{ "format": 1, "language": "fa", "source": { "repo": "jonatasgrosman/wav2vec2-large-xlsr-53-persian", "revision": "234714078a1398a9db88194c5a40fefe6f376dc1", "license": "apache-2.0", "base_model": "facebook/wav2vec2-large-xlsr-53", "datasets": [ "common_voice" ], "files": { "README.md": { "sha256": "b947d7933e6d0b5a94491cc555e80cea18378fba3b65516e83becb17a2235002", "size": 7549 }, "config.json": { "sha256": "a58a11ca6cd6387f97ba4f51ba5621ba5eec26b565c65dda080c573facae61c8", "size": 1565 }, "preprocessor_config.json": { "sha256": "c403ce09975b90dff0dd8302c42d422e9de1f166cd7772df23490069893cb0cf", "size": 158 }, "pytorch_model.bin": { "sha256": "3b859c7f562a2cc3c6002c2eb5178b66777406c4fccf53f196ead46a4f6c4796", "size": 1262208535 }, "special_tokens_map.json": { "sha256": "bb7068de1150661a10b55f9e4b12a0e77af8bf91f5e45e1b58afaf1d0e17f675", "size": 85 }, "vocab.json": { "sha256": "d1691d3d4a072da1ddc1703ca8ee142f8e0fc0f64fa50f4b76963008262c9f94", "size": 656 } } }, "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": 67, "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": 0.000247955322265625, "frame_argmax_agreement_fp32_vs_torch": 1.0, "frame_argmax_agreement_int8": 1.0, "fp32_size": 1262463839, "int8_size": 356319798, "frames": 49, "runtime": { "1.30.0": { "provider": "CPUExecutionProvider", "runs": { "0.5": { "samples": 8000, "shape": [ 1, 24, 67 ], "seconds": 0.063064 }, "60.0": { "samples": 960000, "shape": [ 1, 2999, 67 ], "seconds": 9.383666 } } }, "1.23.2": { "provider": "CPUExecutionProvider", "runs": { "0.5": { "samples": 8000, "shape": [ 1, 24, 67 ], "seconds": 0.045775 }, "60.0": { "samples": 960000, "shape": [ 1, 2999, 67 ], "seconds": 8.072696 } } } }, "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": 5.826939, "peak_rss_bytes": 2907074560 }, "synthetic_sentence": { "voice": "fa", "caption": "\u0633\u0644\u0627\u0645 \u062f\u0646\u06cc\u0627", "samples": 18657, "note": "pipeline/monotonicity check only; not an accuracy claim" } }, "files": { "model.int8.onnx": { "sha256": "b1f2c0687378a71a27567bcaa8e7b374dd81a5adbe12352fc68e43f83e67129f", "size": 356319798 } }, "elapsed_seconds": 40.263 }