423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-15 03:10:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

tiny fix about badam

Browse Source
This commit is contained in:
hiyouga
2024-06-25 01:54:53 +08:00
parent efb81b25ec
commit 095fab58d3
17 changed files with 31 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/llamafactory/train/ppo/trainer.py
View File

@@ -166,7 +166,8 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
self.reward_model = self.accelerator.prepare_model(self.reward_model, evaluation_mode=True)
if finetuning_args.use_badam:
from badam import clip_grad_norm_old_version, BAdamCallback
from badam import BAdamCallback, clip_grad_norm_old_version
self.accelerator.clip_grad_norm_ = MethodType(clip_grad_norm_old_version, self.accelerator)
self.callback_handler.add_callback(BAdamCallback)