dataset_path: shiwk24/MathCanvas-Bench
dataset_kwargs:
  token: False
test_split: test
output_type: generate_until
doc_to_visual: !function utils.mathcanvas_doc_to_visual
doc_to_text: !function utils.mathcanvas_doc_to_text
doc_to_target: "answer"
generation_kwargs:
  max_new_tokens: 8192
  temperature: 0
  do_sample: false
process_results: !function utils.mathcanvas_process_results
metric_list:
  - metric: mathcanvas_weighted_accuracy
    aggregation: mean
    higher_is_better: true
  - metric: mathcanvas_complete_accuracy
    aggregation: mean
    higher_is_better: true
lmms_eval_specific_kwargs:
  default:
    pre_prompt: ""
    post_prompt: "\nYour task is to answer the question above. Give step by step reasoning, and conclude all the answers (include sub-questions) at the end of your solution. If there are multiple sub-questions, format the final answers as <1>answer1</1><2>answer2</2>..."
metadata:
  - version: 0.0
