tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

fix: GRPO config not accept max_prompt_length (#3390) [skip ci]

Browse Source
This commit is contained in:
NanoCode012
2026-02-10 17:52:09 +07:00
committed by GitHub
parent b6d3653f74
commit ed7105dba7
1 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/axolotl/core/trainers/grpo/__init__.py
View File

@@ -164,7 +164,12 @@ class GRPOStrategy:
@classmethod @classmethod
def get_blocklist_args_kwargs(cls) -> list[str]: def get_blocklist_args_kwargs(cls) -> list[str]:
return ["dataset_num_proc", "max_length", "include_tokens_per_second"] return [
"dataset_num_proc",
"max_length",
"include_tokens_per_second",
"max_prompt_length",
]
@classmethod @classmethod
def get_reward_func(cls, reward_func_fqn: str) -> RewardFunc: def get_reward_func(cls, reward_func_fqn: str) -> RewardFunc: