Irokobench: Benchmark Dataset for African languages (#2042)

* add afrixnli to task

* add chat completion

* remove chat completion -untested

* afrimmlu added

* afrimmlu folder update

* afrimmlu folder update

* updated prompt

* remove print

* add afrimgsm -direct

* add squad metric

* fix bash script

* remove direct util, update common yaml

* remove print

* add few show. metric fixes

* fix direct path, add bash script for gpt models

* added transate test

* update afrixnli tasks

* update afrixnli tasks

* update metrics for afrixnli

* prompt translations fix

* prompt translations fix

* filter and metric fix -mgsm

* remove squad metric

* remove squad metric

* add f1 score to mgsm

* add f1 score to mgsm

* update native-direct with lin

* change f1 function

* add lin to utils

* add utils

* remove test limit

* remove test configs

* add swahili to mmlu

* change eng to ewe in ewe yaml mmlu

* add squad metric to mgsm, remove whitespace filter

* added translate test

* added afrixnli_translate

* fix exact match valueError

* fix exact match valueError

* restructure mmlu folder

* spacing

* remove afrimmlu_translate folder

* add utility

* format task name, clean ups

* modefied mgsm

* update on afrimgsm

* update on afrimgsm

* removed utils

* other mgsm varieties

* other mgsm varieties

* adding trasnslate direct

* Update translate_direct_yaml

* add manual xnli prompt, add multichoice for openai models, and adapt multichoice metric for openai model

* edit for open models

* Update translate_direct_yaml

* add verbalizer for xnli

* change xnli from multiple choice to generate

* add manual accuracy scores

* revert xnli to multiple choice

* change afrimgsm utils

* revert xnli to multiple_choice

* cleanups and readmes

* remove openai fixes and unused regex

* pr review changes

* revert metrics.py, task.py and extraction.py to main version

---------
Co-authored-by: Israel Abebe Azime <azime@cg.uni-saarland.de>
Co-authored-by: Israel Abebe Azime <se.israel.abebe@gmail.com>

lm_eval/tasks/afrimgsm/en_cot/afrimgsm_en_cot_zul.yaml

+# Generated by utils.py
+dataset_name: zul
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nStep-by-Step Answer:"}}{% else %}{{"Question: "+question+"\nStep-by-Step Answer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: cot_yaml
+task: afrimgsm_en_cot_zul

lm_eval/tasks/afrimgsm/en_cot/cot_yaml

+# This file will be included in the generated language-specific task configs.
+# It doesn't have a yaml file extension as it is not meant to be imported directly by the harness.
+group:
+    - mgsm
+    - afrimgsm
+    - afrimgsm_en_cot
+dataset_path: masakhane/afrimgsm
+dataset_name: null  # Overridden by language-specific config.
+output_type: generate_until
+training_split: train
+test_split: test
+generation_kwargs:
+  until:
+    - "\n\n"
+    - "\n"
+  do_sample: false
+  temperature: 0.0
+target_delimiter: " "
+metric_list:
+  - metric: exact_match
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+filter_list:
+  - name: "strict-match"
+    filter:
+      - function: "regex"
+        regex_pattern: "The answer is (\\-?[0-9\\.\\,]+)"
+      - function: "take_first"
+  - filter:
+    - function: regex
+      group_select: -1
+      regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
+    - function: take_first
+    name: flexible-extract
+metadata:
+  version: 2.0

lm_eval/tasks/afrimgsm/gen_yaml.sh

+#!/bin/bash
+
+# python utils.py --overwrite --output-dir direct --mode direct
+# python utils.py --overwrite --output-dir direct_native --mode direct-native
+# python utils.py --overwrite --output-dir en_cot --mode en-cot
+# python utils.py --overwrite --output-dir native_cot --mode native-cot
+python utils.py --overwrite --output-dir translate_direct --mode translate-direct

lm_eval/tasks/afrimgsm/run.sh

+lm_eval --model hf   \
+        --model_args pretrained="google/gemma-7b"  --tasks afrimgsm_en_cot_eng,mgsm_en_cot_en,afrimgsm_native_cot_eng,mgsm_native_cot_en,afrimgsm_direct_eng,mgsm_direct_en,afrimgsm_direct_native_eng  \
+        --device cuda:0     \
+        --batch_size 1  \
+        --verbosity DEBUG \
+        --limit 5
\ No newline at end of file

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_amh.yaml

+# Generated by utils.py
+dataset_name: amh
+doc_to_target: '{% if answer is not none %}{{answer[15:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_amh

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_eng.yaml

+# Generated by utils.py
+dataset_name: eng
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_eng

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_ewe.yaml

+# Generated by utils.py
+dataset_name: ewe
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_ewe

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_fra.yaml

+# Generated by utils.py
+dataset_name: fra
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_fra

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_hau.yaml

+# Generated by utils.py
+dataset_name: hau
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_hau

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_ibo.yaml

+# Generated by utils.py
+dataset_name: ibo
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_ibo

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_kin.yaml

+# Generated by utils.py
+dataset_name: kin
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_kin

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_lin.yaml

+# Generated by utils.py
+dataset_name: lin
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_lin

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_lug.yaml

+# Generated by utils.py
+dataset_name: lug
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_lug

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_orm.yaml

+# Generated by utils.py
+dataset_name: orm
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_orm

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_sna.yaml

+# Generated by utils.py
+dataset_name: sna
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_sna

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_sot.yaml

+# Generated by utils.py
+dataset_name: sot
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_sot

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_swa.yaml

+# Generated by utils.py
+dataset_name: swa
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_swa

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_twi.yaml

+# Generated by utils.py
+dataset_name: twi
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_twi

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_wol.yaml

+# Generated by utils.py
+dataset_name: wol
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_wol

lm_eval/tasks/afrimgsm/translate/afrimgsm_translate_xho.yaml

+# Generated by utils.py
+dataset_name: xho
+doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
+doc_to_text: '{% if answer is not none %}{{question+"\nAnswer:"}}{% else %}{{"Question: "+question+"\nAnswer:"}}{% endif %}'
+generation_kwargs:
+  do_sample: false
+  until:
+  - 'Question:'
+  - </s>
+  - <|im_end|>
+include: translate_direct_yaml
+task: afrimgsm_translate_direct_xho