eval_asvspoof2015:
  data:
    data:
    - ASVSpoof2015/test/asvspoof_eval_000{00..193}.tar.gz
    key: binary_spoof
    prompt: Detect if the audio is genuine or a spoof.
    repo_id: mispeech/xares_llm_data
  batch_size: 128
  num_workers: 2
  metric: Accuracy
eval_cremad:
  data:
    data:
    - CremaD/test/valid_cremad_with_text_0000{00..01}.tar.gz
    key: label;labels
    prompt: Identify the emotion expressed in the speech.
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 0
  metric: Accuracy
eval_esc-50:
  data:
    data:
    - ESC-50/test/esc50_fold_5_0000000.tar.gz
    key: text;caption;captions;label;labels
    prompt: Classify the environmental sound.
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 0
  metric: Accuracy
eval_fluentspeechcommands:
  data:
    data:
    - FluentSpeechCommands/test/test_intent_with_text_0000{00..03}.tar.gz
    key: text;caption;captions;label;labels
    prompt: Identify the command, action, and object from the speech.
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 2
  metric: Accuracy
eval_freemusicarchive:
  data:
    data:
    - FreeMusicArchive/test/fma_small_test_0000{00..07}.tar.gz
    key: genre
    prompt: Classify the music genre (FMA).
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 2
  metric: Accuracy
eval_fsd50k:
  data:
    data:
    - FSD50k/test/eval_with_text_0000{00..31}.tar.gz
    key: label;labels
    prompt: Multi-label classification for FSD50k
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 2
  metric: mAP
  metric_args:
    num_classes: 200
eval_fsdkaggle2018:
  data:
    data:
    - FSDKaggle2018/test/test_with_text_0000{00..02}.tar.gz
    key: label;labels
    prompt: Multi-label classification for FSDKaggle2018
    repo_id: mispeech/xares_llm_data
  batch_size: 4
  num_workers: 2
  metric: mAP
  metric_args:
    num_classes: 41
eval_gtzan:
  data:
    data:
    - GTZAN/test/gtzan_fold_9_0000000.tar.gz
    key: genre
    prompt: Classify the music genre (GTZAN).
    repo_id: mispeech/xares_llm_data
  batch_size: 8
  num_workers: 2
  metric: Accuracy
eval_libricount:
  data:
    data:
    - LibriCount/test/libricount_fold04.tar.gz
    key: num_speakers
    prompt: Count the amount of speakers.
    repo_id: mispeech/xares_llm_data
  batch_size: 128
  num_workers: 0
  metric: Accuracy
eval_nsynth:
  data:
    data:
    - NSynth/test/test_0000{00..05}.tar.gz
    key: instrument_family_str
    prompt: Classify the music instrument.
    repo_id: mispeech/xares_llm_data
  batch_size: 8
  num_workers: 0
  metric: Accuracy
eval_speechcommandsv1:
  data:
    data:
    - SpeechCommandsV1/test/test_renamed_with_text_0000{00..06}.tar.gz
    key: text;caption;captions;label;labels
    prompt: Classify the keyword.
    repo_id: mispeech/xares_llm_data
  batch_size: 128
  num_workers: 2
  metric: Accuracy
eval_urbansound8k:
  data:
    data:
    - Urbansound8k/test/urbansound_fold10_0000000.tar.gz
    key: soundevent
    prompt: Classify the urban sound.
    repo_id: mispeech/xares_llm_data
  batch_size: 128
  num_workers: 0
  metric: Accuracy
eval_vocalsound:
  data:
    data:
    - VocalSound/test/test_0000{00..04}.tar.gz
    key: labels
    prompt: Classify the vocal sound.
    repo_id: mispeech/xares_llm_data
  batch_size: 32
  num_workers: 2
  metric: Accuracy
eval_voxceleb1:
  data:
    data:
    - VoxCeleb1/test/test_0000{00..18}.tar.gz
    key: text;caption;captions;label;labels
    prompt: Are the two speakers the same or different?
    repo_id: mispeech/xares_llm_data
  batch_size: 32
  num_workers: 2
  metric: Accuracy
eval_voxlingua33:
  data:
    data:
    - VoxLingua33/test/dev_0000{00..05}.tar.gz
    key: text;caption;captions;label;labels
    prompt: Classify the language.
    repo_id: mispeech/xares_llm_data
  batch_size: 64
  num_workers: 2
  metric: Accuracy