formatting (#2104)

lm_eval/api/metrics.py

@@ -565,4 +565,4 @@ def aggregate_subtask_metrics(metrics, sizes, weight_by_size=True):
 
     assert len(metrics) == len(sizes)
 
-    return sum([metric * size for metric, size in zip(metrics, sizes)]) / sum(sizes)
\ No newline at end of file
+    return sum([metric * size for metric, size in zip(metrics, sizes)]) / sum(sizes)

lm_eval/api/task.py

@@ -1665,4 +1665,4 @@ class PerplexityTask(Task):
     @classmethod
     def count_words(cls, doc) -> int:
         """Downstream tasks with custom word boundaries should override this!"""
-        return len(re.split(r"\s+", doc))
\ No newline at end of file
+        return len(re.split(r"\s+", doc))

lm_eval/filters/extraction.py

@@ -181,4 +181,4 @@ class MultiChoiceRegexFilter(RegexFilter):
                 filtered.append(match)
             filtered_resps.append(filtered)
 
-        return filtered_resps
\ No newline at end of file
+        return filtered_resps

lm_eval/tasks/afrimgsm/README.md

@@ -5,8 +5,8 @@
 IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models
 https://arxiv.org/pdf/2406.03368
 
-IrokoBench is a human-translated benchmark dataset for 16 typologically diverse 
-low-resource African languages covering three tasks: natural language inference (AfriXNLI), 
+IrokoBench is a human-translated benchmark dataset for 16 typologically diverse
+low-resource African languages covering three tasks: natural language inference (AfriXNLI),
 mathematical reasoning (AfriMGSM), and multi-choice knowledge-based QA (AfriMMLU).
 
 
@@ -14,13 +14,13 @@ mathematical reasoning (AfriMGSM), and multi-choice knowledge-based QA (AfriMMLU
 
 ```
 @misc{adelani2024irokobenchnewbenchmarkafrican,
-      title={IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models}, 
+      title={IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models},
       author={David Ifeoluwa Adelani and Jessica Ojo and Israel Abebe Azime and Jian Yun Zhuang and Jesujoba O. Alabi and Xuanli He and Millicent Ochieng and Sara Hooker and Andiswa Bukula and En-Shiun Annie Lee and Chiamaka Chukwuneke and Happy Buzaaba and Blessing Sibanda and Godson Kalipe and Jonathan Mukiibi and Salomon Kabongo and Foutse Yuehgoh and Mmasibidi Setaka and Lolwethu Ndolela and Nkiruka Odu and Rooweither Mabuya and Shamsuddeen Hassan Muhammad and Salomey Osei and Sokhar Samb and Tadesse Kebede Guge and Pontus Stenetorp},
       year={2024},
       eprint={2406.03368},
       archivePrefix={arXiv},
       primaryClass={cs.CL},
-      url={https://arxiv.org/abs/2406.03368}, 
+      url={https://arxiv.org/abs/2406.03368},
 }
 ```
 

lm_eval/tasks/afrimgsm/run.sh

@@ -3,4 +3,4 @@ lm_eval --model hf   \
         --device cuda:0     \
         --batch_size 1  \
         --verbosity DEBUG \
-        --limit 5
\ No newline at end of file
+        --limit 5

lm_eval/tasks/afrimgsm/utils.py

@@ -2,51 +2,74 @@ import argparse
 
 import yaml
 
-languages = ['eng', 'amh', 'ibo', 'fra', 'sna', 'lin', 'wol', 'ewe', 'lug', 'xho', 'kin', 'twi', 'zul', 'orm', 'yor',
-             'hau', 'sot', 'swa']
-
-languages_REGEX = {"eng": "The answer is (\\-?[0-9\\.\\,]+)",
-                   "amh": "መልሱ (\\-?[0-9\\.\\,]+)",
-                   "ibo": "Azịza ya bụ (\\-?[0-9\\.\\,]+)",
-                   'fra': "La réponse est(\\-?[0-9\\.\\,]+)",
-                   'sna': "Mhinduro kumubvunzo ndi (\\-?[0-9\\.\\,]+)",
-                   'lin': "Eyano ezali (\\-?[0-9\\.\\,]+)",
-                   'wol': "Tontu li (\\-?[0-9\\.\\,]+)",
-                   'ewe': "ŋuɖoɖoae nye (\\-?[0-9\\.\\,]+)",
-                   'lug': "Ansa eri (\\-?[0-9\\.\\,]+)",
-                   'xho': "Impendulo ngu (\\-?[0-9\\.\\,]+)",
-                   'kin': "Igisubizo ni (\\-?[0-9\\.\\,]+)",
-                   'twi': "Ne nnyiano yɛ (\\-?[0-9\\.\\,]+)",
-                   'zul': "Impendulo ithi (\\-?[0-9\\.\\,]+)",
-                   'orm': "Deebiin isaa (\\-?[0-9\\.\\,]+)",
-                   'yor': "Ìdáhùn náà ni (\\-?[0-9\\.\\,]+)",
-                   'hau': "Amsar ita ce (\\-?[0-9\\.\\,]+)",
-                   'sot': "Karabo ke (\\-?[0-9\\.\\,]+)",
-                   'swa': "Jibu ni (\\-?[0-9\\.\\,]+)",
-                   }
+
+languages = [
+    "eng",
+    "amh",
+    "ibo",
+    "fra",
+    "sna",
+    "lin",
+    "wol",
+    "ewe",
+    "lug",
+    "xho",
+    "kin",
+    "twi",
+    "zul",
+    "orm",
+    "yor",
+    "hau",
+    "sot",
+    "swa",
+]
+
+languages_REGEX = {
+    "eng": "The answer is (\\-?[0-9\\.\\,]+)",
+    "amh": "መልሱ (\\-?[0-9\\.\\,]+)",
+    "ibo": "Azịza ya bụ (\\-?[0-9\\.\\,]+)",
+    "fra": "La réponse est(\\-?[0-9\\.\\,]+)",
+    "sna": "Mhinduro kumubvunzo ndi (\\-?[0-9\\.\\,]+)",
+    "lin": "Eyano ezali (\\-?[0-9\\.\\,]+)",
+    "wol": "Tontu li (\\-?[0-9\\.\\,]+)",
+    "ewe": "ŋuɖoɖoae nye (\\-?[0-9\\.\\,]+)",
+    "lug": "Ansa eri (\\-?[0-9\\.\\,]+)",
+    "xho": "Impendulo ngu (\\-?[0-9\\.\\,]+)",
+    "kin": "Igisubizo ni (\\-?[0-9\\.\\,]+)",
+    "twi": "Ne nnyiano yɛ (\\-?[0-9\\.\\,]+)",
+    "zul": "Impendulo ithi (\\-?[0-9\\.\\,]+)",
+    "orm": "Deebiin isaa (\\-?[0-9\\.\\,]+)",
+    "yor": "Ìdáhùn náà ni (\\-?[0-9\\.\\,]+)",
+    "hau": "Amsar ita ce (\\-?[0-9\\.\\,]+)",
+    "sot": "Karabo ke (\\-?[0-9\\.\\,]+)",
+    "swa": "Jibu ni (\\-?[0-9\\.\\,]+)",
+}
 
 LANGUAGES = {}
 
 for lang in languages:
-    if lang == 'amh':
+    if lang == "amh":
         LANGUAGES[lang] = {  # English
             "QUESTION": "ጥያቄ:",
             "ANSWER": "በቅደም ተከተል መልስ:",
             "DIRECT": "Answer:",
-            "REGEX": languages_REGEX[lang]}
-    elif lang == 'yor':
+            "REGEX": languages_REGEX[lang],
+        }
+    elif lang == "yor":
         LANGUAGES[lang] = {  # English
             "QUESTION": "Ìbéèrè:",
             "ANSWER": "Ìdáhùn lẹ́sẹsẹ:",
             "DIRECT": "Answer:",
-            "REGEX": languages_REGEX[lang]}
+            "REGEX": languages_REGEX[lang],
+        }
 
     else:
         LANGUAGES[lang] = {  # English
             "QUESTION": "Question:",
             "ANSWER": "Step-by-Step Answer:",
             "DIRECT": "Answer:",
-            "REGEX": languages_REGEX[lang]}
+            "REGEX": languages_REGEX[lang],
+        }
 
 
 def add_regex_pattern(regex_pattern):
@@ -93,13 +116,12 @@ def gen_lang_yamls(output_dir: str, overwrite: bool, mode: str) -> None:
     err = []
     for lang in LANGUAGES.keys():
         try:
-
             yaml_template = "cot_yaml"
             filter_list = {}
             DELIMITER = None
             if mode == "direct":
-                ANSWER = LANGUAGES['eng']["DIRECT"]
-                QUESTION = LANGUAGES['eng']["QUESTION"]
+                ANSWER = LANGUAGES["eng"]["DIRECT"]
+                QUESTION = LANGUAGES["eng"]["QUESTION"]
                 REGEX = None
                 task_name = f"afrimgsm_direct_{lang}"
                 yaml_template = "direct_yaml"
@@ -122,8 +144,8 @@ def gen_lang_yamls(output_dir: str, overwrite: bool, mode: str) -> None:
                 QUESTION = LANGUAGES["eng"]["QUESTION"]
                 task_name = f"afrimgsm_en_cot_{lang}"
             elif mode == "translate-direct":
-                ANSWER = LANGUAGES['eng']["DIRECT"]
-                QUESTION = LANGUAGES['eng']["QUESTION"]
+                ANSWER = LANGUAGES["eng"]["DIRECT"]
+                QUESTION = LANGUAGES["eng"]["QUESTION"]
                 REGEX = None
                 task_name = f"afrimgsm_translate_direct_{lang}"
                 yaml_template = "translate_direct_yaml"
@@ -131,7 +153,7 @@ def gen_lang_yamls(output_dir: str, overwrite: bool, mode: str) -> None:
             file_name = f"{task_name}.yaml"
             ANSWER_TO_SKIP = len(LANGUAGES[lang]["ANSWER"]) + 1
             with open(
-                    f"{output_dir}/{file_name}", "w" if overwrite else "x", encoding="utf8"
+                f"{output_dir}/{file_name}", "w" if overwrite else "x", encoding="utf8"
             ) as f:
                 f.write("# Generated by utils.py\n")
                 yaml.dump(
@@ -140,15 +162,15 @@ def gen_lang_yamls(output_dir: str, overwrite: bool, mode: str) -> None:
                         "dataset_name": lang,
                         "task": f"{task_name}",
                         "doc_to_text": f"""{{% if answer is not none %}}"""
-                                       f"""{{{{question+"\\n{ANSWER}"}}}}"""
-                                       f"""{{% else %}}"""
-                                       f"""{{{{"{QUESTION} "+question+"\\n{ANSWER}"}}}}"""
-                                       f"""{{% endif %}}""",
+                        f"""{{{{question+"\\n{ANSWER}"}}}}"""
+                        f"""{{% else %}}"""
+                        f"""{{{{"{QUESTION} "+question+"\\n{ANSWER}"}}}}"""
+                        f"""{{% endif %}}""",
                         "doc_to_target": f"""{{% if answer is not none %}}"""
-                                         f"""{{{{answer[{ANSWER_TO_SKIP}:]}}}}"""
-                                         f"""{{% else %}}"""
-                                         f"""{{{{answer_number|string}}}}"""
-                                         f"""{{% endif %}}""",
+                        f"""{{{{answer[{ANSWER_TO_SKIP}:]}}}}"""
+                        f"""{{% else %}}"""
+                        f"""{{{{answer_number|string}}}}"""
+                        f"""{{% endif %}}""",
                         **filter_list,
                         "generation_kwargs": {
                             "until": [QUESTION, "</s>", "<|im_end|>"],
@@ -194,4 +216,4 @@ def main() -> None:
 
 
 if __name__ == "__main__":
-    main()
\ No newline at end of file
+    main()

lm_eval/tasks/afrimmlu/README.md

@@ -5,8 +5,8 @@
 IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models
 https://arxiv.org/pdf/2406.03368
 
-IrokoBench is a human-translated benchmark dataset for 16 typologically diverse 
-low-resource African languages covering three tasks: natural language inference (AfriXNLI), 
+IrokoBench is a human-translated benchmark dataset for 16 typologically diverse
+low-resource African languages covering three tasks: natural language inference (AfriXNLI),
 mathematical reasoning (AfriMGSM), and multi-choice knowledge-based QA (AfriMMLU).
 
 
@@ -14,13 +14,13 @@ mathematical reasoning (AfriMGSM), and multi-choice knowledge-based QA (AfriMMLU
 
 ```
 @misc{adelani2024irokobenchnewbenchmarkafrican,
-      title={IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models}, 
+      title={IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models},
       author={David Ifeoluwa Adelani and Jessica Ojo and Israel Abebe Azime and Jian Yun Zhuang and Jesujoba O. Alabi and Xuanli He and Millicent Ochieng and Sara Hooker and Andiswa Bukula and En-Shiun Annie Lee and Chiamaka Chukwuneke and Happy Buzaaba and Blessing Sibanda and Godson Kalipe and Jonathan Mukiibi and Salomon Kabongo and Foutse Yuehgoh and Mmasibidi Setaka and Lolwethu Ndolela and Nkiruka Odu and Rooweither Mabuya and Shamsuddeen Hassan Muhammad and Salomey Osei and Sokhar Samb and Tadesse Kebede Guge and Pontus Stenetorp},
       year={2024},
       eprint={2406.03368},
       archivePrefix={arXiv},
       primaryClass={cs.CL},
-      url={https://arxiv.org/abs/2406.03368}, 
+      url={https://arxiv.org/abs/2406.03368},
 }
 ```
 

lm_eval/tasks/afrimmlu/direct/afrimmlu_common_yaml

@@ -9,18 +9,18 @@ output_type: multiple_choice
 validation_split: validation
 test_split: test
 fewshot_split: validation
-doc_to_text: !function utils.doc_to_text 
+doc_to_text: !function utils.doc_to_text
 doc_to_target: "{{['A', 'B', 'C', 'D'].index(answer)}}"
 doc_to_choice: !function utils.doc_to_choice
 should_decontaminate: true
 doc_to_decontamination_query: "Question: {{question}}\nAnswer:"
 metric_list:
-  - metric: f1 
-    aggregation: !function utils.weighted_f1_score 
+  - metric: f1
+    aggregation: !function utils.weighted_f1_score
     # aggregation: mean
-    average: weighted 
-    hf_evaluate: true 
-    higher_is_better: True 
+    average: weighted
+    hf_evaluate: true
+    higher_is_better: True
     ignore_case: true
     ignore_punctuation: true
     regexes_to_ignore:

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_eng.yaml

 dataset_name: eng
 include: afrimmlu_common_yaml
 task: afrimmlu_direct_eng
-

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_ewe.yaml

 dataset_name: ewe
 include: afrimmlu_common_yaml
 task: afrimmlu_direct_ewe
-

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_fra.yaml

 dataset_name: fra
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_fra
\ No newline at end of file
+task: afrimmlu_direct_fra

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_hau.yaml

 dataset_name: hau
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_hau
\ No newline at end of file
+task: afrimmlu_direct_hau

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_ibo.yaml

 dataset_name: ibo
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_ibo
\ No newline at end of file
+task: afrimmlu_direct_ibo

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_kin.yaml

 dataset_name: kin
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_kin
\ No newline at end of file
+task: afrimmlu_direct_kin

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_lin.yaml

 dataset_name: lin
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_lin
\ No newline at end of file
+task: afrimmlu_direct_lin

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_lug.yaml

 dataset_name: lug
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_lug
\ No newline at end of file
+task: afrimmlu_direct_lug

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_orm.yaml

 dataset_name: orm
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_orm
\ No newline at end of file
+task: afrimmlu_direct_orm

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_sna.yaml

 dataset_name: sna
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_sna
\ No newline at end of file
+task: afrimmlu_direct_sna

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_sot.yaml

 dataset_name: sot
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_sot
\ No newline at end of file
+task: afrimmlu_direct_sot

lm_eval/tasks/afrimmlu/direct/afrimmlu_direct_swa.yaml

 dataset_name: swa
 include: afrimmlu_common_yaml
-task: afrimmlu_direct_swa
\ No newline at end of file
+task: afrimmlu_direct_swa