{
  "variant": "b6369a24",
  "batch_size": 1,
  "sample_rate": 24000,
  "frame_rate": 12.5,
  "frame_size": 1920,
  "ldim": 32,
  "dim": 1024,
  "vocab_size": 4000,
  "eos_threshold": -4.0,
  "temperature": 0.7,
  "lsd_decode_steps": 1,
  "max_token_per_chunk": 50,
  "chunk_token_budget": 80,
  "max_gen_len": 359,
  "flow_lm_capacity": 608,
  "max_voice_prompt_frames": 161,
  "emb_mean": [
    -0.10009765625,
    -0.0260009765625,
    -0.036376953125,
    -0.341796875,
    0.06103515625,
    -0.80078125,
    0.047119140625,
    -0.0186767578125,
    -0.10009765625,
    -0.0693359375,
    -0.0703125,
    -0.1650390625,
    0.419921875,
    -0.0029144287109375,
    -0.031982421875,
    0.0673828125,
    0.021484375,
    -0.04443359375,
    0.0238037109375,
    0.0625,
    -0.13671875,
    -0.1787109375,
    0.294921875,
    0.000644683837890625,
    -0.033203125,
    0.11083984375,
    0.023681640625,
    -0.232421875,
    -0.2060546875,
    -0.6953125,
    -0.05712890625,
    0.17578125
  ],
  "emb_std": [
    0.96484375,
    0.9921875,
    1.0390625,
    0.92578125,
    1.015625,
    0.59375,
    0.98046875,
    0.9765625,
    0.9609375,
    0.984375,
    0.90625,
    0.91015625,
    0.9765625,
    1.0,
    1.0,
    1.0,
    1.015625,
    0.984375,
    1.015625,
    1.0,
    1.0078125,
    1.34375,
    0.84375,
    1.0390625,
    0.9453125,
    0.96484375,
    0.953125,
    0.82421875,
    0.9140625,
    0.9453125,
    0.9921875,
    0.97265625
  ],
  "voices": [
    "alba",
    "marius",
    "javert",
    "jean",
    "fantine",
    "cosette",
    "eponine",
    "azelma"
  ],
  "tokenizer": "tokenizer.model",
  "tokenizer_json": "tokenizer.json.gz",
  "graphs": {
    "text_embed": "text_embed.onnx",
    "latent_embed": "latent_embed.onnx",
    "flow_lm_step": "flow_lm_step.onnx",
    "flow_match": "flow_match.onnx",
    "mimi_decode": "mimi_decode.onnx"
  },
  "flow_lm": {
    "n_layers": 6,
    "num_heads": 16,
    "head_dim": 64,
    "capacity": 608,
    "cache_shape": [
      2,
      1,
      608,
      16,
      64
    ],
    "input_names": [
      "embedding",
      "offset",
      "cache_in_0",
      "cache_in_1",
      "cache_in_2",
      "cache_in_3",
      "cache_in_4",
      "cache_in_5"
    ],
    "output_names": [
      "c",
      "eos",
      "cache_out_0",
      "cache_out_1",
      "cache_out_2",
      "cache_out_3",
      "cache_out_4",
      "cache_out_5"
    ]
  },
  "mimi": {
    "context": 250,
    "frames_per_step": 16,
    "state_names": [
      "decoder.model.0/previous",
      "decoder.model.11/previous",
      "decoder.model.2/partial",
      "decoder.model.3.block.1/previous",
      "decoder.model.3.block.3/previous",
      "decoder.model.5/partial",
      "decoder.model.6.block.1/previous",
      "decoder.model.6.block.3/previous",
      "decoder.model.8/partial",
      "decoder.model.9.block.1/previous",
      "decoder.model.9.block.3/previous",
      "decoder_transformer.transformer.layers.0.self_attn/cache",
      "decoder_transformer.transformer.layers.1.self_attn/cache",
      "upsample.convtr/partial"
    ],
    "state_shapes": [
      [
        1,
        512,
        6
      ],
      [
        1,
        64,
        2
      ],
      [
        1,
        256,
        6
      ],
      [
        1,
        256,
        2
      ],
      [
        1,
        128,
        0
      ],
      [
        1,
        128,
        5
      ],
      [
        1,
        128,
        2
      ],
      [
        1,
        64,
        0
      ],
      [
        1,
        64,
        4
      ],
      [
        1,
        64,
        2
      ],
      [
        1,
        32,
        0
      ],
      [
        2,
        1,
        8,
        250,
        64
      ],
      [
        2,
        1,
        8,
        250,
        64
      ],
      [
        1,
        512,
        16
      ]
    ],
    "attn_state_indices": [
      11,
      12
    ],
    "input_names": [
      "latent",
      "offset",
      "state_in_0",
      "state_in_1",
      "state_in_2",
      "state_in_3",
      "state_in_4",
      "state_in_5",
      "state_in_6",
      "state_in_7",
      "state_in_8",
      "state_in_9",
      "state_in_10",
      "state_in_11",
      "state_in_12",
      "state_in_13"
    ],
    "output_names": [
      "pcm",
      "state_out_0",
      "state_out_1",
      "state_out_2",
      "state_out_3",
      "state_out_4",
      "state_out_5",
      "state_out_6",
      "state_out_7",
      "state_out_8",
      "state_out_9",
      "state_out_10",
      "state_out_11",
      "state_out_12",
      "state_out_13"
    ]
  }
}