Skip to content

5.5. random 1k battles elo sampling#58

Open
ErlisLushtaku wants to merge 1 commit into
pr32-split-v3/04-thinking-model-supportfrom
pr32-split-v3/05.5-elo-sampling
Open

5.5. random 1k battles elo sampling#58
ErlisLushtaku wants to merge 1 commit into
pr32-split-v3/04-thinking-model-supportfrom
pr32-split-v3/05.5-elo-sampling

Conversation

@ErlisLushtaku

@ErlisLushtaku ErlisLushtaku commented Jun 1, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Add Elo sampling controls and CLI wiring.

@ErlisLushtaku ErlisLushtaku changed the base branch from main to pr32-split-v3/04-thinking-model-support June 1, 2026 14:52
@ErlisLushtaku ErlisLushtaku requested a review from kargibora June 1, 2026 14:53
@ErlisLushtaku ErlisLushtaku force-pushed the pr32-split-v3/04-thinking-model-support branch from 673b5fb to d7f4232 Compare June 1, 2026 22:03
@ErlisLushtaku ErlisLushtaku force-pushed the pr32-split-v3/05.5-elo-sampling branch from cb8ea5b to 524f0f4 Compare June 1, 2026 22:04

@kargibora kargibora left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks Erlis! I left some comments, everything looks good to me but just some minor issues.

judge_extra_kwargs["max_model_len"] = args.max_judge_model_len
if args.chat_template is not None:
judge_extra_kwargs["chat_template"] = args.chat_template
judge_extra_kwargs.update(args.engine_kwargs)

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why we delete this? Is there a specific reason?

return slug or "model"


def _jsonable(value):

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Take a look at

def _to_jsonable(value: Any) -> Any:
. I think this is duplicate?

"index": int(index)
if isinstance(index, int | np.integer)
else str(index),
"question_id": str(row.get("question_id", "")),

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we can remove the default value, as without question_idwe fail anyways are select_seeded_random_arena_battles.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants