diff --git a/paddlespeech/s2t/exps/u2_st/model.py b/paddlespeech/s2t/exps/u2_st/model.py
index 2dbbdcd30180418c523765c68c0869eb31e4d77c..144eed9d85f8ec6214c6263eb24059b18a8b96f0 100644
--- a/paddlespeech/s2t/exps/u2_st/model.py
+++ b/paddlespeech/s2t/exps/u2_st/model.py
@@ -432,6 +432,7 @@ class U2STTester(U2STTrainer):
         """"E2E translation from extracted audio feature"""
         cfg = self.config.decoding
         text_feature = self.test_loader.collate_fn.text_feature
+        self.model.eval()
 
         hyps = self.model.decode(
             audio,