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 6ebeb46df..cb5a7e3ec 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 d35875c37..6b02a5156 100755
--- a/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py
+++ b/egs/librispeech/ASR/pruned_transducer_stateless_gtrans/train.py
@@ -1021,7 +1021,7 @@ def run(rank, world_size, args, wb=None):
     model.to(device)
     if world_size > 1:
         logging.info("Using DDP")
-        model = DDP(model, device_ids=[rank])
+        model = DDP(model, device_ids=[rank], find_unused_parameters=True)
 
     optimizer = Eve(model.parameters(), lr=params.initial_lr)