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 dc672b2a1..f8e0962d1 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 afd8713e6..16792323e 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_d2v_v2/train.py
@@ -1036,11 +1036,15 @@ def train_one_epoch(
             elif not params.multi_optim and (batch_idx+1) % params.accum_grads == 0:
                 logging.info('3')
                 set_batch_count(model, params.batch_idx_train)
-                scheduler.step_batch(params.batch_idx_train)
-                scaler.step(optimizer)
-                scaler.update()
-                optimizer.zero_grad()
                 logging.info('4')
+                scheduler.step_batch(params.batch_idx_train)
+                logging.info('5')
+                scaler.step(optimizer)
+                logging.info('6')
+                scaler.update()
+                logging.info('7')
+                optimizer.zero_grad()
+                logging.info('8')
 
         except:  # noqa
             display_and_save_batch(batch, params=params, sp=sp)