config.json

{
  "architectures": [
    "WhisperPhonemePredictionModel"
  ],
  "blank": 0,
  "dropout_rate": 0.1,
  "freeze_encoder": true,
  "hidden_dim": 512,
  "id_2_phoneme": {
    "0": "\u2205",
    "1": "spn",
    "2": "\u3131",
    "3": "\u3132",
    "4": "\u3134",
    "5": "\u3137",
    "6": "\u3138",
    "7": "\u3139",
    "8": "\u3141",
    "9": "\u3142",
    "10": "\u3143",
    "11": "\u3145",
    "12": "\u3146",
    "13": "\u3147",
    "14": "\u3148",
    "15": "\u3149",
    "16": "\u314a",
    "17": "\u314b",
    "18": "\u314c",
    "19": "\u314d",
    "20": "\u314e",
    "21": "\u314f",
    "22": "\u3150",
    "23": "\u3151",
    "24": "\u3152",
    "25": "\u3153",
    "26": "\u3154",
    "27": "\u3155",
    "28": "\u3156",
    "29": "\u3157",
    "30": "\u3158",
    "31": "\u3159",
    "32": "\u315a",
    "33": "\u315b",
    "34": "\u315c",
    "35": "\u315d",
    "36": "\u315e",
    "37": "\u315f",
    "38": "\u3160",
    "39": "\u3161",
    "40": "\u3162",
    "41": "\u3163",
    "42": "\u3133",
    "43": "\u3135",
    "44": "\u3136",
    "45": "\u313a",
    "46": "\u313b",
    "47": "\u313c",
    "48": "\u313d",
    "49": "\u313e",
    "50": "\u313f",
    "51": "\u3140",
    "52": "\u3144"
  },
  "model_name": "openai/whisper-large-v3-turbo",
  "model_type": "whisper_phoneme",
  "phoneme_vocab_size": 53,
  "torch_dtype": "float32",
  "transformers_version": "4.51.3"
}