diff --git a/egs/librispeech/ASR/incremental_transf/.model.py.swp b/egs/librispeech/ASR/incremental_transf/.model.py.swp
index 3c123591e..4783f02d6 100644
Binary files a/egs/librispeech/ASR/incremental_transf/.model.py.swp and b/egs/librispeech/ASR/incremental_transf/.model.py.swp differ
diff --git a/egs/librispeech/ASR/incremental_transf/model.py b/egs/librispeech/ASR/incremental_transf/model.py
index ccf97aff4..22a2e413a 100644
--- a/egs/librispeech/ASR/incremental_transf/model.py
+++ b/egs/librispeech/ASR/incremental_transf/model.py
@@ -256,6 +256,7 @@ class Interformer(nn.Module):
         with torch.cuda.amp.autocast(enabled=False):
             for inter_output, pt_output in zip(inter_layer_outputs, pt_layer_outputs):
                 mse_loss = self.mse(inter_output, pt_output)
+                loss += mse_loss
             '''
             simple_loss, (px_grad, py_grad) = k2.rnnt_loss_smoothed(
                 lm=lm.float(),