From 22eeee212d855514a1a9ea237e63615901db458e Mon Sep 17 00:00:00 2001 From: Daniel Han Date: Fri, 14 Mar 2025 06:11:19 -0700 Subject: [PATCH] Update rl_replacements.py --- unsloth/models/rl_replacements.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py index b94fc55cf0..4071ef835a 100644 --- a/unsloth/models/rl_replacements.py +++ b/unsloth/models/rl_replacements.py @@ -237,9 +237,9 @@ grpo_compute_loss_slow = RL_REPLACEMENTS["grpo_compute_loss_slow"] UnslothEfficientGRPO = RL_REPLACEMENTS["UnslothEfficientGRPO"] grpo_accumulated_loss = RL_REPLACEMENTS["grpo_accumulated_loss"] RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss)) -RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss_slow)) RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(UnslothEfficientGRPO)) RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_accumulated_loss)) +RL_PRE_ITEMS["grpo_trainer"].append(grpo_compute_loss_slow) # Edit _get_per_token_logps to handle mixed precision def grpo_trainer_compute_loss(function_name, function):