{
  "format": 1,
  "language": "it",
  "source": {
    "repo": "jonatasgrosman/wav2vec2-large-xlsr-53-italian",
    "revision": "dab04a3e00d8326052f3fb22a6ff276b822f6131",
    "license": "apache-2.0",
    "base_model": "facebook/wav2vec2-large-xlsr-53",
    "datasets": [
      "common_voice"
    ],
    "files": {
      "pytorch_model.bin": {
        "sha256": "14040a54d1e0caecd8b2271fdd2d96653be45702c7bc8765354140d864fb445a",
        "size": 1262114199
      },
      "README.md": {
        "sha256": "c7cfa3035f0fdc73dd350a65c118f930e60a5999b549a5c570119a0bede29b77",
        "size": 5531
      },
      "config.json": {
        "sha256": "7e69e919b00b2044006c69cea10a0b15d35d65b6acec9f7180afef2edc7df1fb",
        "size": 1782
      },
      "preprocessor_config.json": {
        "sha256": "ca5999a45e98bb76ea87a461ba28a23ad32a5bb9f733b8e0f6546ff38b6c612d",
        "size": 262
      },
      "special_tokens_map.json": {
        "sha256": "bb7068de1150661a10b55f9e4b12a0e77af8bf91f5e45e1b58afaf1d0e17f675",
        "size": 85
      },
      "vocab.json": {
        "sha256": "deee9194b91e4e4f628ff37da6226a9c9cfe564810c917ae545920458723c58e",
        "size": 410
      }
    }
  },
  "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": 44,
    "word_delimiter": "|",
    "text_case": "preserve",
    "unit": "char"
  },
  "text": {
    "spaces_separate_words": true,
    "unknown_characters": "wildcard"
  },
  "verify": {
    "tools": {
      "onnxruntime": [
        "1.30.0",
        "1.23.2"
      ]
    },
    "logit_maxabs_fp32_vs_torch": 0.0001678466796875,
    "frame_argmax_agreement_int8": 1.0,
    "fp32_size": 1262369539,
    "int8_size": 356296099,
    "frames": 49,
    "runtime": {
      "1.30.0": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              44
            ],
            "seconds": 0.060175
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              44
            ],
            "seconds": 7.573283
          }
        }
      },
      "1.23.2": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              44
            ],
            "seconds": 0.045074
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              44
            ],
            "seconds": 7.905604
          }
        }
      }
    },
    "word_start_delta_ms_int8_vs_fp32": {
      "median": 0.0,
      "p95": 0.0,
      "max": 0.0,
      "spans": 2,
      "test_audio": "espeak-ng generated sentence"
    },
    "sanity": {
      "monotonic_sentence_spans": 2,
      "unknown_wildcard_spans": 2
    },
    "cost": {
      "audio_seconds": 30,
      "alignment_seconds": 4.674252,
      "peak_rss_bytes": 2874867712
    },
    "synthetic_sentence": {
      "voice": "it",
      "caption": "ciao mondo",
      "samples": 15187,
      "note": "pipeline/monotonicity check only; not an accuracy claim"
    },
    "ground_truth_generated_boundaries": {
      "words": [
        "ciao",
        "mondo"
      ],
      "silence_ms": 500,
      "segment_start_errors_ms": [
        20.125686813186814,
        51.91483516483527
      ],
      "median": 36.02026098901104,
      "p95": 50.325377747252844,
      "max": 51.91483516483527,
      "note": "espeak-ng segments with known 500 ms join; segment boundaries, not human speech labels"
    }
  },
  "files": {
    "model.int8.onnx": {
      "sha256": "94009e742d5fa40976eb5594e038199e3a3ca81eee21de28c9b86493e7325da5",
      "size": 356296099
    }
  },
  "elapsed_seconds": 41.274
}