diff --git a/src/axolotl/monkeypatch/trainer_grad_accum.py b/src/axolotl/monkeypatch/trainer_grad_accum.py
index d45ca5ed3..600187c63 100644
--- a/src/axolotl/monkeypatch/trainer_grad_accum.py
+++ b/src/axolotl/monkeypatch/trainer_grad_accum.py
@@ -49,6 +49,7 @@ ORIGINAL_LLAMA_FCLM_CODE = """
         cache_position=cache_position,
         **kwargs,
     )
+
     hidden_states = outputs[0]
     # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
     logits = self.lm_head(hidden_states[:, -num_logits_to_keep:, :])