changed dataset to parquet version (#2845)

lm_eval/tasks/bbh/README.md

@@ -51,3 +51,6 @@ None.
 - [ ] Variant with Calculator (see https://github.com/openai/grade-school-math/blob/master/grade_school_math/calculator.py for example implementation)
 - [ ] Using Verifiers
 - [ ] Majority voting "without CoT"
+
+### Changelog
+no version change: changed dataset to `SaylorTwift/bbh`. Do not expect any change in the results.

lm_eval/tasks/bbh/_generate_configs.py

@@ -11,8 +11,6 @@ import requests
 import yaml
 from tqdm import tqdm
 
-from lm_eval import utils
-
 
 def parse_args():
     parser = argparse.ArgumentParser()
@@ -51,9 +49,10 @@ if __name__ == "__main__":
                 for shot in few_shot:
                     try:
                         answer = answer_regex.search(shot)[0]
-                    except Exception:
+                    except Exception as e:
                         print("task", task)
                         print(shot)
+                        raise e
                     example = shot.split("Let's think step by step.")[0]
                     prefix_doc_to_text += f"{example}{answer}\n\n"
 
@@ -70,7 +69,7 @@ if __name__ == "__main__":
         }
 
         file_save_path = args.save_prefix_path + f"/{task}.yaml"
-        utils.eval_logger.info(f"Saving yaml for subset {task} to {file_save_path}")
+        print(f"Saving yaml for subset {task} to {file_save_path}")
         with open(file_save_path, "w", encoding="utf-8") as yaml_file:
             yaml.dump(
                 yaml_dict,

lm_eval/tasks/bbh/cot_fewshot/_cot_fewshot_template_yaml

-dataset_path: lukaemon/bbh
+dataset_path: SaylorTwift/bbh
 output_type: generate_until
 test_split: test
 doc_to_target: "{{target}}"

lm_eval/tasks/bbh/cot_zeroshot/_cot_zeroshot_template_yaml

-dataset_path: lukaemon/bbh
+dataset_path: SaylorTwift/bbh
 output_type: generate_until
 test_split: test
 doc_to_target: "{{target}}"

lm_eval/tasks/bbh/fewshot/_fewshot_template_yaml

-dataset_path: lukaemon/bbh
+dataset_path: SaylorTwift/bbh
 output_type: generate_until
 test_split: test
 doc_to_target: "{{target}}"

lm_eval/tasks/bbh/zeroshot/_zeroshot_template_yaml

-dataset_path: lukaemon/bbh
+dataset_path: SaylorTwift/bbh
 output_type: generate_until
 test_split: test
 doc_to_target: "{{target}}"