{
  "assamese": {
    "n_samples": 500,
    "wer_raw": 16.26,
    "wer_norm": 16.1,
    "wer_numcanon": 16.1,
    "space_norm_wer": 13.58,
    "mer": 6.94,
    "cer_norm": 6.54,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.16,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 2.52,
      "norm_to_mer": 9.16
    }
  },
  "bengali": {
    "n_samples": 500,
    "wer_raw": 17.56,
    "wer_norm": 17.05,
    "wer_numcanon": 17.05,
    "space_norm_wer": 10.83,
    "mer": 3.32,
    "cer_norm": 3.54,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.51,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.22,
      "norm_to_mer": 13.73
    }
  },
  "english": {
    "n_samples": 500,
    "wer_raw": 30.59,
    "wer_norm": 12.52,
    "wer_numcanon": 12.52,
    "space_norm_wer": 7.86,
    "mer": 8.3,
    "cer_norm": 8.15,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 18.07,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.66,
      "norm_to_mer": 4.22
    }
  },
  "gujarati": {
    "n_samples": 500,
    "wer_raw": 16.26,
    "wer_norm": 16.26,
    "wer_numcanon": 16.26,
    "space_norm_wer": 11.61,
    "mer": 4.02,
    "cer_norm": 4.04,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.64,
      "norm_to_mer": 12.23
    }
  },
  "hindi": {
    "n_samples": 500,
    "wer_raw": 13.55,
    "wer_norm": 13.55,
    "wer_numcanon": 13.55,
    "space_norm_wer": 9.38,
    "mer": 4.05,
    "cer_norm": 3.98,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 4.17,
      "norm_to_mer": 9.5
    }
  },
  "kannada": {
    "n_samples": 500,
    "wer_raw": 25.8,
    "wer_norm": 25.79,
    "wer_numcanon": 25.79,
    "space_norm_wer": 16.71,
    "mer": 4.52,
    "cer_norm": 4.95,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.02,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.08,
      "norm_to_mer": 21.26
    }
  },
  "malayalam": {
    "n_samples": 500,
    "wer_raw": 39.54,
    "wer_norm": 39.52,
    "wer_numcanon": 39.52,
    "space_norm_wer": 29.94,
    "mer": 7.07,
    "cer_norm": 7.15,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.02,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.58,
      "norm_to_mer": 32.45
    }
  },
  "marathi": {
    "n_samples": 500,
    "wer_raw": 22.09,
    "wer_norm": 22.09,
    "wer_numcanon": 22.09,
    "space_norm_wer": 15.2,
    "mer": 5.38,
    "cer_norm": 5.6,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 6.89,
      "norm_to_mer": 16.71
    }
  },
  "odia": {
    "n_samples": 500,
    "wer_raw": 22.46,
    "wer_norm": 22.37,
    "wer_numcanon": 22.37,
    "space_norm_wer": 13.12,
    "mer": 4.0,
    "cer_norm": 4.46,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.09,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.25,
      "norm_to_mer": 18.36
    }
  },
  "punjabi": {
    "n_samples": 500,
    "wer_raw": 16.39,
    "wer_norm": 16.09,
    "wer_numcanon": 16.09,
    "space_norm_wer": 12.49,
    "mer": 4.59,
    "cer_norm": 4.34,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.29,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 3.6,
      "norm_to_mer": 11.5
    }
  },
  "tamil": {
    "n_samples": 500,
    "wer_raw": 29.39,
    "wer_norm": 29.17,
    "wer_numcanon": 29.17,
    "space_norm_wer": 19.58,
    "mer": 5.3,
    "cer_norm": 5.55,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.22,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.59,
      "norm_to_mer": 23.87
    }
  },
  "telugu": {
    "n_samples": 500,
    "wer_raw": 24.57,
    "wer_norm": 24.57,
    "wer_numcanon": 24.57,
    "space_norm_wer": 15.04,
    "mer": 4.25,
    "cer_norm": 4.64,
    "empty_hypotheses": 0,
    "normalization_delta": {
      "raw_to_norm": 0.0,
      "norm_to_numcanon": 0.0,
      "norm_to_space_norm": 9.53,
      "norm_to_mer": 20.32
    }
  },
  "__overall__": {
    "n_samples": 6000,
    "wer_raw": 22.56,
    "wer_norm": 21.03,
    "wer_numcanon": 21.03,
    "space_norm_wer": 14.54,
    "mer": 5.23,
    "cer_norm": 5.34
  },
  "__macro_avg__": {
    "n_languages": 12,
    "wer_raw": 22.87,
    "wer_norm": 21.26,
    "wer_numcanon": 21.26,
    "space_norm_wer": 14.61,
    "mer": 5.14,
    "cer_norm": 5.25
  },
  "__meta__": {
    "checkpoint": "/home/ubuntu/training/checkpoints/qwen3-asr-mixed-v2-ckpt-200000",
    "checkpoint_name": "ckpt-200000",
    "model_id": "qwen3-asr-mixed-v2",
    "model_type": "qwen3-asr-1.7B-mixed-v2",
    "dataset": "BayAreaBoys/indic-asr-benchmark-6k",
    "batch_size": 128,
    "inference_time_sec": 679.05,
    "total_audio_sec": 40354.46,
    "rtf": 0.0168,
    "timestamp": "2026-04-05T19:06:50Z",
    "gpu": "NVIDIA A100-SXM4-80GB",
    "framework": "transformers",
    "normalization_version": "v1"
  }
}