diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp
index 7341b56de..2ac3e2edb 100644
Binary files a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp and b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_audio.py.swp differ
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp
index e83ed7901..62735843d 100644
Binary files a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp and b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.data2vec_encoder.py.swp differ
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
index 819667c41..b3c482c1d 100644
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_encoder.py
@@ -65,7 +65,11 @@ class FairSeqData2VecEncoder(EncoderInterface):
         )
         model = models[0]
         model.feature_grad_mult = 0.0 ## for conv network freeze
-        model.mask_prob = 0.0 ## for conv network freeze
+        ## prevent overfitting
+        model.mask_prob = 0.65
+        model.mask_channel_prob = 0.5
+        model.mask_channel_length = 64
+        model.activation_dropout = 0.1 
         
         self.encoders = model
         self.pretrained_params = copy.deepcopy(model.state_dict())