diff --git a/requirements.txt b/requirements.txt
index ded0ee752..7127bf47d 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1 +1,2 @@
 sympy
+packaging
diff --git a/test_autoscan/onnx/auto_scan_test.py b/test_autoscan/onnx/auto_scan_test.py
index 50e1ee2e5..a2c157385 100644
--- a/test_autoscan/onnx/auto_scan_test.py
+++ b/test_autoscan/onnx/auto_scan_test.py
@@ -30,6 +30,8 @@
 paddle.set_device("cpu")
 
 logging.basicConfig(level=logging.INFO, format="%(message)s")
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
 
 settings.register_profile("ci",
                           max_examples=100,
@@ -97,37 +99,37 @@ def run_test(configs):
         loop_func = given(generator())(run_test)
         if reproduce is not None:
             loop_func = reproduce(loop_func)
-        logging.info("Start to running test of {}".format(type(self)))
+        logger.info("Start to running test of {}".format(type(self)))
 
         paddle.disable_static()
         loop_func()
 
-        logging.info(
+        logger.info(
             "===================Statistical Information===================")
-        logging.info("Number of Generated Programs: {}".format(
+        logger.info("Number of Generated Programs: {}".format(
             self.num_ran_tests))
-        logging.info("Number of Ignore Programs: {}".format(
+        logger.info("Number of Ignore Programs: {}".format(
             self.num_ignore_tests))
         successful_ran_programs = int(self.num_ran_tests -
                                       self.num_ignore_tests)
         if successful_ran_programs < min_success_num:
-            logging.warning("satisfied_programs = ran_programs")
-            logging.error(
+            logger.warning("satisfied_programs = ran_programs")
+            logger.error(
                 "At least {} programs need to ran successfully, but now only about {} programs satisfied."
                 .format(min_success_num, successful_ran_programs))
             assert False
         used_time = time.time() - start_time
-        logging.info("Used time: {} s".format(round(used_time, 2)))
+        logger.info("Used time: {} s".format(round(used_time, 2)))
         if max_duration > 0 and used_time > max_duration:
-            logging.error(
+            logger.error(
                 "The duration exceeds {} seconds, if this is neccessary, try to set a larger number for parameter `max_duration`."
                 .format(max_duration))
             assert False
 
     def run_test(self, configs):
         config, attrs = configs
-        logging.info("Run configs: {}".format(config))
-        logging.info("Run attrs: {}".format(attrs))
+        logger.info("Run configs: {}".format(config))
+        logger.info("Run attrs: {}".format(attrs))
 
         assert "op_names" in config.keys(
         ), "config must include op_names in dict keys"
@@ -220,9 +222,9 @@ def run_test(self, configs):
                             randtool("float", -2, 2,
                                      shape).astype(input_type[j]))
                 obj.set_input_data("input_data", tuple(input_data))
-                logging.info("Now Run >>> dtype: {}, op_name: {}".format(
+                logger.info("Now Run >>> dtype: {}, op_name: {}".format(
                     input_type, op_names[i]))
                 obj.run()
             if len(input_type_list) == 0:
                 obj.run()
-        logging.info("Run Successfully!")
+        logger.info("Run Successfully!")
diff --git a/test_autoscan/onnx/onnxbase.py b/test_autoscan/onnx/onnxbase.py
index 240512d19..40d656bf8 100644
--- a/test_autoscan/onnx/onnxbase.py
+++ b/test_autoscan/onnx/onnxbase.py
@@ -26,6 +26,10 @@
 from onnx import TensorProto
 from onnxruntime import InferenceSession
 
+logging.basicConfig(level=logging.INFO, format="%(message)s")
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
+
 DTYPE_ONNX_STR_MAP = {
     'float32': TensorProto.FLOAT,
     'float64': TensorProto.DOUBLE,
@@ -44,6 +48,9 @@ def compare(result, expect, delta=1e-10, rtol=1e-10):
     delta: absolute error
     rtol: relative error
     """
+
+    logger.info(">>> compare ...")
+
     if type(result) == np.ndarray:
         if type(expect) == list:
             expect = expect[0]
@@ -167,6 +174,9 @@ def _mkdir(self):
         """
         make dir to save all
         """
+
+        logger.info(">>> _mkdir ...")
+
         save_path = os.path.join(self.pwd, self.name)
         if not os.path.exists(save_path):
             os.mkdir(save_path)
@@ -175,21 +185,33 @@ def _onnx_to_paddle(self, ver):
         """
         convert onnx to paddle
         """
+        logger.info(">>> _onnx_to_paddle ...")
+
         from x2paddle.convert import onnx2paddle
+
+        logger.info(">>> from x2paddle.convert import onnx2paddle ...")
+
         onnx_path = os.path.join(self.pwd, self.name,
                                  self.name + '_' + str(ver) + '.onnx')
         paddle_path = os.path.join(self.pwd, self.name,
                                    self.name + '_' + str(ver) + '_paddle')
+
+        logger.info(">>> onnx2paddle ...")
+
         onnx2paddle(onnx_path,
                     paddle_path,
                     convert_to_lite=False,
                     enable_onnx_checker=self.enable_onnx_checker,
                     disable_feedback=True)
 
+        logger.info(">>> onnx2paddle finished ...")
+
     def _mk_paddle_res(self, ver):
         """
         make paddle res
         """
+        logger.info(">>> _mk_paddle_res ...")
+
         # input data
         paddle_tensor_feed = list()
         result = list()
@@ -201,7 +223,12 @@ def _mk_paddle_res(self, ver):
         if "float64" in self.inputs_dtype:
             self.run_dynamic = True
 
+        # TODO(megemini): create_predictor stuck
+        self.run_dynamic = True
+
         if self.run_dynamic:
+            logger.info(">>> self.run_dynamic...")
+
             paddle_path = os.path.join(self.pwd, self.name,
                                        self.name + '_' + str(ver) + '_paddle/')
             restore = paddle.load(os.path.join(paddle_path, "model.pdparams"))
@@ -213,7 +240,12 @@ def _mk_paddle_res(self, ver):
             model.set_dict(restore)
             model.eval()
             result = model(*paddle_tensor_feed)
+
+            logger.info(">>> self.run_dynamic finished...")
+
         else:
+            logger.info(">>> NOT self.run_dynamic...")
+
             paddle_model_path = os.path.join(
                 self.pwd, self.name, self.name + '_' + str(ver) +
                 '_paddle/inference_model/model.pdmodel')
@@ -224,22 +256,52 @@ def _mk_paddle_res(self, ver):
             config.set_prog_file(paddle_model_path)
             if os.path.exists(paddle_param_path):
                 config.set_params_file(paddle_param_path)
+
+            logger.info(">>> config.enable_use_gpu...")
+
             # initial GPU memory(M), device ID
             config.enable_use_gpu(200, 0)
+
+            logger.info(">>> config.enable_use_gpu finished...")
+
             # optimize graph and fuse op
             config.switch_ir_optim(False)
             config.enable_memory_optim()
+
+            logger.info(">>> enable_memory_optim finished...")
+
             # disable feed, fetch OP, needed by zero_copy_run
             config.switch_use_feed_fetch_ops(False)
+
+            logger.info(">>> config.disable_glog_info...")
+
             config.disable_glog_info()
+
+            logger.info(">>> config.pass_builder...")
+
             pass_builder = config.pass_builder()
+
+            logger.info(">>> create_predictor(config)...")
+
             predictor = create_predictor(config)
+
+            logger.info(">>> predictor.get_input_names...")
+
             input_names = predictor.get_input_names()
             output_names = predictor.get_output_names()
+
+            logger.info(">>> copy_from_cpu...")
+
             for i in range(len(input_names)):
                 input_tensor = predictor.get_input_handle(input_names[i])
                 input_tensor.copy_from_cpu(self.input_feed[self.inputs_name[i]])
+
+            logger.info(">>> predictor.run...")
+
             predictor.run()
+
+            logger.info(">>> predictor.run finished...")
+
             for output_name in output_names:
                 output_tensor = predictor.get_output_handle(output_name)
                 result.append(output_tensor.copy_to_cpu())
@@ -257,15 +319,24 @@ def _mk_paddle_res(self, ver):
                 result = (result, )
             else:
                 result = (result.numpy(), )
+
+        logger.info(">>> _mk_paddle_res finished ...")
+
         return result
 
     def _mk_onnx_res(self, ver):
         """
         make onnx res
         """
+
+        logger.info('>>> _mk_onnx_res InferenceSession...')
+
         sess = InferenceSession(
             os.path.join(self.pwd, self.name,
                          self.name + '_' + str(ver) + '.onnx'))
+
+        logger.info('>>> sess.run ...')
+
         ort_outs = sess.run(output_names=None, input_feed=self.input_feed)
         return ort_outs
 
@@ -291,6 +362,8 @@ def _mk_onnx_graph(self, ver):
         """
         make onnx graph
         """
+        logger.info(">>> _mk_onnx_graph ... make_node")
+
         node = onnx.helper.make_node(
             self.op_type,
             inputs=self.inputs_name,
@@ -324,8 +397,14 @@ def run(self):
         3. use onnx to make res
         4. compare diff
         """
+
+        logger.info(">>> run ...")
+
         self._mkdir()
         for place in self.places:
+
+            logger.info(">>> place ..." + str(place))
+
             paddle.set_device(place)
             onnx_res = {}
             paddle_res = {}
diff --git a/test_autoscan/onnx/test_auto_scan_abs.py b/test_autoscan/onnx/test_auto_scan_abs.py
index fe750ca0e..3773bfac2 100644
--- a/test_autoscan/onnx/test_auto_scan_abs.py
+++ b/test_autoscan/onnx/test_auto_scan_abs.py
@@ -19,6 +19,12 @@
 import numpy as np
 import unittest
 
+import logging
+
+logging.basicConfig(level=logging.INFO, format="%(message)s")
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
+
 
 class TestAbsConvert(OPConvertAutoScanTest):
     """
@@ -51,6 +57,8 @@ def sample_convert_config(self, draw):
         return (config, attrs)
 
     def test(self):
+        logger.info('>>> test_auto_scan_abs.py::test')
+
         self.run_and_statis(max_examples=30)
 
 
diff --git a/test_autoscan/run_autoscan_onnx.sh b/test_autoscan/run_autoscan_onnx.sh
index df8d1aec9..0e711e3f2 100644
--- a/test_autoscan/run_autoscan_onnx.sh
+++ b/test_autoscan/run_autoscan_onnx.sh
@@ -24,7 +24,11 @@ for var in ${file_arr[@]}
 do
   log_name=${logs_path}/${var}.log
   echo " Now start test: ${var}"
+
+  echo " >>> begin test ..."
+
   python ${var} > ${log_name} 2>&1
+  # python ${var}
 
 done
 
diff --git a/test_autoscan/torch/auto_scan_test.py b/test_autoscan/torch/auto_scan_test.py
index 8b91ed052..6701d3d74 100644
--- a/test_autoscan/torch/auto_scan_test.py
+++ b/test_autoscan/torch/auto_scan_test.py
@@ -32,6 +32,8 @@
 paddle.set_device("cpu")
 
 logging.basicConfig(level=logging.INFO, format="%(message)s")
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
 
 settings.register_profile("ci",
                           max_examples=100,
@@ -112,36 +114,36 @@ def run_test(configs):
         loop_func = given(generator())(run_test)
         if reproduce is not None:
             loop_func = reproduce(loop_func)
-        logging.info("Start to running test of {}".format(type(self)))
+        logger.info("Start to running test of {}".format(type(self)))
 
         paddle.disable_static()
         loop_func()
 
-        logging.info(
+        logger.info(
             "===================Statistical Information===================")
-        logging.info("Number of Generated Programs: {}".format(
+        logger.info("Number of Generated Programs: {}".format(
             self.num_ran_tests))
-        logging.info("Number of Ignore Programs: {}".format(
+        logger.info("Number of Ignore Programs: {}".format(
             self.num_ignore_tests))
         successful_ran_programs = int(self.num_ran_tests -
                                       self.num_ignore_tests)
         if successful_ran_programs < min_success_num:
-            logging.warning("satisfied_programs = ran_programs")
-            logging.error(
+            logger.warning("satisfied_programs = ran_programs")
+            logger.error(
                 "At least {} programs need to ran successfully, but now only about {} programs satisfied."
                 .format(min_success_num, successful_ran_programs))
             assert False
         used_time = time.time() - start_time
-        logging.info("Used time: {} s".format(round(used_time, 2)))
+        logger.info("Used time: {} s".format(round(used_time, 2)))
         if max_duration > 0 and used_time > max_duration:
-            logging.error(
+            logger.error(
                 "The duration exceeds {} seconds, if this is neccessary, try to set a larger number for parameter `max_duration`."
                 .format(max_duration))
             assert False
 
     def run_test(self, configs):
         config, models = configs
-        logging.info("Run configs: {}".format(config))
+        logger.info("Run configs: {}".format(config))
 
         assert "op_names" in config.keys(
         ), "config must include op_names in dict keys"
@@ -218,9 +220,9 @@ def run_test(self, configs):
                             randtool("float", -2, 2,
                                      shape).astype(input_type[j]))
                 obj.set_input_data("input_data", tuple(input_data))
-                logging.info("Now Run >>> dtype: {}, op_name: {}".format(
+                logger.info("Now Run >>> dtype: {}, op_name: {}".format(
                     input_type, op_names[i]))
                 obj.run()
             if len(input_type_list) == 0:
                 obj.run()
-        logging.info("Run Successfully!")
+        logger.info("Run Successfully!")
diff --git a/test_autoscan/torch/torchbase.py b/test_autoscan/torch/torchbase.py
index 435313271..0f66f513c 100644
--- a/test_autoscan/torch/torchbase.py
+++ b/test_autoscan/torch/torchbase.py
@@ -180,6 +180,9 @@ def _mk_paddle_res(self, ):
         if "float64" in self.inputs_dtype:
             self.run_dynamic = True
 
+        # TODO(megemini): create_predictor stuck
+        self.run_dynamic = True
+
         if self.run_dynamic:
             paddle_path = os.path.join(self.pwd, self.name,
                                        self.name + '_paddle/')
diff --git a/test_benchmark/Caffe/convert.sh b/test_benchmark/Caffe/convert.sh
index 190166823..a60db59e7 100644
--- a/test_benchmark/Caffe/convert.sh
+++ b/test_benchmark/Caffe/convert.sh
@@ -10,7 +10,7 @@ find . -name "run.log" | xargs rm -rf
 
 # use black.list to control CI tests
 filename="black.list"
-models=$(ls -d */ | grep -v -F -f "$filename")
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
 num_of_models=$(ls -d */ | grep -v -F -f "$filename" | wc -l)
 
 counter=1
@@ -18,6 +18,11 @@ for model in $models
 do
     echo "[X2Paddle-Caffe] ${counter}/${num_of_models} $model ..."
     cd $model
+
+    # make default result is `Failed` in case of `result.txt` not generated
+    touch result.txt
+    echo $model ">>>Failed"> result.txt
+
     sh run_convert.sh $model 1>run.log 2>run.err &
     cd ..
     counter=$(($counter+1))
diff --git a/test_benchmark/Caffe/tools/log_summary.sh b/test_benchmark/Caffe/tools/log_summary.sh
index 7c0228980..4dc318a61 100644
--- a/test_benchmark/Caffe/tools/log_summary.sh
+++ b/test_benchmark/Caffe/tools/log_summary.sh
@@ -7,7 +7,12 @@ fi
 mkdir ../output
 
 cd ..
-for model in `ls -d */ | grep -v 'tools' | grep -v 'output' | awk -F '/' '{print $1}'`
+
+# use black.list to control CI tests
+filename="black.list"
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
+
+for model in $models
 do
     cp ${model}/run.log output/${model}_run.log
     cp ${model}/run.err output/${model}_run.err
diff --git a/test_benchmark/ONNX/convert.sh b/test_benchmark/ONNX/convert.sh
index 8fc61701a..7b25275b9 100644
--- a/test_benchmark/ONNX/convert.sh
+++ b/test_benchmark/ONNX/convert.sh
@@ -10,7 +10,7 @@ find . -name "run.err" | xargs rm -rf
 
 # use black.list to control CI tests
 filename="black.list"
-models=$(ls -d */ | grep -v -F -f "$filename")
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
 num_of_models=$(ls -d */ | grep -v -F -f "$filename" | wc -l)
 
 counter=1
@@ -18,6 +18,11 @@ for model in $models
 do
     echo "[X2Paddle-ONNX] ${counter}/${num_of_models} $model ..."
     cd $model
+
+    # make default result is `Failed` in case of `result.txt` not generated
+    touch result.txt
+    echo $model ">>>Failed"> result.txt
+
     sh run_convert.sh $model 1>run.log 2>run.err &
     cd ..
     counter=$(($counter+1))
diff --git a/test_benchmark/ONNX/tools/log_summary.sh b/test_benchmark/ONNX/tools/log_summary.sh
index 7c0228980..4dc318a61 100644
--- a/test_benchmark/ONNX/tools/log_summary.sh
+++ b/test_benchmark/ONNX/tools/log_summary.sh
@@ -7,7 +7,12 @@ fi
 mkdir ../output
 
 cd ..
-for model in `ls -d */ | grep -v 'tools' | grep -v 'output' | awk -F '/' '{print $1}'`
+
+# use black.list to control CI tests
+filename="black.list"
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
+
+for model in $models
 do
     cp ${model}/run.log output/${model}_run.log
     cp ${model}/run.err output/${model}_run.err
diff --git a/test_benchmark/PyTorch/convert.sh b/test_benchmark/PyTorch/convert.sh
index 0f8f67873..4c80dfed0 100644
--- a/test_benchmark/PyTorch/convert.sh
+++ b/test_benchmark/PyTorch/convert.sh
@@ -19,7 +19,7 @@ find . -name "run.err" | xargs rm -rf
 
 # use black.list to control CI tests
 filename="black.list"
-models=$(ls -d */ | grep -v -F -f "$filename")
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
 num_of_models=$(ls -d */ | grep -v -F -f "$filename" | wc -l)
 
 counter=1
@@ -27,6 +27,11 @@ for model in $models
 do
     echo "[X2Paddle-PyTorch] ${counter}/${num_of_models} $model ..."
     cd $model
+
+    # make default result is `Failed` in case of `result.txt` not generated
+    touch result.txt
+    echo $model ">>>Failed"> result.txt
+
     sh run_convert.sh 1>run.log 2>run.err &
     cd ..
     counter=$(($counter+1))
diff --git a/test_benchmark/PyTorch/tools/log_summary.sh b/test_benchmark/PyTorch/tools/log_summary.sh
index 7c0228980..4dc318a61 100644
--- a/test_benchmark/PyTorch/tools/log_summary.sh
+++ b/test_benchmark/PyTorch/tools/log_summary.sh
@@ -7,7 +7,12 @@ fi
 mkdir ../output
 
 cd ..
-for model in `ls -d */ | grep -v 'tools' | grep -v 'output' | awk -F '/' '{print $1}'`
+
+# use black.list to control CI tests
+filename="black.list"
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
+
+for model in $models
 do
     cp ${model}/run.log output/${model}_run.log
     cp ${model}/run.err output/${model}_run.err
diff --git a/test_benchmark/TensorFlow/convert.sh b/test_benchmark/TensorFlow/convert.sh
index 2259fce14..25e476bd8 100644
--- a/test_benchmark/TensorFlow/convert.sh
+++ b/test_benchmark/TensorFlow/convert.sh
@@ -10,7 +10,7 @@ find . -name "run.err" | xargs rm -rf
 
 # use black.list to control CI tests
 filename="black.list"
-models=$(ls -d */ | grep -v -F -f "$filename")
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
 num_of_models=$(ls -d */ | grep -v -F -f "$filename" | wc -l)
 
 counter=1
@@ -18,6 +18,11 @@ for model in $models
 do
     echo "[X2Paddle-TensorFlow] ${counter}/${num_of_models} $model ..."
     cd $model
+
+    # make default result is `Failed` in case of `result.txt` not generated
+    touch result.txt
+    echo $model ">>>Failed"> result.txt
+
     sh run_convert.sh $model 1>run.log 2>run.err &
     cd ..
     counter=$(($counter+1))
diff --git a/test_benchmark/TensorFlow/tools/log_summary.sh b/test_benchmark/TensorFlow/tools/log_summary.sh
index 7c0228980..4dc318a61 100644
--- a/test_benchmark/TensorFlow/tools/log_summary.sh
+++ b/test_benchmark/TensorFlow/tools/log_summary.sh
@@ -7,7 +7,12 @@ fi
 mkdir ../output
 
 cd ..
-for model in `ls -d */ | grep -v 'tools' | grep -v 'output' | awk -F '/' '{print $1}'`
+
+# use black.list to control CI tests
+filename="black.list"
+models=$(ls -d */ | grep -v -F -f "$filename" | awk -F '/' '{print $1}')
+
+for model in $models
 do
     cp ${model}/run.log output/${model}_run.log
     cp ${model}/run.err output/${model}_run.err
diff --git a/x2paddle/convert.py b/x2paddle/convert.py
index 3372c5bc1..56ff967c5 100644
--- a/x2paddle/convert.py
+++ b/x2paddle/convert.py
@@ -14,12 +14,16 @@
 
 from six import text_type as _text_type
 from x2paddle import program
-from x2paddle.utils import ConverterCheck
+from x2paddle.utils import ConverterCheck, check_version
 import argparse
 import sys
 import logging
 import time
 
+logging.basicConfig(level=logging.INFO, format="%(message)s")
+logger = logging.getLogger(__name__)
+logger.setLevel(logging.INFO)
+
 
 def arg_parser():
     parser = argparse.ArgumentParser()
@@ -275,6 +279,9 @@ def onnx2paddle(model_path,
                 lite_model_type="naive_buffer",
                 disable_feedback=False,
                 enable_onnx_checker=True):
+
+    logger.info(">>> onnx2paddle ...")
+
     # for convert_id
     time_info = int(time.time())
     if not disable_feedback:
@@ -287,27 +294,27 @@ def onnx2paddle(model_path,
         v0, v1, v2 = version.split('.')
         version_sum = int(v0) * 100 + int(v1) * 10 + int(v2)
         if version_sum < 160:
-            logging.info("[ERROR] onnx>=1.6.0 is required")
+            logger.info("[ERROR] onnx>=1.6.0 is required")
             return
     except:
-        logging.info(
+        logger.info(
             "[ERROR] onnx is not installed, use \"pip install onnx==1.6.0\".")
         return
-    logging.info("Now translating model from onnx to paddle.")
+    logger.info("Now translating model from onnx to paddle.")
 
     # Do optimizer
     if enable_optim:
         from onnxsim import simplify
         onnx_net_opt_path = model_path[:-5] + '_opt.onnx'
-        logging.info("ONNX Model optimizing ...")
+        logger.info("ONNX Model optimizing ...")
         # load your predefined ONNX model
         model = onnx.load(model_path)
         # convert model
         model_simp, check = simplify(model)
         assert check, "Simplified ONNX model could not be validated"
-        logging.info("Export optimized onnx model:{}".format(onnx_net_opt_path))
+        logger.info("Export optimized onnx model:{}".format(onnx_net_opt_path))
         onnx.save(model_simp, onnx_net_opt_path)
-        logging.info("ONNX Model optimized!")
+        logger.info("ONNX Model optimized!")
         model_path = onnx_net_opt_path
 
     from x2paddle.decoder.onnx_decoder import ONNXDecoder
@@ -315,36 +322,36 @@ def onnx2paddle(model_path,
     model = ONNXDecoder(model_path, enable_onnx_checker, input_shape_dict)
     mapper = ONNXOpMapper(model)
     mapper.paddle_graph.build()
-    logging.info("Model optimizing ...")
+    logger.info("Model optimizing ...")
     from x2paddle.optimizer.optimizer import GraphOptimizer
     graph_opt = GraphOptimizer(source_frame="onnx")
     graph_opt.optimize(mapper.paddle_graph)
-    logging.info("Model optimized.")
+    logger.info("Model optimized.")
     mapper.paddle_graph.gen_model(save_dir)
-    logging.info("Successfully exported Paddle static graph model!")
+    logger.info("Successfully exported Paddle static graph model!")
     if not disable_feedback:
         ConverterCheck(task="ONNX",
                        time_info=time_info,
                        convert_state="Success").start()
     if convert_to_lite:
-        logging.info("Now translating model from Paddle to Paddle Lite ...")
+        logger.info("Now translating model from Paddle to Paddle Lite ...")
         if not disable_feedback:
             ConverterCheck(task="ONNX", time_info=time_info,
                            lite_state="Start").start()
         convert2lite(save_dir, lite_valid_places, lite_model_type)
-        logging.info("Successfully exported Paddle Lite support model!")
+        logger.info("Successfully exported Paddle Lite support model!")
         if not disable_feedback:
             ConverterCheck(task="ONNX",
                            time_info=time_info,
                            lite_state="Success").start()
     # for convert survey
-    logging.info("================================================")
-    logging.info("")
-    logging.info(
+    logger.info("================================================")
+    logger.info("")
+    logger.info(
         "Model Converted! Fill this survey to help X2Paddle better, https://iwenjuan.baidu.com/?code=npyd51 "
     )
-    logging.info("")
-    logging.info("================================================")
+    logger.info("")
+    logger.info("================================================")
 
 
 def pytorch2paddle(module,
@@ -449,20 +456,15 @@ def main():
     assert args.save_dir is not None, "--save_dir is not defined"
 
     try:
-        import platform
-        v0, v1, v2 = platform.python_version().split('.')
-        if not (int(v0) >= 3 and int(v1) >= 5):
-            logging.info("[ERROR] python>=3.5 is required")
+        if not sys.version_info >= (3, 8):
+            logging.error("[ERROR] python>=3.8 is required")
             return
+
         import paddle
-        v0, v1, v2 = paddle.__version__.split('.')
-        logging.info("paddle.__version__ = {}".format(paddle.__version__))
-        if v0 == '0' and v1 == '0' and v2 == '0':
-            logging.info(
-                "[WARNING] You are use develop version of paddlepaddle")
-        elif int(v0) != 2 or int(v1) < 0:
-            logging.info("[ERROR] paddlepaddle>=2.0.0 is required")
+        if not check_version('2.0.0'):
+            logging.error("[ERROR] paddlepaddle>=2.0.0 is required")
             return
+
     except:
         logging.info(
             "[ERROR] paddlepaddle not installed, use \"pip install paddlepaddle\""
diff --git a/x2paddle/utils.py b/x2paddle/utils.py
index 2117a40c7..b51dda7f4 100644
--- a/x2paddle/utils.py
+++ b/x2paddle/utils.py
@@ -13,6 +13,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import logging
+
+from packaging.version import Version
+
 import paddle
 import x2paddle
 import hashlib
@@ -30,15 +34,26 @@ def string(param):
     return "\'{}\'".format(param)
 
 
-def check_version():
-    version = paddle.__version__
-    v0, v1, v2 = version.split('.')
-    if not ((v0 == '0' and v1 == '0' and v2 == '0') or
-            (int(v0) >= 2 and int(v1) >= 1)):
-        return False
-    else:
+def check_version(base_version: str = '2.1.0') -> bool:
+    """
+    Return `True` if the current version is equal or bigger than `base_version`.
+    The default version `2.1.0` is used for checking `is_new_version`.
+    """
+    is_new = False
+
+    dev_version = Version('0.0.0')
+    cur_version = Version(paddle.__version__)
+
+    if cur_version == dev_version:
+        logging.info("[WARNING] You are use develop version of paddlepaddle")
+
         return True
 
+    if cur_version >= Version(base_version):
+        return True
+
+    return False
+
 
 def _md5(text: str):
     '''Calculate the md5 value of the input text.'''