test=develop (#3205)

.gitmodules

@@ -4,3 +4,6 @@
 [submodule "PaddleNLP/language_representations_kit/ERNIE"]
 	path = PaddleNLP/language_representations_kit/ERNIE
 	url = https://github.com/PaddlePaddle/ERNIE
+[submodule "PaddleRL"]
+	path = PaddleRL
+	url = https://github.com/PaddlePaddle/PARL

PaddleRL @ a8846355

+Subproject commit a884635519c529c69c34e1134ca6c9d99f2c0007

PaddleRL/DeepQNetwork/DQN_agent.py → legacy/PaddleRL/DeepQNetwork/DQN_agent.py

@@ -71,13 +71,13 @@ class DQNModel(object):
             optimizer.minimize(cost)
 
         vars = list(self.train_program.list_vars())
-        target_vars = list(filter(
-            lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
+        target_vars = list(
+            filter(lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
 
         policy_vars_name = [
-                x.name.replace('target', 'policy') for x in target_vars]
-        policy_vars = list(filter(
-            lambda x: x.name in policy_vars_name, vars))
+            x.name.replace('target', 'policy') for x in target_vars
+        ]
+        policy_vars = list(filter(lambda x: x.name in policy_vars_name, vars))
 
         policy_vars.sort(key=lambda x: x.name)
         target_vars.sort(key=lambda x: x.name)
@@ -153,7 +153,6 @@ class DQNModel(object):
             bias_attr=ParamAttr(name='{}_fc1_b'.format(variable_field)))
         return out
 
-
     def act(self, state, train_or_test):
         sample = np.random.random()
         if train_or_test == 'train' and sample < self.exploration:

PaddleRL/DeepQNetwork/DoubleDQN_agent.py → legacy/PaddleRL/DeepQNetwork/DoubleDQN_agent.py

@@ -64,9 +64,11 @@ class DoubleDQNModel(object):
             greedy_action = fluid.layers.argmax(next_s_predcit_value, axis=1)
             greedy_action = fluid.layers.unsqueeze(greedy_action, axes=[1])
 
-            predict_onehot = fluid.layers.one_hot(greedy_action, self.action_dim)
+            predict_onehot = fluid.layers.one_hot(greedy_action,
+                                                  self.action_dim)
             best_v = fluid.layers.reduce_sum(
-                fluid.layers.elementwise_mul(predict_onehot, targetQ_predict_value),
+                fluid.layers.elementwise_mul(predict_onehot,
+                                             targetQ_predict_value),
                 dim=1)
             best_v.stop_gradient = True
 
@@ -79,13 +81,13 @@ class DoubleDQNModel(object):
             optimizer.minimize(cost)
 
         vars = list(self.train_program.list_vars())
-        target_vars = list(filter(
-            lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
+        target_vars = list(
+            filter(lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
 
         policy_vars_name = [
-                x.name.replace('target', 'policy') for x in target_vars]
-        policy_vars = list(filter(
-            lambda x: x.name in policy_vars_name, vars))
+            x.name.replace('target', 'policy') for x in target_vars
+        ]
+        policy_vars = list(filter(lambda x: x.name in policy_vars_name, vars))
 
         policy_vars.sort(key=lambda x: x.name)
         target_vars.sort(key=lambda x: x.name)
@@ -161,7 +163,6 @@ class DoubleDQNModel(object):
             bias_attr=ParamAttr(name='{}_fc1_b'.format(variable_field)))
         return out
 
-
     def act(self, state, train_or_test):
         sample = np.random.random()
         if train_or_test == 'train' and sample < self.exploration:

PaddleRL/DeepQNetwork/DuelingDQN_agent.py → legacy/PaddleRL/DeepQNetwork/DuelingDQN_agent.py

@@ -71,13 +71,13 @@ class DuelingDQNModel(object):
             optimizer.minimize(cost)
 
         vars = list(self.train_program.list_vars())
-        target_vars = list(filter(
-            lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
+        target_vars = list(
+            filter(lambda x: 'GRAD' not in x.name and 'target' in x.name, vars))
 
         policy_vars_name = [
-                x.name.replace('target', 'policy') for x in target_vars]
-        policy_vars = list(filter(
-            lambda x: x.name in policy_vars_name, vars))
+            x.name.replace('target', 'policy') for x in target_vars
+        ]
+        policy_vars = list(filter(lambda x: x.name in policy_vars_name, vars))
 
         policy_vars.sort(key=lambda x: x.name)
         target_vars.sort(key=lambda x: x.name)
@@ -163,7 +163,6 @@ class DuelingDQNModel(object):
             advantage, dim=1, keep_dim=True))
         return Q
 
-
     def act(self, state, train_or_test):
         sample = np.random.random()
         if train_or_test == 'train' and sample < self.exploration: