Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add custom task (bac-fr) for evaluation of models in french #518

Merged
merged 7 commits into from
Feb 3, 2025
Merged

Add custom task (bac-fr) for evaluation of models in french #518

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
73bc6ae
Add custom task (bac-fr) for evaluation of models in french
mdiazmel Jan 27, 2025
c592de7
Update prompt function for the bac-fr task
mdiazmel Jan 27, 2025
6b4de76
Add metrics for the evaluation of bac-fr
mdiazmel Jan 28, 2025
c791903
Fix function name (as_list)
mdiazmel Jan 29, 2025
fa123e8
Fix prompt for multichoice
mdiazmel Jan 29, 2025
b31f854
Merge branch 'main' into main
mdiazmel Jan 30, 2025
93a3452
Merge branch 'main' into main
mdiazmel Jan 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 31 additions & 1 deletion community_tasks/french_evals.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,18 @@ def prompt_gpqa_fr(line, task_name: str = None):
)


# BAC-fr prompt function
def prompt_bac_fr(line, task_name: str = None):
return Doc(
task_name=task_name,
query=line["prompt"],
choices=[""],
gold_index=0,
instruction="",
specific={"instruction": line["instruction"], "enonce": line["enonce"]},
)
mdiazmel marked this conversation as resolved.
Show resolved Hide resolved


# IFEVal-fr task


Expand Down Expand Up @@ -117,5 +129,23 @@ def prompt_gpqa_fr(line, task_name: str = None):
version=0,
)

# BAC-fr task
bac_fr_task = LightevalTaskConfig(
name="bac-fr",
suite=["community"],
prompt_function=prompt_bac_fr,
hf_repo="fr-gouv-coordination-ia/bac-fr",
hf_subset="default",
hf_avail_splits=["train"],
evaluation_splits=["train"],
few_shots_split=None,
few_shots_select="random_sampling",
generation_size=1,
metric=[], # To be defined
Copy link
Member

Choose a reason for hiding this comment

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

If we feel like the instructions constrain the answer enough, we could go for an exact match. We can also look at @hynky1999 's parser for math equations.

Copy link
Collaborator

@hynky1999 hynky1999 Jan 27, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

This is math bench ?

Copy link
Member

Choose a reason for hiding this comment

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

Includes math questions

Copy link
Member

Choose a reason for hiding this comment

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

(but not math only, I'd say 1/3 to 1/2 are math questions?)

Copy link
Contributor Author

@mdiazmel mdiazmel Jan 27, 2025 •
edited
Loading

Choose a reason for hiding this comment

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

I would say 2/3 math questions!

stop_sequence=["\n"],
trust_dataset=True,
version=0,
)

# STORE YOUR EVALS
TASKS_TABLE = [ifeval_fr_task, gpqa_fr_task]
TASKS_TABLE = [ifeval_fr_task, gpqa_fr_task, bac_fr_task]