diff --git a/fluid/neural_machine_translation/transformer/model.py b/fluid/neural_machine_translation/transformer/model.py
index 083dff5b76b2bf549fc0241663dc3dca977b5d10..73b77cf9d8d1e050a9b77fc14d4405baa20924cd 100644
--- a/fluid/neural_machine_translation/transformer/model.py
+++ b/fluid/neural_machine_translation/transformer/model.py
@@ -203,6 +203,7 @@ def prepare_encoder(src_word,
         size=[src_max_len, src_emb_dim],
         param_attr=fluid.ParamAttr(
             name=pos_enc_param_name, trainable=False))
+    src_pos_enc.stop_gradient = True
     enc_input = src_word_emb + src_pos_enc
     return layers.dropout(
         enc_input,