meta.json

{
  "format": 1,
  "language": "de",
  "source": {
    "repo": "jonatasgrosman/wav2vec2-large-xlsr-53-german",
    "revision": "4b8a02957378d0f2da2ef74091156b032c485a89",
    "license": "apache-2.0",
    "base_model": "facebook/wav2vec2-large-xlsr-53",
    "datasets": [
      "common_voice"
    ],
    "files": {
      "pytorch_model.bin": {
        "sha256": "bff6d75ab89d8ca9cd103df9beb9c10f547501cf5a34aeabea1c8d736c1b81cb",
        "size": 1262089623
      },
      "README.md": {
        "sha256": "cbcb4a266a15873663377fd9904abf646401a2df0b11e44456b035bb7aa936c4",
        "size": 5714
      },
      "config.json": {
        "sha256": "42175a918795a7343c58dbcad995603912a0d85c729cc4d7b8eb089a7a569183",
        "size": 1782
      },
      "preprocessor_config.json": {
        "sha256": "ca5999a45e98bb76ea87a461ba28a23ad32a5bb9f733b8e0f6546ff38b6c612d",
        "size": 262
      },
      "special_tokens_map.json": {
        "sha256": "bb7068de1150661a10b55f9e4b12a0e77af8bf91f5e45e1b58afaf1d0e17f675",
        "size": 85
      },
      "vocab.json": {
        "sha256": "31d0914432f4492a24ce83abfd767421844afe5cef20346a48c6c877dee28378",
        "size": 350
      }
    }
  },
  "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": 38,
    "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": 0.000301361083984375,
    "frame_argmax_agreement_int8": 1.0,
    "fp32_size": 1262344939,
    "int8_size": 356289918,
    "frames": 49,
    "runtime": {
      "1.30.0": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              38
            ],
            "seconds": 0.071937
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              38
            ],
            "seconds": 9.554073
          }
        }
      },
      "1.23.2": {
        "provider": "CPUExecutionProvider",
        "runs": {
          "0.5": {
            "samples": 8000,
            "shape": [
              1,
              24,
              38
            ],
            "seconds": 0.042747
          },
          "60.0": {
            "samples": 960000,
            "shape": [
              1,
              2999,
              38
            ],
            "seconds": 9.145823
          }
        }
      }
    },
    "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": 5.648164,
      "peak_rss_bytes": 2879397888
    },
    "synthetic_sentence": {
      "voice": "de",
      "caption": "hallo welt",
      "samples": 15531,
      "note": "pipeline/monotonicity check only; not an accuracy claim"
    }
  },
  "files": {
    "model.int8.onnx": {
      "sha256": "df581a4ef6059fb6058060b06af271c9096b82fad4f0b8ba1f33c8eea027df69",
      "size": 356289918
    }
  },
  "elapsed_seconds": 41.488
}