diff --git a/docs/api/paddle/optimizer/Adadelta_cn.rst b/docs/api/paddle/optimizer/Adadelta_cn.rst
index db0d2270b87..f4c3b8fda71 100644
--- a/docs/api/paddle/optimizer/Adadelta_cn.rst
+++ b/docs/api/paddle/optimizer/Adadelta_cn.rst
@@ -41,19 +41,7 @@ Adadelta 优化器出自 `DECOUPLED WEIGHT DECAY REGULARIZATION 论文 <https://
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    adadelta = paddle.optimizer.Adadelta(learning_rate=0.0003, epsilon=1e-06, rho=0.95,
-            parameters=linear.parameters())
-    out.backward()
-    adadelta.step()
-    adadelta.clear_grad()
+COPY-FROM: paddle.optimizer.Adadelta
 
 
 方法
@@ -73,20 +61,7 @@ step()
 
 
 
-**代码示例**
-
-.. code-block:: python
-
-    import paddle
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    adadelta = paddle.optimizer.Adadelta(learning_rate=0.0003, epsilon=1e-06, rho=0.95,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    adadelta.step()
-    adadelta.clear_grad()
+COPY-FROM: paddle.optimizer.Adadelta.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -107,23 +82,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adadelta = paddle.optimizer.Adadelta(learning_rate=0.0003, epsilon=1e-06, rho=0.95,
-            parameters=linear.parameters())
-    out.backward()
-    adadelta.minimize(loss)
-    adadelta.clear_grad()
+COPY-FROM: paddle.optimizer.Adadelta.minimize
 
 clear_grad()
 '''''''''
@@ -137,19 +96,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Adadelta(learning_rate=0.0003, epsilon=1e-06, rho=0.95,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Adadelta.clear_grad
 
 set_lr(value)
 '''''''''
@@ -170,26 +117,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    linear = paddle.nn.Linear(10, 10)
-
-    adadelta = paddle.optimizer.Adadelta(weight_decay=0.01,
-                                 learning_rate=0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        adadelta.set_lr(lr_list[i])
-        lr = adadelta.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.Adadelta.set_lr
 
 get_lr()
 '''''''''
@@ -207,37 +135,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-    import numpy as np
-    import paddle
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    adadelta = paddle.optimizer.Adadelta(learning_rate=0.001, parameters = emb.parameters(),weight_decay=0.01)
-    lr = adadelta.get_lr()
-    print(lr) # 0.001
-
-    # example2: PiecewiseDecay is used, return the step learning rate
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.PiecewiseDecay(bd, value, 0)
-    adadelta = paddle.optimizer.Adadelta(scheduler,
-                           parameters=linear.parameters(),
-                           weight_decay=0.01)
-
-    # first step: learning rate is 0.2
-    np.allclose(adadelta.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        adadelta.step()
-        lr = adadelta.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.Adadelta.get_lr
diff --git a/docs/api/paddle/optimizer/AdamW_cn.rst b/docs/api/paddle/optimizer/AdamW_cn.rst
index 6cd79791d19..c72fcdcea37 100755
--- a/docs/api/paddle/optimizer/AdamW_cn.rst
+++ b/docs/api/paddle/optimizer/AdamW_cn.rst
@@ -49,49 +49,7 @@ AdamW 优化器出自 `DECOUPLED WEIGHT DECAY REGULARIZATION <https://arxiv.org/
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.rand([10,10], dtype="float32")
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.AdamW(learning_rate=0.1,
-            parameters=linear.parameters(),
-            beta1=beta1,
-            beta2=beta2,
-            weight_decay=0.01)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
-
-    # Note that the learning_rate of linear_2 is 0.01.
-    linear_1 = paddle.nn.Linear(10, 10)
-    linear_2 = paddle.nn.Linear(10, 10)
-    inp = paddle.uniform(shape=[10, 10], min=-0.1, max=0.1)
-    out = linear_1(inp)
-    out = linear_2(out)
-    loss = paddle.mean(out)
-    adam = paddle.optimizer.AdamW(
-        learning_rate=0.1,
-        parameters=[{
-            'params': linear_1.parameters()
-        }, {
-            'params': linear_2.parameters(),
-            'weight_decay': 0.001,
-            'learning_rate': 0.1,
-            'beta1': 0.8
-        }],
-        weight_decay=0.01,
-        beta1=0.9)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.AdamW
 
 方法
 ::::::::::::
@@ -110,18 +68,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    a = paddle.rand(shape=[2,13], dtype="float32")
-    linear = paddle.nn.Linear(13, 5)
-    adam = paddle.optimizer.AdamW(learning_rate = 0.01,
-                                weight_decay = 0.01,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.AdamW.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -142,24 +89,7 @@ tuple(optimize_ops, params_grads)，其中 optimize_ops 为参数优化 OP 列
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.randn(shape=[10,10], dtype="float32")
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.AdamW(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    out.backward()
-    adam.minimize(loss)
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.AdamW.minimize
 
 clear_grad()
 '''''''''
@@ -172,19 +102,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    a = paddle.rand(shape=[2,13], dtype="float32")
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.AdamW(weight_decay=0.01,
-                                     learning_rate=0.02,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.AdamW.clear_grad
 
 set_lr(value)
 '''''''''
@@ -204,26 +122,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    linear = paddle.nn.Linear(10, 10)
-
-    adam = paddle.optimizer.AdamW(weight_decay=0.01,
-                                 learning_rate=0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        adam.set_lr(lr_list[i])
-        lr = adam.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.AdamW.set_lr
 
 get_lr()
 '''''''''
@@ -240,35 +139,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    adam = paddle.optimizer.AdamW(learning_rate=0.001, parameters = emb.parameters(),weight_decay=0.01)
-    lr = adam.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.randn([10,10], dtype="float32")
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    adam = paddle.optimizer.AdamW(scheduler,
-                           parameters=linear.parameters(),
-                           weight_decay=0.01)
-
-    # learning rate is 0.2
-    print(adam.get_lr())
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        adam.step()
-        lr = adam.get_lr()
-        scheduler.step()
-        print(lr, ret[i])
+COPY-FROM: paddle.optimizer.AdamW.get_lr
diff --git a/docs/api/paddle/optimizer/Adam_cn.rst b/docs/api/paddle/optimizer/Adam_cn.rst
index 1bf2e8eba55..f8e8bc1ba87 100755
--- a/docs/api/paddle/optimizer/Adam_cn.rst
+++ b/docs/api/paddle/optimizer/Adam_cn.rst
@@ -47,45 +47,9 @@ Adam 优化器出自 `Adam 论文 <https://arxiv.org/abs/1412.6980>`_ 的第二
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    adam = paddle.optimizer.Adam(learning_rate=0.1,
-            parameters=linear.parameters())
-    out.backward()
-    adam.step()
-    adam.clear_grad()
-
-.. code-block:: python
-
-    # Adam with beta1/beta2 as Tensor and weight_decay as float
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.Adam(learning_rate=0.1,
-            parameters=linear.parameters(),
-            beta1=beta1,
-            beta2=beta2,
-            weight_decay=0.01)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Adam:code-example1
+
+COPY-FROM: paddle.optimizer.Adam:code-example2
 
 方法
 ::::::::::::
@@ -104,20 +68,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    adam = paddle.optimizer.Adam(learning_rate = 0.01,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Adam.step
 
 append_regularization_ops(parameters_and_grads, regularization=None)
 '''''''''
@@ -155,26 +106,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.Adam(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    out.backward()
-    adam.minimize(loss)
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Adam.minimize
 
 clear_grad()
 '''''''''
@@ -187,20 +119,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Adam(learning_rate=0.02,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Adam.clear_grad
 
 set_lr(value)
 '''''''''
@@ -221,26 +140,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-
-    adam = paddle.optimizer.Adam(0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        adam.set_lr(lr_list[i])
-        lr = adam.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.Adam.set_lr
 
 get_lr()
 '''''''''
@@ -257,39 +157,7 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-    import numpy as np
-    import paddle
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    adam = paddle.optimizer.Adam(0.001, parameters = emb.parameters())
-    lr = adam.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    adam = paddle.optimizer.Adam(scheduler,
-                           parameters=linear.parameters())
-
-    # first step: learning rate is 0.2
-    np.allclose(adam.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        adam.step()
-        lr = adam.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.Adam.get_lr
 
 set_state_dict(state_dict)
 '''''''''
@@ -306,25 +174,7 @@ set_state_dict(state_dict)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    emb = paddle.nn.Embedding(10, 10)
-
-    layer_state_dict = emb.state_dict()
-    paddle.save(layer_state_dict, "emb.pdparams")
-
-    scheduler = paddle.optimizer.lr.NoamDecay(
-        d_model=0.01, warmup_steps=100, verbose=True)
-    adam = paddle.optimizer.Adam(
-        learning_rate=scheduler,
-        parameters=emb.parameters())
-    opt_state_dict = adam.state_dict()
-    paddle.save(opt_state_dict, "adam.pdopt")
-
-    opti_state_dict = paddle.load("adam.pdopt")
-    adam.set_state_dict(opti_state_dict)
+COPY-FROM: paddle.optimizer.Adam.set_state_dict
 
 state_dict(state_dict)
 '''''''''
@@ -344,10 +194,4 @@ state_dict(dict)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    emb = paddle.nn.Embedding(10, 10)
-
-    adam = paddle.optimizer.Adam(0.001, parameters=emb.parameters())
-    state_dict = adam.state_dict()
+COPY-FROM: paddle.optimizer.Adam.state_dict
diff --git a/docs/api/paddle/optimizer/Adamax_cn.rst b/docs/api/paddle/optimizer/Adamax_cn.rst
index 7125659ec43..51e28f354f1 100755
--- a/docs/api/paddle/optimizer/Adamax_cn.rst
+++ b/docs/api/paddle/optimizer/Adamax_cn.rst
@@ -48,22 +48,7 @@ Adamax 优化器是参考 `Adam 论文 <https://arxiv.org/abs/1412.6980>`_ 第 7
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    adam = paddle.optimizer.Adamax(learning_rate=0.1,
-            parameters=linear.parameters())
-    out.backward()
-    adam.step()
-    adam.clear_grad()
-
+COPY-FROM: paddle.optimizer.Adamax
 
 方法
 ::::::::::::
@@ -83,20 +68,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    adam = paddle.optimizer.Adam(learning_rate = 0.01,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Adamax.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -116,26 +88,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.Adamax(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    out.backward()
-    adam.minimize(loss)
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Adamax.minimize
 
 
 clear_grad()
@@ -150,20 +103,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Adamax(learning_rate=0.02,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Adamax.clear_grad
 
 set_lr(value)
 '''''''''
@@ -184,26 +124,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-
-    adam = paddle.optimizer.Adamax(0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        adam.set_lr(lr_list[i])
-        lr = adam.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.Adamax.set_lr
 
 get_lr()
 '''''''''
@@ -221,37 +142,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-
-    import numpy as np
-    import paddle
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    adam = paddle.optimizer.Adamax(0.001, parameters = emb.parameters())
-    lr = adam.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    adam = paddle.optimizer.Adamax(scheduler,
-                           parameters=linear.parameters())
-
-    # first step: learning rate is 0.2
-    np.allclose(adam.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        adam.step()
-        lr = adam.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.Adamax.get_lr
diff --git a/docs/api/paddle/optimizer/Lamb_cn.rst b/docs/api/paddle/optimizer/Lamb_cn.rst
index 10b7c8aabf9..a732c78915c 100755
--- a/docs/api/paddle/optimizer/Lamb_cn.rst
+++ b/docs/api/paddle/optimizer/Lamb_cn.rst
@@ -45,20 +45,7 @@ LAMB（Layer-wise Adaptive Moments optimizer for Batching training）优化器
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-     import paddle
-
-     inp = paddle.uniform(shape=[10, 10], dtype='float32', min=-0.1, max=0.1)
-     linear = paddle.nn.Linear(10, 10)
-     out = linear(inp)
-     loss = paddle.mean(out)
-     beta1 = paddle.to_tensor([0.9], dtype="float32")
-     beta2 = paddle.to_tensor([0.85], dtype="float32")
-     lamb = paddle.optimizer.Lamb(learning_rate=0.002, parameters=linear.parameters(), lamb_weight_decay=0.01)
-     back = out.backward()
-     lamb.step()
-     lamb.clear_grad()
+COPY-FROM: paddle.optimizer.Lamb
 
 方法
 ::::::::::::
@@ -77,20 +64,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    value = paddle.arange(26, dtype='float32')
-    value = paddle.reshape(value, [2, 13])
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    lamb = paddle.optimizer.Lamb(learning_rate = 0.01,
-                                 parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    lamb.step()
-    lamb.clear_grad()
+COPY-FROM: paddle.optimizer.Lamb.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -110,26 +84,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(shape=[10, 10], dtype="float32", min=-0.1, max=0.1)
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    lamb = paddle.optimizer.Lamb(learning_rate=0.1,
-            lamb_weight_decay=0.01,
-            parameters=linear.parameters())
-    out.backward()
-    lamb.minimize(loss)
-    lamb.clear_grad()
-
+COPY-FROM: paddle.optimizer.Lamb.minimize
 
 clear_grad()
 '''''''''
@@ -142,20 +97,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    value = paddle.arange(26, dtype="float32")
-    value = paddle.reshape(value, [2, 13])
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Lamb(learning_rate=0.02,
-                                      parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Lamb.clear_grad
 
 set_lr(value)
 '''''''''
@@ -176,26 +118,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-
-    lamb = paddle.optimizer.Lamb(0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        lamb.set_lr(lr_list[i])
-        lr = lamb.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.Lamb.set_lr
 
 get_lr()
 '''''''''
@@ -213,38 +136,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-
-    import paddle
-    import numpy as np
-
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    lamb = paddle.optimizer.Lamb(0.001, parameters = emb.parameters())
-    lr = lamb.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    inp = paddle.uniform(shape=[10, 10], dtype="float32", min=-0.1, max=0.1)
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    lamb = paddle.optimizer.Lamb(scheduler,
-                           parameters=linear.parameters())
-
-    # first step: learning rate is 0.2
-    np.allclose(lamb.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        lamb.step()
-        lr = lamb.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.Lamb.get_lr
diff --git a/docs/api/paddle/optimizer/Momentum_cn.rst b/docs/api/paddle/optimizer/Momentum_cn.rst
index 470a0f2ab5e..4bb914722ac 100644
--- a/docs/api/paddle/optimizer/Momentum_cn.rst
+++ b/docs/api/paddle/optimizer/Momentum_cn.rst
@@ -37,18 +37,7 @@ Momentum
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    momentum = paddle.optimizer.Momentum(learning_rate=0.1, parameters=linear.parameters(), weight_decay=0.01)
-    out.backward()
-    momentum.step()
-    momentum.clear_grad()
+COPY-FROM: paddle.optimizer.Momentum
 
 
 方法
@@ -69,17 +58,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    momentum = paddle.optimizer.Momentum(learning_rate=0.0003, parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    momentum.step()
-    momentum.clear_grad()
+COPY-FROM: paddle.optimizer.Momentum.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -100,22 +79,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    momentum = paddle.optimizer.Momentum(learning_rate=0.1, parameters=linear.parameters(), weight_decay=0.01)
-    out.backward()
-    momentum.minimize(loss)
-    momentum.clear_grad()
+COPY-FROM: paddle.optimizer.Momentum.minimize
 
 clear_grad()
 '''''''''
@@ -129,18 +93,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Momentum(learning_rate=0.1, parameters=linear.parameters(), weight_decay=0.01)
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Momentum.clear_grad
 
 set_lr(value)
 '''''''''
diff --git a/docs/api/paddle/optimizer/Optimizer_cn.rst b/docs/api/paddle/optimizer/Optimizer_cn.rst
index 0cdb2933bc9..5a81ad86d24 100755
--- a/docs/api/paddle/optimizer/Optimizer_cn.rst
+++ b/docs/api/paddle/optimizer/Optimizer_cn.rst
@@ -25,22 +25,7 @@ Optimizer
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    #以子类 Adam 为例
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    adam = paddle.optimizer.Adam(learning_rate=0.1,
-            parameters=linear.parameters())
-    loss.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Optimizer
 
 方法
 ::::::::::::
@@ -60,21 +45,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    # This can be any optimizer supported by dygraph.
-    adam = paddle.optimizer.Adam(learning_rate = 0.01,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    adam.step()
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Optimizer.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -95,26 +66,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    adam = paddle.optimizer.Adam(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    loss.backward()
-    adam.minimize(loss)
-    adam.clear_grad()
+COPY-FROM: paddle.optimizer.Optimizer.minimize
 
 clear_grad()
 '''''''''
@@ -128,20 +80,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    value = np.arange(26).reshape(2, 13).astype("float32")
-    a = paddle.to_tensor(value)
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.Adam(learning_rate=0.02,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.Optimizer.clear_grad
 
 set_lr(value)
 '''''''''
@@ -162,26 +101,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-
-    adam = paddle.optimizer.Adam(0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        adam.set_lr(lr_list[i])
-        lr = adam.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.Optimizer.set_lr
 
 get_lr()
 '''''''''
@@ -199,36 +119,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-    import numpy as np
-    import paddle
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    adam = paddle.optimizer.Adam(0.001, parameters = emb.parameters())
-    lr = adam.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    inp = np.random.uniform(-0.1, 0.1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    adam = paddle.optimizer.Adam(scheduler,
-                           parameters=linear.parameters())
-
-    # first step: learning rate is 0.2
-    np.allclose(adam.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        adam.step()
-        lr = adam.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.Optimizer.get_lr
diff --git a/docs/api/paddle/optimizer/RMSProp_cn.rst b/docs/api/paddle/optimizer/RMSProp_cn.rst
index e1c06d497a6..f93e61a2c8e 100755
--- a/docs/api/paddle/optimizer/RMSProp_cn.rst
+++ b/docs/api/paddle/optimizer/RMSProp_cn.rst
@@ -48,21 +48,7 @@ RMSProp
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.rand([10,10], dtype="float32")
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    rmsprop = paddle.optimizer.RMSProp(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    out.backward()
-    rmsprop.step()
-    rmsprop.clear_grad()
+COPY-FROM: paddle.optimizer.RMSProp
 
 方法
 ::::::::::::
@@ -82,17 +68,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    a = paddle.rand([2,13], dtype="float32")
-    linear = paddle.nn.Linear(13, 5)
-    rmsprop = paddle.optimizer.RMSProp(learning_rate = 0.01,
-                                parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    rmsprop.step()
-    rmsprop.clear_grad()
+COPY-FROM: paddle.optimizer.RMSProp.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -113,23 +89,9 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.rand([10,10], dtype="float32")
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    rmsprop = paddle.optimizer.RMSProp(learning_rate=0.1,
-            parameters=linear.parameters(),
-            weight_decay=0.01)
-    out.backward()
-    rmsprop.step()
-    rmsprop.clear_grad()
+COPY-FROM: paddle.optimizer.RMSProp.minimize
 
-clear_gradients()
+clear_grad(set_to_zero=True)
 '''''''''
 
 .. note::
@@ -141,18 +103,7 @@ clear_gradients()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    a = paddle.rand([2,13], dtype="float32")
-    linear = paddle.nn.Linear(13, 5)
-    rmsprop = paddle.optimizer.RMSProp(learning_rate=0.02,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    rmsprop.step()
-    rmsprop.clear_gradients()
+COPY-FROM: paddle.optimizer.RMSProp.clear_grad
 
 set_lr(value)
 '''''''''
@@ -173,26 +124,7 @@ set_lr(value)
 
 **代码示例**
 
-.. code-block:: python
-
-
-    import paddle
-
-    linear = paddle.nn.Linear(10, 10)
-    rmsprop = paddle.optimizer.RMSProp(0.1, parameters=linear.parameters())
-
-    # set learning rate manually by python float value
-    lr_list = [0.2, 0.3, 0.4, 0.5, 0.6]
-    for i in range(5):
-        rmsprop.set_lr(lr_list[i])
-        lr = rmsprop.get_lr()
-        print("current lr is {}".format(lr))
-    # Print:
-    #    current lr is 0.2
-    #    current lr is 0.3
-    #    current lr is 0.4
-    #    current lr is 0.5
-    #    current lr is 0.6
+COPY-FROM: paddle.optimizer.RMSProp.set_lr
 
 get_lr()
 '''''''''
@@ -210,35 +142,4 @@ float，当前步骤的学习率。
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-    # example1: _LRScheduler is not used, return value is all the same
-    emb = paddle.nn.Embedding(10, 10, sparse=False)
-    rmsprop = paddle.optimizer.RMSProp(0.001, parameters = emb.parameters())
-    lr = rmsprop.get_lr()
-    print(lr) # 0.001
-
-    # example2: StepDecay is used, return the step learning rate
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.rand([10,10], dtype="float32")
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    bd = [2, 4, 6, 8]
-    value = [0.2, 0.4, 0.6, 0.8, 1.0]
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=2, gamma=0.1)
-    rmsprop = paddle.optimizer.RMSProp(scheduler,
-                           parameters=linear.parameters())
-
-    # first step: learning rate is 0.2
-    np.allclose(rmsprop.get_lr(), 0.2, rtol=1e-06, atol=0.0) # True
-
-    # learning rate for different steps
-    ret = [0.2, 0.2, 0.4, 0.4, 0.6, 0.6, 0.8, 0.8, 1.0, 1.0, 1.0, 1.0]
-    for i in range(12):
-        rmsprop.step()
-        lr = rmsprop.get_lr()
-        scheduler.step()
-        np.allclose(lr, ret[i], rtol=1e-06, atol=0.0) # True
+COPY-FROM: paddle.optimizer.RMSProp.get_lr
diff --git a/docs/api/paddle/optimizer/SGD_cn.rst b/docs/api/paddle/optimizer/SGD_cn.rst
index 20d7df05516..48dfec1b2d7 100644
--- a/docs/api/paddle/optimizer/SGD_cn.rst
+++ b/docs/api/paddle/optimizer/SGD_cn.rst
@@ -28,20 +28,7 @@ SGD
 代码示例
 ::::::::::::
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    inp = paddle.to_tensor(inp)
-    out = linear(inp)
-    loss = paddle.mean(out)
-    sgd = paddle.optimizer.SGD(learning_rate=0.1, parameters=linear.parameters(), weight_decay=0.01)
-    out.backward()
-    sgd.step()
-    sgd.clear_grad()
-
+COPY-FROM: paddle.optimizer.SGD
 
 方法
 ::::::::::::
@@ -60,17 +47,7 @@ step()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    sgd = paddle.optimizer.SGD(learning_rate=0.0003, parameters = linear.parameters())
-    out = linear(a)
-    out.backward()
-    sgd.step()
-    sgd.clear_grad()
+COPY-FROM: paddle.optimizer.SGD.step
 
 minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 '''''''''
@@ -91,22 +68,7 @@ minimize(loss, startup_program=None, parameters=None, no_grad_set=None)
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    inp = paddle.uniform(min=-0.1, max=0.1, shape=[10, 10], dtype='float32')
-    linear = paddle.nn.Linear(10, 10)
-    out = linear(inp)
-    loss = paddle.mean(out)
-
-    beta1 = paddle.to_tensor([0.9], dtype="float32")
-    beta2 = paddle.to_tensor([0.99], dtype="float32")
-
-    sgd = paddle.optimizer.SGD(learning_rate=0.0003, parameters=linear.parameters())
-    out.backward()
-    sgd.minimize(loss)
-    sgd.clear_grad()
+COPY-FROM: paddle.optimizer.SGD.minimize
 
 clear_grad()
 '''''''''
@@ -120,19 +82,7 @@ clear_grad()
 
 **代码示例**
 
-.. code-block:: python
-
-    import paddle
-
-    value = paddle.arange(26, dtype='float32')
-    a = paddle.reshape(value, [2, 13])
-    linear = paddle.nn.Linear(13, 5)
-    optimizer = paddle.optimizer.SGD(learning_rate=0.0003,
-                                     parameters=linear.parameters())
-    out = linear(a)
-    out.backward()
-    optimizer.step()
-    optimizer.clear_grad()
+COPY-FROM: paddle.optimizer.SGD.clear_grad
 
 set_lr(value)
 '''''''''
diff --git a/docs/api/paddle/optimizer/lr/LRScheduler_cn.rst b/docs/api/paddle/optimizer/lr/LRScheduler_cn.rst
index 4448c498e64..7ec4bc52444 100644
--- a/docs/api/paddle/optimizer/lr/LRScheduler_cn.rst
+++ b/docs/api/paddle/optimizer/lr/LRScheduler_cn.rst
@@ -56,34 +56,7 @@ LRScheduler
 代码示例
 ::::::::::::
 
-这里提供了重载基类 ``LRScheduler`` 并实现 ``StepLR`` 的示例，你可以根据你的需求来实现任意子类。
-
-.. code-block:: python
-
-    import paddle
-    from paddle.optimizer.lr import LRScheduler
-
-    class StepDecay(LRScheduler):
-        def __init__(self,
-                    learning_rate,
-                    step_size,
-                    gamma=0.1,
-                    last_epoch=-1,
-                    verbose=False):
-            if not isinstance(step_size, int):
-                raise TypeError(
-                    "The type of 'step_size' must be 'int', but received %s." %
-                    type(step_size))
-            if gamma >= 1.0:
-                raise ValueError('gamma should be < 1.0.')
-
-            self.step_size = step_size
-            self.gamma = gamma
-            super().__init__(learning_rate, last_epoch, verbose)
-
-        def get_lr(self):
-            i = self.last_epoch // self.step_size
-            return self.base_lr * (self.gamma**i)
+COPY-FROM: paddle.optimizer.lr.LRScheduler
 
 方法
 ::::::::::::
@@ -102,27 +75,7 @@ step 函数需要在优化器的 `optimizer.step()` 函数之后调用，调用
 
 **代码示例**
 
-请参考 ``基类 LRScheduler`` 的任意子类实现，这里以 ``StepLR`` 为例进行了示例：
-
-.. code-block:: python
-
-    import paddle
-    import numpy as np
-
-    x = np.random.uniform(-1, 1, [10, 10]).astype("float32")
-    linear = paddle.nn.Linear(10, 10)
-    scheduler = paddle.optimizer.lr.StepDecay(learning_rate=0.5, step_size=5, gamma=0.8, verbose=True)
-    sgd = paddle.optimizer.SGD(learning_rate=scheduler, parameters=linear.parameters())
-    for epoch in range(20):
-        for batch_id in range(2):
-            x = paddle.to_tensor(x)
-            out = linear(x)
-            loss = paddle.mean(out)
-            loss.backward()
-            sgd.step()
-            sgd.clear_gradients()
-            scheduler.step()    # If you update learning rate each step
-      # scheduler.step()        # If you update learning rate each epoch
+COPY-FROM: paddle.optimizer.lr.StepLR
 
 get_lr()
 '''''''''