423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 06:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #6050

Browse Source 
Former-commit-id: dc828218726704ff0453a2d13535663ac6ad7833
This commit is contained in:
hoshi-hiyouga 2024-11-16 16:11:16 +08:00 committed by GitHub
parent e3a0640659
commit 9815d1712c
1 changed files with 4 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/llamafactory/train/dpo/trainer.py
View File

@ -255,10 +255,10 @@ class CustomDPOTrainer(DPOTrainer):
metrics[f"{prefix}rewards/rejected"] = rejected_rewards.mean().item()
metrics[f"{prefix}rewards/accuracies"] = (chosen_rewards > rejected_rewards).float().mean().item()
metrics[f"{prefix}rewards/margins"] = (chosen_rewards - rejected_rewards).mean().item()
metrics[f"{prefix}logps/rejected"] = policy_chosen_logps.mean().item()
metrics[f"{prefix}logps/chosen"] = policy_rejected_logps.mean().item()
metrics[f"{prefix}logits/rejected"] = policy_chosen_logits.mean().item()
metrics[f"{prefix}logits/chosen"] = policy_rejected_logits.mean().item()
metrics[f"{prefix}logps/chosen"] = policy_chosen_logps.mean().item()
metrics[f"{prefix}logps/rejected"] = policy_rejected_logps.mean().item()
metrics[f"{prefix}logits/chosen"] = policy_chosen_logits.mean().item()
metrics[f"{prefix}logits/rejected"] = policy_rejected_logits.mean().item()
if self.loss_type == "orpo":
metrics[f"{prefix}sft_loss"] = sft_loss.mean().item()
metrics[f"{prefix}odds_ratio_loss"] = ((losses - sft_loss) / self.beta).mean().item()