dataset_path: lmms-lab/Omni_Bench_fix
test_split: train
output_type: generate_until

generation_kwargs:
  max_new_tokens: 128
  temperature: 0
  top_p: 0.7
  num_beams: 1
  do_sample: false

metric_list:
  - metric: accuracy
    aggregation: !function utils.omni_bench_aggregate_results
    higher_is_better: true

lmms_eval_specific_kwargs:
  default:
    pre_prompt: ""
    post_prompt: "Answer with the option's letter from the given choices directly."

metadata:
  version: 0.0