++ b/PaddleNLP/neural_machine_translation/transformer/train.py
@@ -11,6 +11,9 @@ if os.environ.get('FLAGS_eager_delete_tensor_gb', None) is None:
 
 import six
 import sys
+if sys.version[0] == '2':
+    reload(sys)
+    sys.setdefaultencoding("utf-8")
 sys.path.append("../../")
 sys.path.append("../../models/neural_machine_translation/transformer/")
 import time
diff --git a/PaddleNLP/unarchived/neural_machine_translation/transformer/infer.py b/PaddleNLP/unarchived/neural_machine_translation/transformer/infer.py
index 9f168199b7a89697489c047d0e3b724b187421fc..62e7ce39531c683d3e2a30e51257e0be6e2aa258 100644
--- a/PaddleNLP/unarchived/neural_machine_translation/transformer/infer.py
+++ b/PaddleNLP/unarchived/neural_machine_translation/transformer/infer.py
@@ -3,6 +3,10 @@ import ast
 import multiprocessing
 import numpy as np
 import os
+import sys
+if sys.version[0] == '2':
+    reload(sys)
+    sys.setdefaultencoding("utf-8")
 from functools import partial
 
 import paddle
@@ -303,13 +307,13 @@ def fast_infer(args):
                     for j in range(end - start):  # for each candidate
                         sub_start = seq_ids.lod()[1][start + j]
                         sub_end = seq_ids.lod()[1][start + j + 1]
-                        hyps[i].append(" ".join([
+                        hyps[i].append(b" ".join([
                             trg_idx2word[idx]
                             for idx in post_process_seq(
                                 np.array(seq_ids)[sub_start:sub_end])
                         ]))
                         scores[i].append(np.array(seq_scores)[sub_end - 1])
-                        print(hyps[i][-1])
+                        print(hyps[i][-1].decode("utf8"))
                         if len(hyps[i]) >= InferTaskConfig.n_best:
                             break
         except (StopIteration, fluid.core.EOFException):
diff --git a/PaddleNLP/unarchived/neural_machine_translation/transformer/train.py b/PaddleNLP/unarchived/neural_machine_translation/transformer/train.py
index ade645997c8698c783372cf1cfd7c87a2c67effa..73213de6777455084f1c99d02c014deab1f0bfb4 100644
--- a/PaddleNLP/unarchived/neural_machine_translation/transformer/train.py
+++ b/PaddleNLP/unarchived/neural_machine_translation/transformer/train.py
@@ -6,6 +6,9 @@ import multiprocessing
 import os
 import six
 import sys
+if sys.version[0] == '2':
+    reload(sys)
+    sys.setdefaultencoding("utf-8")
 import time
 
 import numpy as np