eval_aishell-1:
  data:
    prompt: Transcribe the Chinese speech into text.
    data:
    - AISHELL-1/test/test_000{000..011}.tar.gz
    key: text
    repo_id: mispeech/xares_llm_data
  batch_size: 16
  num_workers: 2
  metric: iCER
eval_clotho:
  data:
    data:
    - Clotho/test/test_0000{00..04}.tar.gz
    key: caption;captions
    prompt: Generate a short caption for the audio.
    repo_id: mispeech/xares_llm_data
  batch_size: 16
  num_workers: 2
  metric: FENSE
eval_librispeech:
  data:
    data:
    - LibriSpeech/test/test-clean-0000{00..19}.tar.gz
    key: trans
    prompt: Transcribe the English speech into text
    repo_id: mispeech/xares_llm_data
  batch_size: 32
  num_workers: 4
  metric: iWER
eval_mecat:
  data:
    data:
    - 0M0/test_0000-0000000.tar.gz
    - 000/test_0000-0000000.tar.gz
    - SMA/test_0000-0000000.tar.gz
    - S00/test_0000-0000000.tar.gz
    - S00/test_0001-0000000.tar.gz
    - 0MA/test_0000-0000000.tar.gz
    - S0A/test_0000-0000000.tar.gz
    - SM0/test_0000-0000000.tar.gz
    - SM0/test_0001-0000000.tar.gz
    - 00A/test_0000-0000000.tar.gz
    key: long
    prompt: Write a general audiocaption.
    repo_id: mispeech/MECAT-Caption
  batch_size: 64
  num_workers: 4
  metric: DATE
eval_songdescriber:
  data:
    data:
    - Songdescriber/test/valid_song_describer_0000{00..01}.tar.gz
    key: captions
    prompt: Generate a caption for the music
    repo_id: mispeech/xares_llm_data
  batch_size: 16
  num_workers: 0
  metric: FENSE