diff --git a/doc/BAIDU_KUNLUN_XPU_SERVING.md b/doc/BAIDU_KUNLUN_XPU_SERVING.md
index 7bc501d634ae6ac6537642e41a43160ca866de49..c57ce515096253678c9222c96a3e57fcd9dd91e7 100644
--- a/doc/BAIDU_KUNLUN_XPU_SERVING.md
+++ b/doc/BAIDU_KUNLUN_XPU_SERVING.md
@@ -22,7 +22,7 @@ cmake -DPYTHON_INCLUDE_DIR=/usr/include/python3.7m/ \
     -DSERVER=ON ..
 make -j10
 ```
-You can run `make install` to produce the target in `./output` directory. Add `-DCMAKE_INSTALL_PREFIX=./output` to specify the output path to CMake command shown above。
+You can run `make install` to produce the target in `./output` directory. Add `-DCMAKE_INSTALL_PREFIX=./output` to specify the output path to CMake command shown above. Please specify `-DWITH_MKL=ON` on Intel CPU platform with AVX2 support.  
 * Compile the Serving Client
 ```
 mkdir -p client-build-arm && cd client-build-arm
diff --git a/python/examples/xpu/bert/README.md b/python/examples/xpu/bert/README.md
index fd8e756603db9298ffa66e4c79feabb3d47e2b3d..676ac361dde376b29ae73cae98bd34a28f55d469 100644
--- a/python/examples/xpu/bert/README.md
+++ b/python/examples/xpu/bert/README.md
@@ -26,7 +26,7 @@ this script will download Chinese Dictionary File vocab.txt and Chinese Sample D
 ### Start Service
 
 ```
-python3 bert_web_service.py serving_server 7703
+python3 -m paddle_serving_server.serve --model serving_server --port 7703 --use_lite --use_xpu --ir_optim
 ```
 
 ### Client Prediction
diff --git a/python/examples/xpu/bert/bert_client.py b/python/examples/xpu/bert/bert_client.py
index 66c46d7cdc8da21b06b877f5c550da2a58ac24f6..db10425baf1e803d7866dc839bc268cb95882eb0 100644
--- a/python/examples/xpu/bert/bert_client.py
+++ b/python/examples/xpu/bert/bert_client.py
@@ -31,7 +31,7 @@ client.connect(endpoint_list)
 for line in sys.stdin:
     feed_dict = reader.process(line)
     for key in feed_dict.keys():
-        feed_dict[key] = np.array(feed_dict[key]).reshape((128, 1))
+        feed_dict[key] = np.array(feed_dict[key]).reshape((1, 128))
     #print(feed_dict)
-    result = client.predict(feed=feed_dict, fetch=fetch, batch=False)
+    result = client.predict(feed=feed_dict, fetch=fetch, batch=True)
 print(result)