meta.json

{
  "format": 1,
  "language": "ja",
  "source": {
    "repo": "reazon-research/japanese-wav2vec2-large-rs35kh",
    "revision": "2162c9edb97f362f7623d6b3a2e417458e879bca",
    "license": "apache-2.0",
    "base_model": "reazon-research/japanese-wav2vec2-large",
    "datasets": [
      "reazonspeech"
    ],
    "files": {
      "README.md": {
        "sha256": "9f0276f386a3ced3211d1c9ceeb9d717edc380c565a1a0f4b1cf388df398c95c",
        "size": 3345
      },
      "added_tokens.json": {
        "sha256": "4de0ae97b4959a8d86c933b0ec9afaf72a917ebf02d1693c1ccba1d34c575599",
        "size": 51
      },
      "config.json": {
        "sha256": "07c76a314b3968cb69f17139cd31eecfabaccc61d0224ae2e0500a7288665d84",
        "size": 2044
      },
      "model.safetensors": {
        "sha256": "cc0228dfe1d939e541624c78911eafc7041698e1b2f4853de8a6364fb4557364",
        "size": 1274078500
      },
      "preprocessor_config.json": {
        "sha256": "ffc72f475cd79680d5d659fbaefc691cdbcee5c9703a15faf708a5e91452a8d8",
        "size": 256
      },
      "special_tokens_map.json": {
        "sha256": "fe941423230c28a1b96e319effd1cb23982cadf1d27e2a62f6054313b6d3fc3f",
        "size": 550
      },
      "tokenizer_config.json": {
        "sha256": "c36b3a74819e8290b923327b9e1cddd2331cc08c2aca29cb379e0aad0cfb6ea9",
        "size": 89573
      },
      "vocab.json": {
        "sha256": "223300c979885b08b6a44f18b6b76cae49099c6de1a0bc2d5694a32827b3a7ab",
        "size": 46399
      }
    }
  },
  "model": {
    "file": "model.int8.onnx",
    "precision": "int8-dynamic",
    "input": "input_values",
    "output": "logits",
    "sample_rate": 16000,
    "frame_samples": 320,
    "min_samples": 400,
    "normalize": "zero-mean-unit-variance",
    "blank_id": 0,
    "vocab_size": 3003,
    "word_delimiter": null,
    "text_case": "preserve",
    "unit": "char"
  },
  "text": {
    "spaces_separate_words": false,
    "unknown_characters": "wildcard"
  },
  "verify": {
    "tools": {
      "onnxruntime": [
        "1.30.0",
        "1.23.2"
      ]
    },
    "logit_maxabs_fp32_vs_torch": 9.918212890625e-05,
    "frame_argmax_agreement_fp32_vs_torch": 1.0,
    "frame_argmax_agreement_int8": 1.0,
    "fp32_size": 1274460764,
    "int8_size": 359313382,
    "frames": 49,
    "runtime": {
      "1.30.0": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              3003
            ],
            "seconds": 0.069697
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              3003
            ],
            "seconds": 13.355701
          }
        }
      },
      "1.23.2": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              3003
            ],
            "seconds": 0.033885
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              3003
            ],
            "seconds": 7.733478
          }
        }
      }
    },
    "word_start_delta_ms_int8_vs_fp32": {
      "median": 0.0,
      "p95": 0.0,
      "max": 0.0,
      "spans": 7,
      "test_audio": "espeak-ng generated sentence"
    },
    "sanity": {
      "monotonic_sentence_spans": 7,
      "unknown_wildcard_spans": 1
    },
    "cost": {
      "audio_seconds": 30,
      "alignment_seconds": 4.144841,
      "peak_rss_bytes": 2903474176
    },
    "synthetic_sentence": {
      "voice": "ja",
      "caption": "\u3053\u3093\u306b\u3061\u306f\u4e16\u754c",
      "samples": 45277,
      "note": "pipeline/monotonicity check only; not an accuracy claim"
    }
  },
  "files": {
    "model.int8.onnx": {
      "sha256": "90b0e01385f90809f7818ffcc14d7ba18070b63e58af0aa64cc86cb3e703a24c",
      "size": 359313382
    }
  },
  "elapsed_seconds": 37.214
}