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 2025-07-17 06:24:33 -07:00
commit 31df063c28
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

@ -237,7 +237,7 @@ RL_FUNCTIONS["grpo_trainer"].append(grpo_trainer__prepare_inputs)
# Fix incorrect special tokens handling and truncation in older TRL versions
def grpo_trainer__generate_and_score_completions(function_name, function):
if function_name != "_prepare_inputs": return function
if function_name != "_generate_and_score_completions": return function
# TRL 0.19.0 did skip_special_tokens = True which should be False
function = function.replace(