diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py
index 3d97b90df9..17215bafb8 100644
--- a/unsloth/models/rl_replacements.py
+++ b/unsloth/models/rl_replacements.py
@@ -201,6 +201,7 @@ RL_FUNCTIONS["grpo_trainer"].append(grpo_trainer__get_per_token_logps)
 grpo_compute_loss     = RL_REPLACEMENTS["grpo_compute_loss"]
 grpo_accumulated_loss = RL_REPLACEMENTS["grpo_accumulated_loss"]
 RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss))
+RL_PRE_ITEMS["grpo_accumulated_loss"].append(inspect.getsource(grpo_accumulated_loss))
 
 # Edit _get_per_token_logps to handle mixed precision
 def grpo_trainer_compute_loss(function_name, function):