Preprocess answer choices of hellaswag task

lm_eval/tasks/hellaswag/hellaswag.yaml

@@ -7,8 +7,8 @@ output_type: multiple_choice
 training_split: train
 validation_split: validation
 test_split: null
-template_aliases: "{% set answer_choices = endings %}{% set gold = label %}"
-doc_to_text: !function preprocess_hellaswag.doc_to_text
+template_aliases: "{% set gold = label %}{% set answer_choices = endings|map('regex_replace', '\\[.*?\\]', '')|map('replace', ' [title]', '. ')|map('replace', '  ', ' ')|list %}" 
+doc_to_text: "{% set text = activity_label ~ ': ' ~ ctx_a ~ ' ' ~ ctx_b.capitalize() %}{{text|regex_replace('\\[.*?\\]', '')|replace(' [title]', '. ')|replace('  ', ' ')}}"
 doc_to_target: "{{gold}}"
 metric_list:
   - metric: acc

lm_eval/tasks/hellaswag/preprocess_hellaswag.py

-import re
-
-def preprocess(text):
-    text = text.strip()
-    # NOTE: Brackets are artifacts of the WikiHow dataset portion of HellaSwag.
-    text = text.replace(" [title]", ". ")
-    text = re.sub("\\[.*?\\]", "", text)
-    text = text.replace("  ", " ")
-    return text
-
-def doc_to_text(doc):
-    ctx = doc["ctx_a"] + " " + doc["ctx_b"].capitalize()
-    query = preprocess(doc["activity_label"] + ": " + ctx)
-    return query