{
  "models": [
    {
      "id": "baseline",
      "label": "Baseline",
      "params_m": 157.965668,
      "unique_blocks": 18,
      "color": "#59636f"
    },
    {
      "id": "full2",
      "label": "CYCLE 9 × 2",
      "params_m": 83.557988,
      "unique_blocks": 9,
      "color": "#4878a8"
    },
    {
      "id": "sequence",
      "label": "SEQUENCE 9 × 2",
      "params_m": 83.557988,
      "unique_blocks": 9,
      "color": "#8064aa"
    },
    {
      "id": "full3",
      "label": "Full ×3",
      "params_m": 58.755428,
      "unique_blocks": 6,
      "color": "#65864f"
    },
    {
      "id": "prefix",
      "label": "Prefix",
      "params_m": 108.360548,
      "unique_blocks": 12,
      "color": "#15877f"
    },
    {
      "id": "middle",
      "label": "Middle",
      "params_m": 108.360548,
      "unique_blocks": 12,
      "color": "#b98223"
    },
    {
      "id": "suffix",
      "label": "Suffix",
      "params_m": 108.360548,
      "unique_blocks": 12,
      "color": "#d56655"
    }
  ],
  "quality": {
    "seedtts": {
      "4": {
        "baseline": {
          "wer": 10.58884784033038,
          "sim": 0.4984570203570697,
          "utmos": 2.8460626198417125
        },
        "full2": {
          "wer": 9.373141089504372,
          "sim": 0.49645202707581443,
          "utmos": 3.083739211782813
        },
        "full3": {
          "wer": 11.560091689804125,
          "sim": 0.494610849653648,
          "utmos": 3.033040461678277
        },
        "prefix": {
          "wer": 6.874375319342261,
          "sim": 0.4952553632165826,
          "utmos": 3.2380951946765624
        },
        "middle": {
          "wer": 8.231033685469601,
          "sim": 0.49632832749493366,
          "utmos": 3.09709899300052
        },
        "suffix": {
          "wer": 10.31027984291083,
          "sim": 0.4988814713792903,
          "utmos": 3.074193566078868
        },
        "sequence": {
          "wer": 9.321615222853826,
          "sim": 0.4983567358679412,
          "utmos": 3.068016666754642
        }
      },
      "32": {
        "baseline": {
          "wer": 2.225220154671117,
          "sim": 0.5666470663580934,
          "utmos": 3.678098540260073
        },
        "full2": {
          "wer": 2.6008526849418123,
          "sim": 0.5616399050751866,
          "utmos": 3.900995242190273
        },
        "full3": {
          "wer": 2.956179982354205,
          "sim": 0.5565226366401048,
          "utmos": 3.859037739169948
        },
        "prefix": {
          "wer": 2.3565941512051483,
          "sim": 0.5631083216953009,
          "utmos": 3.881577420678428
        },
        "middle": {
          "wer": 2.0355973668176306,
          "sim": 0.5662922857719583,
          "utmos": 3.8441630077603106
        },
        "suffix": {
          "wer": 2.1865639504250556,
          "sim": 0.5695269287218039,
          "utmos": 3.9081346195958115
        },
        "sequence": {
          "wer": 2.0331598248445135,
          "sim": 0.572706805949565,
          "utmos": 3.7537908492993344
        }
      }
    },
    "lspc": {
      "4": {
        "baseline": {
          "wer": 17.80511213229663,
          "sim": 0.43342347415144833,
          "utmos": 2.22885447410887
        },
        "full2": {
          "wer": 11.355325634079795,
          "sim": 0.4442398167487715,
          "utmos": 2.592673526999161
        },
        "full3": {
          "wer": 16.683303588700625,
          "sim": 0.4562968442878165,
          "utmos": 2.797687915153526
        },
        "prefix": {
          "wer": 11.195717858384018,
          "sim": 0.4146613728825782,
          "utmos": 2.8083827510030974
        },
        "middle": {
          "wer": 10.830549247802953,
          "sim": 0.44532673799230904,
          "utmos": 2.69871173790664
        },
        "suffix": {
          "wer": 17.17059474343516,
          "sim": 0.4231252598715905,
          "utmos": 2.412381087489045
        },
        "sequence": {
          "wer": 11.460233943816213,
          "sim": 0.42821091138701295,
          "utmos": 2.5578398501439334
        }
      },
      "32": {
        "baseline": {
          "wer": 2.2907938923170534,
          "sim": 0.592414708452724,
          "utmos": 3.893866200413234
        },
        "full2": {
          "wer": 2.1428268284900067,
          "sim": 0.5999040717787377,
          "utmos": 4.061304483178169
        },
        "full3": {
          "wer": 2.6476590768353376,
          "sim": 0.5925562393247499,
          "utmos": 4.062871402059565
        },
        "prefix": {
          "wer": 2.2376698133965105,
          "sim": 0.5694668717688972,
          "utmos": 4.087487536152239
        },
        "middle": {
          "wer": 2.0669970533549544,
          "sim": 0.5825002649785217,
          "utmos": 4.044638044557004
        },
        "suffix": {
          "wer": 2.0987010665940877,
          "sim": 0.5932266958381819,
          "utmos": 4.020698982560046
        },
        "sequence": {
          "wer": 2.2645461150878554,
          "sim": 0.6025599951893575,
          "utmos": 3.934810112228552
        }
      }
    }
  },
  "curves": {
    "steps": [
      4,
      8,
      16,
      32
    ],
    "seedtts": {
      "baseline": [
        10.373048922194684,
        2.938872120046428,
        2.277973025214961,
        2.2631779459607095
      ],
      "full2": [
        9.329426025603578,
        3.624675565027533,
        2.7002632753876266,
        2.7015380947988
      ],
      "full3": [
        11.720389814746632,
        3.876729150625498,
        3.1562419342743113,
        3.228403139368953
      ],
      "prefix": [
        6.787808693679522,
        3.09793706149675,
        2.559455312145083,
        2.7022077802651983
      ],
      "middle": [
        7.94932222497642,
        2.5463498004182474,
        2.0153747891502,
        1.9371319535544231
      ],
      "suffix": [
        10.30620414849295,
        2.8049913189943467,
        1.9545986050906035,
        2.148447852248674
      ],
      "sequence": [
        9.699006959495714,
        2.998688378621166,
        2.214925634786686,
        2.024128253042613
      ]
    },
    "lspc": {
      "baseline": [
        18.37392368787688,
        3.300632532147961,
        2.3046390329972604,
        2.315340435636308
      ],
      "full2": [
        11.35087199198877,
        2.8305572347780346,
        2.276961745142516,
        2.2200460123271157
      ],
      "full3": [
        15.710047341622632,
        3.668381840703574,
        2.783053385973875,
        2.6591941283851597
      ],
      "prefix": [
        11.539910043760225,
        2.9317582434802376,
        2.247306312544426,
        2.190876153284804
      ],
      "middle": [
        10.63518037949794,
        2.711595931246045,
        2.0783155827519804,
        2.0210871154900802
      ],
      "suffix": [
        18.087357712583206,
        2.6132890924201875,
        2.130923971325759,
        1.931206533944556
      ],
      "sequence": [
        11.24490706468484,
        3.142069726231833,
        2.3433126122476327,
        2.4496135591581214
      ]
    }
  },
  "resources": {
    "baseline": {
      "inference_mb": 876.68736,
      "rtf": 0.14282193133216617,
      "training_allocated_gib": 13.666626930236816,
      "training_reserved_gib": 14.74609375,
      "training_gpu": "H100",
      "seconds_per_update": 0.1976313792541623,
      "inference_reserved_mb": null
    },
    "full2": {
      "inference_mb": 574.338048,
      "rtf": 0.14317677912792337,
      "training_allocated_gib": 12.270578384399414,
      "training_reserved_gib": 19.771484375,
      "training_gpu": "H100",
      "seconds_per_update": 0.1941220007278025,
      "inference_reserved_mb": null
    },
    "full3": {
      "inference_mb": 473.554944,
      "rtf": 0.14341695590769837,
      "training_allocated_gib": 11.807664394378662,
      "training_reserved_gib": 12.380859375,
      "training_gpu": "H100",
      "seconds_per_update": 0.19418209415860474,
      "inference_reserved_mb": null
    },
    "prefix": {
      "inference_mb": 675.121152,
      "rtf": 0.14259059190962003,
      "training_allocated_gib": 12.730461597442627,
      "training_reserved_gib": 13.544921875,
      "training_gpu": "H200",
      "seconds_per_update": 0.19893719430547208,
      "inference_reserved_mb": null
    },
    "middle": {
      "inference_mb": 675.121152,
      "rtf": 0.1438628096426699,
      "training_allocated_gib": 12.744056224822998,
      "training_reserved_gib": 13.267578125,
      "training_gpu": "H100",
      "seconds_per_update": 0.19918101634830238,
      "inference_reserved_mb": null
    },
    "suffix": {
      "inference_mb": 675.121152,
      "rtf": 0.1439875271842122,
      "training_allocated_gib": 12.743486881256104,
      "training_reserved_gib": 13.396484375,
      "training_gpu": "H200",
      "seconds_per_update": 0.199657229995355,
      "inference_reserved_mb": null
    },
    "sequence": {
      "inference_mb": 574.338048,
      "rtf": 0.1425890567180732,
      "training_allocated_gib": 12.261483669281006,
      "training_reserved_gib": 13.080078125,
      "training_gpu": "H100",
      "seconds_per_update": 0.19815983222797512,
      "inference_reserved_mb": null
    }
  },
  "protocol": {
    "checkpoint_update": 500000,
    "weights": "EMA",
    "block_calls_per_network_evaluation": 18,
    "solver": "Euler",
    "cfg": 2,
    "sway": -1,
    "quality_seeds": {
      "seedtts": 4,
      "lspc": 3
    },
    "quality_seed_values": {
      "seedtts": [
        0,
        1,
        2,
        666
      ],
      "lspc": [
        0,
        1,
        2
      ]
    },
    "curve_seeds": {
      "seedtts": 666,
      "lspc": 0
    },
    "utterances_per_seed": {
      "seedtts": 1088,
      "lspc": 1127
    },
    "wer": "Mean per-utterance word error rate, in percent; lower is better.",
    "sim": "SIM-o: WavLM-Large/ECAPA-TDNN speaker similarity; higher is better.",
    "utmos": "Predicted naturalness score; higher is better. This is not a human MOS.",
    "curves": "WER at a fixed inference seed across all steps. Curve endpoints therefore differ from the multi-seed quality table.",
    "inference": {
      "gpu": "H100",
      "precision": "FP32",
      "batch_size": 1,
      "steps": 32,
      "dataset": "Seed-TTS test-en",
      "seed": 666,
      "memory": "Peak allocated GPU memory in decimal MB; includes the resident vocoder and its allocations.",
      "rtf": "Total sampling time divided by total generated audio duration. Timing includes sampling and finite/prompt checks, and excludes vocoding, preprocessing, loading, and file I/O.",
      "timing_scope": "One recorded run per model; small timing differences are not a controlled speed ranking.",
      "reserved_memory": "Not recorded by the inference logger; no value is imputed."
    },
    "training": {
      "world_size": 8,
      "precision": "BF16 mixed precision; FP32 parameters, EMA, and Adam states",
      "updates": [
        400000,
        500000
      ],
      "memory": "Late-training records of rank-0 process cumulative peaks, in GiB; not the maximum over all ranks or a minimum GPU-capacity requirement.",
      "seconds_per_update": "Median over 1,000 logged windows, each covering 100 updates.",
      "hardware": "Baseline, CYCLE 9 × 2, SEQUENCE 9 × 2, Loop 6 × 3, and Middle used H100; Prefix and Suffix used H200."
    },
    "scope": "One training run per layout. Equal 500k-update checkpoints are compared; the baseline's best earlier four-step checkpoint is not substituted. No current-stack human MOS has been completed.",
    "sequence": "SEQUENCE repeats adjacent blocks (1,1,2,2,...,9,9). It uses the same 83,557,988 parameters and 18 block calls as CYCLE (1,...,9,1,...,9). Fixed-seed WER curves cover 4, 8, 16, and 32 steps on both datasets; quality-table values average the full inference-seed sets."
  }
}