{
  "assamese": {
    "n_samples": 500,
    "wer": 17.66,
    "cer": 7.23,
    "wer_normalized": 14.87,
    "cer_normalized": 7.01,
    "empty_hypotheses": 0
  },
  "bengali": {
    "n_samples": 500,
    "wer": 16.31,
    "cer": 3.23,
    "wer_normalized": 12.48,
    "cer_normalized": 2.83,
    "empty_hypotheses": 0
  },
  "english": {
    "n_samples": 500,
    "wer": 26.54,
    "cer": 11.19,
    "wer_normalized": 12.41,
    "cer_normalized": 8.15,
    "empty_hypotheses": 0
  },
  "gujarati": {
    "n_samples": 500,
    "wer": 15.5,
    "cer": 3.74,
    "wer_normalized": 11.19,
    "cer_normalized": 3.16,
    "empty_hypotheses": 0
  },
  "hindi": {
    "n_samples": 500,
    "wer": 13.29,
    "cer": 3.79,
    "wer_normalized": 8.84,
    "cer_normalized": 2.74,
    "empty_hypotheses": 0
  },
  "kannada": {
    "n_samples": 500,
    "wer": 25.02,
    "cer": 4.77,
    "wer_normalized": 21.01,
    "cer_normalized": 4.3,
    "empty_hypotheses": 0
  },
  "malayalam": {
    "n_samples": 500,
    "wer": 39.0,
    "cer": 6.99,
    "wer_normalized": 34.7,
    "cer_normalized": 6.29,
    "empty_hypotheses": 0
  },
  "marathi": {
    "n_samples": 500,
    "wer": 21.39,
    "cer": 5.56,
    "wer_normalized": 16.57,
    "cer_normalized": 4.87,
    "empty_hypotheses": 0
  },
  "odia": {
    "n_samples": 500,
    "wer": 22.86,
    "cer": 4.52,
    "wer_normalized": 17.29,
    "cer_normalized": 3.69,
    "empty_hypotheses": 0
  },
  "punjabi": {
    "n_samples": 500,
    "wer": 16.07,
    "cer": 4.18,
    "wer_normalized": 9.44,
    "cer_normalized": 2.89,
    "empty_hypotheses": 0
  },
  "tamil": {
    "n_samples": 500,
    "wer": 28.67,
    "cer": 5.33,
    "wer_normalized": 26.37,
    "cer_normalized": 5.13,
    "empty_hypotheses": 0
  },
  "telugu": {
    "n_samples": 500,
    "wer": 23.2,
    "cer": 4.22,
    "wer_normalized": 18.52,
    "cer_normalized": 3.81,
    "empty_hypotheses": 0
  },
  "__overall__": {
    "n_samples": 6000,
    "wer": 21.88,
    "cer": 5.44,
    "wer_normalized": 16.81,
    "cer_normalized": 4.79
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer": 22.13,
    "cer": 5.4,
    "wer_normalized": 16.97,
    "cer_normalized": 4.57
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-v2-ckpt-300000",
    "checkpoint_name": "mixed-v2-ckpt-300000",
    "backend": "transformers",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 128,
    "force_language": true,
    "inference_time_sec": 680.8,
    "total_audio_sec": 40354.46,
    "rtf": 0.0169
  }
}