From 648183a96037a0d9e758154f98e1e1b8004eea0b Mon Sep 17 00:00:00 2001 From: BlueRum <70618399+ht-zhou@users.noreply.github.com> Date: Thu, 16 Feb 2023 10:25:17 +0800 Subject: [PATCH] [chatgpt]fix train_rm bug with lora (#2741) --- applications/ChatGPT/chatgpt/trainer/rm.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/applications/ChatGPT/chatgpt/trainer/rm.py b/applications/ChatGPT/chatgpt/trainer/rm.py index c24289502..b76ae5373 100644 --- a/applications/ChatGPT/chatgpt/trainer/rm.py +++ b/applications/ChatGPT/chatgpt/trainer/rm.py @@ -43,7 +43,7 @@ class RewardModelTrainer(ABC): # train if use_lora > 0: print("Using Lora") - lora.mark_only_lora_as_trainable(self.model) + lora.mark_only_lora_as_trainable(self.model.model) else: self.model.train() for chosen_ids, c_mask, reject_ids, r_mask in self.train_dataloader: