diff --git a/fluid/PaddleCV/faster_rcnn/.run_ce.sh b/fluid/PaddleCV/faster_rcnn/.run_ce.sh
index 0c26c4e0dea608e85747b3adf79768c7af99f8e8..af2f21da73fa14589258b565a74f78e25dbd4e84 100755
--- a/fluid/PaddleCV/faster_rcnn/.run_ce.sh
+++ b/fluid/PaddleCV/faster_rcnn/.run_ce.sh
@@ -7,11 +7,11 @@ export OMP_NUM_THREADS=1
 cudaid=${face_detection:=0} # use 0-th card as default
 export CUDA_VISIBLE_DEVICES=$cudaid
 
-FLAGS_benchmark=true  python train.py --model_save_dir=output/ --data_dir=dataset/coco/ --max_iter=20 --enable_ce --pretrained_model=./imagenet_resnet50_fusebn | python _ce.py
+FLAGS_benchmark=true  python train.py --model_save_dir=output/ --data_dir=dataset/coco/ --max_iter=10 --enable_ce --pretrained_model=./imagenet_resnet50_fusebn | python _ce.py
 
 
 cudaid=${face_detection_m:=0,1,2,3} # use 0,1,2,3 card as default
 export CUDA_VISIBLE_DEVICES=$cudaid
 
-FLAGS_benchmark=true  python train.py --model_save_dir=output/ --data_dir=dataset/coco/ --max_iter=20 --enable_ce --pretrained_model=./imagenet_resnet50_fusebn | python _ce.py
+FLAGS_benchmark=true  python train.py --model_save_dir=output/ --data_dir=dataset/coco/ --max_iter=10 --enable_ce --pretrained_model=./imagenet_resnet50_fusebn | python _ce.py
 
diff --git a/fluid/PaddleCV/faster_rcnn/config.py b/fluid/PaddleCV/faster_rcnn/config.py
index 71deba6de33bc539dc8f37afa22c853a6f591c52..e776836d89e88447471adbe2f9a9c6ff8301cbe0 100644
--- a/fluid/PaddleCV/faster_rcnn/config.py
+++ b/fluid/PaddleCV/faster_rcnn/config.py
@@ -206,7 +206,6 @@ _C.bbox_clip = np.log(1000. / 16.)
 
 # ce
 _C.enable_ce = False
-_C.num_devices = 1
 
 # dataset path
 _C.train_file_list = 'annotations/instances_train2017.json'
diff --git a/fluid/PaddleCV/faster_rcnn/utility.py b/fluid/PaddleCV/faster_rcnn/utility.py
index 34e4c4a6daf889612dec2dedc3f863afb9a82e31..3687cc76bc21e00ccfdfbaa52270412df4586f9c 100644
--- a/fluid/PaddleCV/faster_rcnn/utility.py
+++ b/fluid/PaddleCV/faster_rcnn/utility.py
@@ -132,8 +132,6 @@ def parse_args():
     # ce
     parser.add_argument(
             '--enable_ce', action='store_true', help='If set, run the task with continuous evaluation logs.')
-    parser.add_argument(
-            '--num_devices', type=int, default=1, help='Number of GPU devices')
     # yapf: enable
     args = parser.parse_args()
     file_name = sys.argv[0]