{ "schema_version": 1, "status": "release", "variant": "dynamic-qint8-perchannel-speed", "source": { "repository": "nvidia/parakeet-tdt_ctc-0.6b-ja", "revision": "44edb27eea9317daf89333e75eb830db4b1cc298", "checkpoint_filename": "parakeet-tdt_ctc-0.6b-ja.nemo", "checkpoint_bytes": 2490951680, "checkpoint_sha256": "a2ed2aab9c82cce4e9b699dca669184d7525c7b5e1eb8c3f7b789421bc3637e8", "license": "CC-BY-4.0" }, "modification": { "export": "faithful shared encoder extracted from the monolithic CTC export; FP32 fused TDT decoder/joint and CTC head", "quantization": "onnxruntime.quantization.quantize_dynamic; MatMul only; QInt8; per-channel; external data", "fine_tuned": false }, "environment": { "export": { "nemo": "3.1.0+b331a3488", "torch": "2.12.0+cpu", "onnx": "1.22.0" }, "quantization": { "python": "3.12.6", "onnxruntime": "1.27.0" }, "validated_runtime": { "onnxruntime": "1.24.4", "provider": "CPUExecutionProvider", "threads": 4 } }, "payload_bytes": 940746826, "payload_files": { "encoder-model.int8.onnx": { "role": "shared quantized encoder graph", "bytes": 3387454, "sha256": "3c4f14996c134b1e6ca2853230ccc0d36c66085af0b6272e002dfba0f25c5f5a" }, "encoder-model.int8.onnx.data": { "role": "shared quantized encoder weights", "bytes": 878444544, "sha256": "a39578a7c16db2d7024dbe8f90ae99892cef69a8bb833b5d1223f9122a7e87bc" }, "decoder_joint-model.onnx": { "role": "FP32 fused TDT predictor and joint network", "bytes": 46286012, "sha256": "64e64c50b7df62707dc7ac1e8b1ea804b9f590db88d84e4abcdf4d1188cf9b8a" }, "ctc-head-model.onnx": { "role": "FP32 CTC head graph", "bytes": 959, "sha256": "23f65702c75a984fa16c95bd284b66b60f9d6693bf1169dc2947b56734cc72d9" }, "ctc-head-model.onnx_data": { "role": "FP32 CTC head weights", "bytes": 12599300, "sha256": "662b5f6b91dd4987ac9d139cc185501979b4de51947b5ba47b4d09cae75287b9" }, "vocab.txt": { "role": "token vocabulary", "bytes": 28557, "sha256": "732f64c53909f2620c713f4106b487d92e6f54a6915b3cd3d1dbd32f9f4f392a" } } }