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: 08f7e0862b9df353a0e4d8274617c1a5e6fa6619
This commit is contained in:
hiyouga 2023-05-28 21:48:33 +08:00
parent a4384e442c
commit 684afbf32c
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/utils/ppo.py
View File

@ -157,8 +157,8 @@ class PPOTrainerForLLaMA(PPOTrainer, PeftTrainer):
stats = self.step(queries, responses, rewards) stats = self.step(queries, responses, rewards)
loss_meter.update(stats["ppo/loss/total"]) loss_meter.update(stats["ppo/loss/total"], n=len(rewards))
reward_meter.update(torch.tensor(rewards).sum().item(), n=len(rewards)) reward_meter.update(torch.stack(rewards).mean().item(), n=len(rewards))
if steps_trained == len_dataloader: if steps_trained == len_dataloader:
dataiter = iter(self.dataloader) dataiter = iter(self.dataloader)