anodyine/vllm-omni
1
0 Fork 0
Code Issues 4 Pull Requests 2 Actions Packages Projects Releases Wiki Activity

Audit SA8775P Cosmos placement and Windows TTS startup

Browse Source
This commit is contained in:
tzhouam committed 2026-09-24 09:36:08 +08:00
1 parent 698f8c13d9
commit 2ff0a783d4
13 files changed
+1993 -11
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

benchmarks/edge_harness/results/e2e_expansion_20260924/README.md
+1 -1
View File
@@ -10,4 +10,4 @@ The [MiniCPM-o speech-head study](evidence/minicpmo_tts_qualcomm/README.md) reco
The [native Windows Qwen3-TTS sustained analysis](evidence/qwen_tts_sustained_drift/README.md) finds a sharp playback throughput collapse around minute 26–27 in the historical 30-minute RTX 5090 Laptop stream. The last 17 serial requests all had simulated underruns with median RTF 1.331; the separate WSL run did not show that behavior. A revised local profiler now records timestamped device-wide NVML clocks, power, temperature, throttling and utilization to investigate the cause in a repeat run. The historical evidence cannot identify the cause or qualify native Windows sustained playback.
The [InternVLA Cosmos S25 attempt](evidence/internvla_cosmos_qualcomm/README.md) repackages the real policy's image encoder from its parity-tested ONNX export. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one pattern. Injecting it into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. S25 six-frame component placement profile returned 100 samples at p50/p95 49.665/51.382 ms, with all 121 execution-detail rows on NPU. The same encoder compiled to FP16 QNN DLC for X Elite CRD; same-fixture NPU-requested inference returned a finite latent at 0.0608% relative L2 / 64.32 dB SNR versus the source, with its 100-sample six-frame profile at p50/p95 82.939/83.837 ms and all 121 execution-detail rows on NPU. An SA8775P FP16 QNN DLC inference returned a finite same-fixture latent at 0.0613% relative L2 / 64.25 dB SNR versus source; placement profiling is pending. Device-local action-policy support remains unqualified. The checkpoint-named Place_Markpen dataset returned HTTP 404, and the official vLLM-Omni example archive in InternData-A1 returned HTTP 403 gated-access denial under the current authenticated account; [real-observation access evidence](evidence/internvla_real_observation_access/README.md) records the blocker without treating synthetic actions as task validation.
The [InternVLA Cosmos S25 attempt](evidence/internvla_cosmos_qualcomm/README.md) repackages the real policy's image encoder from its parity-tested ONNX export. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one pattern. Injecting it into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. S25 six-frame component placement profile returned 100 samples at p50/p95 49.665/51.382 ms, with all 121 execution-detail rows on NPU. The same encoder compiled to FP16 QNN DLC for X Elite CRD; same-fixture NPU-requested inference returned a finite latent at 0.0608% relative L2 / 64.32 dB SNR versus the source, with its 100-sample six-frame profile at p50/p95 82.939/83.837 ms and all 121 execution-detail rows on NPU. An SA8775P FP16 QNN DLC inference returned a finite same-fixture latent at 0.0613% relative L2 / 64.25 dB SNR versus source; its 100-sample six-frame profile measured p50/p95 116.021/116.327 ms with all 121 execution-detail rows on NPU. Device-local action-policy support remains unqualified. The checkpoint-named Place_Markpen dataset returned HTTP 404, and the official vLLM-Omni example archive in InternData-A1 returned HTTP 403 gated-access denial under the current authenticated account; [real-observation access evidence](evidence/internvla_real_observation_access/README.md) records the blocker without treating synthetic actions as task validation.
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md
+2 -2
View File
@@ -6,8 +6,8 @@ Workbench source model `mn4okj3rq` was [submitted for Galaxy S25 QNN DLC compila
The same source model also [compiled to FP16 QNN DLC on Snapdragon X Elite CRD](xelite/compile_report.json), producing target `mmxj0w8kq` for Windows ARM. [Same-fixture NPU-requested inference](xelite/inference_report.json) on dataset `d26qo8p57` returned a finite FP32 `[6,16,32,32]` latent. Its [audit](xelite/audit_report.json) measured **0.0608% relative L2 / 64.32 dB SNR** versus the pinned source pattern, with cosine 0.999999815. The [placement profile](xelite/profile_report.json) returned 100 six-frame component samples at nearest-rank p50/p95 **82.939/83.837 ms**, with all **121 execution-detail rows on NPU**. The 82.446 ms estimated-time field is the sample minimum. Reported inference peak memory was 24,059,904 bytes and first-load peak 692,891,648 bytes. This is component evidence only; neither action quality nor a complete device-local policy was tested.
The same source also [compiled to FP16 QNN DLC on SA8775P ADP](sa8775p/compile_report.json), producing target mn750wz8m for Android. [Same-fixture NPU-requested inference](sa8775p/inference_report.json) using dataset d26qo8p57 returned a finite FP32 [6,16,32,32] latent. Its [audit](sa8775p/audit_report.json) measured **0.0613% relative L2 / 64.25 dB SNR** versus the pinned source pattern, with cosine 0.999999812. Actual NPU node assignment and timing await the [submitted placement profile](sa8775p/profile_submission.json); no complete local policy follows.
The same source also [compiled to FP16 QNN DLC on SA8775P ADP](sa8775p/compile_report.json), producing target mn750wz8m for Android. [Same-fixture NPU-requested inference](sa8775p/inference_report.json) using dataset d26qo8p57 returned a finite FP32 [6,16,32,32] latent. Its [audit](sa8775p/audit_report.json) measured **0.0613% relative L2 / 64.25 dB SNR** versus the pinned source pattern, with cosine 0.999999812. The [placement profile](sa8775p/profile_report.json) returned 100 six-frame component samples at nearest-rank p50/p95 **116.021/116.327 ms**, with all **121 execution-detail rows on NPU**. The 114.779 ms estimated-time field is the sample minimum. Reported inference peak memory was 116,924,416 bytes and first-load peak 659,845,120 bytes. No complete local policy follows.
The [generic collector](../../../../experiments/collect_aihub_component.py) preserves terminal jobs, and the [Cosmos auditor](../../../../experiments/audit_internvla_cosmos_qualcomm.py) checks source/fixture hashes, target/device/dataset identity, latent shape and finite numerical difference against the pinned source reference. A separate [synthetic action sensitivity check](s25/action_sensitivity.json) injected either the pinned source or S25 latent into the same real Place_Markpen policy on **local CPU**, with zero state/noise. The [reproducer](../../../../experiments/probe_internvla_hosted_latent_actions.py) returned finite `[1,50,32]` actions; hosted-vs-source action relative L2 was **0.742%**, maximum absolute difference **0.0115**, cosine **0.999973**. Repeating the source policy gave bitwise identical actions. This is a cross-device numerical handoff experiment, not a device-local policy or robot-task pass; no real-observation action tolerance is known.
Next: audit SA8775P placement, test representative real observations and action tolerances, then build and validate a device-local stateful policy. Full multimodal observation processing, admission, cancellation and sustained profiling remain open.
Next: test representative real observations and action tolerances, then build and validate a device-local stateful policy. Full multimodal observation processing, admission, cancellation and sustained profiling remain open.
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/sa8775p/audit_report.json
+19 -4
View File
@@ -1,6 +1,6 @@
{
"scope": "one six-frame Cosmos image encoder on synthetic pattern; no complete InternVLA policy",
"status": "component_inference_numeric_measured",
"status": "component_inference_numeric_and_placement_measured",
"device": {
"name": "SA8775P ADP",
"os": "14",
@@ -48,7 +48,22 @@
},
"limits": [
"The fixture is synthetic; no real-observation or action-quality tolerance is established.",
"The action policy, state lifecycle, admission and complete request were not run on this device.",
"Requested NPU is not actual per-node placement without a device profile."
]
"The action policy, state lifecycle, admission and complete request were not run on this device."
],
"profile": {
"job_id": "jpe7z7ov5",
"sample_count": 100,
"sample_min_us": 114779,
"nearest_rank_p50_us": 116021,
"nearest_rank_p95_us": 116327,
"reported_estimated_inference_time_us": 114779,
"reported_peak_memory_bytes": 116924416,
"reported_first_load_time_us": 105214319,
"reported_warm_load_time_us": 469698,
"reported_first_load_peak_memory_bytes": 659845120,
"reported_warm_load_peak_memory_bytes": 268201984,
"compute_unit_row_counts": {
"NPU": 121
}
}
}
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/sa8775p/profile_report.json
+1043
View File
File diff suppressed because it is too large. Load diff
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/README.md
+2
View File
@@ -15,4 +15,6 @@ The reproducible [analysis script](../../../../experiments/analyze_tts_sustained
The Windows stream changed regime around minute 26–27. In its last window, all 17 requests had RTF above real time at the median and an underrun at first-audio playback. A small jitter buffer cannot cure this late sustained throughput deficit. The recorded profile sampled memory but not a timestamped GPU clock, temperature, power or throttling timeline, so its cause is unknown. The updated [local TTS profiler](../../../../profile_local_tts.py) can sample NVML and host telemetry with `--gpu-telemetry-interval-s 1`; repeat the Windows run with that option and correlate the device-wide samples with request timestamps. GPU telemetry is whole-device and cannot attribute power or utilization to the model alone.
The first instrumented retry [failed before model startup](windows_startup_failures/unc_cwd_report.json). A second retry from a native Windows working directory still imported an older editable Omni checkout; its [runtime log](windows_startup_failures/stale_checkout.log.gz) shows the stage worker trying the POSIX-only `os.fchmod` path on Windows, and its [report](windows_startup_failures/stale_checkout_report.json) has zero requests. The current fork source already has a Windows path for that lock. With the current fork first on `PYTHONPATH` and native `C:\Users\zhout\w2` as working directory, a [new smoke report](windows_current_source_smoke/report.json) completed startup in 69.38 s, one warmup, a slow-consumer request, cancellation acknowledgement and an after-cancel request. Its [request events](windows_current_source_smoke/requests.jsonl) and 71 [device-wide telemetry samples](windows_current_source_smoke/gpu_telemetry.jsonl) are retained. This smoke validates the measurement path, not 30-minute playback. A matching sustained repeat is running separately and will supersede this startup-only evidence when terminal.
This evidence does not qualify uninterrupted native Windows playback, speech quality, restart after cancellation, or mobile/embedded TTS. It does not establish thermal throttling or an Omni software regression as the cause.
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_current_source_smoke/gpu_telemetry.jsonl
+71
View File
@@ -0,0 +1,71 @@
{"unix": 1790213199.4832008, "gpu_temperature_c": 55, "gpu_power_mw": 20639, "gpu_clock_sm_mhz": 795, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 6, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29135712256, "host_cpu_utilization_pct": 0.0}
{"unix": 1790213200.5007305, "gpu_temperature_c": 55, "gpu_power_mw": 19386, "gpu_clock_sm_mhz": 757, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29094313984, "host_cpu_utilization_pct": 8.0}
{"unix": 1790213201.5018215, "gpu_temperature_c": 54, "gpu_power_mw": 19568, "gpu_clock_sm_mhz": 787, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29095579648, "host_cpu_utilization_pct": 9.1}
{"unix": 1790213202.5088906, "gpu_temperature_c": 54, "gpu_power_mw": 21677, "gpu_clock_sm_mhz": 780, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29093814272, "host_cpu_utilization_pct": 9.9}
{"unix": 1790213203.517305, "gpu_temperature_c": 54, "gpu_power_mw": 20605, "gpu_clock_sm_mhz": 840, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29100527616, "host_cpu_utilization_pct": 6.8}
{"unix": 1790213204.6159062, "gpu_temperature_c": 54, "gpu_power_mw": 22439, "gpu_clock_sm_mhz": 802, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 10, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29043507200, "host_cpu_utilization_pct": 12.5}
{"unix": 1790213205.6573763, "gpu_temperature_c": 54, "gpu_power_mw": 21371, "gpu_clock_sm_mhz": 810, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 10, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 29044187136, "host_cpu_utilization_pct": 13.4}
{"unix": 1790213206.7041438, "gpu_temperature_c": 54, "gpu_power_mw": 19150, "gpu_clock_sm_mhz": 832, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 28750356480, "host_cpu_utilization_pct": 15.1}
{"unix": 1790213207.7247946, "gpu_temperature_c": 54, "gpu_power_mw": 18296, "gpu_clock_sm_mhz": 787, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 28349960192, "host_cpu_utilization_pct": 16.5}
{"unix": 1790213208.7327693, "gpu_temperature_c": 54, "gpu_power_mw": 19546, "gpu_clock_sm_mhz": 772, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3570622464, "host_available_bytes": 28251353088, "host_cpu_utilization_pct": 21.5}
{"unix": 1790213209.7421014, "gpu_temperature_c": 55, "gpu_power_mw": 22072, "gpu_clock_sm_mhz": 915, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 13, "gpu_memory_used_bytes": 3575996416, "host_available_bytes": 28110458880, "host_cpu_utilization_pct": 27.0}
{"unix": 1790213210.7809682, "gpu_temperature_c": 54, "gpu_power_mw": 20365, "gpu_clock_sm_mhz": 765, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3575996416, "host_available_bytes": 28011036672, "host_cpu_utilization_pct": 18.4}
{"unix": 1790213211.8100326, "gpu_temperature_c": 55, "gpu_power_mw": 20358, "gpu_clock_sm_mhz": 825, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3601588224, "host_available_bytes": 28017262592, "host_cpu_utilization_pct": 23.1}
{"unix": 1790213212.8141804, "gpu_temperature_c": 55, "gpu_power_mw": 21057, "gpu_clock_sm_mhz": 967, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 11, "gpu_memory_used_bytes": 3603095552, "host_available_bytes": 27432693760, "host_cpu_utilization_pct": 37.0}
{"unix": 1790213213.8234646, "gpu_temperature_c": 55, "gpu_power_mw": 22226, "gpu_clock_sm_mhz": 787, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 13, "gpu_memory_used_bytes": 3646390272, "host_available_bytes": 26278273024, "host_cpu_utilization_pct": 33.7}
{"unix": 1790213214.8353975, "gpu_temperature_c": 55, "gpu_power_mw": 27358, "gpu_clock_sm_mhz": 742, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3641016320, "host_available_bytes": 26117877760, "host_cpu_utilization_pct": 30.0}
{"unix": 1790213215.8378766, "gpu_temperature_c": 55, "gpu_power_mw": 22878, "gpu_clock_sm_mhz": 990, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 78, "gpu_memory_used_bytes": 3642261504, "host_available_bytes": 26117926912, "host_cpu_utilization_pct": 26.0}
{"unix": 1790213216.858904, "gpu_temperature_c": 56, "gpu_power_mw": 26852, "gpu_clock_sm_mhz": 1072, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 59, "gpu_memory_used_bytes": 3663233024, "host_available_bytes": 26057072640, "host_cpu_utilization_pct": 28.4}
{"unix": 1790213217.9053948, "gpu_temperature_c": 56, "gpu_power_mw": 31184, "gpu_clock_sm_mhz": 1852, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 3659890688, "host_available_bytes": 25986555904, "host_cpu_utilization_pct": 22.3}
{"unix": 1790213218.9857621, "gpu_temperature_c": 56, "gpu_power_mw": 35164, "gpu_clock_sm_mhz": 1995, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 25, "gpu_memory_used_bytes": 3613818880, "host_available_bytes": 25973886976, "host_cpu_utilization_pct": 35.8}
{"unix": 1790213220.041663, "gpu_temperature_c": 56, "gpu_power_mw": 31785, "gpu_clock_sm_mhz": 1957, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 20, "gpu_memory_used_bytes": 3627712512, "host_available_bytes": 25889935360, "host_cpu_utilization_pct": 24.4}
{"unix": 1790213221.083642, "gpu_temperature_c": 56, "gpu_power_mw": 28363, "gpu_clock_sm_mhz": 780, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 14, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25836867584, "host_cpu_utilization_pct": 26.7}
{"unix": 1790213222.1192544, "gpu_temperature_c": 56, "gpu_power_mw": 22691, "gpu_clock_sm_mhz": 705, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 4, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25782816768, "host_cpu_utilization_pct": 20.5}
{"unix": 1790213223.163932, "gpu_temperature_c": 56, "gpu_power_mw": 19904, "gpu_clock_sm_mhz": 802, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 6, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25722703872, "host_cpu_utilization_pct": 22.2}
{"unix": 1790213224.195849, "gpu_temperature_c": 56, "gpu_power_mw": 22283, "gpu_clock_sm_mhz": 885, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 11, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25714479104, "host_cpu_utilization_pct": 21.6}
{"unix": 1790213225.260356, "gpu_temperature_c": 56, "gpu_power_mw": 23106, "gpu_clock_sm_mhz": 885, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 11, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25688887296, "host_cpu_utilization_pct": 18.0}
{"unix": 1790213226.2967477, "gpu_temperature_c": 56, "gpu_power_mw": 22306, "gpu_clock_sm_mhz": 870, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 9, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25670811648, "host_cpu_utilization_pct": 20.6}
{"unix": 1790213227.3697987, "gpu_temperature_c": 56, "gpu_power_mw": 22520, "gpu_clock_sm_mhz": 870, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 10, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25707016192, "host_cpu_utilization_pct": 29.5}
{"unix": 1790213228.4130785, "gpu_temperature_c": 56, "gpu_power_mw": 21084, "gpu_clock_sm_mhz": 795, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 7, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25708683264, "host_cpu_utilization_pct": 26.6}
{"unix": 1790213229.4297194, "gpu_temperature_c": 56, "gpu_power_mw": 18147, "gpu_clock_sm_mhz": 847, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 10, "gpu_memory_used_bytes": 3660677120, "host_available_bytes": 25703862272, "host_cpu_utilization_pct": 20.7}
{"unix": 1790213230.502961, "gpu_temperature_c": 56, "gpu_power_mw": 20550, "gpu_clock_sm_mhz": 787, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 3677454336, "host_available_bytes": 25680760832, "host_cpu_utilization_pct": 17.8}
{"unix": 1790213231.5587397, "gpu_temperature_c": 56, "gpu_power_mw": 26204, "gpu_clock_sm_mhz": 1860, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 4202303488, "host_available_bytes": 25402011648, "host_cpu_utilization_pct": 32.5}
{"unix": 1790213232.6357276, "gpu_temperature_c": 57, "gpu_power_mw": 40989, "gpu_clock_sm_mhz": 2085, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 16, "gpu_memory_used_bytes": 6348148736, "host_available_bytes": 23917826048, "host_cpu_utilization_pct": 34.7}
{"unix": 1790213233.7703621, "gpu_temperature_c": 57, "gpu_power_mw": 33395, "gpu_clock_sm_mhz": 1935, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 6359617536, "host_available_bytes": 24904699904, "host_cpu_utilization_pct": 37.7}
{"unix": 1790213234.9093769, "gpu_temperature_c": 56, "gpu_power_mw": 22149, "gpu_clock_sm_mhz": 1725, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 6208557056, "host_available_bytes": 24895795200, "host_cpu_utilization_pct": 14.6}
{"unix": 1790213236.0116951, "gpu_temperature_c": 56, "gpu_power_mw": 21971, "gpu_clock_sm_mhz": 1717, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 6171332608, "host_available_bytes": 24919101440, "host_cpu_utilization_pct": 13.6}
{"unix": 1790213237.1057372, "gpu_temperature_c": 56, "gpu_power_mw": 23349, "gpu_clock_sm_mhz": 1792, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 4, "gpu_memory_used_bytes": 6613344256, "host_available_bytes": 24639811584, "host_cpu_utilization_pct": 23.4}
{"unix": 1790213238.2306685, "gpu_temperature_c": 56, "gpu_power_mw": 27090, "gpu_clock_sm_mhz": 1852, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 7309598720, "host_available_bytes": 23933394944, "host_cpu_utilization_pct": 32.7}
{"unix": 1790213239.3501265, "gpu_temperature_c": 58, "gpu_power_mw": 51753, "gpu_clock_sm_mhz": 2287, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 23, "gpu_memory_used_bytes": 9271558144, "host_available_bytes": 23618150400, "host_cpu_utilization_pct": 24.8}
{"unix": 1790213240.3548572, "gpu_temperature_c": 57, "gpu_power_mw": 32660, "gpu_clock_sm_mhz": 1717, "gpu_clock_memory_mhz": 14001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 9871343616, "host_available_bytes": 23575003136, "host_cpu_utilization_pct": 15.1}
{"unix": 1790213241.3781776, "gpu_temperature_c": 57, "gpu_power_mw": 29482, "gpu_clock_sm_mhz": 1717, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 9871343616, "host_available_bytes": 23580078080, "host_cpu_utilization_pct": 17.9}
{"unix": 1790213242.3898137, "gpu_temperature_c": 56, "gpu_power_mw": 21525, "gpu_clock_sm_mhz": 1717, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 9871343616, "host_available_bytes": 23433216000, "host_cpu_utilization_pct": 23.1}
{"unix": 1790213243.3960268, "gpu_temperature_c": 56, "gpu_power_mw": 22147, "gpu_clock_sm_mhz": 1717, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 9877635072, "host_available_bytes": 23356010496, "host_cpu_utilization_pct": 18.9}
{"unix": 1790213244.4523852, "gpu_temperature_c": 56, "gpu_power_mw": 22588, "gpu_clock_sm_mhz": 495, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 9866813440, "host_available_bytes": 23321116672, "host_cpu_utilization_pct": 28.0}
{"unix": 1790213245.5123162, "gpu_temperature_c": 56, "gpu_power_mw": 22293, "gpu_clock_sm_mhz": 892, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 7, "gpu_memory_used_bytes": 9868910592, "host_available_bytes": 23358095360, "host_cpu_utilization_pct": 23.2}
{"unix": 1790213246.5181894, "gpu_temperature_c": 56, "gpu_power_mw": 47864, "gpu_clock_sm_mhz": 592, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 14103060480, "host_available_bytes": 23275102208, "host_cpu_utilization_pct": 19.9}
{"unix": 1790213247.5792825, "gpu_temperature_c": 57, "gpu_power_mw": 38254, "gpu_clock_sm_mhz": 1672, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 10, "gpu_memory_used_bytes": 14151294976, "host_available_bytes": 23046995968, "host_cpu_utilization_pct": 19.4}
{"unix": 1790213248.6410213, "gpu_temperature_c": 56, "gpu_power_mw": 19941, "gpu_clock_sm_mhz": 442, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 14153392128, "host_available_bytes": 23047921664, "host_cpu_utilization_pct": 10.7}
{"unix": 1790213249.655187, "gpu_temperature_c": 56, "gpu_power_mw": 20953, "gpu_clock_sm_mhz": 495, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 14146314240, "host_available_bytes": 23050997760, "host_cpu_utilization_pct": 21.0}
{"unix": 1790213250.6775556, "gpu_temperature_c": 56, "gpu_power_mw": 20395, "gpu_clock_sm_mhz": 510, "gpu_clock_memory_mhz": 810, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 2, "gpu_memory_used_bytes": 14146314240, "host_available_bytes": 23054508032, "host_cpu_utilization_pct": 14.3}
{"unix": 1790213251.7742314, "gpu_temperature_c": 56, "gpu_power_mw": 15782, "gpu_clock_sm_mhz": 472, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 2, "gpu_memory_used_bytes": 14146314240, "host_available_bytes": 23066025984, "host_cpu_utilization_pct": 12.8}
{"unix": 1790213252.7879412, "gpu_temperature_c": 55, "gpu_power_mw": 15875, "gpu_clock_sm_mhz": 480, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14146314240, "host_available_bytes": 23091122176, "host_cpu_utilization_pct": 12.3}
{"unix": 1790213253.832425, "gpu_temperature_c": 55, "gpu_power_mw": 14752, "gpu_clock_sm_mhz": 570, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 14149591040, "host_available_bytes": 23080730624, "host_cpu_utilization_pct": 14.9}
{"unix": 1790213254.936086, "gpu_temperature_c": 55, "gpu_power_mw": 17915, "gpu_clock_sm_mhz": 562, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 14149828608, "host_available_bytes": 23076450304, "host_cpu_utilization_pct": 16.0}
{"unix": 1790213255.9399188, "gpu_temperature_c": 55, "gpu_power_mw": 16551, "gpu_clock_sm_mhz": 532, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 4, "gpu_memory_used_bytes": 14149828608, "host_available_bytes": 23093075968, "host_cpu_utilization_pct": 14.4}
{"unix": 1790213256.955558, "gpu_temperature_c": 55, "gpu_power_mw": 16688, "gpu_clock_sm_mhz": 540, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14183448576, "host_available_bytes": 22961127424, "host_cpu_utilization_pct": 16.6}
{"unix": 1790213257.9702487, "gpu_temperature_c": 55, "gpu_power_mw": 16743, "gpu_clock_sm_mhz": 480, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14183448576, "host_available_bytes": 22828597248, "host_cpu_utilization_pct": 17.0}
{"unix": 1790213259.0051956, "gpu_temperature_c": 55, "gpu_power_mw": 19517, "gpu_clock_sm_mhz": 645, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 14183448576, "host_available_bytes": 22954094592, "host_cpu_utilization_pct": 20.2}
{"unix": 1790213260.1248965, "gpu_temperature_c": 55, "gpu_power_mw": 18757, "gpu_clock_sm_mhz": 630, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 5, "gpu_memory_used_bytes": 14183448576, "host_available_bytes": 22931677184, "host_cpu_utilization_pct": 12.4}
{"unix": 1790213261.1317537, "gpu_temperature_c": 55, "gpu_power_mw": 23817, "gpu_clock_sm_mhz": 982, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 6, "gpu_memory_used_bytes": 14185545728, "host_available_bytes": 22937169920, "host_cpu_utilization_pct": 12.1}
{"unix": 1790213262.190852, "gpu_temperature_c": 55, "gpu_power_mw": 17946, "gpu_clock_sm_mhz": 570, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 4, "gpu_memory_used_bytes": 14185545728, "host_available_bytes": 22944727040, "host_cpu_utilization_pct": 12.5}
{"unix": 1790213263.2524807, "gpu_temperature_c": 55, "gpu_power_mw": 18175, "gpu_clock_sm_mhz": 615, "gpu_clock_memory_mhz": 405, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14185545728, "host_available_bytes": 22910607360, "host_cpu_utilization_pct": 20.8}
{"unix": 1790213264.2618542, "gpu_temperature_c": 58, "gpu_power_mw": 49257, "gpu_clock_sm_mhz": 1912, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 31, "gpu_memory_used_bytes": 14233780224, "host_available_bytes": 22751629312, "host_cpu_utilization_pct": 17.0}
{"unix": 1790213265.2645442, "gpu_temperature_c": 55, "gpu_power_mw": 20959, "gpu_clock_sm_mhz": 510, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 14233780224, "host_available_bytes": 22716305408, "host_cpu_utilization_pct": 24.4}
{"unix": 1790213266.3272357, "gpu_temperature_c": 55, "gpu_power_mw": 30621, "gpu_clock_sm_mhz": 742, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14236147712, "host_available_bytes": 22694354944, "host_cpu_utilization_pct": 15.5}
{"unix": 1790213267.8191726, "gpu_temperature_c": 55, "gpu_power_mw": 29425, "gpu_clock_sm_mhz": 615, "gpu_clock_memory_mhz": 9001, "gpu_throttle_reasons": 0, "gpu_utilization_pct": 3, "gpu_memory_used_bytes": 14243291136, "host_available_bytes": 22747131904, "host_cpu_utilization_pct": 23.9}
{"unix": 1790213270.0743954, "gpu_temperature_c": 59, "gpu_power_mw": 65943, "gpu_clock_sm_mhz": 2212, "gpu_clock_memory_mhz": 10820, "gpu_throttle_reasons": 4, "gpu_utilization_pct": 87, "gpu_memory_used_bytes": 14247550976, "host_available_bytes": 22630240256, "host_cpu_utilization_pct": 27.7}
{"unix": 1790213271.0953062, "gpu_temperature_c": 58, "gpu_power_mw": 91914, "gpu_clock_sm_mhz": 1762, "gpu_clock_memory_mhz": 14001, "gpu_throttle_reasons": 4, "gpu_utilization_pct": 87, "gpu_memory_used_bytes": 14247550976, "host_available_bytes": 22642855936, "host_cpu_utilization_pct": 35.7}
{"unix": 1790213272.1024942, "gpu_temperature_c": 59, "gpu_power_mw": 103590, "gpu_clock_sm_mhz": 1815, "gpu_clock_memory_mhz": 14001, "gpu_throttle_reasons": 4, "gpu_utilization_pct": 87, "gpu_memory_used_bytes": 14242496512, "host_available_bytes": 22641094656, "host_cpu_utilization_pct": 29.1}
{"unix": 1790213273.1253207, "gpu_temperature_c": 57, "gpu_power_mw": 29238, "gpu_clock_sm_mhz": 472, "gpu_clock_memory_mhz": 10820, "gpu_throttle_reasons": 4, "gpu_utilization_pct": 1, "gpu_memory_used_bytes": 14244593664, "host_available_bytes": 22686302208, "host_cpu_utilization_pct": 53.5}
{"unix": 1790213274.1981916, "gpu_temperature_c": 57, "gpu_power_mw": 27460, "gpu_clock_sm_mhz": 660, "gpu_clock_memory_mhz": 14001, "gpu_throttle_reasons": 4, "gpu_utilization_pct": 2, "gpu_memory_used_bytes": 3666354176, "host_available_bytes": 26011693056, "host_cpu_utilization_pct": 21.9}
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_current_source_smoke/report.json
+302
View File
@@ -0,0 +1,302 @@
{
"status": "completed",
"argv": [
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\benchmarks\\edge_harness\\profile_local_tts.py",
"--model",
"C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"--out",
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_repo_smoke_detached_20260924",
"--repeats",
"0",
"--length-band",
"medium",
"--concurrency",
"1",
"--sustained-seconds",
"0",
"--gpu-telemetry-interval-s",
"1"
],
"python": "C:\\Users\\zhout\\w2\\omni029venv\\Scripts\\python.exe",
"cwd": "C:\\Users\\zhout\\w2",
"loaded_omni_path": "\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\vllm_omni\\__init__.py",
"host_environment": {
"VLLM_CUDART_SO_PATH": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\torch\\lib\\cudart64_13.dll",
"VLLM_CACHE_ROOT": "C:\\Users\\zhout\\c29",
"PYTHONUTF8": "1"
},
"platform": "Windows-11-10.0.26200-SP0",
"runtime": {
"python": "3.12.10",
"platform": "Windows-11-10.0.26200-SP0",
"torch": "2.13.0+cu130",
"vllm": "0.29.0",
"vllm_omni": "0.29.0rc2.dev71+g00510ce38.d20260922",
"vllm_omni_file": "\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\vllm_omni\\__init__.py",
"vllm_file": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\vllm\\__init__.py",
"checkouts": {}
},
"settings": {
"model": "C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"out": "\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_repo_smoke_detached_20260924",
"repeats": 0,
"sustained_seconds": 0.0,
"gpu_telemetry_interval_s": 1.0,
"length_band": "medium",
"concurrency": 1
},
"start_unix": 1790213199.4230263,
"requests": 3,
"cache_condition": "existing disk/JIT caches; not cold disk",
"quality_gate": "finite PCM checks only; ASR/speaker/listening reference gates pending",
"playback_scope": "arrival-based simulated playback; no sound-device measurement",
"terminal_audio": "DELTA output required; terminal audio is included once in complete playback metrics and archived separately",
"prompts": {
"short": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Hello, this is a speech test."
],
"language": [
"English"
],
"speaker": [
"Vivian"
],
"global_request_id": [
"after_cancel-short-1-3-8424a630"
]
}
},
"medium": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"The garden has trees, flowers, and a small pond. We walk along the path and listen to the birds."
],
"language": [
"English"
],
"speaker": [
"Vivian"
],
"global_request_id": [
"slow_consumer-medium-1-2-9ec16a39"
]
}
},
"long": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. "
],
"language": [
"English"
],
"speaker": [
"Vivian"
],
"global_request_id": [
"cancel-probe-a9894041"
]
}
}
},
"startup_s": 69.3754797000438,
"sampling_params": [
"SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.05, temperature=0.9, top_p=1.0, top_k=50, min_p=0.0, seed=None, stop=[], stop_token_ids=[2150], bad_words=[], thinking_token_budget=None, include_stop_str_in_output=False, ignore_eos=False, max_tokens=4096, min_tokens=2, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, structured_outputs=None, extra_args=None)",
"SamplingParams(n=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.0, top_p=1.0, top_k=0, min_p=0.0, seed=None, stop=[], stop_token_ids=[], bad_words=[], thinking_token_budget=None, include_stop_str_in_output=False, ignore_eos=False, max_tokens=65536, min_tokens=0, logprobs=None, prompt_logprobs=None, skip_special_tokens=True, spaces_between_special_tokens=True, structured_outputs=None, extra_args=None)"
],
"memory_after_load": {
"samples": 189,
"interval_s": 0.25,
"note": "peaks are lower bounds: anything shorter than interval_s is missed. nvml_device_used is the whole card including other processes; tree_pss includes a share of file-backed pages. These three counters are different accountings and must not be summed.",
"nvml_device_used_peak": 14243291136,
"nvml_device_used_min": 3566690304,
"tree_rss_peak": 6339624960,
"tree_pss_peak": 0,
"host_available_min": 22691536896
},
"sustained_wall_s": 0.0,
"cancel": {
"abort_ack_s": 0.007381099974736571,
"scope": "abort acknowledgement and subsequent request; late-event gate not verified"
},
"memory": {
"samples": 202,
"interval_s": 0.25,
"note": "peaks are lower bounds: anything shorter than interval_s is missed. nvml_device_used is the whole card including other processes; tree_pss includes a share of file-backed pages. These three counters are different accountings and must not be summed.",
"nvml_device_used_peak": 14247550976,
"nvml_device_used_min": 3566690304,
"tree_rss_peak": 6505013248,
"tree_pss_peak": 0,
"host_available_min": 22616309760
},
"gpu_telemetry": {
"path": "gpu_telemetry.jsonl",
"samples": 71,
"interval_s": 1.0,
"error": null,
"scope": "whole GPU/host; not attributable to this process; sampled values may miss peaks"
},
"end_unix": 1790213274.3247113
}
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_current_source_smoke/requests.jsonl
+3
View File
@@ -0,0 +1,3 @@
{"request_id": "warmup-medium-1-1", "phase": "warmup", "length_band": "medium", "submitted_unix": 1790213269.0747316, "finished_unix": 1790213271.118368, "concurrency": 1, "sr": 24000, "chunks_all": [{"idx": 0, "t_ms": 797.8608999401331, "samples": 3840, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "aeb469aa6da8e9d4476a94dbcb92a11a6050c1c91e81318fbbb7facdc8bd9f95", "finite": true}, {"idx": 1, "t_ms": 801.3278000289574, "samples": 7680, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "b5414113d4ca9d7b99de9160cdd651f685adb51b2d013cb46e62c13ca044512b", "finite": true}, {"idx": 2, "t_ms": 801.6816999297589, "samples": 15360, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "519adb0aeb2cf628c6ba6c45f64ddb08280d1d332487c57b262bdc9bdba306ab", "finite": true}, {"idx": 3, "t_ms": 1015.4290000209585, "samples": 30720, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "b533a446dc5bc4740437f0363da1ceb070e7cee387744320364166bd992b49ed", "finite": true}, {"idx": 4, "t_ms": 1309.366500005126, "samples": 48000, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "dcdf91dc053e6d19c242fb1dc140fac5fb880f497f5a2f062d60bbbee39408e3", "finite": true}, {"idx": 5, "t_ms": 1684.077400015667, "samples": 48000, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": false, "sha256": "351ad07df82c837a95cfd1c22cf553636b23ad47cb5c4a5bb42ffa106e30a6b1", "finite": true}, {"idx": 6, "t_ms": 2043.140499969013, "samples": 32640, "sr": 24000, "request_id": "warmup-medium-1-1", "terminal": true, "sha256": "5ba83aea7ab71c94851ae824f78c4e04282adbba74b93bc5d8726fe4a956b21d", "finite": true}], "finished": true, "terminal_audio_s": 1.36, "prefix_metrics": {"ttfa_ms": 797.8608999401331, "playback_start_ms": 797.8608999401331, "stall_at_ttfa_ms": 0, "n_chunks": 6, "streamed_audio_s": 6.4, "tail_audio_s": 0.0, "total_wall_s": 2.0438366999151185, "rtf_streamed": 0.263137093752448, "rtf_total": 0.31934948436173727, "gap_ms": [3.4669000888243318, 0.3538999008014798, 213.74730009119958, 293.9374999841675, 374.7109000105411]}, "ttfa_ms": 797.8608999401331, "playback_start_ms": 797.8608999401331, "stall_at_ttfa_ms": 0, "n_chunks": 7, "streamed_audio_s": 7.76, "tail_audio_s": 0.0, "total_wall_s": 2.0438366999151185, "rtf_streamed": 0.2632913015423986, "rtf_total": 0.2633810180302988, "gap_ms": [3.4669000888243318, 0.3538999008014798, 213.74730009119958, 293.9374999841675, 374.7109000105411, 359.06309995334595], "underruns": []}
{"request_id": "slow_consumer-medium-1-2", "phase": "slow_consumer", "length_band": "medium", "submitted_unix": 1790213271.1528645, "finished_unix": 1790213272.9485905, "concurrency": 1, "sr": 24000, "chunks_all": [{"idx": 0, "t_ms": 89.14759999606758, "samples": 3840, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "ed4c0ef2343900c8dcb8703ada282010c959e96285cd578d164289c1ff0cee18", "finite": true}, {"idx": 1, "t_ms": 349.19750003609806, "samples": 7680, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "4781bf3b3378db46727b07b9660389974750af34bf1583948e8ce007c5aeab8e", "finite": true}, {"idx": 2, "t_ms": 526.2864000396803, "samples": 15360, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "42a30cdc018a3b20313a0039361a4d65c84dbb5f955a516a132723e53ff31045", "finite": true}, {"idx": 3, "t_ms": 751.6142000677064, "samples": 30720, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "b6b9de0c46a33182b3316894ec734ecc96b920d4e398a1ef71facb1855b8f028", "finite": true}, {"idx": 4, "t_ms": 933.0112000461668, "samples": 48000, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "115d7794ee94e84f075ded1cb7849872c10a3bd3f7d0482a71d65a5f816aafe1", "finite": true}, {"idx": 5, "t_ms": 1188.879600027576, "samples": 48000, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": false, "sha256": "ef98b9f64d29bbe3ad3e4f47e0c31fe544f8eab4e0e8c3d6df4ffd5c3ba8e809", "finite": true}, {"idx": 6, "t_ms": 1606.7554000765085, "samples": 44160, "sr": 24000, "request_id": "slow_consumer-medium-1-2", "terminal": true, "sha256": "13374cff5bc5c52ba5f0d402d507e91c44200f0b593fa7501e27d75098654e38", "finite": true}], "finished": true, "terminal_audio_s": 1.84, "prefix_metrics": {"ttfa_ms": 89.14759999606758, "playback_start_ms": 189.19750003609806, "stall_at_ttfa_ms": 100.04990004003048, "n_chunks": 6, "streamed_audio_s": 6.4, "tail_audio_s": 0.0, "total_wall_s": 1.795640700031072, "rtf_streamed": 0.18576243750430876, "rtf_total": 0.280568859379855, "gap_ms": [260.0499000400305, 177.08890000358224, 225.3278000280261, 181.39699997846037, 255.8683999814093]}, "ttfa_ms": 89.14759999606758, "playback_start_ms": 189.19750003609806, "stall_at_ttfa_ms": 100.04990004003048, "n_chunks": 7, "streamed_audio_s": 8.24, "tail_audio_s": 0.0, "total_wall_s": 1.795640700031072, "rtf_streamed": 0.1949945873879258, "rtf_total": 0.2179175606833825, "gap_ms": [260.0499000400305, 177.08890000358224, 225.3278000280261, 181.39699997846037, 255.8683999814093, 417.87580004893243], "underruns": [[1, 100.04990004003048]]}
{"request_id": "after_cancel-short-1-3", "phase": "after_cancel", "length_band": "short", "submitted_unix": 1790213273.20442, "finished_unix": 1790213273.7606058, "concurrency": 1, "sr": 24000, "chunks_all": [{"idx": 0, "t_ms": 102.57940005976707, "samples": 3840, "sr": 24000, "request_id": "after_cancel-short-1-3", "terminal": false, "sha256": "c17a50ea811d252330c91d07a282a7e1fb43a94faf3f39481d10dffc6a1b5a93", "finite": true}, {"idx": 1, "t_ms": 163.28460001386702, "samples": 7680, "sr": 24000, "request_id": "after_cancel-short-1-3", "terminal": false, "sha256": "3dfd02017107564eb916b8a1ba9fd341f64186cdd63d7197f8124168f48acbc2", "finite": true}, {"idx": 2, "t_ms": 414.2688000574708, "samples": 15360, "sr": 24000, "request_id": "after_cancel-short-1-3", "terminal": false, "sha256": "4ccb566d55d1491c4a17fbe353487047334929b1a30850e69cb69df285b95fcd", "finite": true}, {"idx": 3, "t_ms": 460.2598000783473, "samples": 30720, "sr": 24000, "request_id": "after_cancel-short-1-3", "terminal": false, "sha256": "67c181988c188cf233128fc9b4ad207c1b571d65b0cafc71da1a68d0428805cb", "finite": true}, {"idx": 4, "t_ms": 555.4441000567749, "samples": 13440, "sr": 24000, "request_id": "after_cancel-short-1-3", "terminal": true, "sha256": "676b453f924e4b7cf74abe1ef9c9364b6b14402b8a5877ef342c1c250688c6dc", "finite": true}], "finished": true, "terminal_audio_s": 0.56, "prefix_metrics": {"ttfa_ms": 102.57940005976707, "playback_start_ms": 102.57940005976707, "stall_at_ttfa_ms": 0, "n_chunks": 4, "streamed_audio_s": 2.4, "tail_audio_s": 0.0, "total_wall_s": 0.5560630999971181, "rtf_streamed": 0.1917749166993114, "rtf_total": 0.23169295833213255, "gap_ms": [60.70519995409995, 250.98420004360378, 45.99100002087653]}, "ttfa_ms": 102.57940005976707, "playback_start_ms": 102.57940005976707, "stall_at_ttfa_ms": 0, "n_chunks": 5, "streamed_audio_s": 2.96, "tail_audio_s": 0.0, "total_wall_s": 0.5560630999971181, "rtf_streamed": 0.1876500338029645, "rtf_total": 0.1878591554044318, "gap_ms": [60.70519995409995, 250.98420004360378, 45.99100002087653, 95.18429997842759], "underruns": []}
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_startup_failures/stale_checkout.log.gz
BIN
View File
Binary file not shown.
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_startup_failures/stale_checkout_report.json
+273
View File
@@ -0,0 +1,273 @@
{
"status": "failed",
"argv": [
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\benchmarks\\edge_harness\\profile_local_tts.py",
"--model",
"C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"--out",
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_nativecwd_smoke_20260924",
"--repeats",
"0",
"--length-band",
"medium",
"--concurrency",
"1",
"--sustained-seconds",
"0",
"--gpu-telemetry-interval-s",
"1"
],
"python": "C:\\Users\\zhout\\w2\\omni029venv\\Scripts\\python.exe",
"host_environment": {
"VLLM_CUDART_SO_PATH": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\torch\\lib\\cudart64_13.dll",
"VLLM_CACHE_ROOT": "C:\\Users\\zhout\\c29",
"PYTHONUTF8": "1"
},
"platform": "Windows-11-10.0.26200-SP0",
"runtime": {
"python": "3.12.10",
"platform": "Windows-11-10.0.26200-SP0",
"torch": "2.13.0+cu130",
"vllm": "0.29.0",
"vllm_omni": "0.29.0",
"vllm_omni_file": "C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\__init__.py",
"vllm_file": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\vllm\\__init__.py",
"checkouts": {}
},
"settings": {
"model": "C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"out": "\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_nativecwd_smoke_20260924",
"repeats": 0,
"sustained_seconds": 0.0,
"gpu_telemetry_interval_s": 1.0,
"length_band": "medium",
"concurrency": 1
},
"start_unix": 1790212654.1643922,
"requests": 0,
"cache_condition": "existing disk/JIT caches; not cold disk",
"quality_gate": "finite PCM checks only; ASR/speaker/listening reference gates pending",
"playback_scope": "arrival-based simulated playback; no sound-device measurement",
"terminal_audio": "DELTA output required; terminal audio is included once in complete playback metrics and archived separately",
"prompts": {
"short": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Hello, this is a speech test."
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
},
"medium": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"The garden has trees, flowers, and a small pond. We walk along the path and listen to the birds."
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
},
"long": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. "
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
}
},
"error": "RuntimeError(\"Orchestrator initialization failed: Engine core initialization failed. See root cause above. Failed proc(s): {'StageEngineCoreProc_stage0_replica0': 1}\")",
"traceback": "Traceback (most recent call last):\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\threading.py\", line 1075, in _bootstrap_inner\n self.run()\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\threading.py\", line 1012, in run\n self._target(*self._args, **self._kwargs)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 462, in _bootstrap_orchestrator\n loop.run_until_complete(_run_orchestrator())\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\asyncio\\base_events.py\", line 691, in run_until_complete\n return future.result()\n ^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 419, in _run_orchestrator\n self._initialize_stages(stage_init_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 345, in _initialize_stages\n self._initialize_stages_impl(stage_init_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 366, in _initialize_stages_impl\n self._runtime.initialize()\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 264, in initialize\n raise exc\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 245, in initialize\n initialized_clients = self._initialize_stage_replicas(stage_plans, self._stage_init_timeout)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 517, in _initialize_stage_replicas\n raise primary_exc\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 478, in _init_group\n client = self._initialize_replica(\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 743, in _initialize_replica\n return self._initialize_local_llm_replica(plan, stage_init_timeout)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 849, in _initialize_local_llm_replica\n launch_stack.close()\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 618, in close\n self.__exit__(None, None, None)\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 610, in __exit__\n raise exc_details[1]\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 595, in __exit__\n if cb(*exc_details):\n ^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 144, in __exit__\n next(self.gen)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_engine_startup.py\", line 1211, in launch_stage_replica\n wait_for_engine_startup(\n File \"C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\vllm\\v1\\engine\\utils.py\", line 1307, in wait_for_engine_startup\n raise RuntimeError(\nRuntimeError: Engine core initialization failed. See root cause above. Failed proc(s): {'StageEngineCoreProc_stage0_replica0': 1}\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\benchmarks\\edge_harness\\profile_local_tts.py\", line 86, in run\n omni = AsyncOmni(model=args.model, deploy_profile=\"edge\", stage_init_timeout=180)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\entrypoints\\async_omni.py\", line 156, in __init__\n OmniBase.__init__(self, model=model, **kwargs)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\entrypoints\\omni_base.py\", line 223, in __init__\n self.engine = AsyncOmniEngine(\n ^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 300, in __init__\n self._wait_for_orchestrator_init(startup_future, startup_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 533, in _wait_for_orchestrator_init\n startup_future.result(\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\concurrent\\futures\\_base.py\", line 456, in result\n return self.__get_result()\n ^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\concurrent\\futures\\_base.py\", lLine truncated
"memory": {
"samples": 72,
"interval_s": 0.25,
"note": "peaks are lower bounds: anything shorter than interval_s is missed. nvml_device_used is the whole card including other processes; tree_pss includes a share of file-backed pages. These three counters are different accountings and must not be summed.",
"nvml_device_used_peak": 3275038720,
"nvml_device_used_min": 3275038720,
"tree_rss_peak": 3044057088,
"tree_pss_peak": 0,
"host_available_min": 28737007616
},
"gpu_telemetry": {
"path": "gpu_telemetry.jsonl",
"samples": 24,
"interval_s": 1.0,
"error": null,
"scope": "whole GPU/host; not attributable to this process; sampled values may miss peaks"
},
"end_unix": 1790212678.0856693
}
benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/windows_startup_failures/unc_cwd_report.json
+273
View File
@@ -0,0 +1,273 @@
{
"status": "failed",
"argv": [
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\benchmarks\\edge_harness\\profile_local_tts.py",
"--model",
"C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"--out",
"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_telemetry_20260924",
"--repeats",
"1",
"--length-band",
"medium",
"--concurrency",
"1",
"--sustained-seconds",
"1800",
"--gpu-telemetry-interval-s",
"1"
],
"python": "C:\\Users\\zhout\\w2\\omni029venv\\Scripts\\python.exe",
"host_environment": {
"VLLM_CUDART_SO_PATH": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\torch\\lib\\cudart64_13.dll",
"VLLM_CACHE_ROOT": "C:\\Users\\zhout\\c29",
"PYTHONUTF8": "1"
},
"platform": "Windows-11-10.0.26200-SP0",
"runtime": {
"python": "3.12.10",
"platform": "Windows-11-10.0.26200-SP0",
"torch": "2.13.0+cu130",
"vllm": "0.29.0",
"vllm_omni": "0.29.0",
"vllm_omni_file": "C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\__init__.py",
"vllm_file": "C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\vllm\\__init__.py",
"checkouts": {}
},
"settings": {
"model": "C:\\Users\\zhout\\w2\\models\\Qwen3-TTS-0.6B-85e237c",
"out": "\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\analysis\\experiments\\tts_windows_telemetry_20260924",
"repeats": 1,
"sustained_seconds": 1800.0,
"gpu_telemetry_interval_s": 1.0,
"length_band": "medium",
"concurrency": 1
},
"start_unix": 1790211568.3760028,
"requests": 0,
"cache_condition": "existing disk/JIT caches; not cold disk",
"quality_gate": "finite PCM checks only; ASR/speaker/listening reference gates pending",
"playback_scope": "arrival-based simulated playback; no sound-device measurement",
"terminal_audio": "DELTA output required; terminal audio is included once in complete playback metrics and archived separately",
"prompts": {
"short": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Hello, this is a speech test."
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
},
"medium": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"The garden has trees, flowers, and a small pond. We walk along the path and listen to the birds."
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
},
"long": {
"prompt_token_ids": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"additional_information": {
"task_type": [
"CustomVoice"
],
"text": [
"Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. Today we are testing a local speech system. It should speak clearly and keep the audio flowing smoothly. "
],
"language": [
"English"
],
"speaker": [
"Vivian"
]
}
}
},
"error": "RuntimeError(\"Orchestrator initialization failed: Engine core initialization failed. See root cause above. Failed proc(s): {'StageEngineCoreProc_stage1_replica0': 1}\")",
"traceback": "Traceback (most recent call last):\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\threading.py\", line 1075, in _bootstrap_inner\n self.run()\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\threading.py\", line 1012, in run\n self._target(*self._args, **self._kwargs)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 462, in _bootstrap_orchestrator\n loop.run_until_complete(_run_orchestrator())\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\asyncio\\base_events.py\", line 691, in run_until_complete\n return future.result()\n ^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 419, in _run_orchestrator\n self._initialize_stages(stage_init_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 345, in _initialize_stages\n self._initialize_stages_impl(stage_init_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 366, in _initialize_stages_impl\n self._runtime.initialize()\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 264, in initialize\n raise exc\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 245, in initialize\n initialized_clients = self._initialize_stage_replicas(stage_plans, self._stage_init_timeout)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 517, in _initialize_stage_replicas\n raise primary_exc\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 478, in _init_group\n client = self._initialize_replica(\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 743, in _initialize_replica\n return self._initialize_local_llm_replica(plan, stage_init_timeout)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_runtime.py\", line 849, in _initialize_local_llm_replica\n launch_stack.close()\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 618, in close\n self.__exit__(None, None, None)\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 610, in __exit__\n raise exc_details[1]\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 595, in __exit__\n if cb(*exc_details):\n ^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\contextlib.py\", line 144, in __exit__\n next(self.gen)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\stage_engine_startup.py\", line 1211, in launch_stage_replica\n wait_for_engine_startup(\n File \"C:\\Users\\zhout\\w2\\omni029venv\\Lib\\site-packages\\vllm\\v1\\engine\\utils.py\", line 1307, in wait_for_engine_startup\n raise RuntimeError(\nRuntimeError: Engine core initialization failed. See root cause above. Failed proc(s): {'StageEngineCoreProc_stage1_replica0': 1}\n\nThe above exception was the direct cause of the following exception:\n\nTraceback (most recent call last):\n File \"\\\\wsl.localhost\\Ubuntu\\home\\zhout\\project\\edge_infer\\vllm-omni-edge\\benchmarks\\edge_harness\\profile_local_tts.py\", line 86, in run\n omni = AsyncOmni(model=args.model, deploy_profile=\"edge\", stage_init_timeout=180)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\entrypoints\\async_omni.py\", line 156, in __init__\n OmniBase.__init__(self, model=model, **kwargs)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\entrypoints\\omni_base.py\", line 223, in __init__\n self.engine = AsyncOmniEngine(\n ^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 300, in __init__\n self._wait_for_orchestrator_init(startup_future, startup_timeout)\n File \"C:\\Users\\zhout\\w2\\vllm-omni-src-0.29.0\\vllm_omni\\engine\\async_omni_engine.py\", line 533, in _wait_for_orchestrator_init\n startup_future.result(\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\concurrent\\futures\\_base.py\", line 456, in result\n return self.__get_result()\n ^^^^^^^^^^^^^^^^^^^\n File \"C:\\Users\\zhout\\AppData\\Local\\Programs\\Python\\Python312\\Lib\\concurrent\\futures\\_base.py\", lLine truncated
"memory": {
"samples": 2681,
"interval_s": 0.25,
"note": "peaks are lower bounds: anything shorter than interval_s is missed. nvml_device_used is the whole card including other processes; tree_pss includes a share of file-backed pages. These three counters are different accountings and must not be summed.",
"nvml_device_used_peak": 3278577664,
"nvml_device_used_min": 3262193664,
"tree_rss_peak": 3045744640,
"tree_pss_peak": 0,
"host_available_min": 26667917312
},
"gpu_telemetry": {
"path": "gpu_telemetry.jsonl",
"samples": 893,
"interval_s": 1.0,
"error": null,
"scope": "whole GPU/host; not attributable to this process; sampled values may miss peaks"
},
"end_unix": 1790212492.9256425
}
benchmarks/edge_harness/results/e2e_profiling_20260922/evidence/summary/README.md
+3 -3
View File
@@ -18,7 +18,7 @@ The identical SA/RB3 TFLite binary matched original ONNX CPU within 3.45e-6 rela
A separate [Qwen3-TTS talker decode-step attempt](../../../e2e_expansion_20260924/evidence/qwen_tts_talker_qualcomm/README.md) recovered a 28-layer ONNX export and synthetic fixed-cache fixture. Historical FP16 QNN context-binary inference executed on S24, S25, SA8775P and X Elite, but all four changed the CPU reference top-1 speech token with 8.57–9.43% logits relative L2; they are numerically unqualified on this fixture. An RB3 compile rejected floating-point input. Separate CPU-requested talker steps on X Elite ONNX and SA8775P TFLite preserved CPU token 80 with logits relative L2 2.07e-5 and 3.33e-5; X Elite had 100 component samples at p50/p95 191.414/309.428 ms with all 1,744 execution-detail rows on CPU; SA had 100 at 52.635/76.925 ms with all 1,800 rows on CPU. No complete talker loop or TTS stream follows.
A [Galaxy S25 InternVLA Cosmos attempt](../../../e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md) packaged the real Place_Markpen image encoder into one ONNX file after two CPU parity fixtures. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one pattern; its 100-sample six-frame component profile had p50/p95 49.665/51.382 ms and all 121 execution-detail rows on NPU. Injecting that latent into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. The same encoder compiled to FP16 QNN DLC for X Elite CRD and SA8775P; same-fixture NPU-requested inferences returned finite latents at 0.0608% and 0.0613% relative L2 versus source, with X Elite profiled on NPU and SA8775P placement pending. This advances component and action-sensitivity evidence, not device-local full-policy qualification. The Place_Markpen checkpoint names dataset a2d_real/a2d_pen_holder, which returned HTTP 404 under the current authenticated Hugging Face account. The official vLLM-Omni example names a different Place_Markpen archive in InternData-A1, but access to that gated dataset returned HTTP 403; [the access record](../../../e2e_expansion_20260924/evidence/internvla_real_observation_access/README.md) leaves real-observation reference-action validation open.
A [Galaxy S25 InternVLA Cosmos attempt](../../../e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md) packaged the real Place_Markpen image encoder into one ONNX file after two CPU parity fixtures. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one pattern; its 100-sample six-frame component profile had p50/p95 49.665/51.382 ms and all 121 execution-detail rows on NPU. Injecting that latent into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. The same encoder compiled to FP16 QNN DLC for X Elite CRD and SA8775P; same-fixture NPU-requested inferences returned finite latents at 0.0608% and 0.0613% relative L2 versus source, with both X Elite and SA8775P profiled on NPU. This advances component and action-sensitivity evidence, not device-local full-policy qualification. The Place_Markpen checkpoint names dataset a2d_real/a2d_pen_holder, which returned HTTP 404 under the current authenticated Hugging Face account. The official vLLM-Omni example names a different Place_Markpen archive in InternData-A1, but access to that gated dataset returned HTTP 403; [the access record](../../../e2e_expansion_20260924/evidence/internvla_real_observation_access/README.md) leaves real-observation reference-action validation open.
A [MiniCPM-o speech-head study](../../../e2e_expansion_20260924/evidence/minicpmo_tts_qualcomm/README.md) recovered the historical 20-layer source and matching S24/S25 fixtures. Both FP16 context-binary outputs preserved the ONNX CPU top-1 speech token on this one decode step with logits relative L2 of 0.346%/0.360%. S24/S25 100-sample component profiles attributed all 1,029 execution-detail rows to NPU, with p50/p95 8.537/9.006 ms and 7.197/7.615 ms. New FP16 QNN DLC compiles passed for SA8775P, X Elite and RB3. SA8775P same-fixture inference preserved CPU top-1 token 1867 with 0.3568% logits relative L2; its 100-sample component profile measured p50/p95 11.901/13.038 ms with all 944 execution-detail rows on NPU. X Elite same-fixture inference also preserved token 1867 with 0.3170% logits relative L2; its 100-sample component profile measured p50/p95 9.905/11.037 ms with all 944 execution-detail rows on NPU; RB3 FP16 QNN inference failed at graph load; its explicit TFLite CPU fallback returned 41 finite outputs, preserved CPU top-1 token 1867 and had logits relative L2 1.26e-6 versus ONNX CPU, with 100 CPU-requested component samples at p50/p95 59.518/71.347 ms and all 1,008 execution-detail rows on CPU. These are speech-head component results, not multimodal MiniCPM-o requests.
@@ -46,7 +46,7 @@ NOT E2E means no complete device/model workload was verified. “Scoped E2E” i
| PC Snapdragon X Elite CRD / AI Hub | NOT E2E (one Spark attention layer on NPU) | NOT E2E (CPU talker/vocoder steps; QNN vocoder numeric) | NOT E2E | NOT E2E (NPU speech-head step profiled) | NOT E2E (Cosmos NPU encoder profiled) |
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | NOT E2E (one Spark attention layer on NPU) | NOT E2E (FP16 talker token mismatch) | NOT E2E | NOT E2E (NPU speech head step profiled) | NOT E2E (Cosmos NPU encoder profiled) |
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | NOT E2E (one Spark attention layer on NPU) | NOT E2E (FP16 talker token mismatch) | NOT E2E | NOT E2E (NPU speech head step profiled) | NOT E2E |
| Embedded SA8775P ADP / AI Hub | NOT E2E (one Spark attention layer on NPU) | NOT E2E (CPU talker/vocoder steps; NPU vocoder failed) | NOT E2E | NOT E2E (NPU speech-head step profiled) | NOT E2E (Cosmos encoder numeric; placement pending) |
| Embedded SA8775P ADP / AI Hub | NOT E2E (one Spark attention layer on NPU) | NOT E2E (CPU talker/vocoder steps; NPU vocoder failed) | NOT E2E | NOT E2E (NPU speech-head step profiled) | NOT E2E (Cosmos NPU encoder profiled) |
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | NOT E2E (Spark CPU layer passed; W8A8 NPU quality open) | NOT E2E (mixed vocoder numeric unqualified) | NOT E2E | NOT E2E (CPU speech-head step profiled) | NOT E2E |
## Per-cell reasons and next steps
@@ -107,7 +107,7 @@ NOT E2E means no complete device/model workload was verified. “Scoped E2E” i
| Embedded SA8775P ADP / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_CPU_VOCODER_PARITY_AND_PLACEMENT_PASS; TALKER_TFLITE_CPU_NUMERIC_PROFILED; QNN_NPU_VOCODER_INFERENCE_FAILED; COMPONENT_GPU_NUMERIC_FAIL; TALKER_FP16_NUMERIC_FAIL; NOT_E2E_PROFILEABLE | On the exact SA8775P, the same TFLite binary's CPU-requested Code2Wav inference matched original ONNX CPU within 4.13e-6 relative L2 / 107.67 dB SNR, without saturation. Its 100-sample CPU-requested component profile had p50/p95 5.384/5.534 s with all 738 execution-detail rows on CPU; the service estimated-time field 5.250 s is the sample minimum. The GPU-requested output saturated 96.63% of samples with 11.832 relative L2; its fresh 100-sample mixed CPU/GPU profile p50/p95 was 1.776/1.880 s, timing failed audio. Even with FP32-preserving GPU options, 84.89% of samples saturated with 11.389 relative L2. The 28-layer FP16 QNN talker step executed but changed CPU top-1 token 80 to 1720 with 9.43% logits relative L2. A separate TFLite CPU-requested talker step preserved CPU top-1 token 80 with logits relative L2 3.33e-5 and maximum K/V relative L2 3.41e-5. Its 100-sample component profile had p50/p95 52.635/76.925 ms with all 1,800 execution-detail rows on CPU. A separate FP16 QNN DLC vocoder compile passed, but same-fixture NPU-requested inference failed after compilation without a waveform or placement samples. No complete pipeline exists. | Isolate the FP16 QNN vocoder runtime failure, GPU delegate failure and FP16 talker numerical drift, then establish listening quality, co-resident streaming, memory and sustained behavior; [vocoder](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md) and [talker](../../../e2e_expansion_20260924/evidence/qwen_tts_talker_qualcomm/README.md) evidence. |
| Embedded SA8775P ADP / AI Hub | Qwen3.8-27B | NOT_E2E_PROFILEABLE | No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device. | Supply and validate the missing matching artifact/backend, then run the complete specified workload. |
| Embedded SA8775P ADP / AI Hub | MiniCPM-o 4.5 | COMPONENT_SPEECH_HEAD_NPU_NUMERIC_PROFILED; NOT_E2E_PROFILEABLE | The real 20-layer MiniCPM-o speech-head FP16 QNN DLC ran on the exact SA8775P ADP. Same-fixture inference returned 41 finite tensors, preserved ONNX CPU top-1 token 1867, and had 0.3568% logits relative L2 and 1.225% maximum K/V relative L2. Its 100-sample component profile had nearest-rank p50/p95 11.901/13.038 ms, with all 944 execution-detail rows on NPU. Reported inference peak memory was 134,696,960 bytes and first-load peak 1,322,835,968 bytes; no complete thinker/encoders/vocoder bundle or device-local stateful controller exists. | Validate complete multimodal generation, state, speech quality, co-resident memory, cancellation and sustained power; [SA8775P raw evidence](../../../e2e_expansion_20260924/evidence/minicpmo_tts_qualcomm/README.md). |
| Embedded SA8775P ADP / AI Hub | InternVLA-A1 | COMPONENT_COSMOS_NUMERIC_MEASURED; PLACEMENT_PENDING; NOT_E2E_PROFILEABLE | The real Place_Markpen Cosmos FP16 QNN image encoder returned a finite FP32 latent for the six-frame synthetic pattern on exact SA8775P ADP. Against the pinned source, relative L2 was 0.0613%, SNR 64.25 dB and cosine 0.999999812. The NPU-requested route awaits actual node-placement/timing profile. No real-observation action tolerance, complete policy or local stateful Omni backend exists. | Audit the [SA8775P Cosmos placement profile](../../../e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md), then validate real observations, action units/timing, state, admission, cancellation and sustained behavior. |
| Embedded SA8775P ADP / AI Hub | InternVLA-A1 | COMPONENT_COSMOS_NPU_NUMERIC_PROFILED; NOT_E2E_PROFILEABLE | The real Place_Markpen Cosmos FP16 QNN image encoder returned a finite FP32 latent for the six-frame synthetic pattern on exact SA8775P ADP. Against the pinned source, relative L2 was 0.0613%, SNR 64.25 dB and cosine 0.999999812. The 100-sample six-frame encoder profile had p50/p95 116.021/116.327 ms, with all 121 execution-detail rows on NPU. Reported component inference peak memory was 116,924,416 bytes and first-load peak 659,845,120 bytes. No real-observation action tolerance, complete policy or local stateful Omni backend exists. | Validate real observations, action units/timing, state, admission, cancellation and sustained behavior; [SA8775P Cosmos raw evidence](../../../e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md). |
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | Spark-X2.5 | COMPONENT_CPU_PARITY_PASS; COMPONENT_NPU_NUMERIC_UNQUALIFIED; NOT_E2E_PROFILEABLE | On the exact hosted Dragonwing RB3 Gen 2 Vision Kit (QCS6490, QC Linux 1.6), the reused S24 W8A16 QNN DLC and a separately RB3-compiled W8A16 DLC each failed inference/profile at QnnModel_composeGraphsFromDlc: MODEL_GRAPH_ERROR. An FP32 ONNX one-layer CPU route passed with all 56 profile rows on CPU, source-relative hidden/K/V L2 of 8.73e-7/2.98e-7/3.37e-7 and 100-sample component p50/p95 19.304/21.113 ms. A calibrated W8A8 QNN DLC ran with all 61 rows on NPU and p50/p95 5.610/5.761 ms, but hidden/K/V relative L2 was 20.914%/1.702%/4.413% versus FP32 CPU, without a next-token quality gate. These are different precisions, not a qualified speedup; no complete 28-layer prefill/decode/KV-ring/sampling, local Omni controller or complete-request timing exists. | Improve representative W8A8 calibration or find another compatible quality-passing artifact; validate token parity, then full device-resident generation, admission/cancel and sustained power; [RB3 raw evidence](../../../e2e_expansion_20260923/evidence/spark_rb3_full_attention/README.md). |
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_MIXED_VOCODER_PROFILED_NUMERIC_UNQUALIFIED; NOT_E2E_PROFILEABLE | Fresh same-fixture TFLite GPU-requested vocoder inference returned a finite unsaturated waveform, but relative L2 versus original ONNX CPU was 15.73% (16.06 dB SNR), without a listening-quality tolerance; it is not a qualified audio path. The fresh 100-sample component profile had nearest-rank p50/p95 3.099/3.133 s and 403 CPU plus 335 GPU execution-detail rows. Its 3.070 s estimated-time field is the sample minimum, not median. The historical NPU talker build was rejected for floating-point inputs; no full model pipeline exists. | Establish waveform quality and a compatible integer talker artifact, then validate complete local execution, memory and sustained behavior; [vocoder evidence](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md). |
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | Qwen3.8-27B | NOT_E2E_PROFILEABLE | No complete qualified target artifact and device-local stateful Omni backend. Hub jobs provide component execution, not an on-device generation/streaming controller; ADB has no attached device. | Supply and validate the missing matching artifact/backend, then run the complete specified workload. |
docs/design/omni_edge_e2e_roadmap.md
+1 -1
View File
@@ -25,7 +25,7 @@ For M4, a [MiniCPM-o 4.5 speech-head study](../../benchmarks/edge_harness/result
The [native Windows Qwen3-TTS sustained audit](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_sustained_drift/README.md) identifies an additional M2 desktop blocker. The historical RTX 5090 Laptop complete-model stream changed regime around minute 26–27; all 17 requests in minutes 27–30 had simulated playback underruns and median RTF 1.331. The separate WSL run had no sustained underruns. Existing logs do not identify the cause; timestamped GPU clocks, power, temperature and throttling are now available in the local profiler for a controlled repeat. Native Windows sustained playback remains unqualified.
For M4, a [Galaxy S25 Cosmos encoder attempt](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md) repackages the real Place_Markpen FP32 ONNX export after retaining two CPU parity fixtures. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one synthetic pattern. Injecting that output into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. S25 six-frame component profile returned 100 samples at p50/p95 49.665/51.382 ms, with all 121 execution-detail rows on NPU; the action policy did not run on S25. The same encoder compiled to FP16 QNN DLC for X Elite CRD; same-fixture NPU-requested inference returned a finite latent at 0.0608% relative L2 versus source, with its 100-sample six-frame profile at p50/p95 82.939/83.837 ms and all 121 execution-detail rows on NPU. An SA8775P FP16 QNN DLC inference returned a finite same-fixture latent at 0.0613% relative L2 versus source; placement profiling is pending. This is component/action-sensitivity evidence, not device-local M4 qualification. The checkpoint train config names a2d_real/a2d_pen_holder; an [authenticated access check](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_real_observation_access/README.md) returned HTTP 404. The official vLLM-Omni example names an InternData-A1 Place_Markpen archive, but that dataset returned HTTP 403 gated-access denial under the current account, so real-observation reference-action quality is still unmeasured.
For M4, a [Galaxy S25 Cosmos encoder attempt](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md) repackages the real Place_Markpen FP32 ONNX export after retaining two CPU parity fixtures. FLOAT16 QNN DLC inference returned a latent within 0.0524% relative L2 of the source on one synthetic pattern. Injecting that output into the real policy on local CPU changed synthetic actions by 0.742% relative L2, without a task tolerance. S25 six-frame component profile returned 100 samples at p50/p95 49.665/51.382 ms, with all 121 execution-detail rows on NPU; the action policy did not run on S25. The same encoder compiled to FP16 QNN DLC for X Elite CRD; same-fixture NPU-requested inference returned a finite latent at 0.0608% relative L2 versus source, with its 100-sample six-frame profile at p50/p95 82.939/83.837 ms and all 121 execution-detail rows on NPU. An SA8775P FP16 QNN DLC inference returned a finite same-fixture latent at 0.0613% relative L2 versus source; its 100-sample six-frame profile measured p50/p95 116.021/116.327 ms with all 121 execution-detail rows on NPU. This is component/action-sensitivity evidence, not device-local M4 qualification. The checkpoint train config names a2d_real/a2d_pen_holder; an [authenticated access check](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/internvla_real_observation_access/README.md) returned HTTP 404. The official vLLM-Omni example names an InternData-A1 Place_Markpen archive, but that dataset returned HTTP 403 gated-access denial under the current account, so real-observation reference-action quality is still unmeasured.
The [WSL CPU expansion](../../benchmarks/edge_harness/results/e2e_expansion_20260923/README.md)
adds constrained three-stage MiniCPM-o text-to-speech requests and a CPU-only