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

Update rl_replacements.py

Browse source
This commit is contained in:
Daniel Han 2025-03-14 06:11:19 -07:00
commit 22eeee212d
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_replacements.py
View file

@ -237,9 +237,9 @@ grpo_compute_loss_slow = RL_REPLACEMENTS["grpo_compute_loss_slow"]
UnslothEfficientGRPO = RL_REPLACEMENTS["UnslothEfficientGRPO"]
grpo_accumulated_loss = RL_REPLACEMENTS["grpo_accumulated_loss"]
RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss))
RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_compute_loss_slow))
RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(UnslothEfficientGRPO))
RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_accumulated_loss))
RL_PRE_ITEMS["grpo_trainer"].append(grpo_compute_loss_slow)
# Edit _get_per_token_logps to handle mixed precision
def grpo_trainer_compute_loss(function_name, function):