diff --git a/python/paddle/hapi/model.py b/python/paddle/hapi/model.py
index 21e3054dde7d7f0e196cfd32298d1a719ea0fa97..4f36effe6dd62e97ccbdf2c6eed422ef3325f64c 100644
--- a/python/paddle/hapi/model.py
+++ b/python/paddle/hapi/model.py
@@ -453,6 +453,12 @@ class StaticGraphAdapter(object):
             if len(name) > 0:
                 rets.insert(i, feed[name])
 
+        # step learning rate scheduler on each batch end
+        if self.model._optimizer and \
+                isinstance(self.model._optimizer._learning_rate,
+                           paddle.optimizer.lr.LRScheduler):
+            self.model._optimizer._learning_rate.step()
+
         # LoDTensor cannot be fetch as numpy directly
         rets = [np.array(v) for v in rets]
         if self.mode == 'test':
@@ -652,6 +658,13 @@ class DynamicGraphAdapter(object):
 
         self.model._optimizer.minimize(final_loss)
         self.model.network.clear_gradients()
+
+        # step learning rate scheduler on each batch end
+        if self.model._optimizer and \
+                isinstance(self.model._optimizer._learning_rate,
+                           paddle.optimizer.lr.LRScheduler):
+            self.model._optimizer._learning_rate.step()
+
         metrics = []
         for metric in self.model._metrics:
             metric_outs = metric.compute(*(to_list(outputs) + labels))
diff --git a/python/paddle/tests/test_model.py b/python/paddle/tests/test_model.py
index 46ea5ec9952f106ae6f04d991050748ca8f5ab98..5cd3e3664160e65b767b7fb551973bc0afae4eb4 100644
--- a/python/paddle/tests/test_model.py
+++ b/python/paddle/tests/test_model.py
@@ -33,7 +33,7 @@ from paddle.nn.layer.loss import CrossEntropyLoss
 from paddle.metric import Accuracy
 from paddle.vision.datasets import MNIST
 from paddle.vision.models import LeNet
-from paddle.io import DistributedBatchSampler
+from paddle.io import DistributedBatchSampler, Dataset
 from paddle.hapi.model import prepare_distributed_context
 from paddle.fluid.dygraph.jit import declarative
 from paddle.fluid.dygraph.dygraph_to_static.program_translator import ProgramTranslator
@@ -295,6 +295,15 @@ class MyModel(paddle.nn.Layer):
         return y
 
 
+class MyDataset(Dataset):
+    def __getitem__(self, idx):
+        return np.random.random(size=(20,)).astype(np.float32), \
+                np.random.randint(0, 10, size=(1,)).astype(np.int64)
+
+    def __len__(self):
+        return 40
+
+
 class TestModelFunction(unittest.TestCase):
     def set_seed(self, seed=1024):
         paddle.manual_seed(seed)
@@ -599,6 +608,56 @@ class TestModelFunction(unittest.TestCase):
         shutil.rmtree(save_dir)
 
 
+class TestModelWithLRScheduler(unittest.TestCase):
+    def test_fit(self):
+        def make_optimizer(parameters=None):
+            base_lr = 1e-3
+            momentum = 0.9
+            weight_decay = 5e-4
+            boundaries = [5, 8]
+            values = [base_lr * (0.1**i) for i in range(len(boundaries) + 1)]
+            learning_rate = paddle.optimizer.lr.PiecewiseDecay(
+                boundaries=boundaries, values=values)
+            learning_rate = paddle.optimizer.lr.LinearWarmup(
+                learning_rate=learning_rate,
+                warmup_steps=4,
+                start_lr=base_lr / 5.,
+                end_lr=base_lr,
+                verbose=True)
+            optimizer = paddle.optimizer.Momentum(
+                learning_rate=learning_rate,
+                weight_decay=weight_decay,
+                momentum=momentum,
+                parameters=parameters)
+            return optimizer
+
+        # dynamic test
+        device = paddle.set_device('cpu')
+        fluid.enable_dygraph(device)
+        net = MyModel()
+        inputs = [InputSpec([None, 20], 'float32', 'x')]
+        labels = [InputSpec([None, 1], 'int64', 'label')]
+        optim = make_optimizer(net.parameters())
+        model = Model(net, inputs, labels)
+        model.prepare(optimizer=optim, loss=CrossEntropyLoss(reduction="sum"))
+
+        dataset = MyDataset()
+        model.fit(dataset, dataset, batch_size=4, epochs=10, num_workers=0)
+
+        # static test
+        paddle.enable_static()
+
+        net = MyModel()
+        inputs = [InputSpec([None, 20], 'float32', 'x')]
+        labels = [InputSpec([None, 1], 'int64', 'label')]
+        optim = make_optimizer(net.parameters())
+        model = Model(net, inputs, labels)
+        model.prepare(optimizer=optim, loss=CrossEntropyLoss(reduction="sum"))
+
+        dataset = MyDataset()
+        model.fit(dataset, dataset, batch_size=4, epochs=10, num_workers=0)
+
+
 class TestRaiseError(unittest.TestCase):
     def test_input_without_name(self):
         net = MyModel()