diff --git a/docs/source/md/quick_start.md b/docs/source/md/quick_start.md
index e22696a60dabfb065b1d3ceb5667ffee7958cab0..e5dd193d150cf7434fa42947899425ba7224fb98 100644
--- a/docs/source/md/quick_start.md
+++ b/docs/source/md/quick_start.md
@@ -89,8 +89,11 @@ while not trainer.stop():
     step_i += 1
     print("batch %d start train" % (step_i))
     trainer.run(feed=data, fetch=[])
-    if step_i % 100 == 0:
+    if trainer_id == 0:
+        print("start saving model")
         trainer.save_inference_program(output_folder)
+    if step_i >= 100:
+       break
 ```
 
 On FL Server Node, a training script is defined as follows: