Add multilingual ARC task (#1419)

lm_eval/tasks/okapi/arc_multilingual/arc_ne.yaml

+include: _arc_yaml
+task: arc_ne
+dataset_path: alexandrainst/m_arc
+dataset_name: ne
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_nl.yaml

+include: _arc_yaml
+task: arc_nl
+dataset_path: alexandrainst/m_arc
+dataset_name: nl
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_pt.yaml

+include: _arc_yaml
+task: arc_pt
+dataset_path: alexandrainst/m_arc
+dataset_name: pt
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_ro.yaml

+include: _arc_yaml
+task: arc_ro
+dataset_path: alexandrainst/m_arc
+dataset_name: ro
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_ru.yaml

+include: _arc_yaml
+task: arc_ru
+dataset_path: alexandrainst/m_arc
+dataset_name: ru
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_sk.yaml

+include: _arc_yaml
+task: arc_sk
+dataset_path: alexandrainst/m_arc
+dataset_name: sk
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_sr.yaml

+include: _arc_yaml
+task: arc_sr
+dataset_path: alexandrainst/m_arc
+dataset_name: sr
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_sv.yaml

+include: _arc_yaml
+task: arc_sv
+dataset_path: alexandrainst/m_arc
+dataset_name: sv
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_ta.yaml

+include: _arc_yaml
+task: arc_ta
+dataset_path: alexandrainst/m_arc
+dataset_name: ta
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_te.yaml

+include: _arc_yaml
+task: arc_te
+dataset_path: alexandrainst/m_arc
+dataset_name: te
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_uk.yaml

+include: _arc_yaml
+task: arc_uk
+dataset_path: alexandrainst/m_arc
+dataset_name: uk
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_vi.yaml

+include: _arc_yaml
+task: arc_vi
+dataset_path: alexandrainst/m_arc
+dataset_name: vi
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/arc_zh.yaml

+include: _arc_yaml
+task: arc_zh
+dataset_path: alexandrainst/m_arc
+dataset_name: zh
+training_split: train
+validation_split: validation
+test_split: test

lm_eval/tasks/okapi/arc_multilingual/utils.py

+import datasets
+import re
+
+
+def preprocess(text):
+    if text is None:
+        return " "
+    text = text.strip()
+    text = text.replace(" [title]", ". ")
+    text = re.sub("\\[.*?\\]", "", text)
+    text = text.replace("  ", " ")
+    return text
+
+
+def process_docs(dataset: datasets.Dataset) -> datasets.Dataset:
+    def _process_doc(doc):
+        # breakpoint()
+        out_doc = {
+            "id": doc["id"],
+            "query": "Question: " + preprocess(doc["instruction"]) + "\nAnswer:",
+            "choices": [preprocess(doc['option_a']), preprocess(doc['option_b']), preprocess(doc['option_c']), preprocess(doc['option_d']), preprocess(doc['option_e'])],
+            "gold": ["A", "B", "C", "D", "E"].index(doc["answer"]),
+        }
+        return out_doc
+
+    return dataset.map(_process_doc)