diff --git a/examples/librispeech/s2/conf/transformer.yaml b/examples/librispeech/s2/conf/transformer.yaml
index d77329f50843e270b52750ef5dcc2e9429bd8617..de1ac347a8f2b57377c3493a804b8d07141f082f 100644
--- a/examples/librispeech/s2/conf/transformer.yaml
+++ b/examples/librispeech/s2/conf/transformer.yaml
@@ -1,6 +1,8 @@
 # https://yaml.org/type/float.html
 # network architecture
 model:
+    cmvn_file:  
+    cmvn_file_type: "json"
     # encoder related
     encoder: transformer
     encoder_conf:
diff --git a/paddlespeech/s2t/models/u2/u2.py b/paddlespeech/s2t/models/u2/u2.py
index fd9982716d26ff3f737173646c9fbaf3f7942255..9977cecc4bb2ecc337f7126cd7c1a38d95424358 100644
--- a/paddlespeech/s2t/models/u2/u2.py
+++ b/paddlespeech/s2t/models/u2/u2.py
@@ -860,7 +860,7 @@ class U2Model(U2DecodeModel):
             int, nn.Layer, nn.Layer, nn.Layer: vocab size, encoder, decoder, ctc
         """
         # cmvn
-        if configs['cmvn_file'] is not None:
+        if 'cmvn_file' in configs and configs['cmvn_file'] is not None:
             mean, istd = load_cmvn(configs['cmvn_file'],
                                    configs['cmvn_file_type'])
             global_cmvn = GlobalCMVN(