ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update rl.py

Browse source
This commit is contained in:
Daniel Han 2025-05-28 05:09:03 -07:00
commit 021267f8ec
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/models/rl.py
View file

@ -494,7 +494,7 @@ def _patch_trl_rl_trainers(trainer_file = "grpo_trainer"):
" elif scale_rewards == True:\n"\
" print('The Dr GRPO paper recommends setting `scale_rewards` to False! Will override. Set it to `None` to force False.')\n"\
" scale_rewards = False\n"\
"elif loss_type.lower() == 'dapo'\n"\
"elif loss_type.lower() == 'dapo':\n"\
" print('The DAPO paper recommends `mask_truncated_completions = True`')\n"\
" print('The DAPO paper recommends `epsilon_high = 0.28`')\n"\
" mask_truncated_completions = True\n"\