diff --git a/python/paddle/framework/io.py b/python/paddle/framework/io.py
index 945c8160b47fbdc38ca387e66bc62ceafc3450ae..d794fce5e378dd71088636d7042d02f776a1fa93 100644
--- a/python/paddle/framework/io.py
+++ b/python/paddle/framework/io.py
@@ -228,7 +228,6 @@ def save(obj, path):
             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(
@@ -320,7 +319,6 @@ def load(path, **configs):
             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(
diff --git a/python/paddle/optimizer/lr.py b/python/paddle/optimizer/lr.py
index e4fb54c229f212f4dfe1378c8141cc7708b2b04e..2d5dc5d998e638eebf573f6d2c3c584052739629 100644
--- a/python/paddle/optimizer/lr.py
+++ b/python/paddle/optimizer/lr.py
@@ -1434,7 +1434,8 @@ class CosineAnnealingDecay(LRScheduler):
                     loss.backward()
                     sgd.step()
                     sgd.clear_gradients()
-                scheduler.step()
+                    scheduler.step()    # If you update learning rate each step
+              # scheduler.step()        # If you update learning rate each epoch
 
             # train on static graph mode
             paddle.enable_static()
@@ -1460,7 +1461,8 @@ class CosineAnnealingDecay(LRScheduler):
                             'y': np.random.randn(3, 4, 5).astype('float32')
                         },
                         fetch_list=loss.name)
-                scheduler.step()
+                    scheduler.step()    # If you update learning rate each step
+              # scheduler.step()        # If you update learning rate each epoch
     """
 
     def __init__(self,