{
  "bos_emb": {
    "shape": [
      32
    ],
    "dtype": "float32",
    "bytes": 128
  },
  "emb_mean": {
    "shape": [
      32
    ],
    "dtype": "float32",
    "bytes": 128
  },
  "emb_std": {
    "shape": [
      32
    ],
    "dtype": "float32",
    "bytes": 128
  },
  "quantizer_weight": {
    "shape": [
      512,
      32,
      1
    ],
    "dtype": "float32",
    "bytes": 65536
  },
  "text_embed_table": {
    "shape": [
      4001,
      1024
    ],
    "dtype": "float32",
    "bytes": 16388096
  },
  "mimi_init_state": {
    "upsample_partial": {
      "shape": [
        1,
        512,
        16
      ],
      "dtype": "float32",
      "bytes": 32768
    },
    "attn0_cache": {
      "shape": [
        2,
        1,
        8,
        256,
        64
      ],
      "dtype": "float32",
      "bytes": 1048576
    },
    "attn0_offset": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "attn0_end_offset": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "attn1_cache": {
      "shape": [
        2,
        1,
        8,
        256,
        64
      ],
      "dtype": "float32",
      "bytes": 1048576
    },
    "attn1_offset": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "attn1_end_offset": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "conv0_prev": {
      "shape": [
        1,
        512,
        6
      ],
      "dtype": "float32",
      "bytes": 12288
    },
    "conv0_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "convtr0_partial": {
      "shape": [
        1,
        256,
        6
      ],
      "dtype": "float32",
      "bytes": 6144
    },
    "res0_conv0_prev": {
      "shape": [
        1,
        256,
        2
      ],
      "dtype": "float32",
      "bytes": 2048
    },
    "res0_conv0_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "res0_conv1_prev": {
      "shape": [
        1,
        128,
        0
      ],
      "dtype": "float32",
      "bytes": 0
    },
    "res0_conv1_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "convtr1_partial": {
      "shape": [
        1,
        128,
        5
      ],
      "dtype": "float32",
      "bytes": 2560
    },
    "res1_conv0_prev": {
      "shape": [
        1,
        128,
        2
      ],
      "dtype": "float32",
      "bytes": 1024
    },
    "res1_conv0_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "res1_conv1_prev": {
      "shape": [
        1,
        64,
        0
      ],
      "dtype": "float32",
      "bytes": 0
    },
    "res1_conv1_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "convtr2_partial": {
      "shape": [
        1,
        64,
        4
      ],
      "dtype": "float32",
      "bytes": 1024
    },
    "res2_conv0_prev": {
      "shape": [
        1,
        64,
        2
      ],
      "dtype": "float32",
      "bytes": 512
    },
    "res2_conv0_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "res2_conv1_prev": {
      "shape": [
        1,
        32,
        0
      ],
      "dtype": "float32",
      "bytes": 0
    },
    "res2_conv1_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "conv_final_prev": {
      "shape": [
        1,
        64,
        2
      ],
      "dtype": "float32",
      "bytes": 512
    },
    "conv_final_first": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    }
  },
  "alba_voice_cache": {
    "cache0": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position0": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "cache1": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position1": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "cache2": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position2": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "cache3": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position3": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "cache4": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position4": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    },
    "cache5": {
      "shape": [
        2,
        1,
        200,
        16,
        64
      ],
      "dtype": "float32",
      "bytes": 1638400
    },
    "position5": {
      "shape": [
        1
      ],
      "dtype": "float32",
      "bytes": 4
    }
  },
  "alba_audio_prompt": {
    "shape": [
      1,
      125,
      1024
    ],
    "dtype": "float32",
    "bytes": 512000
  },
  "tokenizer": {
    "format": "sentencepiece_protobuf",
    "bytes": 59339
  }
}