diff --git a/mindspore/nn/dynamic_lr.py b/mindspore/nn/dynamic_lr.py
index 7d31eee566..81f46029e6 100644
--- a/mindspore/nn/dynamic_lr.py
+++ b/mindspore/nn/dynamic_lr.py
@@ -262,15 +262,20 @@ def polynomial_decay_lr(learning_rate, end_learning_rate, total_step, step_per_e
         (1 - tmp\_epoch / tmp\_decay\_epoch)^{power} + end\_learning\_rate
 
     Where:
+
     .. math::
-        `tmp\_epoch = min(current\_epoch, decay\_epoch),
-        current\_epoch=floor(\frac{i}{step\_per\_epoch})`,
+        tmp\_epoch = min(current\_epoch, decay\_epoch)
+
     .. math::
-        `tmp\_decay\_epoch = decay\_epoch`.
+        current\_epoch=floor(\frac{i}{step\_per\_epoch})
+
+    .. math::
+        tmp\_decay\_epoch = decay\_epoch
 
     If `update_decay_epoch` is true, update the value of `tmp_decay_epoch` every epoch. The formula is:
+
     .. math::
-        `tmp\_decay\_epoch = decay\_epoch * ceil(current\_epoch / decay\_epoch)`
+        tmp\_decay\_epoch = decay\_epoch * ceil(current\_epoch / decay\_epoch)
 
     Args:
         learning_rate (float): The initial value of learning rate.
diff --git a/mindspore/nn/wrap/loss_scale.py b/mindspore/nn/wrap/loss_scale.py
index 1dfc91743c..70464f7f28 100644
--- a/mindspore/nn/wrap/loss_scale.py
+++ b/mindspore/nn/wrap/loss_scale.py
@@ -194,7 +194,7 @@ class TrainOneStepWithLossScaleCell(Cell):
         >>> net_with_loss = Net()
         >>> optimizer = nn.Momentum(net_with_loss.trainable_params(), learning_rate=0.1, momentum=0.9)
         >>> manager = nn.DynamicLossScaleUpdateCell(loss_scale_value=2**12, scale_factor=2, scale_window=1000)
-        >>> train_network = nn.TrainOneStepWithLossScaleCell(net_with_loss, optimizer, scale_update_cell=manager)
+        >>> train_network = nn.TrainOneStepWithLossScaleCell(net_with_loss, optimizer, scale_sense=manager)
         >>> train_network.set_train()
         >>>
         >>> inputs = Tensor(np.ones([16, 16]).astype(np.float32))
diff --git a/tests/st/pynative/loss_scale/test_loss_scale.py b/tests/st/pynative/loss_scale/test_loss_scale.py
index b34fa257b6..c80bc4367f 100644
--- a/tests/st/pynative/loss_scale/test_loss_scale.py
+++ b/tests/st/pynative/loss_scale/test_loss_scale.py
@@ -164,6 +164,31 @@ def test_loss_scale_fp16_lr_overflow():
     assert output_1[0].asnumpy() == output_2[0].asnumpy()
     assert output_1[1].asnumpy() == output_2[1].asnumpy() == True
 
+@pytest.mark.level0
+@pytest.mark.platform_arm_ascend_training
+@pytest.mark.platform_x86_ascend_training
+@pytest.mark.env_onecard
+def test_loss_scale_fp16_lr_overflow_set_sense_scale():
+    inputs = Tensor(np.ones([16, 16]).astype(np.float32))
+    label = Tensor(np.zeros([16, 16]).astype(np.float32))
+    lr = Tensor(np.ones([1], np.float32) * 0.1)
+    net = NetFP16(16, 16)
+    net.set_train()
+
+    loss = MSELoss()
+    optimizer = Momentum(net.trainable_params(), learning_rate=lr, momentum=0.9)
+
+    net_with_loss = WithLossCell(net, loss)
+    train_network = TrainOneStepWithLossScaleCell(net_with_loss, optimizer,
+                                                  scale_sense=Tensor(np.full((1), np.finfo(np.float32).max),
+                                                                     dtype=mstype.float32))
+    output_1 = train_network(inputs, label)
+
+    train_network.set_sense_scale(Tensor(np.full((1), np.finfo(np.float32).max), dtype=mstype.float32))
+    output_2 = train_network(inputs, label)
+    assert output_1[0].asnumpy() == output_2[0].asnumpy()
+    assert output_1[1].asnumpy() == output_2[1].asnumpy() == True
+
 @pytest.mark.level0
 @pytest.mark.platform_arm_ascend_training
 @pytest.mark.platform_x86_ascend_training