diff --git a/demo/slimfacenet/train_eval.py b/demo/slimfacenet/train_eval.py
index 5303bfdf0086919d23e9fe794427c9a25ef0e676..4df84d8b563b98deca7e12bad06640ee22825602 100644
--- a/demo/slimfacenet/train_eval.py
+++ b/demo/slimfacenet/train_eval.py
@@ -150,7 +150,10 @@ def train(exe, train_program, train_out, test_program, test_out, args):
 
 
 def build_program(program, startup, args, is_train=True):
-    num_trainers = len(os.getenv('CUDA_VISIBLE_DEVICES').split(','))
+    if args.use_gpu:
+        num_trainers = len(os.getenv('CUDA_VISIBLE_DEVICES').split(','))
+    else:
+        num_trainers = 1
     places = fluid.cuda_places() if args.use_gpu else fluid.CPUPlace()
 
     train_dataset = CASIA_Face(root=args.train_data_dir)
@@ -166,7 +169,7 @@ def build_program(program, startup, args, is_train=True):
                 image = fluid.data(
                     name='image', shape=[-1, 3, 112, 96], dtype='float32')
                 label = fluid.data(name='label', shape=[-1, 1], dtype='int64')
-                train_reader = paddle.fluid.io.batch(
+                train_reader = fluid.io.batch(
                     train_dataset.reader,
                     batch_size=args.train_batchsize // num_trainers,
                     drop_last=False)
@@ -187,7 +190,7 @@ def build_program(program, startup, args, is_train=True):
             else:
                 nl, nr, flods, flags = parse_filelist(args.test_data_dir)
                 test_dataset = LFW(nl, nr)
-                test_reader = paddle.fluid.io.batch(
+                test_reader = fluid.io.batch(
                     test_dataset.reader,
                     batch_size=args.test_batchsize,
                     drop_last=False)
@@ -231,7 +234,7 @@ def build_program(program, startup, args, is_train=True):
 def quant_val_reader_batch():
     nl, nr, flods, flags = parse_filelist(args.test_data_dir)
     test_dataset = LFW(nl, nr)
-    test_reader = paddle.fluid.io.batch(
+    test_reader = fluid.io.batch(
         test_dataset.reader, batch_size=1, drop_last=False)
     shuffle_reader = fluid.io.shuffle(test_reader, 3)
 
@@ -298,14 +301,16 @@ def main():
         help='The path of the extract features save, must be .mat file')
     args = parser.parse_args()
 
-    num_trainers = len(os.getenv('CUDA_VISIBLE_DEVICES').split(','))
+    if args.use_gpu:
+        num_trainers = len(os.getenv('CUDA_VISIBLE_DEVICES').split(','))
+    else:
+        num_trainers = 1
     print(args)
     print('num_trainers: {}'.format(num_trainers))
     if args.save_ckpt == None:
         args.save_ckpt = 'output'
-    if not os.path.exists(args.save_ckpt):
-        subprocess.call(['mkdir', '-p', args.save_ckpt])
-
+    if not os.path.isdir(args.save_ckpt):
+        os.makedirs(args.save_ckpt)
     with open(os.path.join(args.save_ckpt, 'log.txt'), 'w+') as f:
         f.writelines(str(args) + '\n')
         f.writelines('num_trainers: {}'.format(num_trainers) + '\n')
@@ -346,7 +351,7 @@ def main():
              executor=exe)
         nl, nr, flods, flags = parse_filelist(args.test_data_dir)
         test_dataset = LFW(nl, nr)
-        test_reader = paddle.fluid.io.batch(
+        test_reader = fluid.io.batch(
             test_dataset.reader,
             batch_size=args.test_batchsize,
             drop_last=False)