diff --git a/examples/xformer/sparse_scaled_dot_product_attention.py b/examples/xformer/sparse_scaled_dot_product_attention.py
index 7146179f4dc8ac2edcc50cbf87eee210a417f4b0..1df85ae4e0b2437365a9f49df3676c609f61b37c 100644
--- a/examples/xformer/sparse_scaled_dot_product_attention.py
+++ b/examples/xformer/sparse_scaled_dot_product_attention.py
@@ -38,7 +38,7 @@ def masked_select(input, mask):
 
 
 class BigBirdWrapper(pgl.graph_wrapper.BaseGraphWrapper):
-    """Implement of Edge Drop """
+    """Implement of Big Bird by PGL graph wrapper """
     def __init__(self, input_mask):
         super(BigBirdWrapper, self).__init__()
         max_seqlen = L.shape(input_mask)[1]