diff --git a/doc/fluid/api_cn/nn_cn/SyncBatchNorm_cn.rst b/doc/fluid/api_cn/nn_cn/SyncBatchNorm_cn.rst
index c4d2b742d452e9f702a20a1af6d73f24332fc5bb..6e6b4615c7db330eaf0379762ffd61d0022680d2 100644
--- a/doc/fluid/api_cn/nn_cn/SyncBatchNorm_cn.rst
+++ b/doc/fluid/api_cn/nn_cn/SyncBatchNorm_cn.rst
@@ -62,3 +62,27 @@ SyncBatchNorm
         print(hidden1.numpy())
         # [[[[0.26824948, 1.0936325],[0.26824948, -1.6301316]],[[ 0.8095662, -0.665287],[-1.2744656, 1.1301866 ]]]]
 
+
+方法
+:::::::::
+convert_sync_batchnorm(layer)
+'''''''''''''''''''''''''''''
+
+该接口用于把 ``BatchNorm*d`` 层转换为 ``SyncBatchNorm`` 层。
+
+参数：
+    - **layer** (paddle.nn.Layer) - 包含一个或多个 ``BatchNorm*d`` 层的模型。
+
+返回：
+    如果原始模型中有 ``BatchNorm*d`` 层， 则把 ``BatchNorm*d`` 层转换为 ``SyncBatchNorm`` 层的原始模型。
+
+**代码示例**
+
+.. code-block:: python
+
+    import paddle
+    import paddle.nn as nn
+    paddle.disable_static()
+    model = nn.Sequential(nn.Conv2d(3, 5, 3), nn.BatchNorm2d(5))
+    sync_model = nn.SyncBatchNorm.convert_sync_batchnorm(model)
+
diff --git a/doc/paddle/api/paddle/nn/layer/norm/SyncBatchNorm_cn.rst b/doc/paddle/api/paddle/nn/layer/norm/SyncBatchNorm_cn.rst
index c4d2b742d452e9f702a20a1af6d73f24332fc5bb..84618185a20913de4a73c76c7f9da6d2c32260ec 100644
--- a/doc/paddle/api/paddle/nn/layer/norm/SyncBatchNorm_cn.rst
+++ b/doc/paddle/api/paddle/nn/layer/norm/SyncBatchNorm_cn.rst
@@ -62,3 +62,26 @@ SyncBatchNorm
         print(hidden1.numpy())
         # [[[[0.26824948, 1.0936325],[0.26824948, -1.6301316]],[[ 0.8095662, -0.665287],[-1.2744656, 1.1301866 ]]]]
 
+方法
+:::::::::
+convert_sync_batchnorm(layer)
+'''''''''''''''''''''''''''''
+
+该接口用于把 ``BatchNorm*d`` 层转换为 ``SyncBatchNorm`` 层。
+
+参数：
+    - **layer** (paddle.nn.Layer) - 包含一个或多个 ``BatchNorm*d`` 层的模型。
+
+返回：
+    如果原始模型中有 ``BatchNorm*d`` 层， 则把 ``BatchNorm*d`` 层转换为 ``SyncBatchNorm`` 层的原始模型。
+
+**代码示例**
+
+.. code-block:: python
+
+    import paddle
+    import paddle.nn as nn
+    paddle.disable_static()
+    model = nn.Sequential(nn.Conv2d(3, 5, 3), nn.BatchNorm2d(5))
+    sync_model = nn.SyncBatchNorm.convert_sync_batchnorm(model)
+