diff --git a/dygraph/cycle_gan/train.py b/dygraph/cycle_gan/train.py
index 5147d6324e246aff736c5a79d7a753fc44223cf4..a1422047b0d02f5e6cd9dfaa97e5840d38a7bf69 100644
--- a/dygraph/cycle_gan/train.py
+++ b/dygraph/cycle_gan/train.py
@@ -47,7 +47,7 @@ lambda_identity = 0.5
 tep_per_epoch = 2974
 
 
-def optimizer_setting():
+def optimizer_setting(parameters):
     lr = 0.0002
     optimizer = fluid.optimizer.Adam(
         learning_rate=fluid.layers.piecewise_decay(
@@ -56,6 +56,7 @@ def optimizer_setting():
                 140 * step_per_epoch, 160 * step_per_epoch, 180 * step_per_epoch
             ],
             values=[lr, lr * 0.8, lr * 0.6, lr * 0.4, lr * 0.2, lr * 0.1]),
+        parameter_list=parameters,
         beta1=0.5)
     return optimizer
 
@@ -88,9 +89,14 @@ def train(args):
 
         losses = [[], []]
         t_time = 0
-        optimizer1 = optimizer_setting()
-        optimizer2 = optimizer_setting()
-        optimizer3 = optimizer_setting()
+
+        vars_G = cycle_gan.build_generator_resnet_9blocks_a.parameters() + cycle_gan.build_generator_resnet_9blocks_b.parameters()
+        vars_da = cycle_gan.build_gen_discriminator_a.parameters()
+        vars_db = cycle_gan.build_gen_discriminator_b.parameters()
+
+        optimizer1 = optimizer_setting(vars_G)
+        optimizer2 = optimizer_setting(vars_da)
+        optimizer3 = optimizer_setting(vars_db)
 
         for epoch in range(args.epoch):
             batch_id = 0
@@ -114,9 +120,8 @@ def train(args):
                 g_loss_out = g_loss.numpy()
 
                 g_loss.backward()
-                vars_G = cycle_gan.build_generator_resnet_9blocks_a.parameters() + cycle_gan.build_generator_resnet_9blocks_b.parameters()
 
-                optimizer1.minimize(g_loss, parameter_list=vars_G)
+                optimizer1.minimize(g_loss)
                 cycle_gan.clear_gradients()
 
                 fake_pool_B = B_pool.pool_image(fake_B).numpy()
@@ -137,8 +142,7 @@ def train(args):
                 d_loss_A = fluid.layers.reduce_mean(d_loss_A)
 
                 d_loss_A.backward()
-                vars_da = cycle_gan.build_gen_discriminator_a.parameters()
-                optimizer2.minimize(d_loss_A, parameter_list=vars_da)
+                optimizer2.minimize(d_loss_A)
                 cycle_gan.clear_gradients()
 
                 # optimize the d_B network
@@ -150,8 +154,7 @@ def train(args):
                 d_loss_B = fluid.layers.reduce_mean(d_loss_B)
 
                 d_loss_B.backward()
-                vars_db = cycle_gan.build_gen_discriminator_b.parameters()
-                optimizer3.minimize(d_loss_B, parameter_list=vars_db)
+                optimizer3.minimize(d_loss_B)
 
                 cycle_gan.clear_gradients()
 
diff --git a/dygraph/mnist/train.py b/dygraph/mnist/train.py
index 0398913bc14b5e0679a5038a527a18887c9cbe3f..bb0568c08161afab84a287a53f7cf579fb1f92fb 100644
--- a/dygraph/mnist/train.py
+++ b/dygraph/mnist/train.py
@@ -187,7 +187,7 @@ def train_mnist(args):
         if args.use_data_parallel:
             strategy = fluid.dygraph.parallel.prepare_context()
         mnist = MNIST()
-        adam = AdamOptimizer(learning_rate=0.001)
+        adam = AdamOptimizer(learning_rate=0.001, parameter_list=mnist.parameters())
         if args.use_data_parallel:
             mnist = fluid.dygraph.parallel.DataParallel(mnist, strategy)
 
diff --git a/dygraph/reinforcement_learning/actor_critic.py b/dygraph/reinforcement_learning/actor_critic.py
index 7fce62abe10980ed37e2ad5848f369b16ef193a1..26ff614a588bc5f8206377c1862993312d335120 100644
--- a/dygraph/reinforcement_learning/actor_critic.py
+++ b/dygraph/reinforcement_learning/actor_critic.py
@@ -68,7 +68,7 @@ with fluid.dygraph.guard():
     policy = Policy()
 
     eps = np.finfo(np.float32).eps.item()
-    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=3e-2)
+    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=3e-2, parameter_list=policy.parameters())
 
     def get_mean_and_std(values=[]):
         n = 0.
diff --git a/dygraph/reinforcement_learning/reinforce.py b/dygraph/reinforcement_learning/reinforce.py
index d9d6a15e20225e017753dc30a568de961242e288..e7f4d7e56a6608a8d5027459746806639b06eac7 100644
--- a/dygraph/reinforcement_learning/reinforce.py
+++ b/dygraph/reinforcement_learning/reinforce.py
@@ -67,7 +67,7 @@ with fluid.dygraph.guard():
     policy = Policy()
 
     eps = np.finfo(np.float32).eps.item()
-    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=1e-2)
+    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=1e-2, parameter_list=policy.parameters())
 
     def get_mean_and_std(values=[]):
         n = 0.
diff --git a/dygraph/reinforcement_learning/test_actor_critic_load.py b/dygraph/reinforcement_learning/test_actor_critic_load.py
index c58b9495f2605cdd6b9749db542bbab3c1e3eba7..21699bef7669c33389d52c71545836d7e0e82dee 100644
--- a/dygraph/reinforcement_learning/test_actor_critic_load.py
+++ b/dygraph/reinforcement_learning/test_actor_critic_load.py
@@ -68,7 +68,7 @@ with fluid.dygraph.guard():
     policy = Policy()
 
     eps = np.finfo(np.float32).eps.item()
-    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=3e-2)
+    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=3e-2, parameter_list=policy.parameters())
 
     def get_mean_and_std(values=[]):
         n = 0.
diff --git a/dygraph/reinforcement_learning/test_reinforce_load.py b/dygraph/reinforcement_learning/test_reinforce_load.py
index 9d8aa5dc68eb7d8c5e155c1b6f0762c9b2a00a3b..31edd66b0f8e4dea8907af71c117fb62dfb6ae42 100644
--- a/dygraph/reinforcement_learning/test_reinforce_load.py
+++ b/dygraph/reinforcement_learning/test_reinforce_load.py
@@ -67,7 +67,7 @@ with fluid.dygraph.guard():
     policy = Policy()
 
     eps = np.finfo(np.float32).eps.item()
-    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=1e-2)
+    optimizer = fluid.optimizer.AdamOptimizer(learning_rate=1e-2, parameter_list=policy.parameters())
 
     def get_mean_and_std(values=[]):
         n = 0.