{
  "name": "SenseVoiceSmall-coreml",
  "upstream": {
    "model": "FunAudioLLM/SenseVoiceSmall",
    "code": "https://github.com/FunAudioLLM/SenseVoice",
    "model_revision": "3eb3b4eeffc2f2dde6051b853983753db33e35c3",
    "code_commit": "b97857e5c807a466d0a062cf84dba9fe38a1d66a",
    "assets": [
      {
        "name": "cmvn_am.mvn",
        "url": "https://modelscope.cn/api/v1/models/iic/SenseVoiceSmall/repo?Revision=master&FilePath=am.mvn",
        "sha256": "29b3c740a2c0cfc6b308126d31d7f265fa2be74f3bb095cd2f143ea970896ae5",
        "bytes": 11203
      },
      {
        "name": "spm",
        "url": "https://modelscope.cn/api/v1/models/iic/SenseVoiceSmall/repo?Revision=master&FilePath=chn_jpn_yue_eng_ko_spectok.bpe.model",
        "sha256": "aa87f86064c3730d799ddf7af3c04659151102cba548bce325cf06ba4da4e6a8",
        "bytes": 377341
      }
    ]
  },
  "artifacts": {
    "coreml_zip": "coreml/SenseVoiceSmall.mlmodelc.zip"
  },
  "coreml": {
    "format": "mlmodelc",
    "inputs": [
      {
        "name": "speech",
        "dtype": "float32",
        "shape": [
          1,
          "T",
          560
        ]
      },
      {
        "name": "speech_lengths",
        "dtype": "int32",
        "shape": [
          1
        ]
      },
      {
        "name": "language",
        "dtype": "int32",
        "shape": [
          1
        ]
      },
      {
        "name": "textnorm",
        "dtype": "int32",
        "shape": [
          1
        ]
      }
    ],
    "outputs": [
      {
        "name": "ctc_logits"
      },
      {
        "name": "encoder_out_lens"
      }
    ]
  },
  "decoding": {
    "ctc_blank_id": 0,
    "token_offset": 0
  },
  "notes": [
    "Do not run a single forward pass on extremely long audio; use VAD/chunking."
  ],
  "build": {
    "artifact": "coreml/SenseVoiceSmall.mlmodelc.zip",
    "sha256": "880711fa03577363e6c1b1b6e9321f130ea1a53d5c065d92e1abd8a431bad6be",
    "bytes": 432164139,
    "xcodebuild_version": "Xcode 26.0\nBuild version 17A324"
  }
}