423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-11-05 18:32:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

tiny fix

Browse Source 
Former-commit-id: 98c4a0af6b3e27ae393d2847f48a01d23d9c8780
This commit is contained in:
hiyouga 2024-07-02 23:06:13 +08:00
parent 1aaee45a94
commit 3f2b9e9326
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llamafactory/train/ppo/trainer.py
View File

@ -393,7 +393,7 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
else: else:
reward_model = self.reward_model reward_model = self.reward_model
with unwrap_model_for_generation(reward_model, self.accelerator), self.amp_context: # support bf16 with self.amp_context: # support bf16
_, _, values = reward_model(**batch, return_dict=True, use_cache=False) _, _, values = reward_model(**batch, return_dict=True, use_cache=False)
if self.finetuning_args.reward_model_type == "lora": if self.finetuning_args.reward_model_type == "lora":