lmms_eval/tasks/mmvet/mmvet.yaml

dataset_path: lmms-lab/MMVet
dataset_kwargs:
  token: True
task: "mmvet"
test_split: test
output_type: generate_until
doc_to_visual: !function utils.mmvet_doc_to_visual
doc_to_text: !function utils.doc_to_text # Such that {{question}} will be replaced by doc["question"]
doc_to_target: "{{answer}}"
generation_kwargs:
  until:
    - "ASSISTANT:"
  max_new_tokens: 1024
  temperature: 0
  top_p: 0
  num_beams: 1
  do_sample: false
process_results: !function utils.mmvet_process_results # apply gpt eval here
metric_list:
  - metric: gpt_eval_score
    aggregation: !function utils.mmvet_aggregate_results
    higher_is_better: true
metadata:
  version: 0.0
  gpt_eval_model_name: "gpt-4"
model_specific_prompt_kwargs:
  default:
    pre_prompt: ""
    post_prompt: ""