diff --git a/douzero/dmc/dmc.py b/douzero/dmc/dmc.py
index b2920a0..858f622 100644
--- a/douzero/dmc/dmc.py
+++ b/douzero/dmc/dmc.py
@@ -129,6 +129,7 @@ def train(flags):
     ]
     frames, stats = 0, {k: 0 for k in stat_keys}
     position_frames = {'landlord':0, 'landlord_up':0, 'landlord_front':0, 'landlord_down':0}
+    position_locks = {'landlord': threading.Lock(), 'landlord_up': threading.Lock(), 'landlord_front': threading.Lock(), 'landlord_down': threading.Lock()}
 
     def sync_onnx_model(frames):
         p_path = '%s/%s' % (flags.onnx_model_path, flags.xpid)
@@ -139,17 +140,18 @@ def train(flags):
                 model_path = '%s/%s/model_%s.onnx' % (flags.onnx_model_path, flags.xpid, position)
                 onnx_params = learner_model.get_model(position)\
                     .get_onnx_params(torch.device('cpu') if flags.training_device == 'cpu' else torch.device('cuda:' + flags.training_device))
-                torch.onnx.export(
-                    learner_model.get_model(position),
-                    onnx_params['args'],
-                    model_path,
-                    export_params=True,
-                    opset_version=10,
-                    do_constant_folding=True,
-                    input_names=onnx_params['input_names'],
-                    output_names=onnx_params['output_names'],
-                    dynamic_axes=onnx_params['dynamic_axes']
-                )
+                with position_locks[position]:
+                    torch.onnx.export(
+                        learner_model.get_model(position),
+                        onnx_params['args'],
+                        model_path,
+                        export_params=True,
+                        opset_version=10,
+                        do_constant_folding=True,
+                        input_names=onnx_params['input_names'],
+                        output_names=onnx_params['output_names'],
+                        dynamic_axes=onnx_params['dynamic_axes']
+                    )
         onnx_frame.value = frames
 
     # Load models if any
@@ -214,7 +216,6 @@ def train(flags):
     locks = {}
     for device in device_iterator:
         locks[device] = {'landlord': threading.Lock(), 'landlord_up': threading.Lock(), 'landlord_front': threading.Lock(), 'landlord_down': threading.Lock()}
-    position_locks = {'landlord': threading.Lock(), 'landlord_up': threading.Lock(), 'landlord_front': threading.Lock(), 'landlord_down': threading.Lock()}
 
     for i in range(flags.num_threads):
         for position in ['landlord', 'landlord_up', 'landlord_front', 'landlord_down']: