tag: afrimgsm_tt_cot_tasks
dataset_path: masakhane/afrimgsm-translate-test
dataset_name: null  # Overridden by language-specific config.
output_type: generate_until
test_split: test
doc_to_target: '{% if answer is not none %}{{answer[21:]}}{% else %}{{answer_number|string}}{% endif %}'
doc_to_text: 'Solve the following math question \n\nQuestion: {{question}} \nStep-by-Step Answer: '
generation_kwargs:
  do_sample: false
  until:
  - 'Question:'
  - </s>
  - <|im_end|>
metric_list:
  - metric: exact_match
    aggregation: mean
    higher_is_better: true
    ignore_case: true
    ignore_punctuation: true
filter_list:
  - name: "strict-match"
    filter:
      - function: "regex"
        regex_pattern: "The answer is (\\-?[0-9\\.\\,]+)"
      - function: "take_first"
  - filter:
    - function: regex
      group_select: -1
      regex_pattern: (-?[$0-9.,]{2,})|(-?[0-9]+)
    - function: take_first
    name: flexible-extract
metadata:
  version: 2.0
