{ "v1/mps": { "files": 300, "audio_seconds": 2106, "wall_s": 30.4, "ms_per_audio_second": 14.43, "x_realtime": 69.3 }, "v2/mps": { "files": 300, "audio_seconds": 2106, "wall_s": 6.51, "ms_per_audio_second": 3.09, "x_realtime": 323.6 }, "v1/cpu": { "files": 300, "audio_seconds": 2106, "wall_s": 279.8, "ms_per_audio_second": 132.86, "x_realtime": 7.5 }, "v2/cpu": { "files": 300, "audio_seconds": 2106, "wall_s": 82.92, "ms_per_audio_second": 39.38, "x_realtime": 25.4 }, "v3.1/mps": { "files": 300, "audio_seconds": 2106.0, "wall_s": 13.75, "ms_per_audio_second": 6.53, "x_realtime": 153.2 }, "v3.1/cpu": { "files": 300, "audio_seconds": 2106.0, "wall_s": 49.03, "ms_per_audio_second": 23.28, "x_realtime": 43.0 }, "_setup": { "machine": "Apple MacBook Pro, M3 Pro (6 performance + 6 efficiency CPU cores, 18-core GPU), 36 GB RAM", "os": "macOS 15.7.4", "python": "3.13.3", "torch": "2.15.0", "torch_cpu_threads": 6, "data": "first 300 files of the regenerated LibriSpeech test-clean set (2 106 s of audio, 16 kHz FLAC)", "method": "one process, Detector.predict_file per file, end to end (audio decoding, features, model); model loading and one warm-up file excluded" }, "nisqa/cpu": { "files": 300, "audio_seconds": 2111, "wall_s": 7.95, "ms_per_audio_second": 3.76, "x_realtime": 265.6 }, "v3-lite/cpu": { "files": 300, "audio_seconds": 2111, "wall_s": 24.89, "ms_per_audio_second": 11.79, "x_realtime": 84.8 }, "v3-lite/mps": { "files": 300, "audio_seconds": 2111, "wall_s": 27.74, "ms_per_audio_second": 13.14, "x_realtime": 76.1 } }