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 8f4aef777..c47d9155f 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/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py
index 0abe8b37c..053f1c981 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py
@@ -1073,9 +1073,10 @@ def train_one_epoch(
             # of the grad scaler is configurable, but we can't configure it to have different
             # behavior depending on the current grad scale.
             cur_grad_scale = scaler._scale.item()
+            '''
             if cur_grad_scale < 1.0 or (cur_grad_scale < 8.0 and batch_idx % 400 == 0):
-                continue
-                #scaler.update(cur_grad_scale * 2.0)
+                scaler.update(cur_grad_scale * 2.0)
+            '''
             if cur_grad_scale < 0.01:
                 logging.warning(f"Grad scale is small: {cur_grad_scale}")
             if cur_grad_scale < 1.0e-05: