AfroBench: How Good are Large Language Models on African Languages? (#2825)

* 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

* add afrisenti

* utilities

* pulled from main

* add afrixnli

* add afrimmlu

* update afrixnli prompts

* mising senti language

* fix afrisenti prompt 2

* fix afrisenti prompts

* fix afrisenti prompts

* configure task grouping

* add multiple prompts to afrixnli for irokobench

* add multiple prompts to afrimmlu for irokobench

* Update afrixnli_yaml

* fixes and moves

* fixes and moves

* afrimmlu multiple prompts configs

* remove validation set from afrimmlu

* remove eng from afrimmlu translate test

* correct dataset path

* multiple prompts for mgsm

* file restructure

* afribench grouping

* repo restructuring

* repo restructuring

* update exact match to hugging face exact match and add new mgsm language

* remove decontamination

* update generation kwargs

* update generation kwargs for all mgsm prompts

* remove lang

* update generation kwargs for afrimgsm translatetest

* add afrimgsm cot for direct and translate

* remove eng from translate-cot

* add masakhaPOS tasks

* remove changes from task script

* add masakhanews tasks

* add uhura arc easy

* add afriqa and belebele files

* add tags for easier run. add naija rc

* add new metrics and transformation scripts

* fix afriqa swa fewshot split

* add naijarc

* add afrobench lite tasks

* update afrobench

* update afrobench

* remove unverified files to avoid bugs

* remove files not needed

* add afrobench tasks

* add afrobench tasks

* change to version 1

* change to version 1

* update afrobench

* update afrobench

* restore metric to original script

* update readme instructions

* add individual dataset readmes

* add link to collections

* correct run script

* align with main

* align with main

* align with main

* align with main

* align with main

* align with main

* align with main

* align with main

* failed run fixes

* failed run fixes

* add afrimgsm cot

* Apply precommit fixes

* update mafand dataset name

* pull request fixes

* remove afrihate due to availability

---------
Co-authored-by: Israel Abebe Azime <azime@cg.uni-saarland.de>
Co-authored-by: Israel Abebe Azime <se.israel.abebe@gmail.com>
Co-authored-by: David Adelani <davlanade@gmail.com>
Co-authored-by: theyorubayesian <akin.o.oladipo@gmail.com>

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_amh.yaml

+# Generated by utils.py
+dataset_name: amh
+include: afrisenti
+task: afrisenti_amh_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_arq.yaml

+# Generated by utils.py
+dataset_name: arq
+include: afrisenti
+task: afrisenti_arq_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_ary.yaml

+# Generated by utils.py
+dataset_name: ary
+include: afrisenti
+task: afrisenti_ary_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_hau.yaml

+# Generated by utils.py
+dataset_name: hau
+include: afrisenti
+task: afrisenti_hau_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_ibo.yaml

+# Generated by utils.py
+dataset_name: ibo
+include: afrisenti
+task: afrisenti_ibo_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_kin.yaml

+# Generated by utils.py
+dataset_name: kin
+include: afrisenti
+task: afrisenti_kin_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_orm.yaml

+# Generated by utils.py
+dataset_name: orm
+include: afrisenti
+task: afrisenti_orm_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_pcm.yaml

+# Generated by utils.py
+dataset_name: pcm
+include: afrisenti
+task: afrisenti_pcm_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_por.yaml

+# Generated by utils.py
+dataset_name: por
+include: afrisenti
+task: afrisenti_por_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_swa.yaml

+# Generated by utils.py
+dataset_name: swa
+include: afrisenti
+task: afrisenti_swa_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_tir.yaml

+# Generated by utils.py
+dataset_name: tir
+include: afrisenti
+task: afrisenti_tir_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_tso.yaml

+# Generated by utils.py
+dataset_name: tso
+include: afrisenti
+task: afrisenti_tso_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_twi.yaml

+# Generated by utils.py
+dataset_name: twi
+include: afrisenti
+task: afrisenti_twi_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/afrisenti_yor.yaml

+# Generated by utils.py
+dataset_name: yor
+include: afrisenti
+task: afrisenti_yor_prompt_1

lm_eval/tasks/afrobench/afrisenti/prompt_1/run.sh

+#!/bin/bash
+
+models=(
+
+  "google/gemma-1.1-7b-it"
+  "CohereForAI/aya-101"
+  "meta-llama/Llama-2-7b-chat-hf"
+  "meta-llama/Meta-Llama-3-8B-Instruct"
+  "google/gemma-2-9b-it"
+  "bigscience/mt0-xxl"
+  "google/gemma-2-27b-it"
+  "meta-llama/Meta-Llama-3-70B-Instruct"
+)
+task=afrisenti_amh_prompt_1,afrisenti_arq_prompt_1,afrisenti_ary_prompt_1,afrisenti_hau_prompt_1,afrisenti_ibo_prompt_1,afrisenti_kin_prompt_1,afrisenti_pcm_prompt_1,afrisenti_por_prompt_1,afrisenti_swa_prompt_1,afrisenti_tir_prompt_1,afrisenti_tso_prompt_1,afrisenti_twi_prompt_1,afrisenti_yor_prompt_1
+
+for model in "${models[@]}"
+do
+  echo "Evaluating model: $model"
+  for fewshot in 0 5
+  do
+    export OUTPUT_DIR=results/$fewshot
+
+    mkdir -p "$OUTPUT_DIR"
+
+    lm_eval --model hf \
+            --model_args "pretrained=${model}" \
+            --tasks $task\
+            --device cuda:0 \
+            --batch_size 16 \
+            --output_path "$OUTPUT_DIR" \
+            --num_fewshot $fewshot \
+            --verbosity DEBUG
+  done
+done

lm_eval/tasks/afrobench/afrisenti/prompt_1/utils.py

+from lm_eval.utils import weighted_f1_score

lm_eval/tasks/afrobench/afrisenti/prompt_1/xx.py

+from datasets import load_dataset
+
+
+# ['amh', 'hau', 'ibo', 'arq', 'ary', 'yor', 'por', 'twi', 'tso', 'tir', 'orm', 'pcm', 'kin', 'swa']
+
+data = load_dataset("masakhane/afrisenti", "pcm", trust_remote_code=True)
+print(data)
+print(data["test"][:5])
+#
+# ['Naija', 'Pipo', 'wey', 'dey', 'for', 'inside', 'social', 'Media', 'sef', 'don', 'put', 'hand', 'for', 'ear', 'give',
+#  'federal', 'goment', 'and', 'polical', 'leader', 'dem', 'ova', 'di', 'kilin', '.']
+#
+# [6, 0, 14, 17, 2, 2, 6, 0, 7, 17, 16, 0, 2, 0, 16, 0, 0, 9, 0, 0, 11, 2, 8, 0, 1]

lm_eval/tasks/afrobench/afrisenti/prompt_2/afrisenti

+tag:
+    - afrobench_sentiment_tasks
+    - afrisent_prompt_2
+dataset_path: masakhane/afrisenti
+dataset_name: null
+dataset_kwargs: {trust_remote_code: True}
+output_type: multiple_choice
+validation_split: validation
+test_split: test
+fewshot_split: train
+doc_to_target: label
+doc_to_choice:
+    - "negative"
+    - "positive"
+    - "neutral"
+should_decontaminate: true
+doc_to_decontamination_query: 'text: {{tweet}} \nlabel: '
+metric_list:
+  - metric: f1
+    aggregation: !function utils.weighted_f1_score
+    # aggregation: mean
+    average: weighted
+    hf_evaluate: true
+    higher_is_better: True
+    ignore_case: true
+    ignore_punctuation: true
+    regexes_to_ignore:
+      - ","
+      - "\\$"
+  - metric: acc
+    aggregation: mean
+    higher_is_better: true
+    ignore_case: true
+    ignore_punctuation: true
+    regexes_to_ignore:
+      - ","
+      - "\\$"
+metadata:
+  version: 1.0

lm_eval/tasks/afrobench/afrisenti/prompt_2/afrisenti_amh.yaml

+# Generated by utils.py
+dataset_name: amh
+doc_to_text: Does this Amharic statement; '{{tweet}}' have a Neutral, Positive or
+  Negative sentiment? Labels only
+include: afrisenti
+task: afrisenti_amh_prompt_2

lm_eval/tasks/afrobench/afrisenti/prompt_2/afrisenti_arq.yaml

+# Generated by utils.py
+dataset_name: arq
+doc_to_text: Does this Algerian Arabic statement; '{{tweet}}' have a Neutral, Positive
+  or Negative sentiment? Labels only
+include: afrisenti
+task: afrisenti_arq_prompt_2