diff --git a/egs/librispeech/ASR/.run_v2.sh.swp b/egs/librispeech/ASR/.run_v2.sh.swp
index 79249b15a..33fff072b 100644
Binary files a/egs/librispeech/ASR/.run_v2.sh.swp and b/egs/librispeech/ASR/.run_v2.sh.swp differ
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train.py.swp
index 72e1a41b1..b66bcc0a9 100644
Binary files a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train.py.swp and b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/.train.py.swp differ
diff --git a/icefall/.utils.py.swp b/icefall/.utils.py.swp
index 57ec46015..d34e2c3e6 100644
Binary files a/icefall/.utils.py.swp and b/icefall/.utils.py.swp differ
diff --git a/icefall/utils.py b/icefall/utils.py
index 7eb780a74..cf492c269 100644
--- a/icefall/utils.py
+++ b/icefall/utils.py
@@ -192,7 +192,7 @@ def encode_supervisions(
     each other.
     """
     try: start_frame = supervisions["start_frame"]
-    except: start_frame = [0 for i in range(len(supervisions["cut"]))]
+    except: start_frame = torch.IntTensor([0 for i in range(len(supervisions["cut"]))])
 
     try: num_frames = supervisions["num_frames"]
     except:
@@ -200,6 +200,7 @@ def encode_supervisions(
         for supervision in supervisions['cut']:
             try: num_frames.append(supervision.tracks[0].cut.recording.num_samples)
             except: num_frames.append(supervision.recording.num_samples)
+        num_frames = torch.IntTensor(num_frames)
 
     supervision_segments = torch.stack(
         (