diff --git a/configs/yolov3_reader.yml b/configs/yolov3_reader.yml
index d167c70bc1f97db6dc997c9d535aeca92d4e3473..5e11364b0ad831e68513f60ea68ca511e1018a8b 100644
--- a/configs/yolov3_reader.yml
+++ b/configs/yolov3_reader.yml
@@ -52,7 +52,7 @@ TrainReader:
   drop_last: true
   worker_num: 4
   bufsize: 4
-  use_process: false #true
+  use_process: true
 
 
 EvalReader:
diff --git a/tools/train.py b/tools/train.py
index 5eba5f30c58d9027941f450ce83e1c3507571c62..c37d45f20ecc037ac4bf71200a361041799b1894 100755
--- a/tools/train.py
+++ b/tools/train.py
@@ -87,14 +87,7 @@ def parse_args():
     return args
 
 
-def run():
-    FLAGS = parse_args()
-
-    cfg = load_config(FLAGS.config)
-    merge_config(FLAGS.opt)
-    check_config(cfg)
-    check_gpu(cfg.use_gpu)
-    check_version()
+def run(FLAGS, cfg):
     env = os.environ
 
     FLAGS.dist = 'PADDLE_TRAINER_ID' in env and 'PADDLE_TRAINERS_NUM' in env
@@ -108,6 +101,9 @@ def run():
         random.seed(0)
         np.random.seed(0)
 
+    if dist.ParallelEnv().nranks > 1:
+        paddle.distributed.init_parallel_env()
+
     # Model
     main_arch = cfg.architecture
     model = create(cfg.architecture)
@@ -126,8 +122,7 @@ def run():
 
     # Parallel Model 
     if dist.ParallelEnv().nranks > 1:
-        strategy = paddle.distributed.init_parallel_env()
-        model = paddle.DataParallel(model, strategy)
+        model = paddle.DataParallel(model)
 
     # Data Reader 
     start_iter = 0
@@ -137,7 +132,9 @@ def run():
         devices_num = int(os.environ.get('CPU_NUM', 1))
 
     train_reader = create_reader(
-        cfg.TrainReader, (cfg.max_iters - start_iter), cfg, devices_num=1)
+        cfg.TrainReader, (cfg.max_iters - start_iter),
+        cfg,
+        devices_num=devices_num)
 
     time_stat = deque(maxlen=cfg.log_iter)
     start_time = time.time()
@@ -193,7 +190,15 @@ def run():
 
 
 def main():
-    dist.spawn(run)
+    FLAGS = parse_args()
+
+    cfg = load_config(FLAGS.config)
+    merge_config(FLAGS.opt)
+    check_config(cfg)
+    check_gpu(cfg.use_gpu)
+    check_version()
+
+    run(FLAGS, cfg)
 
 
 if __name__ == "__main__":