dataset_manifest.json

{
  "dataset_name": "emilia-s2s",
  "artifact_kind": "codec_pretok_tts",
  "objective": "named_speaker_text_to_audio",
  "export_format": "codec_only",
  "prompt_format": "speaker_name_colon_text_audio",
  "prompt_template": "{speaker_name}: {transcript}<audio>{mimi_codes}</audio>",
  "lang": "hi_en",
  "speaker_map": {
    "SP_SP010": "Ira",
    "SP_SP061": "Aisha",
    "SP_SP073": "Siya",
    "SP_SP079": "Zoya"
  },
  "reference_conditioning": {
    "enabled": false,
    "static_references_emitted": false,
    "pairing_rule": "none"
  },
  "source": {
    "gcs_prefix": "gs://gp-s2s-data/s2s-data/emilia-s2s/chunk_00036_200k/parquet",
    "shards_requested": 325,
    "shards_processed": 325
  },
  "num_quantizers": 8,
  "selection": {
    "min_seconds": 1.0,
    "max_seconds": 30.0,
    "max_samples": 0,
    "max_shards": 0,
    "start_shard_index": 0,
    "validation_ratio": 0.01,
    "test_ratio": 0.005,
    "split_seed": 42,
    "normalize_whitespace": true,
    "unknown_speaker_policy": "error"
  },
  "audio_codec": {
    "backend": "mimi",
    "source": "hf_pretrained",
    "model_ref": "kyutai/mimi",
    "sample_rate": 24000,
    "codebook_size": 2048,
    "max_codebooks": 32,
    "encode_downsample": 1920
  },
  "counts": {
    "train": 194797,
    "validation": 2055,
    "test": 1032
  },
  "counts_by_speaker": {
    "SP_SP010": {
      "train": 48763,
      "validation": 502,
      "test": 286
    },
    "SP_SP061": {
      "train": 48625,
      "validation": 494,
      "test": 249
    },
    "SP_SP073": {
      "train": 48683,
      "validation": 532,
      "test": 239
    },
    "SP_SP079": {
      "train": 48726,
      "validation": 527,
      "test": 258
    }
  },
  "duration_hours": {
    "train": 830.477273,
    "validation": 8.818905,
    "test": 4.387542
  },
  "dropped": {
    "missing_fields": 0,
    "unknown_speaker": 0,
    "duration": 113,
    "decode_error": 0
  },
  "shard_size": 2048,
  "written_files": {
    "train": [
      "train/part-00000.parquet",
      "train/part-00001.parquet",
      "train/part-00002.parquet",
      "train/part-00003.parquet",
      "train/part-00004.parquet",
      "train/part-00005.parquet",
      "train/part-00006.parquet",
      "train/part-00007.parquet",
      "train/part-00008.parquet",
      "train/part-00009.parquet",
      "train/part-00010.parquet",
      "train/part-00011.parquet",
      "train/part-00012.parquet",
      "train/part-00013.parquet",
      "train/part-00014.parquet",
      "train/part-00015.parquet",
      "train/part-00016.parquet",
      "train/part-00017.parquet",
      "train/part-00018.parquet",
      "train/part-00019.parquet",
      "train/part-00020.parquet",
      "train/part-00021.parquet",
      "train/part-00022.parquet",
      "train/part-00023.parquet",
      "train/part-00024.parquet",
      "train/part-00025.parquet",
      "train/part-00026.parquet",
      "train/part-00027.parquet",
      "train/part-00028.parquet",
      "train/part-00029.parquet",
      "train/part-00030.parquet",
      "train/part-00031.parquet",
      "train/part-00032.parquet",
      "train/part-00033.parquet",
      "train/part-00034.parquet",
      "train/part-00035.parquet",
      "train/part-00036.parquet",
      "train/part-00037.parquet",
      "train/part-00038.parquet",
      "train/part-00039.parquet",
      "train/part-00040.parquet",
      "train/part-00041.parquet",
      "train/part-00042.parquet",
      "train/part-00043.parquet",
      "train/part-00044.parquet",
      "train/part-00045.parquet",
      "train/part-00046.parquet",
      "train/part-00047.parquet",
      "train/part-00048.parquet",
      "train/part-00049.parquet",
      "train/part-00050.parquet",
      "train/part-00051.parquet",
      "train/part-00052.parquet",
      "train/part-00053.parquet",
      "train/part-00054.parquet",
      "train/part-00055.parquet",
      "train/part-00056.parquet",
      "train/part-00057.parquet",
      "train/part-00058.parquet",
      "train/part-00059.parquet",
      "train/part-00060.parquet",
      "train/part-00061.parquet",
      "train/part-00062.parquet",
      "train/part-00063.parquet",
      "train/part-00064.parquet",
      "train/part-00065.parquet",
      "train/part-00066.parquet",
      "train/part-00067.parquet",
      "train/part-00068.parquet",
      "train/part-00069.parquet",
      "train/part-00070.parquet",
      "train/part-00071.parquet",
      "train/part-00072.parquet",
      "train/part-00073.parquet",
      "train/part-00074.parquet",
      "train/part-00075.parquet",
      "train/part-00076.parquet",
      "train/part-00077.parquet",
      "train/part-00078.parquet",
      "train/part-00079.parquet",
      "train/part-00080.parquet",
      "train/part-00081.parquet",
      "train/part-00082.parquet",
      "train/part-00083.parquet",
      "train/part-00084.parquet",
      "train/part-00085.parquet",
      "train/part-00086.parquet",
      "train/part-00087.parquet",
      "train/part-00088.parquet",
      "train/part-00089.parquet",
      "train/part-00090.parquet",
      "train/part-00091.parquet",
      "train/part-00092.parquet",
      "train/part-00093.parquet",
      "train/part-00094.parquet",
      "train/part-00095.parquet"
    ],
    "validation": [
      "validation/part-00000.parquet",
      "validation/part-00001.parquet"
    ],
    "test": [
      "test/part-00000.parquet"
    ]
  },
  "source_scan": {
    "scanned": 197997
  },
  "sample_id_sha256_by_split": {
    "train": "a6b90b2ad678cf08c18661f7219c649cee18053dc19f8b3578437830f3bbf612",
    "validation": "a885f4dcead087172565c42a773d8526220eb6d614de17989b8b7bb52bf79568",
    "test": "58561dd7d1d4a3c7016668fd828b254b2930eb96bf012919ae9cff645aa112a5"
  },
  "fingerprint_sha256": "8cff46ccaae210ea1e6c1fabeebe018b486bee0051da336e61e892a5578b1f28"
}