test_split: test
output_type: generate_until
doc_to_messages: !function utils.doc_to_messages
doc_to_text: !function utils.doc_to_text
doc_to_visual: !function utils.doc_to_visual
doc_to_target: answer
process_results: !function utils.process_results
generation_kwargs:
  max_new_tokens: 32768
  temperature: 0
  do_sample: false
model_specific_generation_kwargs:
  default: {}
  qwen2_5_omni:
    raise_on_error: true
  qwen2_5_omni_chat:
    raise_on_error: true
  qwen3_omni:
    raise_on_error: true
  qwen3_omni_chat:
    raise_on_error: true
lmms_eval_specific_kwargs:
  default:
    supports_separate_audio: false
  openai:
    api_audio_opt_in: true
  qwen2_5_omni:
    supports_separate_audio: true
  qwen2_5_omni_chat:
    supports_separate_audio: true
  qwen3_omni:
    supports_separate_audio: true
  qwen3_omni_chat:
    supports_separate_audio: true
metric_list:
  - metric: text
    aggregation: !function utils.aggregate_text
    higher_is_better: true
  - metric: text_image
    aggregation: !function utils.aggregate_text_image
    higher_is_better: true
  - metric: cs
    aggregation: !function utils.aggregate_cs
    higher_is_better: true
metadata:
  version: 1.0
