diff --git a/benchmark/torch/AlphaZero/Coach.py b/benchmark/torch/AlphaZero/Coach.py
index 01394b076db969db42a7277b5d95f82bd661db3d..ad856b9e96b31881f1af5a60253ac22a7d9c24a5 100644
--- a/benchmark/torch/AlphaZero/Coach.py
+++ b/benchmark/torch/AlphaZero/Coach.py
@@ -54,9 +54,9 @@ class Coach():
 
         self.test_dataset = get_test_dataset()
 
-    def _run_remote_tasks(self, signal_queue):
+    def _run_remote_tasks(self, signal_queue, seed):
         # The remote actor will actually run on the local machine or other machines of xparl cluster
-        remote_actor = Actor(self.game, self.args)
+        remote_actor = Actor(self.game, self.args, seed)
 
         while True:
             # receive running task signal
@@ -91,12 +91,12 @@ class Coach():
         # connect to xparl cluster to submit jobs
         parl.connect(self.args.master_address)
 
-        for i in range(self.args.actors_num):
+        for seed in range(self.args.actors_num):
             signal_queue = queue.Queue()
             self.remote_actors_signal_queues.append(signal_queue)
 
             remote_thread = threading.Thread(
-                target=self._run_remote_tasks, args=(signal_queue, ))
+                target=self._run_remote_tasks, args=(signal_queue, seed))
             remote_thread.setDaemon(True)
             remote_thread.start()
 
diff --git a/benchmark/torch/AlphaZero/actor.py b/benchmark/torch/AlphaZero/actor.py
index 5ed719b92d292903f81f7c92a983927bf5c9cab5..19d684bb47370126af394c07f571882c3b32948c 100644
--- a/benchmark/torch/AlphaZero/actor.py
+++ b/benchmark/torch/AlphaZero/actor.py
@@ -23,7 +23,8 @@ from utils import win_loss_draw
 
 @parl.remote_class
 class Actor(object):
-    def __init__(self, game, args):
+    def __init__(self, game, args, seed):
+        np.random.seed(seed)
         os.environ['OMP_NUM_THREADS'] = "1"
         self.game = game
         self.args = args