diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/.train.py.swp b/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/.train.py.swp
index da6260f33..753c3741b 100644
Binary files a/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/.train.py.swp and b/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/.train.py.swp differ
diff --git a/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py b/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py
index 49f747346..c7b43cbc5 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py
@@ -172,7 +172,7 @@ def add_model_arguments(parser: argparse.ArgumentParser):
         help="How many left context can be seen in chunks when calculating attention.",
     )
     parser.add_argument(
-        "--group-size",
+        "--group-num",
         type=int,
         default=12,
     )