Advance Qualcomm TTS and InternVLA component qualification
This commit is contained in:
1 parent
ea373e3cbf
commit
1949c31f80
20 files changed
+5901
-11
No files matched your search
@@ -0,0 +1,120 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Audit one hosted InternVLA Cosmos encoder output against a pinned source fixture."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import hashlib
|
||||
import json
|
||||
import math
|
||||
from pathlib import Path
|
||||
|
||||
import numpy as np
|
||||
|
||||
|
||||
MODEL_SHA256 = "656a5263f584851a1457ef6dcf60b995f7c67b056b3c90a4f33c814f7caf9ec9"
|
||||
FIXTURE_SHA256 = "dcbaf1391ca914d4bbccc1cb384daf05666e771d9cdae90c4f069ec8fcf05368"
|
||||
COMPILE_OPTIONS = "--target_runtime qnn_dlc --qnn_options default_graph_htp_precision=FLOAT16"
|
||||
|
||||
|
||||
def file_record(path: Path) -> dict:
|
||||
with path.open("rb") as source:
|
||||
digest = hashlib.file_digest(source, "sha256").hexdigest()
|
||||
return {"bytes": path.stat().st_size, "sha256": digest}
|
||||
|
||||
|
||||
def main() -> None:
|
||||
parser = argparse.ArgumentParser(description=__doc__)
|
||||
for name in (
|
||||
"source-onnx", "fixture", "compile-report", "inference-submission",
|
||||
"inference-report", "device-output", "output-report",
|
||||
):
|
||||
parser.add_argument(f"--{name}", type=Path, required=True)
|
||||
args = parser.parse_args()
|
||||
|
||||
compiled = json.loads(args.compile_report.read_text(encoding="utf-8"))
|
||||
submitted = json.loads(args.inference_submission.read_text(encoding="utf-8"))
|
||||
inferred = json.loads(args.inference_report.read_text(encoding="utf-8"))
|
||||
if compiled.get("status") != "SUCCESS" or inferred.get("status") != "SUCCESS":
|
||||
raise ValueError("compile and inference must be terminal successes")
|
||||
if compiled.get("source_model_id") != "mn4okj3rq" or compiled.get("options") != COMPILE_OPTIONS:
|
||||
raise ValueError("compile used a different source or options")
|
||||
if inferred.get("model_id") != compiled.get("target_model_id"):
|
||||
raise ValueError("inference used a different target model")
|
||||
if inferred.get("device") != compiled.get("device"):
|
||||
raise ValueError("compile/inference device metadata differs")
|
||||
if inferred.get("device", {}).get("name") != "Samsung Galaxy S25":
|
||||
raise ValueError("inference did not use the exact Galaxy S25")
|
||||
if inferred.get("options") != "--compute_unit npu":
|
||||
raise ValueError("inference did not request NPU")
|
||||
if (inferred.get("job_id") != submitted.get("job_id")
|
||||
or inferred.get("input_dataset_id") != submitted.get("input_dataset_id")):
|
||||
raise ValueError("inference used a different job or input dataset")
|
||||
if submitted.get("fixture_key") != "pattern_pixels":
|
||||
raise ValueError("submission did not use the pattern fixture")
|
||||
|
||||
files = {
|
||||
name: file_record(path)
|
||||
for name, path in {
|
||||
"source_onnx": args.source_onnx,
|
||||
"fixture": args.fixture,
|
||||
"device_output": args.device_output,
|
||||
}.items()
|
||||
}
|
||||
if files["source_onnx"]["sha256"] != MODEL_SHA256 or files["fixture"]["sha256"] != FIXTURE_SHA256:
|
||||
raise ValueError("source model or fixture hash differs")
|
||||
with np.load(args.fixture, allow_pickle=False) as fixture:
|
||||
if fixture.files != ["ramp_pixels", "ramp_reference", "pattern_pixels", "pattern_reference"]:
|
||||
raise ValueError("fixture layout differs")
|
||||
pixels = np.asarray(fixture["pattern_pixels"])
|
||||
reference = np.asarray(fixture["pattern_reference"])
|
||||
if pixels.shape != (6, 3, 256, 256) or pixels.dtype != np.float32 or not np.isfinite(pixels).all():
|
||||
raise ValueError("input pixels violate fixed source contract")
|
||||
if reference.shape != (6, 16, 32, 32) or reference.dtype != np.float32 or not np.isfinite(reference).all():
|
||||
raise ValueError("source reference violates fixed latent contract")
|
||||
with np.load(args.device_output, allow_pickle=False) as output:
|
||||
if output.files != ["output_0__0"]:
|
||||
raise ValueError("unexpected hosted output layout")
|
||||
actual = np.asarray(output["output_0__0"])
|
||||
if (actual.shape != reference.shape or actual.dtype not in (np.float16, np.float32)
|
||||
or not np.isfinite(actual).all()):
|
||||
raise ValueError("hosted output violates finite latent contract")
|
||||
|
||||
source_values = reference.astype(np.float64).ravel()
|
||||
device_values = actual.astype(np.float64).ravel()
|
||||
error = source_values - device_values
|
||||
source_norm = float(np.linalg.norm(source_values))
|
||||
device_norm = float(np.linalg.norm(device_values))
|
||||
error_norm = float(np.linalg.norm(error))
|
||||
if source_norm == 0 or device_norm == 0:
|
||||
raise ValueError("zero latent norm cannot define numerical similarity")
|
||||
report = {
|
||||
"scope": "one six-frame Cosmos image encoder on synthetic pattern; no complete InternVLA policy",
|
||||
"status": "component_inference_numeric_measured",
|
||||
"device": inferred["device"],
|
||||
"source_model_id": compiled["source_model_id"],
|
||||
"target_model_id": compiled["target_model_id"],
|
||||
"compile_job_id": compiled["job_id"],
|
||||
"inference_job_id": inferred["job_id"],
|
||||
"input_dataset_id": inferred["input_dataset_id"],
|
||||
"requested_run_options": inferred["options"],
|
||||
"files": files,
|
||||
"output_dtype": str(actual.dtype),
|
||||
"source_vs_device": {
|
||||
"relative_l2": error_norm / source_norm,
|
||||
"snr_db": 20 * math.log10(source_norm / error_norm) if error_norm else None,
|
||||
"max_abs_error": float(np.max(np.abs(error))),
|
||||
"cosine": float(np.dot(source_values, device_values) / (source_norm * device_norm)),
|
||||
},
|
||||
"limits": [
|
||||
"The fixture is synthetic; no real-observation or action-quality tolerance is established.",
|
||||
"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.",
|
||||
],
|
||||
}
|
||||
args.output_report.write_text(json.dumps(report, indent=2) + "\n", encoding="utf-8")
|
||||
print(json.dumps({"status": report["status"], "difference": report["source_vs_device"]}, indent=2))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -1,7 +1,9 @@
|
||||
# 2026-09-24 Qualcomm component expansion
|
||||
|
||||
The [Qwen3-TTS vocoder study](evidence/qwen_tts_vocoder_qualcomm/README.md) reuses a retained fixed-shape Code2Wav export and historical Workbench input to test numerical behavior on SA8775P and other Qualcomm settings. The fresh SA8775P GPU inference produces a saturated waveform; RB3 produces an unsaturated but divergent waveform. Neither is numerically qualified as a usable TTS path. X Elite ONNX CPU output matches the local ONNX Runtime reference on the retained fixture; its 94-sample component p50/p95 was 6.384/6.644 s with all execution-detail rows on CPU. Its DirectML retry remains pending. Other Workbench jobs for RB3 profiling and SA8775P controls are also pending; their submission IDs are retained for completion. No complete on-device TTS stream or model-quality qualification is claimed.
|
||||
The [Qwen3-TTS vocoder study](evidence/qwen_tts_vocoder_qualcomm/README.md) reuses a retained fixed-shape Code2Wav export and historical Workbench input to test numerical behavior on SA8775P and other Qualcomm settings. The fresh SA8775P GPU-requested inference produces a saturated waveform, while the same binary's CPU-requested output on the exact SA device matches ONNX CPU within 4.13e-6 relative L2. RB3 produces an unsaturated but 15.73%-divergent waveform; its fresh 100-sample mixed CPU/GPU component p50/p95 was 3.099/3.133 s, not a usable TTS timing claim. X Elite ONNX CPU output matches the local ONNX Runtime reference on the retained fixture; its 94-sample component p50/p95 was 6.384/6.644 s with all execution-detail rows on CPU. Corrected X Elite DirectML inference and profile both failed during provider initialization; separate FP16 QNN DLC compilation passed, with NPU-requested inference pending. SA CPU placement profiling and FP32-preserving GPU control remain pending. No complete on-device TTS stream or model-quality qualification is claimed.
|
||||
|
||||
A local HX370 x86-64 LiteRT CPU control of the identical SA/RB3 TFLite binary matched the original ONNX CPU waveform within 3.45e-6 relative L2. The SA saturation is therefore not universal to the converted artifact; hosted CPU and GPU precision controls are still needed to isolate its device/runtime cause.
|
||||
|
||||
The [talker decode-step attempt](evidence/qwen_tts_talker_qualcomm/README.md) pins a separate 28-layer source and synthetic cache fixture. Historical FP16 QNN context-binary compiles passed for S24, S25, SA8775P and X Elite, with same-fixture inference pending; RB3 rejected that floating-point-input artifact. Compile success is not a numerical or end-to-end pass.
|
||||
|
||||
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 compilation passed; same-fixture NPU-requested inference was pending. Full action-policy support remains unqualified.
|
||||
+9
@@ -0,0 +1,9 @@
|
||||
# InternVLA Cosmos encoder on Galaxy S25 NPU: compile attempt
|
||||
|
||||
This tests the **real Place_Markpen Cosmos CI8x8 image encoder component**, not the action policy or complete InternVLA. The [source export](../../../e2e_expansion_20260923/evidence/internvla_cosmos_amd_npu/README.md) pins the checkpoint, FP32 `[6,3,256,256]` input, `[6,16,32,32]` output, two CPU parity fixtures and the original external-data ONNX files. The external data were embedded into one 125,112,088-byte ONNX file outside Git at `/home/zhout/project/edge_infer/models/InternVLA-cosmos-onnx/cosmos_encoder_embedded.onnx` (SHA256 `656a5263f584851a1457ef6dcf60b995f7c67b056b3c90a4f33c814f7caf9ec9`). The embedded form retained reference-relative ONNX CPU L2 errors of `8.74e-7` and `2.29e-6` on ramp and pattern fixtures.
|
||||
|
||||
Workbench source model `mn4okj3rq` was [submitted for Galaxy S25 QNN DLC compilation](s25/compile_submission.json) as job `jpxl4d1jp`, with `default_graph_htp_precision=FLOAT16`. The [compile](s25/compile_report.json) succeeded and produced target model `mqkp12rxm`. A [same-fixture pattern inference](s25/inference_submission.json) requesting NPU was submitted as `jp8ex43zp` and remained pending at this evidence snapshot. Compile success does not establish NPU execution, output parity, action quality, latency, or full-policy support.
|
||||
|
||||
The [generic collector](../../../../experiments/collect_aihub_component.py) can retrieve the pending output, 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. Neither script converts an encoder-only result to policy support.
|
||||
|
||||
Next: retrieve and audit the device latent, inspect actual NPU placement with a profile if numerical error is acceptable, then test action-policy impact and a device-local stateful policy. Full multimodal observation processing, admission, cancellation and sustained profiling remain open.
|
||||
+32
@@ -0,0 +1,32 @@
|
||||
{
|
||||
"retrieved_utc": "2026-09-23T23:21:02.427625+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"job_id": "jpxl4d1jp",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jpxl4d1jp/",
|
||||
"status": "SUCCESS",
|
||||
"status_message": "",
|
||||
"source_model_id": "mn4okj3rq",
|
||||
"target_model_id": "mqkp12rxm",
|
||||
"device": {
|
||||
"name": "Samsung Galaxy S25",
|
||||
"os": "15",
|
||||
"attributes": [
|
||||
"os:android",
|
||||
"framework:tflite",
|
||||
"framework:onnx",
|
||||
"abi:aarch64-android",
|
||||
"vendor:samsung",
|
||||
"format:phone",
|
||||
"chipset:qualcomm-snapdragon-8-elite-for-galaxy",
|
||||
"chipset:sm8750-ac",
|
||||
"hexagon:v79",
|
||||
"soc-model:69",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"options": "--target_runtime qnn_dlc --qnn_options default_graph_htp_precision=FLOAT16",
|
||||
"hub_version": "aihub-2026.09.11.0"
|
||||
}
|
||||
+35
@@ -0,0 +1,35 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:18:50.383852+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"source_model_id": "mn4okj3rq",
|
||||
"source_artifact": {
|
||||
"name": "cosmos_encoder_embedded.onnx",
|
||||
"bytes": 125112088,
|
||||
"sha256": "656a5263f584851a1457ef6dcf60b995f7c67b056b3c90a4f33c814f7caf9ec9"
|
||||
},
|
||||
"upstream_export_report": "e2e_expansion_20260923/evidence/internvla_cosmos_amd_npu/export_report.json",
|
||||
"device": {
|
||||
"name": "Samsung Galaxy S25",
|
||||
"os": "15",
|
||||
"attributes": [
|
||||
"os:android",
|
||||
"framework:tflite",
|
||||
"framework:onnx",
|
||||
"abi:aarch64-android",
|
||||
"vendor:samsung",
|
||||
"format:phone",
|
||||
"chipset:qualcomm-snapdragon-8-elite-for-galaxy",
|
||||
"chipset:sm8750-ac",
|
||||
"hexagon:v79",
|
||||
"soc-model:69",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"options": "--target_runtime qnn_dlc --qnn_options default_graph_htp_precision=FLOAT16",
|
||||
"job_id": "jpxl4d1jp",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jpxl4d1jp/",
|
||||
"status_at_submission": "OPTIMIZING_MODEL"
|
||||
}
|
||||
+16
@@ -0,0 +1,16 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:21:34.357321+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"source_compile_job_id": "jpxl4d1jp",
|
||||
"target_model_id": "mqkp12rxm",
|
||||
"fixture_sha256": "dcbaf1391ca914d4bbccc1cb384daf05666e771d9cdae90c4f069ec8fcf05368",
|
||||
"fixture_key": "pattern_pixels",
|
||||
"device": "Samsung Galaxy S25",
|
||||
"options": "--compute_unit npu",
|
||||
"kind": "inference",
|
||||
"status": "SUBMITTED",
|
||||
"job_id": "jp8ex43zp",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jp8ex43zp/",
|
||||
"input_dataset_id": "d26qo8p57"
|
||||
}
|
||||
+5
-3
@@ -6,14 +6,16 @@ The historical Workbench source model `mqyer339n` is a 423,655,047-byte ONNX gra
|
||||
|
||||
The [S25 historical inference](s25_reference_inference.json) on the same fixture differed from local FP32 ONNX CPU by **4.24% relative L2 / 27.46 dB SNR** ([comparison](s25_cpu_comparison.json)); that waveform has no listening-quality tolerance. A fresh [SA8775P inference](sa8775p/inference_report.json) used historical compiled TFLite model `mnze95don` and requested GPU. Its [output](sa8775p/device_output.npz) had **96.63%** of samples at absolute amplitude at least 0.999, RMS **0.989** versus CPU **0.0837**, relative L2 **11.832**, and SNR **−21.46 dB**. The fresh [profile](sa8775p/profile_report.json) attributed 403 execution-detail rows to CPU and 335 to GPU; 100 component samples had nearest-rank p50/p95 **1.776/1.880 s** (minimum 1.675 s). The [audit](sa8775p/audit_report.json) pins all output hashes, waveform differences and placement. This is measured execution of a **numerically failed** output, not a usable TTS performance result. The [SA8775P compile record](sa8775p/compile_report.json) pins the artifact and source IDs.
|
||||
|
||||
The fresh [RB3 inference](rb3/inference_report.json) on the same fixture returned a finite, **unsaturated** waveform ([output](rb3/device_output.npz)). Its [comparison](rb3/inference_comparison.json) against original ONNX CPU was **15.73% relative L2 / 16.06 dB SNR**, still without a listening-quality threshold. Thus the bitwise identical SA and RB3 TFLite target downloads produce materially different outputs on their respective devices; target file identity does not establish runtime numerical parity. The fresh RB3 profile was still pending at this snapshot.
|
||||
The fresh [RB3 inference](rb3/inference_report.json) on the same fixture returned a finite, **unsaturated** waveform ([output](rb3/device_output.npz)). Its [comparison](rb3/inference_comparison.json) against original ONNX CPU was **15.73% relative L2 / 16.06 dB SNR**, still without a listening-quality threshold. The fresh [profile](rb3/profile_report.json) returned 100 component samples at nearest-rank p50/p95 **3.099/3.133 s** and 403 CPU plus 335 GPU execution-detail rows; the [audit](rb3/audit_report.json) ties the profile to this numerically unqualified output. Thus the bitwise identical SA and RB3 TFLite target downloads produce materially different outputs on their respective devices; target file identity does not establish runtime numerical parity.
|
||||
|
||||
A separate [local LiteRT 2.2.0 CPU control](local_litert_cpu/report.json) ran that **same pinned TFLite binary** on the WSL x86-64 HX370 host with four interpreter threads. Its [waveform](local_litert_cpu/device_output.npz) was finite and unsaturated, within **3.45e-6 relative L2 / 109.23 dB SNR** of the original ONNX Runtime CPU output. The [reproducer](../../../../experiments/probe_qwen_tts_vocoder_litert_cpu.py) checks binary/fixture/reference hashes and I/O contracts. This narrows the SA saturation to hosted runtime/device behavior rather than a universally bad TFLite conversion. The one local invoke is not a comparable performance profile; it does not prove either embedded device's CPU route or identify the specific delegate operation at fault.
|
||||
|
||||
On the **exact hosted SA8775P**, a [CPU-requested inference](sa8775p/cpu_control/inference_report.json) of that binary also returned a finite unsaturated waveform. The [same-fixture comparison](sa8775p/cpu_control/inference_comparison.json) was **4.13e-6 relative L2 / 107.67 dB SNR** against original ONNX CPU. The GPU-requested SA output differed from this SA CPU-requested output by **11.832 relative L2** and saturated 96.63% of samples. This is strong evidence that the tested GPU-requested route causes the waveform failure on SA, while its exact delegate operation and all CPU placement details still require the [CPU profile](sa8775p/cpu_control/profile_submission.json) and FP32-preserving GPU control.
|
||||
|
||||
The re-fetched *historical* [SA8775P profile](sa8775p/historical_profile_report.json) has 100 component samples, nearest-rank p50/p95 **1.751/1.874 s**, and 403 CPU plus 335 GPU execution-detail rows. The *historical* [RB3 profile](rb3/historical_profile_report.json) has 100 samples, p50/p95 **3.143/3.189 s**, also with 403 CPU plus 335 GPU rows. Their previously quoted 1.701/3.107 s estimated-time fields are sample minima, not medians. Both profiles exclude talker/predictor, stage handoff, streaming and full-request time.
|
||||
|
||||
The X Elite [ONNX CPU inference](xelite/cpu/inference_report.json) succeeded on the pinned fixture. Its [waveform comparison](xelite/cpu/inference_comparison.json) to the original local ONNX Runtime CPU output was **8.09e-7 relative L2 / 121.84 dB SNR**, with no saturation. The fresh [CPU profile](xelite/cpu/profile_report.json) returned **94** component samples, nearest-rank p50/p95 **6.384/6.644 s**, with all **634** execution-detail rows on CPU; the service's 5.917 s estimated-time field is the sample minimum. The [full audit](xelite/cpu/audit_report.json) pins source, fixture, target, waveform, placement and percentiles. This is near-source component parity on one fixture, without a listening-quality or complete-stream gate. The first X Elite [GPU inference](xelite/gpu/inference_report.json) and [profile](xelite/gpu/profile_report.json) jobs failed because Workbench rejects specifying both compute unit and ONNX execution provider; the [DirectML retry](xelite/directml/inference_submission.json) specifies DirectML alone. The [X Elite ONNX compile](xelite/compile_report.json) passed after the older TFLite target was rejected on Windows ARM.
|
||||
The X Elite [ONNX CPU inference](xelite/cpu/inference_report.json) succeeded on the pinned fixture. Its [waveform comparison](xelite/cpu/inference_comparison.json) to the original local ONNX Runtime CPU output was **8.09e-7 relative L2 / 121.84 dB SNR**, with no saturation. The fresh [CPU profile](xelite/cpu/profile_report.json) returned **94** component samples, nearest-rank p50/p95 **6.384/6.644 s**, with all **634** execution-detail rows on CPU; the service's 5.917 s estimated-time field is the sample minimum. The [full audit](xelite/cpu/audit_report.json) pins source, fixture, target, waveform, placement and percentiles. This is near-source component parity on one fixture, without a listening-quality or complete-stream gate. The first X Elite [GPU inference](xelite/gpu/inference_report.json) and [profile](xelite/gpu/profile_report.json) jobs failed because Workbench rejects specifying both compute unit and ONNX execution provider. The corrected [DirectML inference](xelite/directml/inference_report.json) and [profile](xelite/directml/profile_report.json) both failed during DML provider initialization with Windows error `80070057` (`The parameter is incorrect`), without an output or timing samples. A separate [FP16 QNN DLC compile](xelite/qnn_npu/compile_report.json) succeeded and [NPU-requested same-fixture inference](xelite/qnn_npu/inference_submission.json) was pending. The [X Elite ONNX CPU compile](xelite/compile_report.json) passed after the older TFLite target was rejected on Windows ARM. These failures constrain the tested DirectML route, not all GPU or NPU artifacts.
|
||||
|
||||
The fresh [RB3 profile](rb3/profile_submission.json), [SA CPU](sa8775p/cpu_control/inference_submission.json) and [FP32-preserving GPU](sa8775p/gpu_fp32/inference_submission.json) controls, and X Elite [DirectML retry](xelite/directml/inference_submission.json) remained pending in this snapshot. Pending jobs have no numerical or latency conclusion. The [collector](../../../../experiments/collect_aihub_component.py) retrieves their terminal reports without resubmitting jobs, and the [audit script](../../../../experiments/audit_qwen_tts_vocoder_qualcomm.py) checks source/fixture hashes, device and model IDs, actual placement, waveform error, saturation and sample percentiles.
|
||||
The [SA CPU profile](sa8775p/cpu_control/profile_submission.json), [FP32-preserving GPU](sa8775p/gpu_fp32/inference_submission.json) control, and X Elite [QNN NPU inference](xelite/qnn_npu/inference_submission.json) remained pending in this snapshot. Pending jobs have no numerical or latency conclusion. The [collector](../../../../experiments/collect_aihub_component.py) retrieves their terminal reports without resubmitting jobs, and the [audit script](../../../../experiments/audit_qwen_tts_vocoder_qualcomm.py) checks source/fixture hashes, device and model IDs, actual placement, waveform error, saturation and sample percentiles.
|
||||
|
||||
This is one fixed vocoder component with a retained input, not complete Qwen3-TTS. The next gates are a quality-passing SA/RB3/X Elite target, talker and code predictor numerical checks, device-resident handoff/co-residency, playable streaming, cancellation, memory admission and sustained power/thermal profiling.
|
||||
+86
@@ -0,0 +1,86 @@
|
||||
{
|
||||
"scope": "one Qwen3-TTS code2wav 48,000-sample component on one retained fixture; no complete TTS stream",
|
||||
"status": "component_executed_quality_unqualified",
|
||||
"device": {
|
||||
"name": "Dragonwing RB3 Gen 2 Vision Kit",
|
||||
"os": "1.6",
|
||||
"attributes": [
|
||||
"os:qc_linux",
|
||||
"framework:tflite",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-oe-linux-gcc11",
|
||||
"format:iot",
|
||||
"chipset:qualcomm-qcs6490",
|
||||
"chipset:qcs6490",
|
||||
"chipset:qualcomm-dragonwing-qcs6490",
|
||||
"hexagon:v68",
|
||||
"soc-model:93",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"compile_job_id": "j5mmy28w5",
|
||||
"inference_job_id": "jg9zmkv8p",
|
||||
"profile_job_id": "jp1nj707g",
|
||||
"target_model_id": "mn0wyzd3q",
|
||||
"input_dataset_id": "d7m80jl32",
|
||||
"s25_prior_inference_job_id": "jpx428ojg",
|
||||
"requested_run_options": "--compute_unit gpu",
|
||||
"files": {
|
||||
"source_onnx": {
|
||||
"bytes": 423655047,
|
||||
"sha256": "7c07666229c6d404894132e22f784525e3212a49efad55ddd532eda5286c2b2b"
|
||||
},
|
||||
"fixture": {
|
||||
"bytes": 129721,
|
||||
"sha256": "dfca9e8d2a724a568a31b6cd18ff6a52fdc205fd34e13cbf1cd75f0f7937e82d"
|
||||
},
|
||||
"cpu_output": {
|
||||
"bytes": 179942,
|
||||
"sha256": "8de78b94b59f01e49852f7706adbd2508a9976b8e2705cfd3b94d07bb3606cb1"
|
||||
},
|
||||
"s25_output": {
|
||||
"bytes": 115235,
|
||||
"sha256": "4a26486d5e84a6e22d16f9b26f995b8fe645c29bbd70b103615ec35a0fcb10a0"
|
||||
},
|
||||
"device_output": {
|
||||
"bytes": 113399,
|
||||
"sha256": "476cae29892fbd6d93b8b4f3790e26bc6907cc2c8987bab3c141e1a585aea23a"
|
||||
}
|
||||
},
|
||||
"cpu_source_vs_device": {
|
||||
"relative_l2": 0.15731895608191257,
|
||||
"snr_db": 16.06437888182974,
|
||||
"max_abs": 0.11359542608261108,
|
||||
"cosine": 0.988868865169743,
|
||||
"reference_rms": 0.08373439205485074,
|
||||
"output_rms": 0.07852377091606144,
|
||||
"output_saturated_fraction_abs_ge_0_999": 0.0
|
||||
},
|
||||
"s25_gpu_vs_device": {
|
||||
"relative_l2": 0.1529833843922324,
|
||||
"snr_db": 16.307114711572623,
|
||||
"max_abs": 0.11407470703125,
|
||||
"cosine": 0.9896069628698109,
|
||||
"reference_rms": 0.08376787228276501,
|
||||
"output_rms": 0.07852377091606144,
|
||||
"output_saturated_fraction_abs_ge_0_999": 0.0
|
||||
},
|
||||
"profile": {
|
||||
"sample_count": 100,
|
||||
"sample_min_us": 3069620,
|
||||
"nearest_rank_p50_us": 3098543,
|
||||
"nearest_rank_p95_us": 3132926,
|
||||
"reported_estimated_inference_time_us": 3069620,
|
||||
"reported_peak_memory_bytes": 326594560,
|
||||
"compute_unit_row_counts": {
|
||||
"CPU": 403,
|
||||
"GPU": 335
|
||||
}
|
||||
},
|
||||
"limits": [
|
||||
"Historical source export does not attest its exact checkpoint revision.",
|
||||
"This fixed component waveform has no listening or task-quality tolerance.",
|
||||
"Hosted component profiling excludes talker, predictor, stage handoff, streaming, admission and sustained power/thermal behavior."
|
||||
]
|
||||
}
|
||||
+5361
File diff suppressed because it is too large.
Load diff
BIN
Binary file not shown.
+51
@@ -0,0 +1,51 @@
|
||||
{
|
||||
"scope": "one Qwen3-TTS Code2Wav inference on the exact SA8775P ADP CPU; no complete TTS stream",
|
||||
"status": "component_cpu_numeric_parity_on_one_fixture",
|
||||
"compile_job_id": "jp410o3vp",
|
||||
"inference_job_id": "jp1nj7x7g",
|
||||
"target_model_id": "mnze95don",
|
||||
"input_dataset_id": "d7m80jl32",
|
||||
"requested_compute_unit": "cpu",
|
||||
"device": {
|
||||
"name": "SA8775P ADP",
|
||||
"os": "14",
|
||||
"attributes": [
|
||||
"os:android",
|
||||
"framework:tflite",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-android",
|
||||
"format:auto",
|
||||
"chipset:qualcomm-sa8775p",
|
||||
"hexagon:v73",
|
||||
"soc-model:52",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"output_sha256": "853feface7a93f7a22a0ffba8d1e297fe5a8780bd64f0444a6ce07b277246d3e",
|
||||
"source_onnx_cpu_vs_sa_cpu": {
|
||||
"relative_l2": 4.1337705035866275e-06,
|
||||
"snr_db": 107.67307275926692,
|
||||
"max_abs": 6.854534149169922e-06,
|
||||
"cosine": 0.9999999999914584,
|
||||
"reference_rms": 0.08373439205485074,
|
||||
"output_rms": 0.08373439765610562,
|
||||
"output_saturated_fraction_abs_ge_0_999": 0.0
|
||||
},
|
||||
"sa_cpu_vs_sa_gpu_requested": {
|
||||
"relative_l2": 11.832416467511793,
|
||||
"snr_db": -21.46146894384565,
|
||||
"max_abs": 1.4882185459136963,
|
||||
"cosine": 0.017841895688532488,
|
||||
"reference_rms": 0.08373439765610562,
|
||||
"output_rms": 0.988730688363501,
|
||||
"output_saturated_fraction_abs_ge_0_999": 0.9663333333333334
|
||||
},
|
||||
"limits": [
|
||||
"No actual per-node placement profile yet.",
|
||||
"One fixed vocoder waveform has no listening-quality tolerance.",
|
||||
"No talker, predictor, handoff or complete TTS stream tested."
|
||||
]
|
||||
}
|
||||
+32
@@ -0,0 +1,32 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:20:01.293905+00:00",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"job_id": "jp1nj7x7g",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jp1nj7x7g/",
|
||||
"status": "SUCCESS",
|
||||
"status_message": "",
|
||||
"options": "--compute_unit cpu",
|
||||
"hub_version": "aihub-2026.09.11.0",
|
||||
"device": {
|
||||
"name": "SA8775P ADP",
|
||||
"os": "14",
|
||||
"attributes": [
|
||||
"os:android",
|
||||
"framework:tflite",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-android",
|
||||
"format:auto",
|
||||
"chipset:qualcomm-sa8775p",
|
||||
"hexagon:v73",
|
||||
"soc-model:52",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"model_id": "mnze95don",
|
||||
"input_dataset_id": "d7m80jl32",
|
||||
"outputs": "device_output.npz"
|
||||
}
|
||||
+12
@@ -0,0 +1,12 @@
|
||||
{
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"source_compile_job_id": "jp410o3vp",
|
||||
"target_model_id": "mnze95don",
|
||||
"device": "SA8775P ADP",
|
||||
"options": "--compute_unit cpu",
|
||||
"kind": "profile",
|
||||
"status": "SUBMITTED",
|
||||
"job_id": "jp2rwv7xg",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jp2rwv7xg/"
|
||||
}
|
||||
+31
@@ -0,0 +1,31 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:24:59.412365+00:00",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"job_id": "jp2rwvy6g",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jp2rwvy6g/",
|
||||
"status": "FAILED",
|
||||
"status_message": "Failed to profile the model: Exception during initialization: C:\\dev\\tetracode\\build\\third_party.build\\onnxruntime\\src\\onnxruntime\\core\\providers\\dml\\DmlExecutionProvider\\src\\MLOperatorAuthorImpl.cpp(2844)\\onnxruntime.dll!00007FFD9D307164: (caller: 00007FFD9D3236B0) Exception(2) tid(2870) 80070057 The parameter is incorrect.\r\n",
|
||||
"options": "--onnx_execution_providers directml",
|
||||
"hub_version": "aihub-2026.09.11.0",
|
||||
"device": {
|
||||
"name": "Snapdragon X Elite CRD",
|
||||
"os": "11",
|
||||
"attributes": [
|
||||
"os:windows",
|
||||
"format:compute",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-windows",
|
||||
"chipset:qualcomm-snapdragon-x-elite",
|
||||
"chipset:sc8380xp",
|
||||
"hexagon:v73",
|
||||
"soc-model:60",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"model_id": "mnz079ezm",
|
||||
"input_dataset_id": "d7m80jl32"
|
||||
}
|
||||
+40
@@ -0,0 +1,40 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:30:03.698535+00:00",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"job_id": "jpyox7305",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jpyox7305/",
|
||||
"status": "FAILED",
|
||||
"status_message": "Failed to profile the model: Exception during initialization: C:\\dev\\tetracode\\build\\third_party.build\\onnxruntime\\src\\onnxruntime\\core\\providers\\dml\\DmlExecutionProvider\\src\\MLOperatorAuthorImpl.cpp(2844)\\onnxruntime.dll!00007FF89C7F7164: (caller: 00007FF89C8136B0) Exception(2) tid(2264) 80070057 The parameter is incorrect.\r\n",
|
||||
"options": "--onnx_execution_providers directml",
|
||||
"hub_version": "aihub-2026.09.11.0",
|
||||
"device": {
|
||||
"name": "Snapdragon X Elite CRD",
|
||||
"os": "11",
|
||||
"attributes": [
|
||||
"os:windows",
|
||||
"format:compute",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-windows",
|
||||
"chipset:qualcomm-snapdragon-x-elite",
|
||||
"chipset:sc8380xp",
|
||||
"hexagon:v73",
|
||||
"soc-model:60",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"model_id": "mnz079ezm",
|
||||
"shapes": {
|
||||
"quantized": [
|
||||
[
|
||||
1,
|
||||
512,
|
||||
97
|
||||
],
|
||||
"float32"
|
||||
]
|
||||
}
|
||||
}
|
||||
+31
@@ -0,0 +1,31 @@
|
||||
{
|
||||
"retrieved_utc": "2026-09-23T23:28:23.406644+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"job_id": "jgdd3zrkg",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jgdd3zrkg/",
|
||||
"status": "SUCCESS",
|
||||
"status_message": "",
|
||||
"source_model_id": "mqyer339n",
|
||||
"target_model_id": "mmxj098kq",
|
||||
"device": {
|
||||
"name": "Snapdragon X Elite CRD",
|
||||
"os": "11",
|
||||
"attributes": [
|
||||
"os:windows",
|
||||
"format:compute",
|
||||
"framework:onnx",
|
||||
"vendor:qualcomm",
|
||||
"abi:aarch64-windows",
|
||||
"chipset:qualcomm-snapdragon-x-elite",
|
||||
"chipset:sc8380xp",
|
||||
"hexagon:v73",
|
||||
"soc-model:60",
|
||||
"htp-supports-fp16:true",
|
||||
"htp-supports-weight-sharing:true",
|
||||
"framework:qnn"
|
||||
]
|
||||
},
|
||||
"options": "--target_runtime qnn_dlc --qnn_options default_graph_htp_precision=FLOAT16",
|
||||
"hub_version": "aihub-2026.09.11.0"
|
||||
}
|
||||
+12
@@ -0,0 +1,12 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:25:18.407088+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"source_model_id": "mqyer339n",
|
||||
"device": "Snapdragon X Elite CRD",
|
||||
"options": "--target_runtime qnn_dlc --qnn_options default_graph_htp_precision=FLOAT16",
|
||||
"kind": "compile",
|
||||
"status": "SUBMITTED",
|
||||
"job_id": "jgdd3zrkg",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jgdd3zrkg/"
|
||||
}
|
||||
+14
@@ -0,0 +1,14 @@
|
||||
{
|
||||
"timestamp_utc": "2026-09-23T23:28:36.325272+00:00",
|
||||
"facility": "Qualcomm AI Hub Workbench",
|
||||
"qai_hub_version": "0.55.0",
|
||||
"source_compile_job_id": "jgdd3zrkg",
|
||||
"target_model_id": "mmxj098kq",
|
||||
"input_dataset_id": "d7m80jl32",
|
||||
"device": "Snapdragon X Elite CRD",
|
||||
"options": "--compute_unit npu",
|
||||
"kind": "inference",
|
||||
"status": "SUBMITTED",
|
||||
"job_id": "jprl0n9ep",
|
||||
"url": "https://workbench.aihub.qualcomm.com/jobs/jprl0n9ep/"
|
||||
}
|
||||
@@ -12,11 +12,13 @@ On a hosted [Galaxy S24 Spark attention-layer experiment](../../../e2e_expansion
|
||||
|
||||
The same W8A16 layer also ran on hosted [Snapdragon X Elite CRD](../../../e2e_expansion_20260923/evidence/spark_xelite_full_attention/README.md) and [SA8775P ADP](../../../e2e_expansion_20260923/evidence/spark_sa8775_full_attention/README.md) NPUs, with pinned-fixture outputs identical to S24/S25; the 100-sample component p50/p95 was 1.433/1.606 ms and 1.837/2.079 ms, respectively. On [RB3 Gen 2 / QCS6490](../../../e2e_expansion_20260923/evidence/spark_rb3_full_attention/README.md), that W8A16 graph failed QNN DLC loading even after target-specific recompilation. An FP32 ONNX CPU layer then passed with source-relative L2 below 9e-7, while a calibrated W8A8 NPU layer executed but had 20.9% hidden-state relative L2 and remains numerically unqualified. These hosted component outcomes do not establish full generation or comparable device speed.
|
||||
|
||||
A [Qwen3-TTS Code2Wav follow-up](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md) recovered source and target artifacts, a retained fixture, and raw historical profiles. On SA8775P, fresh GPU-requested inference saturated 96.63% of waveform samples and differed from original ONNX CPU by 11.832 relative L2; this tested vocoder path is numerically unqualified. Fresh RB3 inference was unsaturated but differed from CPU by 15.73% relative L2 (16.06 dB SNR), also without a listening-quality gate. Historical SA8775P/RB3 profile p50/p95 values were 1.751/1.874 s and 3.143/3.189 s, respectively, with both CPU and GPU rows; earlier estimated-time fields were minima. X Elite ONNX CPU vocoder inference passed on the same fixture with 8.09e-7 relative L2 versus local ONNX CPU; its 94-sample component p50/p95 was 6.384/6.644 s with all execution-detail rows on CPU. DirectML retry was pending. None is a complete TTS stream.
|
||||
A [Qwen3-TTS Code2Wav follow-up](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md) recovered source and target artifacts, a retained fixture, and raw historical profiles. On SA8775P, fresh GPU-requested inference saturated 96.63% of waveform samples and differed from original ONNX CPU by 11.832 relative L2; the same binary's CPU-requested output on the exact SA device matched ONNX CPU within 4.13e-6 relative L2. The GPU-requested path remains numerically failed while CPU placement profiling is pending. Fresh RB3 inference was unsaturated but differed from CPU by 15.73% relative L2 (16.06 dB SNR), without a listening-quality gate; its 100-sample mixed CPU/GPU component p50/p95 was 3.099/3.133 s. X Elite ONNX CPU vocoder inference passed on the same fixture with 8.09e-7 relative L2 versus local ONNX CPU; its 94-sample component p50/p95 was 6.384/6.644 s with all execution-detail rows on CPU. DirectML retry was pending. None is a complete TTS stream.
|
||||
|
||||
The identical SA/RB3 TFLite binary matched original ONNX CPU within 3.45e-6 relative L2 in a separate local HX370 x86-64 LiteRT CPU control. This narrows the SA waveform saturation to hosted device/runtime behavior, without proving either embedded CPU route or a specific GPU-delegate fault.
|
||||
|
||||
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 compiles passed for S24, S25, SA8775P and X Elite, but the same-fixture inference jobs were pending. An RB3 compile rejected floating-point input. No talker runtime or numerical pass is implied by those compile outcomes.
|
||||
|
||||
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 compilation passed, while same-fixture NPU-requested inference was pending. This changes the artifact investigation for that cell, not its NPU or full-policy qualification.
|
||||
|
||||
## Configuration × model
|
||||
|
||||
@@ -83,7 +85,7 @@ NOT E2E means no complete device/model workload was verified. “Scoped E2E” i
|
||||
| PC CPU+iGPU+NPU +/- discrete GPU, joint model execution | MiniCPM-o 4.5 | NOT_E2E_PROFILEABLE | No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline. | Supply and validate the missing matching artifact/backend, then run the complete specified workload. |
|
||||
| PC CPU+iGPU+NPU +/- discrete GPU, joint model execution | InternVLA-A1 | NOT_E2E_PROFILEABLE | No qualifying same-model multi-accelerator deployment/artifact plan or overlap/recovery evidence. Individual routes are not a joint pipeline. | Supply and validate the missing matching artifact/backend, then run the complete specified workload. |
|
||||
| PC Snapdragon X Elite CRD / AI Hub | Spark-X2.5 | COMPONENT_NPU_INFERENCE_PROFILED; NOT_E2E_PROFILEABLE | A real Spark-X2.5-1.7B one-layer W8A16 QDQ attention artifact, compiled as QNN DLC, ran on the exact hosted Snapdragon X Elite CRD (Windows 11 ARM64). Inference returned finite hidden/new-K/new-V tensors bitwise identical to the prior S24/S25 NPU outputs on the same fixture; the profile attributed all 73 execution-detail rows to NPU. Across 100 component samples, nearest-rank p50/p95 was 1.433/1.606 ms; the 1.346 ms estimated-time field is the minimum. Hidden/K/V relative L2 versus original FP32 ONNX CPU was 1.351%/0.377%/0.650%, without a next-token quality tolerance. There is no complete 28-layer prefill/decode/KV-ring/sampling loop, resident Omni controller or complete-request timing. | Qualify a complete X Elite resident artifact/controller with state, token quality, memory admission/cancel and sustained power; [X Elite NPU raw evidence](../../../e2e_expansion_20260923/evidence/spark_xelite_full_attention/README.md). |
|
||||
| PC Snapdragon X Elite CRD / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_ONNX_CPU_VOCODER_PARITY_PROFILED; NOT_E2E_PROFILEABLE | The fresh ONNX vocoder inference on the exact Windows 11 ARM64 X Elite CRD requested CPU and returned an unsaturated waveform on the pinned fixture, matching local ONNX Runtime CPU with 8.09e-7 relative L2 and 121.84 dB SNR. Its 94-sample component profile had nearest-rank p50/p95 6.384/6.644 s and all 634 execution-detail rows on CPU; 5.917 s estimated time is the sample minimum. This is one component, without a listening-quality gate. Corrected DirectML inference/profile jobs were pending; the initial DirectML submissions failed because mutually exclusive run options were specified. Historical QNN talker profiling lacks a same-fixture numerical or full-pipeline gate. The older TFLite vocoder target was rejected on Windows ARM; no complete stream exists. | Audit DirectML retry, qualify talker/predictor and complete local streaming, memory and sustained behavior; [vocoder evidence](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md). |
|
||||
| PC Snapdragon X Elite CRD / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_ONNX_CPU_VOCODER_PARITY_PROFILED; DIRECTML_RUNTIME_FAILED; QNN_NPU_COMPILE_PASS; NOT_E2E_PROFILEABLE | The fresh ONNX vocoder inference on the exact Windows 11 ARM64 X Elite CRD requested CPU and returned an unsaturated waveform on the pinned fixture, matching local ONNX Runtime CPU with 8.09e-7 relative L2 and 121.84 dB SNR. Its 94-sample component profile had nearest-rank p50/p95 6.384/6.644 s and all 634 execution-detail rows on CPU; 5.917 s estimated time is the sample minimum. This is one component, without a listening-quality gate. Corrected DirectML inference and profile both failed at provider initialization with Windows error 80070057, producing no output or timing. A separate FP16 QNN DLC compile passed, but same-fixture NPU-requested inference was pending. Historical QNN talker profiling lacks a same-fixture numerical or full-pipeline gate. The older TFLite vocoder target was rejected on Windows ARM; no complete stream exists. | Audit the explicit QNN NPU inference, investigate the tested DirectML provider failure, and qualify talker/predictor plus complete local streaming, memory and sustained behavior; [vocoder evidence](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md). |
|
||||
| PC Snapdragon X Elite CRD / 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. |
|
||||
| PC Snapdragon X Elite CRD / AI Hub | MiniCPM-o 4.5 | 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. |
|
||||
| PC Snapdragon X Elite CRD / AI Hub | InternVLA-A1 | 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. |
|
||||
@@ -91,19 +93,19 @@ NOT E2E means no complete device/model workload was verified. “Scoped E2E” i
|
||||
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | Qwen3-TTS 0.6B CustomVoice | NOT_E2E_PROFILEABLE | Fresh FP16 predictor inference passed; its one output exactly matches the same compiled artifact/dataset reference. Historical decoder parity exists. No complete co-resident NPU/GPU stream. | Validate the device-local talker/predictor/vocoder pipeline, output quality, recovery and 30-minute thermal behavior. |
|
||||
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | 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. |
|
||||
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | MiniCPM-o 4.5 | NOT_E2E_PROFILEABLE | Fresh FP16 speech-head inference passed; all 41 outputs exactly match the same artifact/dataset reference. Historical FP16 numerical parity passed; tested W8A16 top-1 parity failed. Thinker/encoders/vocoder are not qualified together. | Supply a qualified complete mobile checkpoint/export and validate the full multimodal/stateful pipeline. |
|
||||
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | InternVLA-A1 | 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. |
|
||||
| Mobile Galaxy S25 / Snapdragon 8 Elite for Galaxy | InternVLA-A1 | COMPONENT_NPU_ARTIFACT_COMPILE_PASS; INFERENCE_PENDING; NOT_E2E_PROFILEABLE | The real Place_Markpen Cosmos image encoder FP32 ONNX export matched its source on two six-frame fixtures (relative L2 8.74e-7/2.29e-6) after packaging external weights into a single file. Galaxy S25 FP16 QNN DLC compilation passed, but same-fixture NPU-requested inference was pending. There is no measured NPU placement, latent/action-policy parity, complete qualified target bundle or device-local controller; ADB has no attached device. | Audit the [S25 encoder inference](../../../e2e_expansion_20260924/evidence/internvla_cosmos_qualcomm/README.md), then validate action quality, state, admission, cancellation and sustained behavior. |
|
||||
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | Spark-X2.5 | COMPONENT_NPU_INFERENCE_PROFILED; NOT_E2E_PROFILEABLE | The real 1.7B checkpoint's fixed-shape full-attention decoder **one-layer** W8A16 QDQ artifact compiled as QNN DLC and ran on an exact hosted Galaxy S24 (Android 14). A same-dataset inference returned finite hidden/new-K/new-V outputs identical to the prior S25 NPU outputs; the profile attributed all 73 execution-detail rows to NPU. Across 100 component samples, nearest-rank p50/p95 was 1.109/1.205 ms; the 1.098 ms `estimated_inference_time` field is the minimum, not a median. Relative L2 versus original FP32 ONNX CPU was 1.351%/0.377%/0.650% for hidden/K/V; no token-quality tolerance is known. This is one cache bucket, not 28-layer prefill/decode, KV-ring, sampling, a resident local controller or complete-request timing. ADB has no attached device. | Qualify a complete device-resident artifact/controller with state and token-quality gates, 128-token generation across 512/1024 context transitions, admission/cancel and sustained memory/power; [S24 NPU raw evidence](../../../e2e_expansion_20260923/evidence/spark_s24_full_attention/README.md). |
|
||||
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | Qwen3-TTS 0.6B CustomVoice | NOT_E2E_PROFILEABLE | Historical FP16 decoder parity has SNR 39.13 dB. Tested calibrated INT8 decoder parity failed with negative SNR. Neither result is a full local TTS stream. | Validate the complete local pipeline and a quality-passing artifact set. |
|
||||
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | 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. |
|
||||
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | MiniCPM-o 4.5 | NOT_E2E_PROFILEABLE | Historical FP16 speech-head job re-fetched successfully; SNR 49.22 dB/top-1 match in the recorded parity test. W8A16 top-1 failed. Not complete MiniCPM-o. | Qualify thinker/encoders/vocoder and stateful local integration. |
|
||||
| Mobile Galaxy S24 / Snapdragon 8 Gen 3 | InternVLA-A1 | 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 | Spark-X2.5 | COMPONENT_NPU_INFERENCE_PROFILED; NOT_E2E_PROFILEABLE | A real Spark-X2.5-1.7B one-layer W8A16 QDQ attention artifact, compiled as QNN DLC, ran on the exact hosted SA8775P ADP (Android 14). Inference returned finite hidden/new-K/new-V tensors bitwise identical to the prior S24/S25 NPU outputs on the same fixture; the profile attributed all 73 execution-detail rows to NPU. Across 100 component samples, nearest-rank p50/p95 was 1.837/2.079 ms; the 1.749 ms estimated-time field is the minimum. Hidden/K/V relative L2 versus original FP32 ONNX CPU was 1.351%/0.377%/0.650%, without a next-token quality tolerance. There is no complete 28-layer prefill/decode/KV-ring/sampling loop, resident Omni controller or complete-request timing. | Qualify a complete embedded resident artifact/controller with state, token quality, memory admission/cancel and sustained power; [SA8775P NPU raw evidence](../../../e2e_expansion_20260923/evidence/spark_sa8775_full_attention/README.md). |
|
||||
| Embedded SA8775P ADP / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_VOCODER_NUMERIC_FAIL; NOT_E2E_PROFILEABLE | Historical NPU talker and mixed CPU+GPU vocoder profiles exist without a complete pipeline. A fresh GPU-requested Code2Wav inference on the retained fixture saturated 96.63% of waveform samples, with 11.832 relative L2 and -21.46 dB SNR versus original ONNX CPU; this tested vocoder path is numerically unqualified. The fresh profile has 100 samples at nearest-rank p50/p95 1.776/1.880 s with 403 CPU and 335 GPU execution-detail rows, not an all-GPU stage; it times a numerically failed output. CPU/FP32-GPU control jobs were pending. | Isolate delegate/precision failure, establish waveform and listening quality, then validate co-resident talker/predictor/vocoder, streaming, memory and sustained behavior; [raw vocoder evidence](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md). |
|
||||
| Embedded SA8775P ADP / AI Hub | Qwen3-TTS 0.6B CustomVoice | COMPONENT_CPU_REQUESTED_VOCODER_PARITY_PASS; COMPONENT_GPU_NUMERIC_FAIL; NOT_E2E_PROFILEABLE | On the exact SA8775P, the same TFLite binary's CPU-requested Code2Wav inference matched original ONNX CPU on the retained fixture within 4.13e-6 relative L2 / 107.67 dB SNR, with no saturation. Actual CPU placement/timing profile was pending. The GPU-requested output saturated 96.63% of samples, with 11.832 relative L2 / -21.46 dB SNR; its fresh 100-sample profile p50/p95 was 1.776/1.880 s with 403 CPU and 335 GPU execution-detail rows, timing a numerically failed waveform. Historical NPU talker profiling lacks a full pipeline. | Audit the pending CPU profile and FP32-preserving GPU control, establish listening quality, then validate co-resident talker/predictor/vocoder, streaming, memory and sustained behavior; [raw vocoder evidence](../../../e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/README.md). |
|
||||
| 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 | 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 | InternVLA-A1 | 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 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_VOCODER_INFERENCE_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 historical mixed CPU+GPU profile re-fetched with 100 samples at nearest-rank p50/p95 3.143/3.189 s and 403 CPU plus 335 GPU execution-detail rows. Its estimated-time field 3.107 s is the minimum, not median. The fresh profile was pending. The historical NPU talker build was rejected for floating-point inputs; no full model pipeline exists. | Audit the fresh profile, 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-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. |
|
||||
| Embedded RB3 Gen 2 / QCS6490 / AI Hub | MiniCPM-o 4.5 | 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 RB3 Gen 2 / QCS6490 / AI Hub | InternVLA-A1 | 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. |
|
||||
|
||||
Reference in new issue
Block a user