config.json

{
  "modelName": "kws-zipformer-gigaspeech",
  "source": "pkufool/keyword-spotting-models@v0.11/icefall-kws-zipformer-gigaspeech-20240219",
  "checkpoint": "finetune",
  "language": "en",
  "license": "Apache-2.0",
  "feature": {
    "type": "kaldi-fbank",
    "sampleRate": 16000,
    "numMelBins": 80,
    "frameLengthMs": 25,
    "frameShiftMs": 10,
    "dither": 0.0,
    "snipEdges": false,
    "normalizeSamples": true,
    "highFreq": -400.0
  },
  "encoder": {
    "chunkSize": 16,
    "leftContextFrames": 64,
    "totalInputFrames": 45,
    "outputFrames": 8,
    "joinerDim": 320,
    "layerStateNames": [
      "cache_s0_l0_key",
      "cache_s0_l0_nonlin_attn",
      "cache_s0_l0_val1",
      "cache_s0_l0_val2",
      "cache_s0_l0_conv1",
      "cache_s0_l0_conv2",
      "cache_s1_l0_key",
      "cache_s1_l0_nonlin_attn",
      "cache_s1_l0_val1",
      "cache_s1_l0_val2",
      "cache_s1_l0_conv1",
      "cache_s1_l0_conv2",
      "cache_s2_l0_key",
      "cache_s2_l0_nonlin_attn",
      "cache_s2_l0_val1",
      "cache_s2_l0_val2",
      "cache_s2_l0_conv1",
      "cache_s2_l0_conv2",
      "cache_s3_l0_key",
      "cache_s3_l0_nonlin_attn",
      "cache_s3_l0_val1",
      "cache_s3_l0_val2",
      "cache_s3_l0_conv1",
      "cache_s3_l0_conv2",
      "cache_s4_l0_key",
      "cache_s4_l0_nonlin_attn",
      "cache_s4_l0_val1",
      "cache_s4_l0_val2",
      "cache_s4_l0_conv1",
      "cache_s4_l0_conv2",
      "cache_s5_l0_key",
      "cache_s5_l0_nonlin_attn",
      "cache_s5_l0_val1",
      "cache_s5_l0_val2",
      "cache_s5_l0_conv1",
      "cache_s5_l0_conv2"
    ],
    "layerStateShapes": [
      [
        64,
        1,
        128
      ],
      [
        1,
        1,
        64,
        96
      ],
      [
        64,
        1,
        48
      ],
      [
        64,
        1,
        48
      ],
      [
        1,
        128,
        15
      ],
      [
        1,
        128,
        15
      ],
      [
        32,
        1,
        128
      ],
      [
        1,
        1,
        32,
        96
      ],
      [
        32,
        1,
        48
      ],
      [
        32,
        1,
        48
      ],
      [
        1,
        128,
        15
      ],
      [
        1,
        128,
        15
      ],
      [
        16,
        1,
        128
      ],
      [
        1,
        1,
        16,
        96
      ],
      [
        16,
        1,
        48
      ],
      [
        16,
        1,
        48
      ],
      [
        1,
        128,
        7
      ],
      [
        1,
        128,
        7
      ],
      [
        8,
        1,
        256
      ],
      [
        1,
        1,
        8,
        96
      ],
      [
        8,
        1,
        96
      ],
      [
        8,
        1,
        96
      ],
      [
        1,
        128,
        7
      ],
      [
        1,
        128,
        7
      ],
      [
        16,
        1,
        128
      ],
      [
        1,
        1,
        16,
        96
      ],
      [
        16,
        1,
        48
      ],
      [
        16,
        1,
        48
      ],
      [
        1,
        128,
        7
      ],
      [
        1,
        128,
        7
      ],
      [
        32,
        1,
        128
      ],
      [
        1,
        1,
        32,
        96
      ],
      [
        32,
        1,
        48
      ],
      [
        32,
        1,
        48
      ],
      [
        1,
        128,
        15
      ],
      [
        1,
        128,
        15
      ]
    ],
    "cachedEmbedLeftPadShape": [
      1,
      128,
      3,
      19
    ]
  },
  "decoder": {
    "vocabSize": 500,
    "blankId": 0,
    "contextSize": 2,
    "decoderDim": 320
  },
  "kws": {
    "defaultThreshold": 0.15,
    "defaultContextScore": 0.5,
    "defaultNumTrailingBlanks": 1,
    "autoResetSeconds": 1.5
  }
}