GitLab

sync with remote

* add bertaqa tasks

* rename basquetrivia-->bertaqa ; make template stub not .yaml

* add bertaqa entry to lm_eval/tasks/README.md

---------
Co-authored-by: haileyschoelkopf <hailey@eleuther.ai>

* Added ArabicMMLU

* Rename `ammlu` to `arabicmmlu`

* log fewshot_as_multiturn in general tracker args

* Update evaluator.py

---------
Co-authored-by: Lintang Sutawika <lintang@eleuther.ai>

* init paloma benchmark

* pre-process in utils function

* add `task_alias`

* updated task aliases

* Update paloma_dolma-v1_5.yaml

* Update paloma_twitterAAE_HELM_fixed.yaml

* Update paloma_dolma_100_programing_languages.yaml

* update on names

* fix paloma template issue

---------
Co-authored-by: Zafir Stojanovski <zaf.stojano@gmail.com>
Co-authored-by: Zafir Stojanovski <zafir.stojanovski@icloud.com>
Co-authored-by: Lintang Sutawika <lintang@eleuther.ai>

* init paloma benchmark

* pre-process in utils function

* add `task_alias`

* updated task aliases

* Update paloma_dolma-v1_5.yaml

* Update paloma_twitterAAE_HELM_fixed.yaml

* Update paloma_dolma_100_programing_languages.yaml

---------
Co-authored-by: Lintang Sutawika <lintang@eleuther.ai>

Signed-off-by: changwangss <chang1.wang@intel.com>

* fix: add filter to os.walk to ignore 'ipynb_checkpoints

* Update __init__.py

* Update __init__.py

---------
Co-authored-by: Lintang Sutawika <lintang@eleuther.ai>

Co-authored-by: lintangsutawika <lintang@eleuther.ai>

Resolve conflict.

* Update vllm_causallms.py

* adjust

---------
Co-authored-by: lintangsutawika <lintang@eleuther.ai>

* `samples` is newline delimited

* updated git and pre-commit

* appease pre-commit

* nit

* Revert back for now

* Revert for now

---------
Co-authored-by: Lintang Sutawika <lintang@eleuther.ai>

Fix bug where `self.max_tokens` was not set

* Update README.md

* Delete lm_eval/tasks/ammlu directory

* results filenames handling moved to utils

* zeno results handling fix

* tasks_for_model backward compatibility

* results files logic moved to tasks_for_model

* moved sanitize_model_name to utils

* sort metrics in output table

* update docstring in `consolidate_results`

* add tests for verifying consistency of table output

* update tests to account for floating point inconsistencies

* updated tests based on `pythia-14m`