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 0ccd9b774..769eac71a 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 97a21c4a0..17596e55f 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
@@ -495,7 +495,7 @@ class Data2VecAudioModel(BaseFairseqModel):
                 nonzero = padding_mask[i].nonzero()
                 try:
                     print(nonzero[0])
-                    conv_feat_all = torch.cat([conv_feat_all, features[i][nonzero[0]][:]])
+                    conv_feat_all = torch.cat([conv_feat_all, features[i, nonzero[0], :]])
                 except:
                     conv_feat_all = torch.cat([conv_feat_all, features])
             print(conv_feat_all.size())