Skip to content

Conversation

hynky1999
Copy link
Collaborator

Goal

Add Hellaswag tasks supporting about 36 unique languages.

Since it uses template it support all 3 types of formulation out of the box :)

How to test:

lighteval accelerate --output_dir=./tmp --custom_tasks="lighteval.tasks.multilingual.tasks" --tasks="custom|{task}|0|0" --model_args=pretrained=gpt2 --override_batch_size=1 --max_samples=100 --save_details

where task in hellaswag_tel_cf|0|0 hellaswag_tur_cf|0|0 hellaswag_tha_cf|0|0

Comments

  • We talked about removing the suite all together. Since right now we use suite I decided to use custom suite. I can switch to multilingual or whatever

@hynky1999 hynky1999 changed the base branch from multilang_copa_task to main September 30, 2024 18:16
@hynky1999 hynky1999 changed the title Multilingual Hellaswg tasks Multilingual Hellaswag tasks Oct 1, 2024
Language.SINDHI: TranslationLiterals(language=Language.SINDHI),
Language.TURKMEN: TranslationLiterals(language=Language.TURKMEN),
Language.SOUTH_AZERBAIJANI: TranslationLiterals(language=Language.SOUTH_AZERBAIJANI),
Language.SORANI: TranslationLiterals(language=Language.SORANI),
Copy link
Member

Choose a reason for hiding this comment

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

In the future, would be nice to sort this one alphabetically to avoid duplicates

@hynky1999 hynky1999 merged commit 994fe73 into main Oct 1, 2024
2 checks passed
hynky1999 added a commit that referenced this pull request May 22, 2025
* add multilignaul dynamic generative metrics

* draft

* finish multichoice config

* update tokenizers + install nltk reqs

* use punkt tab

* Update src/lighteval/utils/imports.py

Co-authored-by: Nathan Habib <[email protected]>

* Update src/lighteval/metrics/normalizations.py

Co-authored-by: Nathan Habib <[email protected]>

* fix imports

* remove unused import

* finish implementation of templates + move stuff around

* resolve nits

* when in rome do as romans do (handle error messages the same way)

* fix utils

* nicers tests + fix them

* nicer todo

* add nice doscrings 📃

* add even more docstring

* nit

* fix test

* add multilingual to dev group

* merge nli, add languagees to literals

* translation literals

* add nli

* add copa tasks + fix tranlation literals

* add hellaswag tasks

* remove custom telgu hellaswag

* remove hindi hellaswag

* add rcb + chinese nli

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* add two new tasks + docs

* add nice docs

* update hellaswag with docs

* move hellaswag to lighteval suite

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* enable returning none from templates + better typing

* change unoficial hellaswag names to have community_prefix + unify hellaswag preprocesisng

* let strip be optional in hellaswag

---------

Co-authored-by: Nathan Habib <[email protected]>
Co-authored-by: Hynek Kydlicek <[email protected]>
Co-authored-by: Clémentine Fourrier <[email protected]>
NathanHB added a commit that referenced this pull request Sep 19, 2025
* add multilignaul dynamic generative metrics

* draft

* finish multichoice config

* update tokenizers + install nltk reqs

* use punkt tab

* Update src/lighteval/utils/imports.py

Co-authored-by: Nathan Habib <[email protected]>

* Update src/lighteval/metrics/normalizations.py

Co-authored-by: Nathan Habib <[email protected]>

* fix imports

* remove unused import

* finish implementation of templates + move stuff around

* resolve nits

* when in rome do as romans do (handle error messages the same way)

* fix utils

* nicers tests + fix them

* nicer todo

* add nice doscrings 📃

* add even more docstring

* nit

* fix test

* add multilingual to dev group

* merge nli, add languagees to literals

* translation literals

* add nli

* add copa tasks + fix tranlation literals

* add hellaswag tasks

* remove custom telgu hellaswag

* remove hindi hellaswag

* add rcb + chinese nli

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* add two new tasks + docs

* add nice docs

* update hellaswag with docs

* move hellaswag to lighteval suite

* Update src/lighteval/tasks/multilingual/tasks.py

Co-authored-by: Clémentine Fourrier <[email protected]>

* enable returning none from templates + better typing

* change unoficial hellaswag names to have community_prefix + unify hellaswag preprocesisng

* let strip be optional in hellaswag

---------

Co-authored-by: Nathan Habib <[email protected]>
Co-authored-by: Hynek Kydlicek <[email protected]>
Co-authored-by: Clémentine Fourrier <[email protected]>
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.

3 participants