diff --git a/ppstructure/vqa/eval_ser.py b/ppstructure/vqa/eval_ser.py
index e56aa27c81862be16268694a6d3f0da43a103fef..e0612219504aecc01aa4ccae3e3a499ff3d0c639 100644
--- a/ppstructure/vqa/eval_ser.py
+++ b/ppstructure/vqa/eval_ser.py
@@ -128,12 +128,16 @@ def evaluate(args,
         "f1": f1_score(out_label_list, preds_list),
     }
 
-    with open(os.path.join(args.output_dir, "test_gt.txt"), "w") as fout:
+    with open(
+            os.path.join(args.output_dir, "test_gt.txt"), "w",
+            encoding='utf-8') as fout:
         for lbl in out_label_list:
             for l in lbl:
                 fout.write(l + "\t")
             fout.write("\n")
-    with open(os.path.join(args.output_dir, "test_pred.txt"), "w") as fout:
+    with open(
+            os.path.join(args.output_dir, "test_pred.txt"), "w",
+            encoding='utf-8') as fout:
         for lbl in preds_list:
             for l in lbl:
                 fout.write(l + "\t")
diff --git a/ppstructure/vqa/helper/eval_with_label_end2end.py b/ppstructure/vqa/helper/eval_with_label_end2end.py
index a983b080161fbc349cbb199cbe2cb9d5b4537b2f..3aa439acb269d74165543fac7e0042cfc213f08d 100644
--- a/ppstructure/vqa/helper/eval_with_label_end2end.py
+++ b/ppstructure/vqa/helper/eval_with_label_end2end.py
@@ -37,7 +37,7 @@ def parse_ser_results_fp(fp, fp_type="gt", ignore_background=True):
     assert fp_type in ["gt", "pred"]
     key = "label" if fp_type == "gt" else "pred"
     res_dict = dict()
-    with open(fp, "r") as fin:
+    with open(fp, "r", encoding='utf-8') as fin:
         lines = fin.readlines()
 
     for _, line in enumerate(lines):
diff --git a/ppstructure/vqa/helper/trans_xfun_data.py b/ppstructure/vqa/helper/trans_xfun_data.py
index b5ebd5dfbd8addda0701a7cfd2387133f7a8776b..25b3963d8362d28ea1df4c62d1491095b8c49253 100644
--- a/ppstructure/vqa/helper/trans_xfun_data.py
+++ b/ppstructure/vqa/helper/trans_xfun_data.py
@@ -16,13 +16,13 @@ import json
 
 
 def transfer_xfun_data(json_path=None, output_file=None):
-    with open(json_path, "r") as fin:
+    with open(json_path, "r", encoding='utf-8') as fin:
         lines = fin.readlines()
 
     json_info = json.loads(lines[0])
     documents = json_info["documents"]
     label_info = {}
-    with open(output_file, "w") as fout:
+    with open(output_file, "w", encoding='utf-8') as fout:
         for idx, document in enumerate(documents):
             img_info = document["img"]
             document = document["document"]
diff --git a/ppstructure/vqa/infer_re.py b/ppstructure/vqa/infer_re.py
index ae2f52550294b072179c3bdba28c3572369e11a3..139ade9aae3a430eb7f9c9b5cb0ec30da465ca6e 100644
--- a/ppstructure/vqa/infer_re.py
+++ b/ppstructure/vqa/infer_re.py
@@ -92,7 +92,7 @@ def infer(args):
 def load_ocr(img_folder, json_path):
     import json
     d = []
-    with open(json_path, "r") as fin:
+    with open(json_path, "r", encoding='utf-8') as fin:
         lines = fin.readlines()
         for line in lines:
             image_name, info_str = line.split("\t")
diff --git a/ppstructure/vqa/infer_ser.py b/ppstructure/vqa/infer_ser.py
index 4ad220094a26b330555fbe9122a46fb56e64fe1e..3c7ad03aa362971ab80aa63e92404942b136ca31 100644
--- a/ppstructure/vqa/infer_ser.py
+++ b/ppstructure/vqa/infer_ser.py
@@ -59,7 +59,8 @@ def pad_sentences(tokenizer,
             encoded_inputs["bbox"] = encoded_inputs["bbox"] + [[0, 0, 0, 0]
                                                                ] * difference
         else:
-            assert False, f"padding_side of tokenizer just supports [\"right\"] but got {tokenizer.padding_side}"
+            assert False, "padding_side of tokenizer just supports [\"right\"] but got {}".format(
+                tokenizer.padding_side)
     else:
         if return_attention_mask:
             encoded_inputs["attention_mask"] = [1] * len(encoded_inputs[
@@ -224,7 +225,7 @@ def infer(args):
 
     # load ocr results json
     ocr_results = dict()
-    with open(args.ocr_json_path, "r") as fin:
+    with open(args.ocr_json_path, "r", encoding='utf-8') as fin:
         lines = fin.readlines()
         for line in lines:
             img_name, json_info = line.split("\t")
@@ -234,7 +235,10 @@ def infer(args):
     infer_imgs = get_image_file_list(args.infer_imgs)
 
     # loop for infer
-    with open(os.path.join(args.output_dir, "infer_results.txt"), "w") as fout:
+    with open(
+            os.path.join(args.output_dir, "infer_results.txt"),
+            "w",
+            encoding='utf-8') as fout:
         for idx, img_path in enumerate(infer_imgs):
             print("process: [{}/{}]".format(idx, len(infer_imgs), img_path))
 
diff --git a/ppstructure/vqa/infer_ser_e2e.py b/ppstructure/vqa/infer_ser_e2e.py
index 7cd9907dfbf1f964b0423364c87ae39de7a69932..a93a98ff6d0c2bf35a6bb85a16faacf82ebec1ce 100644
--- a/ppstructure/vqa/infer_ser_e2e.py
+++ b/ppstructure/vqa/infer_ser_e2e.py
@@ -113,7 +113,10 @@ if __name__ == "__main__":
 
     # loop for infer
     ser_engine = SerPredictor(args)
-    with open(os.path.join(args.output_dir, "infer_results.txt"), "w") as fout:
+    with open(
+            os.path.join(args.output_dir, "infer_results.txt"),
+            "w",
+            encoding='utf-8') as fout:
         for idx, img_path in enumerate(infer_imgs):
             print("process: [{}/{}], {}".format(idx, len(infer_imgs), img_path))
 
diff --git a/ppstructure/vqa/infer_ser_re_e2e.py b/ppstructure/vqa/infer_ser_re_e2e.py
index a1d0f52eeecbc6c2ceba5964355008f638f371dd..23737406d1d2f31f79df9ddb1a9a6bcc5976aabe 100644
--- a/ppstructure/vqa/infer_ser_re_e2e.py
+++ b/ppstructure/vqa/infer_ser_re_e2e.py
@@ -112,7 +112,10 @@ if __name__ == "__main__":
 
     # loop for infer
     ser_re_engine = SerReSystem(args)
-    with open(os.path.join(args.output_dir, "infer_results.txt"), "w") as fout:
+    with open(
+            os.path.join(args.output_dir, "infer_results.txt"),
+            "w",
+            encoding='utf-8') as fout:
         for idx, img_path in enumerate(infer_imgs):
             print("process: [{}/{}], {}".format(idx, len(infer_imgs), img_path))
 
diff --git a/ppstructure/vqa/utils.py b/ppstructure/vqa/utils.py
index 7e862e976f4aee91e81eb00284a20734ea93854e..44a6298080a456b514a719edf82ff1f1b60fd077 100644
--- a/ppstructure/vqa/utils.py
+++ b/ppstructure/vqa/utils.py
@@ -32,7 +32,7 @@ def set_seed(seed):
 
 
 def get_bio_label_maps(label_map_path):
-    with open(label_map_path, "r") as fin:
+    with open(label_map_path, "r", encoding='utf-8') as fin:
         lines = fin.readlines()
     lines = [line.strip() for line in lines]
     if "O" not in lines:
diff --git a/ppstructure/vqa/xfun.py b/ppstructure/vqa/xfun.py
index d62cdb5da5514280b62687d80d345ede9484ee90..eb9750dd205e3838fb30cbd5fe3ff419955d19f8 100644
--- a/ppstructure/vqa/xfun.py
+++ b/ppstructure/vqa/xfun.py
@@ -162,7 +162,7 @@ class XFUNDataset(Dataset):
         return encoded_inputs
 
     def read_all_lines(self, ):
-        with open(self.label_path, "r") as fin:
+        with open(self.label_path, "r", encoding='utf-8') as fin:
             lines = fin.readlines()
         return lines