diff --git a/dygraph/transformer/train.py b/dygraph/transformer/train.py
index 658c082ac52abff467a8aa7160446c19cd3a8198..42a34d32545b8b73d3083ce733eb97b82ecaefc4 100644
--- a/dygraph/transformer/train.py
+++ b/dygraph/transformer/train.py
@@ -110,7 +110,7 @@ class ModelHyperParams(object):
     # to process after each sub-layer
     postprocess_cmd = "da"  # dropout + residual connection
     # random seed used in dropout for CE.
-    dropout_seed = None
+    dropout_seed = 0
     # the flag indicating whether to share embedding and softmax weights.
     # vocabularies in source and target should be same for weight sharing.
     weight_sharing = False