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 059862571..dec1fc554 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_audio.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
index 34ce40466..b2fff73d8 100644
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/data2vec_audio.py
@@ -494,7 +494,7 @@ class Data2VecAudioModel(BaseFairseqModel):
             prompt = prompt.expand((features.size()[0], prompt.size()[0], prompt.size()[1]))
             print(prompt.size())
             print(features.size())
-            features = torch.cat([prompt, features])
+            features = torch.stack([prompt, features], dim=1)
             print(features.size())
             prompt_padding_mask = torch.zeros(promt.size()).type(torch.BoolTensor).to(features.device)
             padding_mask = torch.cat([prompt_padding_mask, padding_mask])