diff --git a/PaddleCV/ocr_recognition/eval.py b/PaddleCV/ocr_recognition/eval.py
index bc8be4fcf4767183629c2a9c3dc7ea49af5b20b1..99db6c89aaf98d2539f1f1426d50cbc64f80e9e9 100644
--- a/PaddleCV/ocr_recognition/eval.py
+++ b/PaddleCV/ocr_recognition/eval.py
@@ -1,5 +1,6 @@
 import paddle.fluid as fluid
 from utility import add_arguments, print_arguments, to_lodtensor, get_ctc_feeder_data, get_attention_feeder_data
+from utility import check_gpu
 from attention_model import attention_eval
 from crnn_ctc_model import ctc_eval
 import data_reader
@@ -70,6 +71,7 @@ def evaluate(args):
 def main():
     args = parser.parse_args()
     print_arguments(args)
+    check_gpu(args.use_gpu)
     evaluate(args)
 
 
diff --git a/PaddleCV/ocr_recognition/infer.py b/PaddleCV/ocr_recognition/infer.py
index f7d9fc1416a532557a6ba530306a484b164faaa8..9ad75d17f54ff121e5fbea090659c3c488bfe77a 100755
--- a/PaddleCV/ocr_recognition/infer.py
+++ b/PaddleCV/ocr_recognition/infer.py
@@ -1,6 +1,7 @@
 from __future__ import print_function
 import paddle.fluid as fluid
 from utility import add_arguments, print_arguments, to_lodtensor, get_ctc_feeder_data, get_attention_feeder_for_infer, get_ctc_feeder_for_infer
+from utility import check_gpu
 import paddle.fluid.profiler as profiler
 from crnn_ctc_model import ctc_infer
 from attention_model import attention_infer
@@ -138,6 +139,7 @@ def prune(words, sos, eos):
 def main():
     args = parser.parse_args()
     print_arguments(args)
+    check_gpu(args.use_gpu)
     if args.profile:
         if args.use_gpu:
             with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
diff --git a/PaddleCV/ocr_recognition/train.py b/PaddleCV/ocr_recognition/train.py
index 2e294907a6bbac5f311c420ad22d51eafa972da7..a2370efdb89fec9ed5d824c9b2f933e347196922 100755
--- a/PaddleCV/ocr_recognition/train.py
+++ b/PaddleCV/ocr_recognition/train.py
@@ -4,6 +4,7 @@ from __future__ import division
 from __future__ import print_function
 import paddle.fluid as fluid
 from utility import add_arguments, print_arguments, to_lodtensor, get_ctc_feeder_data, get_attention_feeder_data
+from utility import check_gpu
 import paddle.fluid.profiler as profiler
 from crnn_ctc_model import ctc_train_net
 from attention_model import attention_train_net
@@ -67,7 +68,9 @@ def train(args):
         cycle=args.total_step > 0,
         model=args.model)
     test_reader = data_reader.test(
-        test_images_dir=args.test_images, test_list_file=args.test_list, model=args.model)
+        test_images_dir=args.test_images,
+        test_list_file=args.test_list,
+        model=args.model)
 
     # prepare environment
     place = fluid.CPUPlace()
@@ -115,8 +118,8 @@ def train(args):
         for data in test_reader():
             exe.run(inference_program, feed=get_feeder_data(data, place))
         _, test_seq_error = error_evaluator.eval(exe)
-        print("\nTime: %s; Iter[%d]; Test seq error: %s.\n" % (
-            time.time(), iter_num, str(test_seq_error[0])))
+        print("\nTime: %s; Iter[%d]; Test seq error: %s.\n" %
+              (time.time(), iter_num, str(test_seq_error[0])))
 
         #Note: The following logs are special for CE monitoring.
         #Other situations do not need to care about these logs.
@@ -155,10 +158,10 @@ def train(args):
             iter_num += 1
             # training log
             if iter_num % args.log_period == 0:
-                print("\nTime: %s; Iter[%d]; Avg loss: %.3f; Avg seq err: %.3f" % (
-                    time.time(), iter_num,
-                    total_loss / (args.log_period * args.batch_size),
-                    total_seq_error / (args.log_period * args.batch_size)))
+                print("\nTime: %s; Iter[%d]; Avg loss: %.3f; Avg seq err: %.3f"
+                      % (time.time(), iter_num,
+                         total_loss / (args.log_period * args.batch_size),
+                         total_seq_error / (args.log_period * args.batch_size)))
                 print("kpis	train_cost	%f" % (total_loss / (args.log_period *
                                                             args.batch_size)))
                 print("kpis	train_acc	%f" % (
@@ -203,6 +206,7 @@ def train(args):
 def main():
     args = parser.parse_args()
     print_arguments(args)
+    check_gpu(args.use_gpu)
     if args.profile:
         if args.use_gpu:
             with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
diff --git a/PaddleCV/ocr_recognition/utility.py b/PaddleCV/ocr_recognition/utility.py
index a22a744128e01f97026d1535278566ec8c8f7a66..ab5bbd817d2bbffaba3e6ad58b3acf8b5d2252af 100755
--- a/PaddleCV/ocr_recognition/utility.py
+++ b/PaddleCV/ocr_recognition/utility.py
@@ -140,3 +140,22 @@ def get_attention_feeder_for_infer(data, place):
         "init_ids": init_ids,
         "init_scores": init_scores
     }
+
+
+def check_gpu(use_gpu):
+    """
+    Log error and exit when set use_gpu=true in paddlepaddle
+    cpu version.
+    """
+    err = "Config use_gpu cannot be set as true while you are " \
+          "using paddlepaddle cpu version ! \nPlease try: \n" \
+          "\t1. Install paddlepaddle-gpu to run model on GPU \n" \
+          "\t2. Set use_gpu as false in config file to run " \
+          "model on CPU"
+
+    try:
+        if use_gpu and not fluid.is_compiled_with_cuda():
+            logger.error(err)
+            sys.exit(1)
+    except Exception as e:
+        pass