{
  "configuration": {
    "evaluator": "vocoderbench.mos.MosEvaluator",
    "source_commit": "cc2700db57bb83ee13dc31ebe1b868c254e15d09",
    "checkpoint_revision": "506474f2b33dc77c234d668cc419be1861899cad",
    "checkpoint_sha256": "c8149d988e4bbf3f347e6966b5d769de347a5f8c59ffca1dc4bd4bf5b8585e57",
    "config_sha256": {
      "config.json": "4937977e24d12d1bba70cdce8709c3c04807a8e4ae8ddac4229c48c436ae99ae",
      "preprocessor_config.json": "b225d617c025463b9e157e06afea8b90dc7078fc70b013c533328423e0486b4a"
    },
    "source_sha256": "fb632aa5440245525289962372db4f3eb007cd2b3b493b2d05ecccb878e85f03",
    "fold": 0,
    "extraction_seed": 20261012,
    "extraction_repetitions": 1,
    "device": "cpu",
    "threads": 4,
    "predict_dataset": "bvcc",
    "remove_silent_section": false,
    "input": "native_full_length_float32",
    "nnpack_enabled": false
  },
  "configuration_sha256": "100fb0fae926ac8b022a980e35ec11e2286fedb1fc7d798a1b9647f448aedb65",
  "source_script_sha256": "8087feec9501fc6ca8d953768a8bc1733c28dabc32b8fc719ad9599f4c3ef7cb",
  "manifest_sha256": "90130e77fc19c2263d985fa6e138805ba187487b9fdefa8089cd1d88a5d1da8b",
  "previous_results_sha256": "dea3413f70beef631f758c086c036dfb94e98258e59834e994cf64b31364bf63",
  "environment": {
    "python": "3.12.14",
    "platform": "Linux-6.18.44-x86_64-with-glibc2.41",
    "cpu": "",
    "logical_cpu_count": 9,
    "packages": {
      "numpy": "2.3.5",
      "scipy": "1.17.0",
      "torch": "2.8.0+cpu",
      "torchaudio": "2.8.0+cpu",
      "onnxruntime": "1.22.0",
      "speechbrain": "1.0.3",
      "utmosv2": "1.3.1.dev0",
      "transformers": "4.51.3",
      "timm": "1.0.15",
      "librosa": "0.11.0"
    }
  },
  "previous_mos_provenance": {
    "name": "UTMOSv2",
    "quantity": "predicted_mos_not_human_rating",
    "source_repository": "https://github.com/sarulab-speech/UTMOSv2",
    "source_commit": "cc2700db57bb83ee13dc31ebe1b868c254e15d09",
    "checkpoint_revision": "506474f2b33dc77c234d668cc419be1861899cad",
    "checkpoint_sha256": "c8149d988e4bbf3f347e6966b5d769de347a5f8c59ffca1dc4bd4bf5b8585e57",
    "fold": 0,
    "device": "cpu",
    "configuration": "fusion_stage3",
    "predict_dataset": "bvcc",
    "repetitions": 1,
    "extraction_seed": 20261012,
    "cpu_threads": 4,
    "remove_silent_section": false,
    "nnpack_enabled": false,
    "internal_sample_rate": 16000,
    "ssl_crop_seconds": 3.0,
    "spectrogram_crop_seconds": 1.4,
    "spectrogram_frames_per_repetition": 2,
    "spectrogram_inner_mixup": true,
    "short_input_extension": "tile",
    "code_license": "MIT",
    "weights_license": "MIT",
    "independent_human_rating_calibration": "not_performed_for_this_project",
    "scope": "JVS/LibriSpeechの再合成への適用は未校正。言語・発声別に集計する。",
    "paper_protocol_difference": "論文の5fold×5抽出平均ではなく、fold0と上記反復数。",
    "initialization": "official_architecture_then_complete_strict_state_dict",
    "packages": {
      "torch": "2.8.0+cpu",
      "torchaudio": "2.8.0+cpu",
      "torchvision": "0.23.0+cpu",
      "transformers": "4.51.3",
      "timm": "1.0.15",
      "librosa": "0.11.0",
      "numpy": "2.3.5",
      "utmosv2": "1.3.1.dev0"
    },
    "checkpoint": {
      "url": "https://huggingface.co/sarulab-speech/UTMOSv2/resolve/506474f2b33dc77c234d668cc419be1861899cad/fold0_s42_best_model.pth",
      "sha256": "c8149d988e4bbf3f347e6966b5d769de347a5f8c59ffca1dc4bd4bf5b8585e57",
      "bytes": 818531314
    },
    "feature_configuration": {
      "config.json": {
        "url": "https://huggingface.co/facebook/wav2vec2-base/resolve/0b5b8e868dd84f03fd87d01f9c4ff0f080fecfe8/config.json",
        "sha256": "4937977e24d12d1bba70cdce8709c3c04807a8e4ae8ddac4229c48c436ae99ae",
        "bytes": 1842
      },
      "preprocessor_config.json": {
        "url": "https://huggingface.co/facebook/wav2vec2-base/resolve/0b5b8e868dd84f03fd87d01f9c4ff0f080fecfe8/preprocessor_config.json",
        "sha256": "b225d617c025463b9e157e06afea8b90dc7078fc70b013c533328423e0486b4a",
        "bytes": 159
      }
    },
    "initialization_status": "ready"
  }
}