diff --git a/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/.utils.py.swp b/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/.utils.py.swp
index 261bb3d47..ec4727030 100644
Binary files a/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/.utils.py.swp and b/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/.utils.py.swp differ
diff --git a/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/utils.py b/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/utils.py
new file mode 100644
index 000000000..91ffcd85c
--- /dev/null
+++ b/egs/LJSpeech/ASR/pruned_transducer_stateless_d2v_v2/utils.py
@@ -0,0 +1,17 @@
+import math
+import torch.nn.functional as F
+
+
+def pad_to_multiple(x, multiple, dim=-1, value=0):
+    # Inspired from https://github.com/lucidrains/local-attention/blob/master/local_attention/local_attention.py#L41
+    if x is None:
+        return None, 0
+    tsz = x.size(dim)
+    m = tsz / multiple
+    remainder = math.ceil(m) * multiple - tsz 
+    if m.is_integer():
+        return x, 0
+    pad_offset = (0,) * (-1 - dim) * 2 
+
+    return F.pad(x, (*pad_offset, 0, remainder), value=value), remainder
+~