{ "date": "2026-09-05", "source_commit": "49f04884ae5f9d2dd393f5ed99094987f2311b4e", "measurement_script": "tests/batch_decode/provider_recheck.py", "platforms": { "4090": { "recognizer-trt-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-trt-3s.json" }, "utc": "2026-09-05T11:21:29Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.34765625, "VmRSS_MiB": 43.984375, "Pss_MiB": 42.8134765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93998.19140625, "memory.current_MiB": 10056.6640625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 0, 2535, 69.00" }, "rows": [ { "seconds": 3, "p50_ms": 9.83219500631094, "p95_ms": 9.945117868483067, "samples_ms": [ 9.945117868483067, 9.769310941919684, 9.83219500631094 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 9419.66015625, "VmRSS_MiB": 7545.27734375, "Pss_MiB": 7539.62109375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88220.77734375, "memory.current_MiB": 15480.61328125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11096, 1, 2535, 66.00" }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 5.668936863075942, "ready": { "VmHWM_MiB": 9419.66015625, "VmRSS_MiB": 7535.2734375, "Pss_MiB": 7529.62109375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88229.89453125, "memory.current_MiB": 15478.8125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11076, 1, 2535, 64.71" }, "after_warmup": { "VmHWM_MiB": 9419.66015625, "VmRSS_MiB": 7545.265625, "Pss_MiB": 7539.61328125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88221.01171875, "memory.current_MiB": 15480.59375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11096, 1, 2535, 65.21" }, "end": { "VmHWM_MiB": 9419.66015625, "VmRSS_MiB": 7545.28125, "Pss_MiB": 7539.625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88205.765625, "memory.current_MiB": 15480.78125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11096, 1, 2535, 65.96" } }, "recognizer-cuda-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-cuda-3s.json" }, "utc": "2026-09-05T11:21:22Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.0703125, "VmRSS_MiB": 44.05078125, "Pss_MiB": 42.8828125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94102.03515625, "memory.current_MiB": 10056.3359375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 0, 210, 24.68" }, "rows": [ { "seconds": 3, "p50_ms": 12.316416949033737, "p95_ms": 12.441034894436598, "samples_ms": [ 12.441034894436598, 12.279435060918331, 12.316416949033737 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 1349.44921875, "VmRSS_MiB": 1338.8359375, "Pss_MiB": 1333.130859375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92710.6875, "memory.current_MiB": 11139.91796875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 65.64" }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 1.2943845291156322, "ready": { "VmHWM_MiB": 1349.44921875, "VmRSS_MiB": 1285.0390625, "Pss_MiB": 1279.384765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92787.00390625, "memory.current_MiB": 11133.140625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 63.75" }, "after_warmup": { "VmHWM_MiB": 1349.44921875, "VmRSS_MiB": 1338.82421875, "Pss_MiB": 1333.123046875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92735.7578125, "memory.current_MiB": 11139.85546875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 63.85" }, "end": { "VmHWM_MiB": 1349.44921875, "VmRSS_MiB": 1338.83984375, "Pss_MiB": 1333.134765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92702.60546875, "memory.current_MiB": 11140.25390625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 65.64" } }, "profile-overhead-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100", "warmup": 10, "repeats": 50, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/4090/profile-overhead-cuda.json" }, "utc": "2026-09-05T11:20:16Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.6796875, "VmRSS_MiB": 44.1875, "Pss_MiB": 43.13671875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93913.26171875, "memory.current_MiB": 9749.13671875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 1, 2535, 68.36" }, "rows": [ { "frames": 50, "p50_ms": 19.815590931102633, "p95_ms": 27.880260022357106, "samples_ms": [ 19.896043930202723, 19.520729081705213, 19.733091117814183, 19.640037091448903, 20.090003963559866, 20.683970069512725, 19.486719043925405, 22.75036391802132, 20.187159068882465, 19.594815094023943, 19.631236093118787, 19.40722414292395, 24.65425105765462, 27.880260022357106, 27.882711961865425, 27.68651908263564, 27.85342396236956, 28.622120153158903, 24.61720909923315, 20.64343192614615, 20.97013220191002, 20.505737978965044, 20.043679047375917, 19.600895000621676, 19.92457709275186, 25.274900952354074, 19.845237024128437, 20.04446811042726, 19.601206993684173, 19.462329102680087, 19.60747502744198, 19.825123017653823, 19.7522749658674, 19.799976143985987, 19.655192038044333, 20.979034015908837, 19.806058844551444, 19.478884059935808, 19.939049845561385, 19.65229003690183, 19.399993121623993, 19.47223301976919, 19.20835510827601, 19.587428076192737, 19.61425389163196, 19.839842803776264, 20.576479146257043, 19.779864000156522, 19.459079019725323, 19.400409888476133 ], "memory": { "VmHWM_MiB": 1329.8125, "VmRSS_MiB": 1313.82421875, "Pss_MiB": 1308.232421875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93288.3828125, "memory.current_MiB": 10814.54296875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 41, 2685, 102.88" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 21.635569515638053, "p95_ms": 23.34937802515924, "samples_ms": [ 38.000833010300994, 23.94727780483663, 21.299692802131176, 20.927957026287913, 20.719661843031645, 22.8713050018996, 22.71975320763886, 22.436712170019746, 21.671341033652425, 22.811661008745432, 23.023668909445405, 21.576959174126387, 21.346628898754716, 20.96745721064508, 21.064766915515065, 20.16754588112235, 20.16551815904677, 21.077806130051613, 20.03202703781426, 20.063359988853335, 21.19093295186758, 21.44456305541098, 21.725330036133528, 21.566512063145638, 21.486128913238645, 21.4913550298661, 21.942669991403818, 21.388205932453275, 21.204244112595916, 21.590422838926315, 21.858954103663564, 22.212967975065112, 21.5816309209913, 21.577497012913227, 22.049805149435997, 23.34937802515924, 21.4217659085989, 22.20359817147255, 21.874733036383986, 21.504993084818125, 21.69066690839827, 21.83964801952243, 21.846866933628917, 21.916901925578713, 21.761723095551133, 21.72456099651754, 21.870122058317065, 21.713213995099068, 21.884783869609237, 21.599797997623682 ], "memory": { "VmHWM_MiB": 1531.5390625, "VmRSS_MiB": 1531.5390625, "Pss_MiB": 1525.947265625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92928.18359375, "memory.current_MiB": 11032.3203125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 39, 2715, 126.33" }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" } ], "load_s": 1.2610339389648288, "ready": { "VmHWM_MiB": 1329.8125, "VmRSS_MiB": 1257.6171875, "Pss_MiB": 1252.076171875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92941.83984375, "memory.current_MiB": 10807.96875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 67.85" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "CUDAExecutionProvider": 348960, "CPUExecutionProvider": 94200 }, "end": { "VmHWM_MiB": 3483.21875, "VmRSS_MiB": 3187.12109375, "Pss_MiB": 3181.533203125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 90607.20703125, "memory.current_MiB": 12997.9921875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 70.29" } }, "proof-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/4090/proof-cuda.json" }, "utc": "2026-09-05T11:03:26Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.45703125, "VmRSS_MiB": 44.0546875, "Pss_MiB": 43.025390625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94547.80078125, "memory.current_MiB": 7768.75390625, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "9569, 36, 2715, 129.35" }, "rows": [ { "frames": 50, "p50_ms": 102.02446603216231, "p95_ms": 102.02446603216231, "samples_ms": [ 102.02446603216231 ], "memory": { "VmHWM_MiB": 1331.94140625, "VmRSS_MiB": 1315.3828125, "Pss_MiB": 1309.8125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93594.578125, "memory.current_MiB": 8835.6875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 6, 2715, 79.02" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 1.2600936400704086, "ready": { "VmHWM_MiB": 1331.94140625, "VmRSS_MiB": 1259.19921875, "Pss_MiB": 1253.68359375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93600.125, "memory.current_MiB": 8829.0625, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "10998, 0, 2715, 80.04" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "CUDAExecutionProvider": 2908, "CPUExecutionProvider": 785 }, "end": { "VmHWM_MiB": 1331.94140625, "VmRSS_MiB": 1329.23828125, "Pss_MiB": 1323.66796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93601.73828125, "memory.current_MiB": 8852.23046875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 6, 2715, 77.90" } }, "maximum-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "2000", "warmup": 1, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/maximum-trt.json" }, "utc": "2026-09-05T11:04:16Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 44.9296875, "VmRSS_MiB": 43.859375, "Pss_MiB": 42.8876953125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94635.51953125, "memory.current_MiB": 8155.8515625, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "9569, 0, 2535, 70.47" }, "rows": [ { "frames": 2000, "p50_ms": 103.71513315476477, "p95_ms": 119.16982592083514, "samples_ms": [ 119.16982592083514, 103.71513315476477, 101.45730199292302 ], "memory": { "VmHWM_MiB": 9400.4140625, "VmRSS_MiB": 7527.8359375, "Pss_MiB": 7522.37890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88046.56640625, "memory.current_MiB": 13579.19921875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11398, 83, 2535, 159.42" }, "finite": true, "input_sha256": "8dd6299031bec40b45a4061c4670062fb06238b5da5111ea3e49d85335d81d47" } ], "load_s": 6.318245554110035, "ready": { "VmHWM_MiB": 9400.4140625, "VmRSS_MiB": 7525.31640625, "Pss_MiB": 7519.86328125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88076.828125, "memory.current_MiB": 13578.86328125, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11086, 1, 2535, 69.79" }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 9400.4140625, "VmRSS_MiB": 7527.90234375, "Pss_MiB": 7522.4453125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88041.20703125, "memory.current_MiB": 13579.5, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11398, 83, 2535, 160.96" } }, "proof-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/4090/proof-trt.json" }, "utc": "2026-09-05T11:04:24Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.34765625, "VmRSS_MiB": 43.8046875, "Pss_MiB": 42.77734375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93935.015625, "memory.current_MiB": 8164.43359375, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "9569, 0, 2715, 84.74" }, "rows": [ { "frames": 50, "p50_ms": 36.35196899995208, "p95_ms": 36.35196899995208, "samples_ms": [ 36.35196899995208 ], "memory": { "VmHWM_MiB": 9402.30859375, "VmRSS_MiB": 7527.8125, "Pss_MiB": 7522.30078125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88699.7734375, "memory.current_MiB": 13579.2890625, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11106, 0, 2535, 71.36" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 5.809436520095915, "ready": { "VmHWM_MiB": 9402.30859375, "VmRSS_MiB": 7525.28125, "Pss_MiB": 7519.7734375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88731.51953125, "memory.current_MiB": 13578.76171875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11086, 1, 2535, 71.31" }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "TensorrtExecutionProvider": 1 }, "end": { "VmHWM_MiB": 9402.30859375, "VmRSS_MiB": 7527.921875, "Pss_MiB": 7522.4140625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88408.96484375, "memory.current_MiB": 13579.5, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11106, 0, 2535, 71.34" } }, "environment": { "platform": "Linux-6.8.0-136-generic-x86_64-with-glibc2.36", "python": "3.10.19", "packages": { "onnxruntime-gpu": "1.23.2", "numpy": "2.2.6", "sherpa-onnx": "1.12.36", "torch": "2.9.1", "kitt-core": "0.3.7", "pipecat-ai": "0.0.108", "tensorrt-cu12": "10.13.3.9" }, "sha256": { "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt": "f449eb28dc567533d7fa59be34e2abca8784f771850c78a47fb731a31429a1dc", "/app/src/asr/batch_asr_manager.py": "fb9a00f6cdbc5bd972d3bb82c85b44a70277a593d1694460071bac124ef25815", "/app/src/kitt_stt_service.py": "5283feaee7a213feb529afe1d98308c7fec447ccd18aa9f737a7632ea08b83bc" } }, "recognizer-offline-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-offline-6s.json" }, "utc": "2026-09-05T11:21:55Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "06e5dd3d0564acef6b62a38ac08f597c0b8676949e38fea8ab9fd995d57f6457", "start": { "VmHWM_MiB": 43.9296875, "VmRSS_MiB": 43.9296875, "Pss_MiB": 42.9111328125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93892.6640625, "memory.current_MiB": 10056.37109375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 1, 2535, 66.61" }, "rows": [ { "seconds": 6, "p50_ms": 14.036272186785936, "p95_ms": 16.950801946222782, "samples_ms": [ 16.950801946222782, 14.036272186785936, 13.705335091799498 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 1577.28515625, "VmRSS_MiB": 973.48828125, "Pss_MiB": 967.98828125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93513.75390625, "memory.current_MiB": 10619.52734375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11106, 1, 2535, 65.74" }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 1.6889989480841905, "ready": { "VmHWM_MiB": 1577.28515625, "VmRSS_MiB": 928.2890625, "Pss_MiB": 922.79296875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93539.0, "memory.current_MiB": 10582.00390625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11070, 1, 2535, 65.00" }, "after_warmup": { "VmHWM_MiB": 1577.28515625, "VmRSS_MiB": 960.953125, "Pss_MiB": 955.45703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93524.015625, "memory.current_MiB": 10606.80859375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11090, 1, 2535, 65.74" }, "end": { "VmHWM_MiB": 1577.28515625, "VmRSS_MiB": 973.49609375, "Pss_MiB": 967.99609375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93513.75390625, "memory.current_MiB": 10619.60546875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11106, 1, 2535, 67.04" } }, "recognizer-offline-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-offline-3s.json" }, "utc": "2026-09-05T11:21:52Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "06e5dd3d0564acef6b62a38ac08f597c0b8676949e38fea8ab9fd995d57f6457", "start": { "VmHWM_MiB": 43.6953125, "VmRSS_MiB": 43.6953125, "Pss_MiB": 42.669921875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93623.375, "memory.current_MiB": 10056.3359375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 31, 2535, 68.47" }, "rows": [ { "seconds": 3, "p50_ms": 9.804417146369815, "p95_ms": 10.050446027889848, "samples_ms": [ 10.050446027889848, 9.804417146369815, 9.670068975538015 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 1576.5703125, "VmRSS_MiB": 961.19140625, "Pss_MiB": 955.6845703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93136.42578125, "memory.current_MiB": 10607.5546875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11090, 1, 2535, 66.24" }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 1.7296661280561239, "ready": { "VmHWM_MiB": 1576.5703125, "VmRSS_MiB": 928.125, "Pss_MiB": 922.6220703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93168.34765625, "memory.current_MiB": 10582.02734375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11070, 1, 2535, 64.87" }, "after_warmup": { "VmHWM_MiB": 1576.5703125, "VmRSS_MiB": 960.62890625, "Pss_MiB": 955.1259765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93160.29296875, "memory.current_MiB": 10606.921875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11090, 1, 2535, 65.07" }, "end": { "VmHWM_MiB": 1576.5703125, "VmRSS_MiB": 961.203125, "Pss_MiB": 955.6962890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93107.6328125, "memory.current_MiB": 10607.66015625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11090, 1, 2535, 66.24" } }, "graph-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 50, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/graph-trt.json" }, "utc": "2026-09-05T11:03:47Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 44.7578125, "VmRSS_MiB": 44.0390625, "Pss_MiB": 43.009765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94002.6640625, "memory.current_MiB": 7781.96484375, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "9569, 22, 2535, 70.82" }, "rows": [ { "frames": 50, "p50_ms": 5.870328983291984, "p95_ms": 6.181202130392194, "samples_ms": [ 5.995477084070444, 6.181202130392194, 5.990627920255065, 5.884398939087987, 5.874823080375791, 5.8600930497050285, 5.8989571407437325, 5.982069065794349, 5.889218067750335, 5.878840107470751, 5.897678900510073, 5.968726007267833, 6.167426006868482, 5.916664842516184, 5.827125860378146, 5.80192799679935, 5.80118503421545, 5.797432037070394, 5.843093153089285, 5.823334213346243, 5.803827894851565, 5.774520104750991, 6.1061689630150795, 6.498213857412338, 6.002160022035241, 5.961541086435318, 5.887178936973214, 5.852616857737303, 5.956716137006879, 5.842820042744279, 5.9444650541990995, 5.80316293053329, 5.783254047855735, 5.816357908770442, 5.935645895078778, 5.864623002707958, 5.83229586482048, 5.845745792612433, 5.865834886208177, 6.148942979052663, 5.941423121839762, 5.842474987730384, 5.813851952552795, 5.783188855275512, 6.281176814809442, 5.864342907443643, 5.825690925121307, 5.808993009850383, 5.758647108450532, 5.8880518190562725 ], "memory": { "VmHWM_MiB": 10882.62890625, "VmRSS_MiB": 8945.109375, "Pss_MiB": 8939.5615234375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87707.7421875, "memory.current_MiB": 14940.9609375, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11504, 95, 2715, 109.86" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 5.860850098542869, "p95_ms": 6.049487041309476, "samples_ms": [ 5.961216986179352, 5.935000954195857, 5.871298955753446, 5.851659923791885, 5.782776046544313, 5.8162331115454435, 5.962618160992861, 6.049487041309476, 5.870247958227992, 5.894138943403959, 5.8134200517088175, 5.968550220131874, 5.880054784938693, 5.829411093145609, 5.814214004203677, 5.801074905321002, 6.04877807199955, 6.00055605173111, 5.872046109288931, 5.8604341465979815, 5.856740055605769, 5.7862610556185246, 5.831134971231222, 5.851654103025794, 5.928278900682926, 5.861266050487757, 5.78969088383019, 5.933940876275301, 6.075609941035509, 5.8459690771996975, 5.807780893519521, 5.834975047037005, 5.963525967672467, 5.823779851198196, 5.875423084944487, 5.8369990438222885, 5.913279019296169, 5.83251379430294, 5.977369146421552, 6.077764090150595, 5.8792969211936, 5.876827985048294, 5.830192938446999, 5.96297811716795, 5.82878990098834, 5.8167309034615755, 5.834386916831136, 5.81317488104105, 5.797193851321936, 5.919974064454436 ], "memory": { "VmHWM_MiB": 10882.62890625, "VmRSS_MiB": 8945.2578125, "Pss_MiB": 8939.7099609375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87701.83203125, "memory.current_MiB": 14940.79296875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11520, 95, 2715, 151.51" }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" }, { "frames": 250, "p50_ms": 9.877901058644056, "p95_ms": 10.38077101111412, "samples_ms": [ 9.786374168470502, 9.802610846236348, 9.806396905332804, 9.839836973696947, 9.79762407951057, 9.850073140114546, 9.788010967895389, 9.820964885875583, 10.1941111497581, 10.076764971017838, 9.805353125557303, 10.358612984418869, 10.156973032280803, 9.868319844827056, 9.863507002592087, 9.89880901761353, 9.961461881175637, 9.861113037914038, 9.941444033756852, 9.854090865701437, 9.786400943994522, 9.852737188339233, 9.978126967325807, 9.80105297639966, 10.21820679306984, 10.114829055964947, 9.844722924754024, 10.148921050131321, 10.122086154296994, 9.878817945718765, 9.833561023697257, 9.84923099167645, 9.8028180655092, 9.98988514766097, 9.854567935690284, 9.939749026671052, 10.100500890985131, 10.041940957307816, 9.825702058151364, 9.819163009524345, 10.38077101111412, 10.10015606880188, 9.796522092074156, 10.44334308244288, 10.013247141614556, 9.994515916332603, 10.814161039888859, 10.094942059367895, 9.894102113321424, 9.876984171569347 ], "memory": { "VmHWM_MiB": 10882.62890625, "VmRSS_MiB": 8945.2734375, "Pss_MiB": 8939.7216796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87689.75390625, "memory.current_MiB": 14940.91796875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11552, 97, 2715, 211.32" }, "finite": true, "input_sha256": "726cc736d46a4b8ca3c2d2831abc8c8e3a59489168194774d072d5fc1aa502b5" } ], "load_s": 22.87757252715528, "ready": { "VmHWM_MiB": 10882.62890625, "VmRSS_MiB": 8942.63671875, "Pss_MiB": 8937.0888671875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87488.30078125, "memory.current_MiB": 14940.3046875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11484, 1, 2715, 82.75" }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 10882.62890625, "VmRSS_MiB": 8945.2734375, "Pss_MiB": 8939.7216796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87635.7734375, "memory.current_MiB": 14941.03515625, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11552, 97, 2715, 202.98" } }, "maximum-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "2000", "warmup": 1, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/maximum-cuda.json" }, "utc": "2026-09-05T11:03:24Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.5078125, "VmRSS_MiB": 44.18359375, "Pss_MiB": 43.15234375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94433.11328125, "memory.current_MiB": 7768.796875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "9569, 0, 2715, 127.14" }, "rows": [ { "frames": 2000, "p50_ms": 82.95771200209856, "p95_ms": 120.07294804789126, "samples_ms": [ 82.95771200209856, 120.07294804789126, 79.37704701907933 ], "memory": { "VmHWM_MiB": 1689.70703125, "VmRSS_MiB": 1689.70703125, "Pss_MiB": 1684.1533203125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92742.65625, "memory.current_MiB": 9212.69921875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "12032, 0, 2715, 127.95" }, "finite": true, "input_sha256": "8dd6299031bec40b45a4061c4670062fb06238b5da5111ea3e49d85335d81d47" } ], "load_s": 1.2869792010169476, "ready": { "VmHWM_MiB": 1323.80078125, "VmRSS_MiB": 1252.19921875, "Pss_MiB": 1246.6767578125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93261.9453125, "memory.current_MiB": 8821.99609375, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "10998, 0, 2715, 79.90" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 1689.77734375, "VmRSS_MiB": 1689.77734375, "Pss_MiB": 1684.2236328125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92753.98828125, "memory.current_MiB": 9212.66796875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "12032, 0, 2715, 136.85" } }, "proof-offline": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": null, "cache": "offline-context", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/4090/proof-offline.json" }, "utc": "2026-09-05T11:20:14Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "06e5dd3d0564acef6b62a38ac08f597c0b8676949e38fea8ab9fd995d57f6457", "start": { "VmHWM_MiB": 43.74609375, "VmRSS_MiB": 43.74609375, "Pss_MiB": 42.775390625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93735.44140625, "memory.current_MiB": 9749.296875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 1, 2535, 72.91" }, "rows": [ { "frames": 50, "p50_ms": 38.67384698241949, "p95_ms": 38.67384698241949, "samples_ms": [ 38.67384698241949 ], "memory": { "VmHWM_MiB": 1559.76953125, "VmRSS_MiB": 942.125, "Pss_MiB": 936.673828125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93416.75, "memory.current_MiB": 10289.25, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11100, 1, 2535, 67.98" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 1.5385181810706854, "ready": { "VmHWM_MiB": 1559.76953125, "VmRSS_MiB": 922.7578125, "Pss_MiB": 917.318359375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93453.50390625, "memory.current_MiB": 10272.04296875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11080, 1, 2535, 67.82" }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "TensorrtExecutionProvider": 1 }, "end": { "VmHWM_MiB": 1559.76953125, "VmRSS_MiB": 943.49609375, "Pss_MiB": 938.044921875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93400.5078125, "memory.current_MiB": 10290.79296875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11100, 1, 2535, 67.98" } }, "graph-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 50, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/graph-cuda.json" }, "utc": "2026-09-05T11:03:20Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.1484375, "VmRSS_MiB": 44.19140625, "Pss_MiB": 43.15625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94323.875, "memory.current_MiB": 7768.95703125, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "9569, 0, 210, 25.48" }, "rows": [ { "frames": 50, "p50_ms": 8.099773898720741, "p95_ms": 8.473813068121672, "samples_ms": [ 8.571076905354857, 8.468265878036618, 8.431632071733475, 8.433279115706682, 8.469401160255075, 8.414068957790732, 8.434494957327843, 8.072796044871211, 8.017083862796426, 7.971742190420628, 8.10020393691957, 8.074975106865168, 8.087877882644534, 8.077963953837752, 8.180863922461867, 8.143690880388021, 8.044074987992644, 7.963568903505802, 8.212907006964087, 8.115028031170368, 8.023969130590558, 7.977916160598397, 8.023849921301007, 7.998517947271466, 7.950983010232449, 7.953159045428038, 8.099343860521913, 8.055791025981307, 7.969917031005025, 7.957899011671543, 8.052777964621782, 8.053057827055454, 7.954038912430406, 7.906545186415315, 8.149323053658009, 7.981724804267287, 8.678117068484426, 8.473813068121672, 8.472369983792305, 8.215585025027394, 8.084323955699801, 8.02172115072608, 8.227000012993813, 8.209341205656528, 8.333527948707342, 8.161817910149693, 8.117231074720621, 8.149266941472888, 8.160552941262722, 8.365872083231807 ], "memory": { "VmHWM_MiB": 1343.2578125, "VmRSS_MiB": 1343.2578125, "Pss_MiB": 1337.6884765625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93203.09765625, "memory.current_MiB": 8861.44921875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 96, 2715, 93.09" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 8.199101081117988, "p95_ms": 9.006958222016692, "samples_ms": [ 8.73490795493126, 11.447304161265492, 8.425913052633405, 8.226433070376515, 8.43983911909163, 8.175900904461741, 8.24852311052382, 8.152117021381855, 8.297699969261885, 9.670398198068142, 8.672333089634776, 9.006958222016692, 8.467909879982471, 8.367082104086876, 8.455845061689615, 8.693926967680454, 8.366711903363466, 8.245080942288041, 8.347121998667717, 8.374579017981887, 8.193623973056674, 8.151482092216611, 8.137778844684362, 8.09103506617248, 8.088441099971533, 8.040953893214464, 8.076958823949099, 8.19609989412129, 8.205363992601633, 8.11130995862186, 7.991138147190213, 8.04407405667007, 8.137430995702744, 8.076580939814448, 8.173601003363729, 8.199485018849373, 8.144114864990115, 8.094658143818378, 8.46433499827981, 8.138969074934721, 8.129321038722992, 8.126966189593077, 8.752302965149283, 8.35670018568635, 8.198717143386602, 8.099723840132356, 8.40347702614963, 8.209537016227841, 8.122283034026623, 8.127495180815458 ], "memory": { "VmHWM_MiB": 1353.58203125, "VmRSS_MiB": 1353.58203125, "Pss_MiB": 1348.0087890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93360.67578125, "memory.current_MiB": 8871.296875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 96, 2715, 144.43" }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" }, { "frames": 250, "p50_ms": 10.53515262901783, "p95_ms": 11.255748802796006, "samples_ms": [ 10.45744912698865, 14.259216142818332, 10.458218166604638, 10.369115974754095, 10.323842987418175, 10.644539026543498, 10.557692032307386, 12.684379937127233, 10.44950494542718, 10.352127952501178, 10.490768821910024, 10.530903935432434, 11.00951200351119, 11.255748802796006, 10.69182506762445, 10.572172002866864, 10.667134774848819, 10.59655612334609, 10.59263595379889, 10.504112113267183, 10.626198956742883, 10.77653095126152, 10.571060003712773, 10.585281997919083, 10.53838711231947, 10.55507198907435, 10.578634915873408, 10.468026855960488, 10.473251808434725, 10.623123962432146, 10.567819932475686, 10.564719093963504, 10.574449086561799, 10.53191814571619, 10.51663001999259, 10.576133849099278, 10.409882059320807, 10.449822060763836, 10.47564297914505, 10.468292981386185, 10.378600098192692, 10.375062003731728, 10.30937978066504, 10.478492127731442, 10.517844930291176, 10.424173902720213, 10.948535986244678, 10.553158819675446, 10.45288797467947, 10.414808057248592 ], "memory": { "VmHWM_MiB": 1392.20703125, "VmRSS_MiB": 1392.20703125, "Pss_MiB": 1386.6337890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93323.3828125, "memory.current_MiB": 8910.078125, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 94, 2715, 187.86" }, "finite": true, "input_sha256": "726cc736d46a4b8ca3c2d2831abc8c8e3a59489168194774d072d5fc1aa502b5" } ], "load_s": 1.23971680784598, "ready": { "VmHWM_MiB": 1331.015625, "VmRSS_MiB": 1267.3046875, "Pss_MiB": 1261.7822265625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92958.16796875, "memory.current_MiB": 8835.390625, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 43.74" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 1392.2109375, "VmRSS_MiB": 1392.2109375, "Pss_MiB": 1386.6376953125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93323.3359375, "memory.current_MiB": 8910.01171875, "memory.peak_MiB": 9348.1015625, "gpu_total_memory_util_clock_power": "11008, 94, 2715, 187.86" } }, "threads-0": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100", "warmup": 10, "repeats": 50, "threads": 0, "profile": false, "dump_context": null, "out": "/bench/4090/threads-0.json" }, "utc": "2026-09-05T11:05:49Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.015625, "VmRSS_MiB": 43.8203125, "Pss_MiB": 42.83203125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94291.515625, "memory.current_MiB": 8164.55859375, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "9569, 0, 210, 24.10" }, "rows": [ { "frames": 50, "p50_ms": 8.060825522989035, "p95_ms": 8.63280682824552, "samples_ms": [ 8.616873063147068, 8.509655948728323, 8.637282997369766, 8.63280682824552, 8.51972191594541, 8.598410058766603, 8.602818008512259, 8.694210089743137, 8.278383174911141, 8.180059026926756, 8.322093868628144, 8.342624176293612, 8.099260041490197, 8.047274081036448, 8.31997999921441, 8.129586931318045, 8.050275966525078, 8.139221929013729, 8.107965113595128, 8.031220873817801, 7.962523028254509, 8.072791853919625, 8.109858958050609, 7.985012838616967, 7.9500009305775166, 8.048890857025981, 8.072787895798683, 8.005269104614854, 7.897312985733151, 8.084134897217155, 8.069693110883236, 7.945452118292451, 7.856121053919196, 8.034829050302505, 8.01987899467349, 7.963546086102724, 7.878015050664544, 8.064765948802233, 8.011135971173644, 7.917403010651469, 7.855510106310248, 8.056885097175837, 8.027825970202684, 7.933846907690167, 7.89362913928926, 8.064870024099946, 8.046431932598352, 8.014529943466187, 8.031915174797177, 8.313344791531563 ], "memory": { "VmHWM_MiB": 1329.66796875, "VmRSS_MiB": 1314.609375, "Pss_MiB": 1309.0439453125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93586.203125, "memory.current_MiB": 9232.78125, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 49, 2535, 91.18" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 8.283938630484045, "p95_ms": 8.799452800303698, "samples_ms": [ 8.211379870772362, 10.530577972531319, 8.298075059428811, 8.467761101201177, 8.151942864060402, 8.124914951622486, 8.007480064406991, 8.349568117409945, 8.192567154765129, 8.248487953096628, 8.147320011630654, 8.401099825277925, 8.272247156128287, 8.224858902394772, 8.074200944975019, 8.309891913086176, 8.370294002816081, 8.246963145211339, 8.191049098968506, 8.384463144466281, 8.383968845009804, 8.340099127963185, 8.154294220730662, 8.398585021495819, 8.303694892674685, 8.19485797546804, 8.355834055691957, 8.21699295192957, 8.239621063694358, 8.15508607774973, 8.302135160192847, 8.13956605270505, 8.266152115538716, 8.13264399766922, 8.310295874252915, 8.327320916578174, 8.240215945988894, 8.161770878359675, 8.343128021806479, 8.37245606817305, 8.20759404450655, 8.114352123811841, 10.054668877273798, 8.799452800303698, 8.656781166791916, 8.772105909883976, 8.724011946469545, 8.295630104839802, 8.231825893744826, 8.707279106602073 ], "memory": { "VmHWM_MiB": 1329.66796875, "VmRSS_MiB": 1324.9375, "Pss_MiB": 1319.3720703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93579.0859375, "memory.current_MiB": 9242.8046875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 49, 2535, 141.11" }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" } ], "load_s": 1.2397303429897875, "ready": { "VmHWM_MiB": 1329.66796875, "VmRSS_MiB": 1258.296875, "Pss_MiB": 1252.8056640625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93184.26953125, "memory.current_MiB": 9224.86328125, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 44.99" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 1329.66796875, "VmRSS_MiB": 1324.9375, "Pss_MiB": 1319.3720703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93556.94140625, "memory.current_MiB": 9242.859375, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 49, 2535, 141.11" } }, "threads-4": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100", "warmup": 10, "repeats": 50, "threads": 4, "profile": false, "dump_context": null, "out": "/bench/4090/threads-4.json" }, "utc": "2026-09-05T11:05:52Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.171875, "VmRSS_MiB": 43.734375, "Pss_MiB": 42.740234375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94191.37890625, "memory.current_MiB": 8165.82421875, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "9569, 77, 2715, 109.46" }, "rows": [ { "frames": 50, "p50_ms": 8.246775949373841, "p95_ms": 8.668638998642564, "samples_ms": [ 8.635737001895905, 8.560436079278588, 8.642892818897963, 8.715770905837417, 8.668638998642564, 8.622227003797889, 8.642559871077538, 8.606433868408203, 11.148469988256693, 8.601360954344273, 8.623968809843063, 8.52910103276372, 8.321640081703663, 8.194496855139732, 8.221652125939727, 8.227096870541573, 8.187278872355819, 8.312889840453863, 8.393495809286833, 8.253688924014568, 8.087381953373551, 8.064740104600787, 8.358324179425836, 8.416353026404977, 8.238949114456773, 8.248284924775362, 8.36954009719193, 8.279758971184492, 8.28473805449903, 8.423299062997103, 8.24526697397232, 8.080721832811832, 8.163159945979714, 8.214089088141918, 8.120379177853465, 8.039525942876935, 8.250419981777668, 8.177320007234812, 8.131648879498243, 8.033443940803409, 8.249416016042233, 8.1403199583292, 8.010149002075195, 7.942738011479378, 8.114669006317854, 8.062155917286873, 7.97898811288178, 7.915307069197297, 8.14188807271421, 8.18430888466537 ], "memory": { "VmHWM_MiB": 1328.30859375, "VmRSS_MiB": 1320.328125, "Pss_MiB": 1314.796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93404.828125, "memory.current_MiB": 9238.125, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 59, 2715, 98.12" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 8.165534469299018, "p95_ms": 8.510251063853502, "samples_ms": [ 8.187636034563184, 10.543305892497301, 8.131871931254864, 8.105963934212923, 8.095040917396545, 8.076647063717246, 8.001997834071517, 8.018811931833625, 8.129468187689781, 8.049007039517164, 8.236924884840846, 8.136442862451077, 8.061606902629137, 8.033592021092772, 8.508000988513231, 8.076460100710392, 8.111421018838882, 8.027039933949709, 8.413651958107948, 8.076969999819994, 8.00425885245204, 8.0515849404037, 8.470979053527117, 8.415475953370333, 8.327368879690766, 8.151690009981394, 8.510251063853502, 8.339386200532317, 8.20996006950736, 8.182507939636707, 8.602639194577932, 8.402524050325155, 8.2543829921633, 8.21007788181305, 8.263461058959365, 8.23655305430293, 8.220601128414273, 8.204706013202667, 8.23265896178782, 8.160844910889864, 8.0806789919734, 8.305852068588138, 8.170224027708173, 8.108583046123385, 8.089655078947544, 8.18487792275846, 8.156499126926064, 8.143543032929301, 8.06212704628706, 8.323675952851772 ], "memory": { "VmHWM_MiB": 1330.63671875, "VmRSS_MiB": 1330.63671875, "Pss_MiB": 1325.1015625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93396.14453125, "memory.current_MiB": 9248.14453125, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 94, 2715, 135.00" }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" } ], "load_s": 1.214797917054966, "ready": { "VmHWM_MiB": 1328.30859375, "VmRSS_MiB": 1264.0703125, "Pss_MiB": 1258.5859375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93140.01953125, "memory.current_MiB": 9231.390625, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "10998, 0, 2715, 78.11" }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 1330.640625, "VmRSS_MiB": 1330.640625, "Pss_MiB": 1325.10546875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93380.39453125, "memory.current_MiB": 9248.22265625, "memory.peak_MiB": 16830.52734375, "gpu_total_memory_util_clock_power": "11008, 94, 2715, 141.01" } }, "recognizer-cuda-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-cuda-6s.json" }, "utc": "2026-09-05T11:21:24Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.25, "VmRSS_MiB": 44.09765625, "Pss_MiB": 42.8984375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94397.5390625, "memory.current_MiB": 10056.3359375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 0, 2535, 66.61" }, "rows": [ { "seconds": 6, "p50_ms": 16.98519312776625, "p95_ms": 24.588905973359942, "samples_ms": [ 24.588905973359942, 16.98519312776625, 16.747202957049012 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 1354.48828125, "VmRSS_MiB": 1320.7734375, "Pss_MiB": 1315.0166015625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92923.87109375, "memory.current_MiB": 11119.6015625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 64.75" }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 1.272912859916687, "ready": { "VmHWM_MiB": 1354.48828125, "VmRSS_MiB": 1266.26171875, "Pss_MiB": 1260.572265625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92955.58984375, "memory.current_MiB": 11113.2578125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 63.86" }, "after_warmup": { "VmHWM_MiB": 1354.48828125, "VmRSS_MiB": 1320.6484375, "Pss_MiB": 1314.8955078125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92939.62109375, "memory.current_MiB": 11119.7265625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 64.16" }, "end": { "VmHWM_MiB": 1354.48828125, "VmRSS_MiB": 1320.77734375, "Pss_MiB": 1315.0205078125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92916.2421875, "memory.current_MiB": 11119.703125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 66.75" } }, "recognizer-cuda-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-cuda-15s.json" }, "utc": "2026-09-05T11:21:27Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.44921875, "VmRSS_MiB": 43.86328125, "Pss_MiB": 42.71875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93878.38671875, "memory.current_MiB": 10056.34375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 0, 2535, 67.73" }, "rows": [ { "seconds": 15, "p50_ms": 29.76058516651392, "p95_ms": 39.752352982759476, "samples_ms": [ 39.752352982759476, 29.76058516651392, 29.660990927368402 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 1351.88671875, "VmRSS_MiB": 1333.3984375, "Pss_MiB": 1327.712890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92808.82421875, "memory.current_MiB": 11133.1171875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 68.19" }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 1.3007597599644214, "ready": { "VmHWM_MiB": 1351.88671875, "VmRSS_MiB": 1279.01171875, "Pss_MiB": 1273.376953125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92759.3671875, "memory.current_MiB": 11126.4375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "10998, 0, 2535, 64.06" }, "after_warmup": { "VmHWM_MiB": 1351.88671875, "VmRSS_MiB": 1333.27734375, "Pss_MiB": 1327.595703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92826.9921875, "memory.current_MiB": 11133.11328125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 64.44" }, "end": { "VmHWM_MiB": 1351.88671875, "VmRSS_MiB": 1333.40234375, "Pss_MiB": 1327.716796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 92809.0703125, "memory.current_MiB": 11133.12109375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11008, 0, 2535, 68.19" } }, "recognizer-trt-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-trt-15s.json" }, "utc": "2026-09-05T11:21:44Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.328125, "VmRSS_MiB": 43.94140625, "Pss_MiB": 42.7890625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93946.0546875, "memory.current_MiB": 10056.3984375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 7, 2535, 64.04" }, "rows": [ { "seconds": 15, "p50_ms": 29.02281889691949, "p95_ms": 31.039170920848846, "samples_ms": [ 31.039170920848846, 29.02281889691949, 28.13853695988655 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 9417.7265625, "VmRSS_MiB": 7545.31640625, "Pss_MiB": 7539.646484375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88223.6171875, "memory.current_MiB": 15480.59765625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11128, 31, 2535, 68.67" }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 6.13117758394219, "ready": { "VmHWM_MiB": 9417.7265625, "VmRSS_MiB": 7535.171875, "Pss_MiB": 7529.505859375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88331.07421875, "memory.current_MiB": 15478.43359375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11076, 1, 2535, 65.04" }, "after_warmup": { "VmHWM_MiB": 9417.7265625, "VmRSS_MiB": 7545.3046875, "Pss_MiB": 7539.638671875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88313.91796875, "memory.current_MiB": 15480.9375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11096, 1, 2535, 65.15" }, "end": { "VmHWM_MiB": 9417.7265625, "VmRSS_MiB": 7545.3203125, "Pss_MiB": 7539.650390625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88200.484375, "memory.current_MiB": 15480.64453125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11128, 31, 2535, 68.59" } }, "offline-build": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "offline-context", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": false, "dump_context": "/bench/model_ctx.onnx", "out": "/bench/4090/offline-build.json" }, "utc": "2026-09-05T11:19:43Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.08984375, "VmRSS_MiB": 43.74609375, "Pss_MiB": 42.734375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93754.3203125, "memory.current_MiB": 10157.84375, "memory.peak_MiB": 17912.359375, "gpu_total_memory_util_clock_power": "9569, 0, 210, 26.13" }, "rows": [ { "frames": 50, "p50_ms": 34.83066102489829, "p95_ms": 34.83066102489829, "samples_ms": [ 34.83066102489829 ], "memory": { "VmHWM_MiB": 10886.87109375, "VmRSS_MiB": 8948.24609375, "Pss_MiB": 8942.6806640625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87830.08203125, "memory.current_MiB": 16539.05859375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11504, 1, 2715, 83.60" }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 22.776474329875782, "ready": { "VmHWM_MiB": 10886.87109375, "VmRSS_MiB": 8945.8125, "Pss_MiB": 8940.2548828125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87846.54296875, "memory.current_MiB": 16538.44140625, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11484, 1, 2715, 83.02" }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 10886.87109375, "VmRSS_MiB": 8948.3125, "Pss_MiB": 8942.7470703125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 87823.44140625, "memory.current_MiB": 16539.3359375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11504, 6, 2715, 83.54" } }, "recognizer-offline-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-offline-15s.json" }, "utc": "2026-09-05T11:21:57Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "06e5dd3d0564acef6b62a38ac08f597c0b8676949e38fea8ab9fd995d57f6457", "start": { "VmHWM_MiB": 44.06640625, "VmRSS_MiB": 44.06640625, "Pss_MiB": 43.0478515625, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 94586.07421875, "memory.current_MiB": 10056.3046875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 1, 2535, 67.36" }, "rows": [ { "seconds": 15, "p50_ms": 28.181665111333132, "p95_ms": 35.02555494196713, "samples_ms": [ 35.02555494196713, 28.181665111333132, 27.927712071686983 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 1576.4921875, "VmRSS_MiB": 991.875, "Pss_MiB": 986.37109375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93403.2265625, "memory.current_MiB": 10637.8203125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11122, 1, 2535, 67.96" }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 1.690665485104546, "ready": { "VmHWM_MiB": 1576.4921875, "VmRSS_MiB": 928.41796875, "Pss_MiB": 922.91796875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93501.1328125, "memory.current_MiB": 10581.95703125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11070, 1, 2535, 65.48" }, "after_warmup": { "VmHWM_MiB": 1576.4921875, "VmRSS_MiB": 961.078125, "Pss_MiB": 955.578125, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93453.67578125, "memory.current_MiB": 10606.953125, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11090, 1, 2535, 66.15" }, "end": { "VmHWM_MiB": 1576.4921875, "VmRSS_MiB": 991.87890625, "Pss_MiB": 986.375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93386.24609375, "memory.current_MiB": 10637.88671875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11122, 1, 2535, 67.96" } }, "recognizer-trt-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/4090/recognizer-trt-6s.json" }, "utc": "2026-09-05T11:21:37Z", "machine": "x86_64", "ort": "1.23.2", "numpy": "2.2.6", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.45703125, "VmRSS_MiB": 43.86328125, "Pss_MiB": 42.7333984375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 93923.6015625, "memory.current_MiB": 10056.57421875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "9569, 0, 2535, 64.25" }, "rows": [ { "seconds": 6, "p50_ms": 16.086635179817677, "p95_ms": 16.83194888755679, "samples_ms": [ 16.83194888755679, 15.954064903780818, 16.086635179817677 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 9418.6015625, "VmRSS_MiB": 7545.06640625, "Pss_MiB": 7539.4208984375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88670.65625, "memory.current_MiB": 15480.7421875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11112, 1, 2535, 66.99" }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 6.014078652951866, "ready": { "VmHWM_MiB": 9418.6015625, "VmRSS_MiB": 7535.14453125, "Pss_MiB": 7529.5029296875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88712.3828125, "memory.current_MiB": 15478.546875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11076, 1, 2535, 64.53" }, "after_warmup": { "VmHWM_MiB": 9418.6015625, "VmRSS_MiB": 7545.0546875, "Pss_MiB": 7539.4130859375, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88686.8984375, "memory.current_MiB": 15480.57421875, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11096, 1, 2535, 65.77" }, "end": { "VmHWM_MiB": 9418.6015625, "VmRSS_MiB": 7545.0703125, "Pss_MiB": 7539.4248046875, "MemTotal_MiB": 128413.15234375, "MemAvailable_MiB": 88652.1953125, "memory.current_MiB": 15480.74609375, "memory.peak_MiB": 18432.0, "gpu_total_memory_util_clock_power": "11112, 1, 2535, 66.91" } }, "service": { "service-cuda-1": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:10:00", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 22.37, "server_rtf": 0.01, "server_ttft_ms": 25.11, "client_final_latency_ms": 32.51, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.02, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 30.45, "server_rtf": 0.008, "server_ttft_ms": 43.57, "client_final_latency_ms": 39.66, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 47.44, "server_rtf": 0.003, "server_ttft_ms": 20.69, "client_final_latency_ms": 53.74, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 22.4, "server_rtf": 0.053, "server_ttft_ms": 43.67, "client_final_latency_ms": 25.95, "interim_count": 6, "max_interim_gap_s": 0.526, "feed_secs": 3.17, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 21.0, "server_rtf": 0.048, "server_ttft_ms": 19.88, "client_final_latency_ms": 23.61, "interim_count": 11, "max_interim_gap_s": 1.012, "feed_secs": 6.28, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 39.02, "server_rtf": 0.056, "server_ttft_ms": 19.37, "client_final_latency_ms": 186.43, "interim_count": 28, "max_interim_gap_s": 1.032, "feed_secs": 15.79, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-cuda-2": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:11:09", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 21.81, "server_rtf": 0.009, "server_ttft_ms": 30.37, "client_final_latency_ms": 29.71, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.03, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 27.55, "server_rtf": 0.006, "server_ttft_ms": 31.98, "client_final_latency_ms": 35.56, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 38.79, "server_rtf": 0.003, "server_ttft_ms": 21.29, "client_final_latency_ms": 44.55, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 31.42, "server_rtf": 0.052, "server_ttft_ms": 19.42, "client_final_latency_ms": 34.06, "interim_count": 7, "max_interim_gap_s": 0.522, "feed_secs": 3.16, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 39.39, "server_rtf": 0.048, "server_ttft_ms": 20.11, "client_final_latency_ms": 41.61, "interim_count": 11, "max_interim_gap_s": 1.019, "feed_secs": 6.31, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 67.94, "server_rtf": 0.062, "server_ttft_ms": 20.06, "client_final_latency_ms": 489.38, "interim_count": 27, "max_interim_gap_s": 1.023, "feed_secs": 15.81, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-cuda-3": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:12:16", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 22.36, "server_rtf": 0.011, "server_ttft_ms": 29.78, "client_final_latency_ms": 28.11, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.03, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 27.08, "server_rtf": 0.006, "server_ttft_ms": 34.76, "client_final_latency_ms": 32.67, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 39.06, "server_rtf": 0.003, "server_ttft_ms": 21.46, "client_final_latency_ms": 45.51, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 30.05, "server_rtf": 0.051, "server_ttft_ms": 44.62, "client_final_latency_ms": 33.27, "interim_count": 7, "max_interim_gap_s": 0.521, "feed_secs": 3.15, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 24.36, "server_rtf": 0.047, "server_ttft_ms": 19.43, "client_final_latency_ms": 25.94, "interim_count": 11, "max_interim_gap_s": 1.02, "feed_secs": 6.3, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 28.01, "server_rtf": 0.059, "server_ttft_ms": 373.74, "client_final_latency_ms": 30.5, "interim_count": 27, "max_interim_gap_s": 1.297, "feed_secs": 15.77, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null } ] }, "service-trt-1": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:14:06", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 15.27, "server_rtf": 0.005, "server_ttft_ms": 11.68, "client_final_latency_ms": 21.37, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.03, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 20.79, "server_rtf": 0.003, "server_ttft_ms": 11.25, "client_final_latency_ms": 31.95, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.04, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 40.01, "server_rtf": 0.002, "server_ttft_ms": 11.05, "client_final_latency_ms": 45.92, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 10.56, "server_rtf": 0.021, "server_ttft_ms": 16.86, "client_final_latency_ms": 14.62, "interim_count": 7, "max_interim_gap_s": 0.518, "feed_secs": 3.17, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 23.31, "server_rtf": 0.028, "server_ttft_ms": 26.2, "client_final_latency_ms": 39.72, "interim_count": 11, "max_interim_gap_s": 1.052, "feed_secs": 6.31, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 33.03, "server_rtf": 0.034, "server_ttft_ms": 16.26, "client_final_latency_ms": 42.04, "interim_count": 29, "max_interim_gap_s": 1.039, "feed_secs": 15.76, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-trt-2": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:15:13", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 14.35, "server_rtf": 0.005, "server_ttft_ms": 20.49, "client_final_latency_ms": 36.4, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.05, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 16.34, "server_rtf": 0.003, "server_ttft_ms": 14.12, "client_final_latency_ms": 31.77, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.05, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 34.19, "server_rtf": 0.002, "server_ttft_ms": 11.0, "client_final_latency_ms": 39.9, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.11, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 9.35, "server_rtf": 0.018, "server_ttft_ms": 11.62, "client_final_latency_ms": 22.24, "interim_count": 7, "max_interim_gap_s": 0.606, "feed_secs": 3.16, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 27.79, "server_rtf": 0.031, "server_ttft_ms": 10.48, "client_final_latency_ms": 42.74, "interim_count": 11, "max_interim_gap_s": 1.138, "feed_secs": 6.32, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 26.07, "server_rtf": 0.037, "server_ttft_ms": 11.2, "client_final_latency_ms": 31.16, "interim_count": 28, "max_interim_gap_s": 1.028, "feed_secs": 15.82, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-trt-3": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:16:21", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 15.18, "server_rtf": 0.007, "server_ttft_ms": 20.2, "client_final_latency_ms": 36.6, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.02, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 17.02, "server_rtf": 0.003, "server_ttft_ms": 23.09, "client_final_latency_ms": 33.98, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 35.61, "server_rtf": 0.002, "server_ttft_ms": 21.94, "client_final_latency_ms": 47.25, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.18, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 23.98, "server_rtf": 0.029, "server_ttft_ms": 36.3, "client_final_latency_ms": 36.36, "interim_count": 7, "max_interim_gap_s": 0.537, "feed_secs": 3.16, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 10.85, "server_rtf": 0.025, "server_ttft_ms": 11.81, "client_final_latency_ms": 23.77, "interim_count": 11, "max_interim_gap_s": 1.027, "feed_secs": 6.31, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 22.81, "server_rtf": 0.032, "server_ttft_ms": 9.97, "client_final_latency_ms": 38.15, "interim_count": 28, "max_interim_gap_s": 1.02, "feed_secs": 15.77, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] } } }, "orin": { "recognizer-trt-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-trt-3s.json" }, "utc": "2026-09-05T11:24:58Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.2890625, "VmRSS_MiB": 43.640625, "Pss_MiB": 36.7802734375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17326.34375, "memory.current_MiB": 5453.1015625 }, "rows": [ { "seconds": 3, "p50_ms": 28.338345233350992, "p95_ms": 28.608966153115034, "samples_ms": [ 28.608966153115034, 28.338345233350992, 28.333416674286127 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 8587.9921875, "VmRSS_MiB": 6970.5703125, "Pss_MiB": 5547.9658203125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12136.109375, "memory.current_MiB": 10267.87890625 }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 44.892178926151246, "ready": { "VmHWM_MiB": 8587.9921875, "VmRSS_MiB": 6939.90625, "Pss_MiB": 5537.7138671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12151.265625, "memory.current_MiB": 10236.13671875 }, "after_warmup": { "VmHWM_MiB": 8587.9921875, "VmRSS_MiB": 6970.5703125, "Pss_MiB": 5547.9658203125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12135.94921875, "memory.current_MiB": 10267.87890625 }, "end": { "VmHWM_MiB": 8587.9921875, "VmRSS_MiB": 6970.5703125, "Pss_MiB": 5547.9658203125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12136.109375, "memory.current_MiB": 10267.87890625 } }, "recognizer-cuda-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-cuda-3s.json" }, "utc": "2026-09-05T11:24:44Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.29296875, "VmRSS_MiB": 43.64453125, "Pss_MiB": 36.892578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17324.9609375, "memory.current_MiB": 5440.95703125 }, "rows": [ { "seconds": 3, "p50_ms": 44.9584829621017, "p95_ms": 45.153488870710135, "samples_ms": [ 44.8529152199626, 45.153488870710135, 44.9584829621017 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 2347.0, "VmRSS_MiB": 2347.0, "Pss_MiB": 1195.8037109375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16258.6796875, "memory.current_MiB": 6507.92578125 }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 2.9862349182367325, "ready": { "VmHWM_MiB": 2308.48828125, "VmRSS_MiB": 2308.48828125, "Pss_MiB": 1165.83984375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16262.734375, "memory.current_MiB": 6501.0859375 }, "after_warmup": { "VmHWM_MiB": 2347.00390625, "VmRSS_MiB": 2347.00390625, "Pss_MiB": 1195.7978515625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16258.6796875, "memory.current_MiB": 6508.03515625 }, "end": { "VmHWM_MiB": 2347.0, "VmRSS_MiB": 2347.0, "Pss_MiB": 1195.8037109375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16258.6796875, "memory.current_MiB": 6507.92578125 } }, "proof-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/orin/proof-cuda.json" }, "utc": "2026-09-05T11:03:43Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.1171875, "VmRSS_MiB": 43.46875, "Pss_MiB": 34.9326171875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16287.19921875, "memory.current_MiB": 3328.85546875 }, "rows": [ { "frames": 50, "p50_ms": 270.45236388221383, "p95_ms": 270.45236388221383, "samples_ms": [ 270.45236388221383 ], "memory": { "VmHWM_MiB": 2299.55859375, "VmRSS_MiB": 2265.78515625, "Pss_MiB": 1106.03125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15289.359375, "memory.current_MiB": 4312.45703125 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 2.8636796781793237, "ready": { "VmHWM_MiB": 2299.55859375, "VmRSS_MiB": 2218.921875, "Pss_MiB": 1070.2412109375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15293.0703125, "memory.current_MiB": 4307.95703125 }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "CUDAExecutionProvider": 2908, "CPUExecutionProvider": 785 }, "end": { "VmHWM_MiB": 2299.55859375, "VmRSS_MiB": 2279.7421875, "Pss_MiB": 1119.794921875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15275.33203125, "memory.current_MiB": 4328.71875 } }, "maximum-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "2000", "warmup": 1, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/maximum-trt.json" }, "utc": "2026-09-05T11:04:27Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 43.734375, "VmRSS_MiB": 42.0859375, "Pss_MiB": 33.91015625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16811.77734375, "memory.current_MiB": 2914.796875 }, "rows": [ { "frames": 2000, "p50_ms": 531.4703239127994, "p95_ms": 532.2657362557948, "samples_ms": [ 531.4360843040049, 532.2657362557948, 531.4703239127994 ], "memory": { "VmHWM_MiB": 8581.86328125, "VmRSS_MiB": 7226.0546875, "Pss_MiB": 5524.650390625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11349.07421875, "memory.current_MiB": 8476.32421875 }, "finite": true, "input_sha256": "8dd6299031bec40b45a4061c4670062fb06238b5da5111ea3e49d85335d81d47" } ], "load_s": 16.342918436042964, "ready": { "VmHWM_MiB": 8581.86328125, "VmRSS_MiB": 6948.8203125, "Pss_MiB": 5523.09765625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11625.25390625, "memory.current_MiB": 8200.32421875 }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 8581.86328125, "VmRSS_MiB": 7226.0546875, "Pss_MiB": 5524.708984375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11349.07421875, "memory.current_MiB": 8476.32421875 } }, "proof-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/orin/proof-trt.json" }, "utc": "2026-09-05T11:04:49Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 43.0625, "VmRSS_MiB": 41.4140625, "Pss_MiB": 33.5, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16585.44921875, "memory.current_MiB": 3272.60546875 }, "rows": [ { "frames": 50, "p50_ms": 43.360243551433086, "p95_ms": 43.360243551433086, "samples_ms": [ 43.360243551433086 ], "memory": { "VmHWM_MiB": 8580.4921875, "VmRSS_MiB": 6968.55078125, "Pss_MiB": 5523.158203125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11813.46875, "memory.current_MiB": 8244.546875 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 16.460460057016462, "ready": { "VmHWM_MiB": 8580.4921875, "VmRSS_MiB": 6947.28515625, "Pss_MiB": 5521.443359375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11877.8125, "memory.current_MiB": 8224.421875 }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "TensorrtExecutionProvider": 1 }, "end": { "VmHWM_MiB": 8580.4921875, "VmRSS_MiB": 6968.55078125, "Pss_MiB": 5523.275390625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11813.5390625, "memory.current_MiB": 8244.671875 } }, "environment": { "platform": "Linux-5.15.148-tegra-aarch64-with-glibc2.35", "python": "3.10.12", "packages": { "onnxruntime-gpu": "1.24.0", "numpy": "1.26.4", "sherpa-onnx": "1.13.0+cuda", "torch": "2.8.0", "kitt-core": "0.3.7", "pipecat-ai": "0.0.108" }, "sha256": { "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt": "f449eb28dc567533d7fa59be34e2abca8784f771850c78a47fb731a31429a1dc", "/app/src/asr/batch_asr_manager.py": "fb9a00f6cdbc5bd972d3bb82c85b44a70277a593d1694460071bac124ef25815", "/app/src/kitt_stt_service.py": "5283feaee7a213feb529afe1d98308c7fec447ccd18aa9f737a7632ea08b83bc" } }, "recognizer-offline-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-offline-6s.json" }, "utc": "2026-09-05T11:26:42Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "d9cbddcb2ab713058520f37c4a914a9bb04dc013ce11721cc3193ec5a030531f", "start": { "VmHWM_MiB": 41.94921875, "VmRSS_MiB": 41.93359375, "Pss_MiB": 35.123046875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17358.30078125, "memory.current_MiB": 5076.12109375 }, "rows": [ { "seconds": 6, "p50_ms": 39.39258912578225, "p95_ms": 51.991411950439215, "samples_ms": [ 51.991411950439215, 39.39258912578225, 39.054222870618105 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 2266.18359375, "VmRSS_MiB": 2154.390625, "Pss_MiB": 891.08984375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16618.93359375, "memory.current_MiB": 5816.79296875 }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 8.231064447201788, "ready": { "VmHWM_MiB": 2266.18359375, "VmRSS_MiB": 2080.0390625, "Pss_MiB": 853.857421875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16646.74609375, "memory.current_MiB": 5788.78515625 }, "after_warmup": { "VmHWM_MiB": 2266.18359375, "VmRSS_MiB": 2125.72265625, "Pss_MiB": 878.591796875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16629.39453125, "memory.current_MiB": 5804.16796875 }, "end": { "VmHWM_MiB": 2266.18359375, "VmRSS_MiB": 2154.390625, "Pss_MiB": 891.09765625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16618.93359375, "memory.current_MiB": 5816.79296875 } }, "recognizer-offline-3s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-offline-3s.json" }, "utc": "2026-09-05T11:26:32Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "d9cbddcb2ab713058520f37c4a914a9bb04dc013ce11721cc3193ec5a030531f", "start": { "VmHWM_MiB": 39.95703125, "VmRSS_MiB": 39.94140625, "Pss_MiB": 33.138671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17370.9453125, "memory.current_MiB": 5011.4375 }, "rows": [ { "seconds": 3, "p50_ms": 27.856863103806973, "p95_ms": 28.213052079081535, "samples_ms": [ 28.213052079081535, 27.856863103806973, 27.84412819892168 ], "texts": [ "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。", "帮我导航到内湖科学园区，然后。" ], "memory": { "VmHWM_MiB": 2262.328125, "VmRSS_MiB": 2121.98828125, "Pss_MiB": 874.748046875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16633.87890625, "memory.current_MiB": 5800.4140625 }, "audio_sha256": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c" } ], "load_s": 8.505435324739665, "ready": { "VmHWM_MiB": 2262.328125, "VmRSS_MiB": 2075.953125, "Pss_MiB": 849.76953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16651.72265625, "memory.current_MiB": 5784.7890625 }, "after_warmup": { "VmHWM_MiB": 2262.328125, "VmRSS_MiB": 2121.36328125, "Pss_MiB": 874.1953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16634.37109375, "memory.current_MiB": 5799.7890625 }, "end": { "VmHWM_MiB": 2262.328125, "VmRSS_MiB": 2121.98828125, "Pss_MiB": 874.755859375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16633.87890625, "memory.current_MiB": 5800.4140625 } }, "graph-trt": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 50, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/graph-trt.json" }, "utc": "2026-09-05T11:04:03Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 44.9765625, "VmRSS_MiB": 43.328125, "Pss_MiB": 35.041015625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17439.1015625, "memory.current_MiB": 2731.1171875 }, "rows": [ { "frames": 50, "p50_ms": 19.691878464072943, "p95_ms": 19.752709660679102, "samples_ms": [ 19.70544597133994, 19.757382106035948, 19.752709660679102, 19.701669923961163, 19.700711127370596, 19.696613773703575, 19.690182991325855, 19.672934897243977, 19.69638979062438, 19.682278856635094, 19.696263130754232, 19.683815073221922, 19.676711410284042, 19.707717932760715, 19.695783033967018, 19.68810288235545, 19.684615079313517, 19.695366732776165, 19.69357393682003, 19.682695157825947, 19.674534909427166, 19.66071128845215, 19.680998753756285, 19.681639038026333, 19.7247420437634, 19.685926847159863, 19.702534191310406, 19.73296608775854, 19.670823123306036, 19.686630927026272, 19.69645405188203, 19.69565497711301, 19.765510223805904, 19.6837498806417, 19.68861510977149, 19.680550787597895, 19.730629865080118, 19.722693599760532, 19.671590998768806, 19.7359099984169, 19.673895090818405, 19.69530200585723, 19.698150921612978, 19.67962272465229, 19.6489030495286, 19.665798638015985, 19.683430902659893, 19.616263918578625, 19.70567088574171, 19.717350136488676 ], "memory": { "VmHWM_MiB": 8576.609375, "VmRSS_MiB": 6965.41015625, "Pss_MiB": 5519.8056640625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11627.34765625, "memory.current_MiB": 7938.140625 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 20.0820651371032, "p95_ms": 20.144895650446415, "samples_ms": [ 20.088705234229565, 20.172735676169395, 20.082144998013973, 20.07955266162753, 20.05571313202381, 20.10873705148697, 20.144895650446415, 20.087041426450014, 20.085121039301157, 20.073568914085627, 20.078560803085566, 20.12150501832366, 20.178591832518578, 20.084065850824118, 20.111871883273125, 20.12278512120247, 20.083329174667597, 20.092160906642675, 20.13609604910016, 20.115456078201532, 20.105825271457434, 20.101120695471764, 20.075744949281216, 20.077088847756386, 20.11824119836092, 20.074272993952036, 20.081825088709593, 20.081985276192427, 20.063937176018953, 20.046977791935205, 20.095009356737137, 20.068353042006493, 20.075553096830845, 20.076640881597996, 20.07884904742241, 20.048961974680424, 20.099265035241842, 20.03580890595913, 20.09206498041749, 20.05027374252677, 20.057728979736567, 20.039009861648083, 20.096545107662678, 20.058208610862494, 20.088129211217165, 20.054017193615437, 20.058241672813892, 20.04009671509266, 20.053538028150797, 20.08976135402918 ], "memory": { "VmHWM_MiB": 8576.609375, "VmRSS_MiB": 6981.76953125, "Pss_MiB": 5520.125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11611.19921875, "memory.current_MiB": 7954.23046875 }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" }, { "frames": 250, "p50_ms": 38.85491704568267, "p95_ms": 38.90283638611436, "samples_ms": [ 38.82434172555804, 38.819316774606705, 38.83115714415908, 38.85774873197079, 38.80430990830064, 38.83774904534221, 38.8385490514338, 38.868341129273176, 38.837014231830835, 38.84962107986212, 38.836917374283075, 38.88491587713361, 38.795189931988716, 38.854997139424086, 38.86859631165862, 38.86882122606039, 38.84558938443661, 38.86603657156229, 38.876563776284456, 38.840373046696186, 38.802677765488625, 38.833620958030224, 38.82638877257705, 38.7958618812263, 38.85698085650802, 38.84840523824096, 38.82789285853505, 38.90283638611436, 38.86402118951082, 38.86805288493633, 38.850133307278156, 38.85909216478467, 38.86990901082754, 38.82018104195595, 38.86363701894879, 38.90485176816583, 38.85886911302805, 38.80123794078827, 38.887765258550644, 38.85483695194125, 38.82878925651312, 38.93157187849283, 38.81659731268883, 38.86264469474554, 38.869749289005995, 38.89570105820894, 38.88459596782923, 38.78283780068159, 38.86357322335243, 38.87838777154684 ], "memory": { "VmHWM_MiB": 8576.609375, "VmRSS_MiB": 7013.765625, "Pss_MiB": 5520.138671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11578.984375, "memory.current_MiB": 7986.23046875 }, "finite": true, "input_sha256": "726cc736d46a4b8ca3c2d2831abc8c8e3a59489168194774d072d5fc1aa502b5" } ], "load_s": 16.505027560982853, "ready": { "VmHWM_MiB": 8576.609375, "VmRSS_MiB": 6944.05859375, "Pss_MiB": 5517.76953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11647.34375, "memory.current_MiB": 7917.51953125 }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 8576.609375, "VmRSS_MiB": 7013.765625, "Pss_MiB": 5520.142578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 11578.984375, "memory.current_MiB": 7986.23046875 } }, "maximum-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "2000", "warmup": 1, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/maximum-cuda.json" }, "utc": "2026-09-05T11:03:36Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.1015625, "VmRSS_MiB": 43.453125, "Pss_MiB": 34.9306640625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16270.80078125, "memory.current_MiB": 3317.8828125 }, "rows": [ { "frames": 2000, "p50_ms": 471.621782053262, "p95_ms": 525.1042190939188, "samples_ms": [ 471.58101527020335, 525.1042190939188, 471.621782053262 ], "memory": { "VmHWM_MiB": 3745.40234375, "VmRSS_MiB": 3745.40234375, "Pss_MiB": 1561.6328125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 14811.71875, "memory.current_MiB": 4776.07421875 }, "finite": true, "input_sha256": "8dd6299031bec40b45a4061c4670062fb06238b5da5111ea3e49d85335d81d47" } ], "load_s": 2.9225918371230364, "ready": { "VmHWM_MiB": 2341.40625, "VmRSS_MiB": 2292.796875, "Pss_MiB": 1144.1220703125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15217.96484375, "memory.current_MiB": 4370.86328125 }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 3793.2890625, "VmRSS_MiB": 3745.58203125, "Pss_MiB": 1561.6220703125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 14840.7734375, "memory.current_MiB": 4769.67578125 } }, "proof-offline": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": null, "cache": "offline-context", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": true, "dump_context": null, "out": "/bench/orin/proof-offline.json" }, "utc": "2026-09-05T11:24:12Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "d9cbddcb2ab713058520f37c4a914a9bb04dc013ce11721cc3193ec5a030531f", "start": { "VmHWM_MiB": 41.46875, "VmRSS_MiB": 41.453125, "Pss_MiB": 34.7236328125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17340.40625, "memory.current_MiB": 5439.046875 }, "rows": [ { "frames": 50, "p50_ms": 40.895509999245405, "p95_ms": 40.895509999245405, "samples_ms": [ 40.895509999245405 ], "memory": { "VmHWM_MiB": 2264.5859375, "VmRSS_MiB": 2128.66015625, "Pss_MiB": 872.75, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16622.69921875, "memory.current_MiB": 6154.09765625 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 8.155600168276578, "ready": { "VmHWM_MiB": 2264.5859375, "VmRSS_MiB": 2099.43359375, "Pss_MiB": 862.2578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16633.65234375, "memory.current_MiB": 6145.34765625 }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "node_provider_counts": { "TensorrtExecutionProvider": 1 }, "end": { "VmHWM_MiB": 2264.5859375, "VmRSS_MiB": 2128.8359375, "Pss_MiB": 872.90625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16621.22265625, "memory.current_MiB": 6154.1796875 } }, "graph-cuda": { "arguments": { "self_check": false, "mode": "graph", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "/bench/cache", "audio": null, "length": 3, "frames": "50,100,250", "warmup": 10, "repeats": 50, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/graph-cuda.json" }, "utc": "2026-09-05T11:03:24Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.125, "VmRSS_MiB": 43.4765625, "Pss_MiB": 34.9423828125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16273.375, "memory.current_MiB": 3317.71875 }, "rows": [ { "frames": 50, "p50_ms": 35.26263288222253, "p95_ms": 35.917248111218214, "samples_ms": [ 34.740337170660496, 35.34141601994634, 35.255048889666796, 34.78036727756262, 35.317257046699524, 35.429125651717186, 35.06138827651739, 35.917248111218214, 35.69139540195465, 35.80224188044667, 35.6347868219018, 35.15028301626444, 34.827854949980974, 35.01044400036335, 34.84612703323364, 34.883023239672184, 34.50373187661171, 35.019244998693466, 35.15517804771662, 35.70381086319685, 35.173002164810896, 35.621732007712126, 34.88132590427995, 34.81163224205375, 35.27021687477827, 35.43559089303017, 35.28797719627619, 35.960831213742495, 35.6738599948585, 35.35434417426586, 35.52109329029918, 35.13469872996211, 34.88971013575792, 35.61242017894983, 33.59456127509475, 34.53630721196532, 35.179401747882366, 35.18644208088517, 34.81703996658325, 36.2774976529181, 35.53898120298982, 35.13681096956134, 35.16785008832812, 35.76493030413985, 35.09261878207326, 35.55226093158126, 35.44925432652235, 35.77517019584775, 35.72224220260978, 35.374727100133896 ], "memory": { "VmHWM_MiB": 2302.81640625, "VmRSS_MiB": 2302.77734375, "Pss_MiB": 1143.09765625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15253.33203125, "memory.current_MiB": 4338.3203125 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" }, { "frames": 100, "p50_ms": 36.19095543399453, "p95_ms": 37.28205990046263, "samples_ms": [ 35.2917849086225, 35.29824782162905, 36.245403345674276, 35.71942588314414, 36.515831016004086, 36.00294282659888, 36.164731718599796, 36.02764615789056, 35.41955817490816, 35.645123571157455, 35.71033803746104, 36.326234228909016, 35.850209183990955, 35.69049900397658, 35.87862430140376, 36.07369400560856, 36.22405929490924, 36.49525484070182, 36.05618979781866, 35.27047298848629, 35.37978371605277, 36.930704955011606, 36.71822790056467, 36.434104200452566, 36.429304163903, 37.01467206701636, 37.31556423008442, 36.85944201424718, 37.504168692976236, 36.956561263650656, 37.039375863969326, 37.28205990046263, 37.21620514988899, 36.678549367934465, 36.64424503222108, 36.670580971986055, 36.19033098220825, 36.3377220928669, 36.22898692265153, 36.093692760914564, 36.074141040444374, 36.19157988578081, 36.32821794599295, 36.12665180116892, 36.011581774801016, 35.54150927811861, 35.91347159817815, 35.806016996502876, 36.5569181740284, 36.08140628784895 ], "memory": { "VmHWM_MiB": 2302.94140625, "VmRSS_MiB": 2302.91796875, "Pss_MiB": 1143.201171875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15245.57421875, "memory.current_MiB": 4338.328125 }, "finite": true, "input_sha256": "dda0d977bf96004405c6cbeac7cd57cfb466001cc340939617503e06dff71881" }, { "frames": 250, "p50_ms": 44.99113420024514, "p95_ms": 45.087133068591356, "samples_ms": [ 45.02147017046809, 45.087133068591356, 45.026556588709354, 45.0810850597918, 44.92726316675544, 45.14793213456869, 45.08518008515239, 45.05926230922341, 45.05874961614609, 45.05862109363079, 45.045309234410524, 45.00818997621536, 44.97356619685888, 45.08185340091586, 45.01020582392812, 44.97343814000487, 44.99490978196263, 44.97238202020526, 45.03452591598034, 44.98905409127474, 45.03378923982382, 44.97071914374828, 45.01974210143089, 44.993214309215546, 45.209402684122324, 45.0693410821259, 45.02022219821811, 44.9975342489779, 45.029789209365845, 45.015709940344095, 44.985693879425526, 44.92348711937666, 44.94751896709204, 44.921407010406256, 44.939679093658924, 44.945663306862116, 44.91065535694361, 44.91897486150265, 44.914528261870146, 44.91004720330238, 44.927998911589384, 44.974525924772024, 44.89926295354962, 44.942494947463274, 44.97020598500967, 44.957630801945925, 44.947007205337286, 45.057693030685186, 44.98674999922514, 44.905310962349176 ], "memory": { "VmHWM_MiB": 2303.0859375, "VmRSS_MiB": 2303.0859375, "Pss_MiB": 1143.33203125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15244.85546875, "memory.current_MiB": 4338.328125 }, "finite": true, "input_sha256": "726cc736d46a4b8ca3c2d2831abc8c8e3a59489168194774d072d5fc1aa502b5" } ], "load_s": 2.922731113154441, "ready": { "VmHWM_MiB": 2255.953125, "VmRSS_MiB": 2255.953125, "Pss_MiB": 1107.2763671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15254.30859375, "memory.current_MiB": 4334.0546875 }, "providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 2303.0859375, "VmRSS_MiB": 2303.0859375, "Pss_MiB": 1143.3359375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 15244.85546875, "memory.current_MiB": 4338.328125 } }, "recognizer-cuda-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-cuda-6s.json" }, "utc": "2026-09-05T11:24:48Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 43.26953125, "VmRSS_MiB": 41.62109375, "Pss_MiB": 34.7763671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17324.67578125, "memory.current_MiB": 5450.7578125 }, "rows": [ { "seconds": 6, "p50_ms": 54.28683990612626, "p95_ms": 81.33051730692387, "samples_ms": [ 81.33051730692387, 54.16031228378415, 54.28683990612626 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 2280.67578125, "VmRSS_MiB": 2028.625, "Pss_MiB": 877.107421875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16578.59765625, "memory.current_MiB": 6193.64453125 }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 2.963092844001949, "ready": { "VmHWM_MiB": 2280.67578125, "VmRSS_MiB": 1985.91015625, "Pss_MiB": 843.3251953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16582.5859375, "memory.current_MiB": 6189.13671875 }, "after_warmup": { "VmHWM_MiB": 2280.67578125, "VmRSS_MiB": 2028.2890625, "Pss_MiB": 876.982421875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16578.71875, "memory.current_MiB": 6193.51953125 }, "end": { "VmHWM_MiB": 2280.67578125, "VmRSS_MiB": 2028.625, "Pss_MiB": 877.107421875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16578.59765625, "memory.current_MiB": 6193.64453125 } }, "recognizer-cuda-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "cuda", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-cuda-15s.json" }, "utc": "2026-09-05T11:24:53Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 43.28515625, "VmRSS_MiB": 41.63671875, "Pss_MiB": 34.892578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17323.0546875, "memory.current_MiB": 5450.98046875 }, "rows": [ { "seconds": 15, "p50_ms": 87.3554041609168, "p95_ms": 162.2863090597093, "samples_ms": [ 162.2863090597093, 87.33684523031116, 87.3554041609168 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 2369.8046875, "VmRSS_MiB": 2369.8046875, "Pss_MiB": 1218.4638671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16240.8203125, "memory.current_MiB": 6535.62890625 }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 2.9110557180829346, "ready": { "VmHWM_MiB": 2351.5, "VmRSS_MiB": 2302.828125, "Pss_MiB": 1160.326171875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16265.71484375, "memory.current_MiB": 6506.73828125 }, "after_warmup": { "VmHWM_MiB": 2351.5, "VmRSS_MiB": 2345.2890625, "Pss_MiB": 1194.1044921875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16264.7734375, "memory.current_MiB": 6511.36328125 }, "end": { "VmHWM_MiB": 2369.8046875, "VmRSS_MiB": 2369.8046875, "Pss_MiB": 1218.4638671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16240.8203125, "memory.current_MiB": 6535.62890625 } }, "recognizer-trt-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-trt-15s.json" }, "utc": "2026-09-05T11:26:13Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 41.70703125, "VmRSS_MiB": 40.05859375, "Pss_MiB": 33.333984375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17359.0859375, "memory.current_MiB": 5028.9609375 }, "rows": [ { "seconds": 15, "p50_ms": 81.25538099557161, "p95_ms": 89.33074912056327, "samples_ms": [ 89.33074912056327, 81.25538099557161, 81.13400498405099 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 8589.88671875, "VmRSS_MiB": 7003.2890625, "Pss_MiB": 5548.7578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12121.13671875, "memory.current_MiB": 10245.12109375 }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 15.974479925818741, "ready": { "VmHWM_MiB": 8589.88671875, "VmRSS_MiB": 6941.40234375, "Pss_MiB": 5539.2734375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12175.8671875, "memory.current_MiB": 10191.09375 }, "after_warmup": { "VmHWM_MiB": 8589.88671875, "VmRSS_MiB": 6971.29296875, "Pss_MiB": 5548.751953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12153.6328125, "memory.current_MiB": 10213.12109375 }, "end": { "VmHWM_MiB": 8589.88671875, "VmRSS_MiB": 7003.2890625, "Pss_MiB": 5548.7578125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12121.13671875, "memory.current_MiB": 10245.12109375 } }, "offline-build": { "arguments": { "self_check": false, "mode": "graph", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": null, "cache": "offline-context", "audio": null, "length": 3, "frames": "50", "warmup": 0, "repeats": 1, "threads": 2, "profile": false, "dump_context": "/bench/model_ctx.onnx", "out": "/bench/orin/offline-build.json" }, "utc": "2026-09-05T11:22:28Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 43.33203125, "VmRSS_MiB": 41.68359375, "Pss_MiB": 34.9228515625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 18838.32421875, "memory.current_MiB": 2837.01171875 }, "rows": [ { "frames": 50, "p50_ms": 44.593106023967266, "p95_ms": 44.593106023967266, "samples_ms": [ 44.593106023967266 ], "memory": { "VmHWM_MiB": 8641.109375, "VmRSS_MiB": 7024.60546875, "Pss_MiB": 5594.0576171875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12140.38671875, "memory.current_MiB": 9652.37890625 }, "finite": true, "input_sha256": "472ba979ad46ce462cc2cceeb0d218314362876e03b63f06a136a1c19b3dbab7" } ], "load_s": 69.14930559229106, "ready": { "VmHWM_MiB": 8641.109375, "VmRSS_MiB": 7003.52734375, "Pss_MiB": 5593.2001953125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12160.26171875, "memory.current_MiB": 9632.00390625 }, "providers": [ "TensorrtExecutionProvider", "CUDAExecutionProvider", "CPUExecutionProvider" ], "end": { "VmHWM_MiB": 8641.109375, "VmRSS_MiB": 7024.60546875, "Pss_MiB": 5594.2900390625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12140.38671875, "memory.current_MiB": 9652.50390625 } }, "recognizer-offline-15s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/bench/model_ctx.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "offline-context", "audio": "/bench/audio", "length": 15, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-offline-15s.json" }, "utc": "2026-09-05T11:26:51Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "d9cbddcb2ab713058520f37c4a914a9bb04dc013ce11721cc3193ec5a030531f", "start": { "VmHWM_MiB": 41.94921875, "VmRSS_MiB": 41.93359375, "Pss_MiB": 35.275390625, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17362.984375, "memory.current_MiB": 5076.31640625 }, "rows": [ { "seconds": 15, "p50_ms": 81.05635596439242, "p95_ms": 87.37560827285051, "samples_ms": [ 87.37560827285051, 81.05635596439242, 80.73318982496858 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥，接着把冷气打开，温度调到24度，风量开大。帮我导航到内湖科学园区，然后选最快的路线。不要走。" ], "memory": { "VmHWM_MiB": 2264.3359375, "VmRSS_MiB": 2188.25390625, "Pss_MiB": 909.109375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16589.83984375, "memory.current_MiB": 5835.4296875 }, "audio_sha256": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f" } ], "load_s": 8.243862583767623, "ready": { "VmHWM_MiB": 2264.3359375, "VmRSS_MiB": 2078.0234375, "Pss_MiB": 851.830078125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16639.8046875, "memory.current_MiB": 5787.4609375 }, "after_warmup": { "VmHWM_MiB": 2264.3359375, "VmRSS_MiB": 2124.4609375, "Pss_MiB": 877.29296875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16621.46875, "memory.current_MiB": 5803.5546875 }, "end": { "VmHWM_MiB": 2264.3359375, "VmRSS_MiB": 2188.25390625, "Pss_MiB": 909.109375, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 16589.83984375, "memory.current_MiB": 5835.4296875 } }, "recognizer-trt-6s": { "arguments": { "self_check": false, "mode": "recognizer", "provider": "trt", "model": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx", "tokens": "/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt", "cache": "/bench/cache", "audio": "/bench/audio", "length": 6, "frames": "50,100,250", "warmup": 10, "repeats": 3, "threads": 2, "profile": false, "dump_context": null, "out": "/bench/orin/recognizer-trt-6s.json" }, "utc": "2026-09-05T11:25:46Z", "machine": "aarch64", "ort": "1.24.0", "numpy": "1.26.4", "model_sha256": "719559b0512db928e18eb3c130a122a55dc10166fb6bd2c8c90f2b6f51064994", "start": { "VmHWM_MiB": 45.69921875, "VmRSS_MiB": 44.05078125, "Pss_MiB": 37.20703125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 17377.265625, "memory.current_MiB": 5053.48046875 }, "rows": [ { "seconds": 6, "p50_ms": 39.50392780825496, "p95_ms": 53.87589568272233, "samples_ms": [ 53.87589568272233, 39.48588017374277, 39.50392780825496 ], "texts": [ "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。", "帮我导航到内湖科学园区，然后选最快的路线，不要走高架桥。" ], "memory": { "VmHWM_MiB": 8595.42578125, "VmRSS_MiB": 6992.83203125, "Pss_MiB": 5554.2763671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12113.0625, "memory.current_MiB": 10268.3046875 }, "audio_sha256": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" } ], "load_s": 23.767939010169357, "ready": { "VmHWM_MiB": 8595.42578125, "VmRSS_MiB": 6946.9375, "Pss_MiB": 5544.7939453125, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12150.87890625, "memory.current_MiB": 10230.1796875 }, "after_warmup": { "VmHWM_MiB": 8595.42578125, "VmRSS_MiB": 6976.83203125, "Pss_MiB": 5554.2763671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12129.1875, "memory.current_MiB": 10252.3046875 }, "end": { "VmHWM_MiB": 8595.42578125, "VmRSS_MiB": 6992.83203125, "Pss_MiB": 5554.2763671875, "MemTotal_MiB": 62840.6796875, "MemAvailable_MiB": 12113.0703125, "memory.current_MiB": 10268.3046875 } }, "service": { "service-cuda-1": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:13:28", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 87.09, "server_rtf": 0.035, "server_ttft_ms": 68.85, "client_final_latency_ms": 104.92, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 104.43, "server_rtf": 0.021, "server_ttft_ms": 69.48, "client_final_latency_ms": 122.94, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.13, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 235.31, "server_rtf": 0.009, "server_ttft_ms": 69.05, "client_final_latency_ms": 364.83, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.31, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 93.02, "server_rtf": 0.117, "server_ttft_ms": 68.62, "client_final_latency_ms": 96.54, "interim_count": 5, "max_interim_gap_s": 0.511, "feed_secs": 3.06, "text": "幫我導航到內湖科學園區。然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 93.99, "server_rtf": 0.122, "server_ttft_ms": 67.29, "client_final_latency_ms": 97.66, "interim_count": 11, "max_interim_gap_s": 1.026, "feed_secs": 6.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 217.87, "server_rtf": 0.145, "server_ttft_ms": 67.91, "client_final_latency_ms": 277.08, "interim_count": 26, "max_interim_gap_s": 1.031, "feed_secs": 15.32, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-cuda-2": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:14:38", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 87.55, "server_rtf": 0.018, "server_ttft_ms": 157.91, "client_final_latency_ms": 104.94, "interim_count": 0, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 96.58, "server_rtf": 0.02, "server_ttft_ms": 69.13, "client_final_latency_ms": 114.5, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 225.19, "server_rtf": 0.009, "server_ttft_ms": 75.32, "client_final_latency_ms": 243.13, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.34, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 76.68, "server_rtf": 0.091, "server_ttft_ms": 66.77, "client_final_latency_ms": 124.32, "interim_count": 5, "max_interim_gap_s": 1.006, "feed_secs": 3.07, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 79.93, "server_rtf": 0.117, "server_ttft_ms": 66.36, "client_final_latency_ms": 90.37, "interim_count": 11, "max_interim_gap_s": 1.01, "feed_secs": 6.11, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 180.07, "server_rtf": 0.13, "server_ttft_ms": 67.22, "client_final_latency_ms": 184.14, "interim_count": 27, "max_interim_gap_s": 1.022, "feed_secs": 15.29, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null } ] }, "service-cuda-3": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:15:49", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 90.5, "server_rtf": 0.037, "server_ttft_ms": 688.04, "client_final_latency_ms": 742.46, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 98.23, "server_rtf": 0.02, "server_ttft_ms": 70.0, "client_final_latency_ms": 116.11, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.13, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 220.57, "server_rtf": 0.009, "server_ttft_ms": 68.51, "client_final_latency_ms": 238.09, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.34, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 81.9, "server_rtf": 0.111, "server_ttft_ms": 67.49, "client_final_latency_ms": 85.41, "interim_count": 6, "max_interim_gap_s": 0.513, "feed_secs": 3.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 82.68, "server_rtf": 0.117, "server_ttft_ms": 67.16, "client_final_latency_ms": 92.21, "interim_count": 10, "max_interim_gap_s": 1.266, "feed_secs": 6.13, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 189.43, "server_rtf": 0.123, "server_ttft_ms": 67.18, "client_final_latency_ms": 193.51, "interim_count": 24, "max_interim_gap_s": 1.192, "feed_secs": 15.33, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-trt-1": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:19:12", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 54.3, "server_rtf": 0.02, "server_ttft_ms": 40.55, "client_final_latency_ms": 78.09, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 69.86, "server_rtf": 0.012, "server_ttft_ms": 42.53, "client_final_latency_ms": 124.52, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.13, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 234.66, "server_rtf": 0.008, "server_ttft_ms": 41.19, "client_final_latency_ms": 252.17, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.34, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 48.81, "server_rtf": 0.061, "server_ttft_ms": 37.78, "client_final_latency_ms": 59.44, "interim_count": 6, "max_interim_gap_s": 0.511, "feed_secs": 3.05, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 45.49, "server_rtf": 0.064, "server_ttft_ms": 39.54, "client_final_latency_ms": 79.92, "interim_count": 11, "max_interim_gap_s": 1.021, "feed_secs": 6.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 211.04, "server_rtf": 0.077, "server_ttft_ms": 37.81, "client_final_latency_ms": 394.73, "interim_count": 26, "max_interim_gap_s": 1.129, "feed_secs": 15.32, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null } ] }, "service-trt-2": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:20:21", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 55.73, "server_rtf": 0.02, "server_ttft_ms": 41.14, "client_final_latency_ms": 78.23, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 61.15, "server_rtf": 0.011, "server_ttft_ms": 40.72, "client_final_latency_ms": 102.01, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.13, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 206.14, "server_rtf": 0.006, "server_ttft_ms": 40.21, "client_final_latency_ms": 223.65, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.35, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 49.96, "server_rtf": 0.061, "server_ttft_ms": 37.7, "client_final_latency_ms": 59.74, "interim_count": 5, "max_interim_gap_s": 0.673, "feed_secs": 3.06, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 45.99, "server_rtf": 0.064, "server_ttft_ms": 37.9, "client_final_latency_ms": 80.94, "interim_count": 11, "max_interim_gap_s": 1.019, "feed_secs": 6.11, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 170.87, "server_rtf": 0.076, "server_ttft_ms": 37.77, "client_final_latency_ms": 175.05, "interim_count": 26, "max_interim_gap_s": 1.022, "feed_secs": 15.3, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。", "error": null } ] }, "service-trt-3": { "environment": { "target": "ws://127.0.0.1:19242/stt/sensevoice", "uri": "ws://127.0.0.1:19242/stt/sensevoice", "warmup_runs": 3, "single_connection_series": true }, "generated": "2026-09-05T11:21:31", "results": [ { "feed": "bulk", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 53.97, "server_rtf": 0.01, "server_ttft_ms": 163.55, "client_final_latency_ms": 196.52, "interim_count": 0, "max_interim_gap_s": null, "feed_secs": 0.04, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "bulk", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 56.96, "server_rtf": 0.01, "server_ttft_ms": 40.22, "client_final_latency_ms": 105.91, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.12, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "bulk", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 205.27, "server_rtf": 0.006, "server_ttft_ms": 40.14, "client_final_latency_ms": 222.84, "interim_count": 1, "max_interim_gap_s": null, "feed_secs": 0.33, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null }, { "feed": "realtime", "target_s": 3.0, "duration_s": 3.0, "status": "ok", "server_latency_ms": 46.81, "server_rtf": 0.06, "server_ttft_ms": 38.02, "client_final_latency_ms": 58.25, "interim_count": 6, "max_interim_gap_s": 0.51, "feed_secs": 3.05, "text": "幫我導航到內湖科學園區，然後。", "error": null }, { "feed": "realtime", "target_s": 6.0, "duration_s": 6.0, "status": "ok", "server_latency_ms": 46.95, "server_rtf": 0.06, "server_ttft_ms": 37.7, "client_final_latency_ms": 81.88, "interim_count": 10, "max_interim_gap_s": 1.288, "feed_secs": 6.11, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋。", "error": null }, { "feed": "realtime", "target_s": 15.0, "duration_s": 15.0, "status": "ok", "server_latency_ms": 167.32, "server_rtf": 0.071, "server_ttft_ms": 37.81, "client_final_latency_ms": 171.54, "interim_count": 25, "max_interim_gap_s": 1.182, "feed_secs": 15.32, "text": "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "error": null } ] } } } }, "audio_source": "tests/functional/audios/multiuser-10s.wav; repeat PCM and trim to 3/6/15 seconds, 16kHz mono PCM16", "audio_sha256": { "clip-3s.wav": "11034b66e2ae87bdec98dfe7401642387b85262b2e9c222b9e201030dc397a1c", "clip-15s.wav": "4950a411664da365f4b8cfe13838f881430f01f2fed8b2c3335719f6b791ff9f", "clip-6s.wav": "9566be6ccf6c7307830514e0fe5be86250c493efaef2800f27bfccc1ca1850b0" }, "method": { "graph": "seed=42; threads=2; warmup=10; samples=50; T=50/100/250; profiling OFF", "maximum": "T=2000; warmup=1; samples=3; profiling OFF", "recognizer": "new process per provider/length; threads=2; one 3s warmup; 3 decodes; no temporary server running", "service": "49f0488 source; streaming_interim=true; profile_dbg=true; warmup=3; 3 series per provider; bulk+realtime", "memory": "MiB=2^20 bytes; same temporary container reused across processes; cgroup includes accumulated page cache, not comparable to fresh-Pod historical cgroup", "offline": "new empty cache; context metadata copied from source model; original ONNX temporarily renamed during recognizer/profiler proof, restored after" }, "limits": [ "Shared GPUs, existing workloads not stopped; 4090 idle residency initially 9569 MiB.", "Historical source WAV and benchmark scripts absent; exact historical audio/hash/thread/profiler settings not known.", "Legacy MB notation is not assumed equivalent to MiB; no exact legacy memory percentage claimed.", "Service is real STT WebSocket frame transport, excludes browser/LiveKit/audio-device latency.", "Service text varies in some repeated runs; no accuracy metric or TensorRT regression inferred from this sample." ], "validation": { "offline": "625 passed, 28 skipped; 15.49s; full tests/test.sh; live suite skipped on port19999", "live_benchmark_cases": 72, "recognizer_case_decodes": 54, "profile_proofs": "both platforms, cache-warm and offline only TensorRT nodes", "documentation": "39 passed; report and traceability checks", "cleanup": "temporary Docker and Orin Pod removed; original Orin Pod Running 1/1 with zero restarts; original four 4090 compute PIDs remain", "render": "Playwright screenshots visually inspected for graph/service/memory sections" }, "commands": { "run-clean-memory.sh": "#!/bin/sh\nset -eu\nsh /bench/run-recognizers.sh\ncd /bench\nBENCH_PY=${BENCH_PY:-python3}\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nfor length in 3 6 15; do\n \"$BENCH_PY\" /bench/provider_recheck.py --mode recognizer --provider trt --model /bench/model_ctx.onnx --tokens /app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/tokens.txt --audio /bench/audio --length \"$length\" --cache offline-context --repeats 3 --out \"/bench/$BENCH_PLATFORM/recognizer-offline-${length}s.json\" > \"/bench/$BENCH_PLATFORM/recognizer-offline-${length}s.log\" 2>&1\ndone\n", "run-models.sh": "#!/bin/sh\nset -eu\ncd /app\nBENCH_PY=${BENCH_PY:-python3}\nBENCH_PLATFORM=${BENCH_PLATFORM:?}\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nBUNDLE=/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1\nOUT=/bench/$BENCH_PLATFORM\nmkdir -p \"$OUT\"\nfor provider in cuda trt; do\n \"$BENCH_PY\" /bench/provider_recheck.py --provider \"$provider\" --model \"$BUNDLE/model.onnx\" --cache /bench/cache --out \"$OUT/graph-$provider.json\" > \"$OUT/graph-$provider.log\" 2>&1\n \"$BENCH_PY\" /bench/provider_recheck.py --provider \"$provider\" --model \"$BUNDLE/model.onnx\" --cache /bench/cache --frames 2000 --warmup 1 --repeats 3 --out \"$OUT/maximum-$provider.json\" > \"$OUT/maximum-$provider.log\" 2>&1\n \"$BENCH_PY\" /bench/provider_recheck.py --provider \"$provider\" --model \"$BUNDLE/model.onnx\" --cache /bench/cache --frames 50 --warmup 0 --repeats 1 --profile --out \"$OUT/proof-$provider.json\" > \"$OUT/proof-$provider.log\" 2>&1\n for length in 3 6 15; do\n \"$BENCH_PY\" /bench/provider_recheck.py --mode recognizer --provider \"$provider\" --model \"$BUNDLE/model.onnx\" --tokens \"$BUNDLE/tokens.txt\" --audio /bench/audio --length \"$length\" --cache /bench/cache --repeats 3 --out \"$OUT/recognizer-$provider-${length}s.json\" > \"$OUT/recognizer-$provider-${length}s.log\" 2>&1\n done\ndone\nprintf 'model measurements complete: %s\\n' \"$BENCH_PLATFORM\"\n", "run-threads.sh": "#!/bin/sh\nset -eu\ncd /app\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nfor threads in 0 4; do\n /app/.venv/bin/python /bench/provider_recheck.py --provider cuda --model /app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx --cache /bench/cache --frames 50,100 --threads \"$threads\" --out \"/bench/4090/threads-$threads.json\" > \"/bench/4090/threads-$threads.log\" 2>&1\ndone\n", "run-offline.sh": "#!/bin/sh\nset -eu\ncd /bench\nBENCH_PY=${BENCH_PY:-python3}\nBENCH_PLATFORM=${BENCH_PLATFORM:?}\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nBUNDLE=/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1\nOUT=/bench/$BENCH_PLATFORM\nmkdir -p /bench/offline-context\n\"$BENCH_PY\" /bench/provider_recheck.py --provider trt --model \"$BUNDLE/model.onnx\" --cache offline-context --dump-context /bench/model_ctx.onnx --frames 50 --warmup 0 --repeats 1 --out \"$OUT/offline-build.json\" > \"$OUT/offline-build.log\" 2>&1\n\"$BENCH_PY\" - <<'PY'\nimport onnx\nfrom pathlib import Path\nsource=onnx.load('/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1/model.onnx')\np=Path('/bench/model_ctx.onnx')\nmodel=onnx.load(str(p))\nexisting={x.key for x in model.metadata_props}\nfor item in source.metadata_props:\n if item.key not in existing: model.metadata_props.append(item)\nonnx.save(model,str(p))\nprint('EP context prepared:',p.stat().st_size,'bytes')\nPY\nmv \"$BUNDLE/model.onnx\" \"$BUNDLE/model.onnx.recheck-hidden\"\ntrap 'mv \"$BUNDLE/model.onnx.recheck-hidden\" \"$BUNDLE/model.onnx\"' EXIT INT TERM\nfor length in 3 6 15; do\n \"$BENCH_PY\" /bench/provider_recheck.py --mode recognizer --provider trt --model /bench/model_ctx.onnx --tokens \"$BUNDLE/tokens.txt\" --audio /bench/audio --length \"$length\" --cache offline-context --repeats 3 --out \"$OUT/recognizer-offline-${length}s.json\" > \"$OUT/recognizer-offline-${length}s.log\" 2>&1\ndone\n\"$BENCH_PY\" /bench/provider_recheck.py --provider trt --model /bench/model_ctx.onnx --cache offline-context --frames 50 --warmup 0 --repeats 1 --profile --out \"$OUT/proof-offline.json\" > \"$OUT/proof-offline.log\" 2>&1\n\nmv \"$BUNDLE/model.onnx.recheck-hidden\" \"$BUNDLE/model.onnx\"\ntrap - EXIT INT TERM\nif [ \"$BENCH_PLATFORM\" = 4090 ]; then\n \"$BENCH_PY\" /bench/provider_recheck.py --provider cuda --model \"$BUNDLE/model.onnx\" --cache /bench/cache --frames 50,100 --profile --out \"$OUT/profile-overhead-cuda.json\" > \"$OUT/profile-overhead-cuda.log\" 2>&1\nfi\n", "run-service.sh": "#!/bin/sh\nset -eu\ncd /app\nBENCH_PY=${BENCH_PY:-python3}\nBENCH_PLATFORM=${BENCH_PLATFORM:?}\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nOUT=/bench/$BENCH_PLATFORM\nfor provider in cuda trt; do\n provider_value=$provider\n if [ \"$provider\" = trt ]; then provider_value=\"trt:$OUT/recognizer-trt-3s.config\"; fi\n cat > \"$OUT/service-$provider.yaml\" < \"$OUT/service-$provider.log\" 2>&1 &\n service_pid=$!\n trap 'kill -KILL \"$service_pid\" 2>/dev/null || true' EXIT INT TERM\n \"$BENCH_PY\" - <<'PY'\nimport time, urllib.request\nfor _ in range(180):\n try:\n urllib.request.urlopen('http://127.0.0.1:19242/stt/sensevoice/health',timeout=2)\n break\n except Exception:\n time.sleep(1)\nelse:\n raise SystemExit('temporary service startup timeout')\nPY\n for repetition in 1 2 3; do\n \"$BENCH_PY\" tests/batch_decode/hardware_profile.py --target ws://127.0.0.1:19242/stt/sensevoice --lengths 3,6,15 --warmup 3 --out \"$OUT/service-$provider-$repetition\" > \"$OUT/service-$provider-$repetition.log\" 2>&1\n done\n kill \"$service_pid\"\n sleep 2\n kill -KILL \"$service_pid\" 2>/dev/null || true\n wait \"$service_pid\" || true\n trap - EXIT INT TERM\ndone\n", "run-recognizers.sh": "#!/bin/sh\nset -eu\ncd /app\nBENCH_PY=${BENCH_PY:-python3}\nBENCH_PLATFORM=${BENCH_PLATFORM:?}\nfor lib in /app/.venv/lib/python*/site-packages/tensorrt_libs /app/.venv/lib/python*/site-packages/nvidia/*/lib; do\n export LD_LIBRARY_PATH=\"$lib:${LD_LIBRARY_PATH:-}\"\ndone\nBUNDLE=/app/models/sherpa-onnx-sense-voice-kitt-wake-lora-v1\nOUT=/bench/$BENCH_PLATFORM\nmkdir -p \"$OUT\"\nfor provider in cuda trt; do\n for length in 3 6 15; do\n \"$BENCH_PY\" /bench/provider_recheck.py --mode recognizer --provider \"$provider\" --model \"$BUNDLE/model.onnx\" --tokens \"$BUNDLE/tokens.txt\" --audio /bench/audio --length \"$length\" --cache /bench/cache --repeats 3 --out \"$OUT/recognizer-$provider-${length}s.json\" > \"$OUT/recognizer-$provider-${length}s.log\" 2>&1\n done\ndone\nprintf 'model measurements complete: %s\\n' \"$BENCH_PLATFORM\"\n" }, "service_text_variations": [ { "platform": "4090", "feed": "bulk", "length": 15, "texts": { "cuda": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ], "trt": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ] } }, { "platform": "4090", "feed": "realtime", "length": 15, "texts": { "cuda": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ], "trt": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ] } }, { "platform": "orin", "feed": "bulk", "length": 15, "texts": { "cuda": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ], "trt": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ] } }, { "platform": "orin", "feed": "realtime", "length": 3, "texts": { "cuda": [ "幫我導航到內湖科學園區。然後。", "幫我導航到內湖科學園區，然後。" ], "trt": [ "幫我導航到內湖科學園區，然後。" ] } }, { "platform": "orin", "feed": "realtime", "length": 15, "texts": { "cuda": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ], "trt": [ "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開。幫我導航到內湖科學園區，然後選最快的路線，不要走。", "幫我導航到內湖科學園區，然後選最快的路線，不要走高架橋，接着把冷氣打開，溫度調到24度，風量開大。幫我導航到內湖科學園區，然後選最快的路線。不要走。" ] } } ] }