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 2025-02-19 22:03:47 -08:00
commit 18dfabd8e5
1 changed files with 2 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

2
unsloth/models/rl_replacements.py
View file

@ -200,8 +200,10 @@ pass
RL_FUNCTIONS["grpo_trainer"].append(grpo_trainer__get_per_token_logps)
grpo_compute_loss = RL_REPLACEMENTS["grpo_compute_loss"]
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(UnslothEfficientGRPO))
RL_PRE_ITEMS["grpo_trainer"].append(inspect.getsource(grpo_accumulated_loss))
# Edit _get_per_token_logps to handle mixed precision