Record pinned S25 and X Elite vocoder waveform evidence

This commit is contained in:
tzhouam committed 2026-09-24 14:38:51 +08:00
1 parent 203039e054
commit 4386562f9b
13 files changed
+291 -25

No files matched your search

+8 -4
View File
@@ -45,9 +45,11 @@ waveform at 1.89e-6 relative L2 on the same fixture, narrowing the model
continuity question without proving the historical export revision.
A [fresh revision-pinned 25-frame export](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/s25/full_pinned_tflite/README.md)
now matches local eager CPU at 1.99e-6 relative L2 and has compiled for S25
TFLite. Its GPU-requested numerical inference and node-placement profile are
pending. This avoids promoting the historical artifact's timing to a
revision-qualified deployment claim.
TFLite. Its GPU-requested inference returned an unsaturated waveform at
4.239% relative L2 versus pinned local ONNX CPU, bitwise identical to the
historical target's output on the same fixture. The new target's own
node-placement profile is pending, so the historical timing does not yet
qualify this revision-pinned artifact.
On [Galaxy S24](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/s24/tflite_short_chunk/README.md),
the same pinned two-frame TFLite source passed CPU-requested waveform parity
at 5.75e-6 relative L2, but the GPU-requested waveform saturated completely
@@ -56,7 +58,9 @@ at 24.218 relative L2. The 100-sample short CPU component p50/p95 was
measured 0.601/0.895 s but failed the waveform gate. A FP32-preserving GPU
option returned the same saturated waveform bitwise, so it did not repair
the tested delegate artifact. A [Snapdragon X Elite short QNN graph](../../benchmarks/edge_harness/results/e2e_expansion_20260924/evidence/qwen_tts_vocoder_qualcomm/xelite/qnn_short_chunk/README.md)
compiled, with NPU-requested inference pending. Neither changes M2 support
compiled and returned an unsaturated NPU-requested waveform at 0.930%
relative L2 versus local ONNX CPU on one synthetic fixture. Its node-placement
profile is pending. Neither changes M2 support
depth before full stream and quality gates.
The [Spark hosted Qualcomm component expansion](../../benchmarks/edge_harness/results/e2e_expansion_20260923/evidence/spark_s24_full_attention/README.md) advances M1 only to one real fixed-shape decoder attention layer. The same W8A16 QNN DLC ran on S24, Snapdragon X Elite CRD and SA8775P ADP NPUs with identical outputs on one pinned fixture; S25 had prior separate component evidence. On [RB3 Gen 2](../../benchmarks/edge_harness/results/e2e_expansion_20260923/evidence/spark_rb3_full_attention/README.md), the tested W8A16 graph failed QNN loading even after an exact-device compile. FP32 ONNX ran on CPU with near-source parity, while calibrated W8A8 ran on NPU but had 20.9% hidden-state relative L2 and remains numerically unqualified. These are component C results, not full M1: device-local 28-layer prefill/continuous decode, KV/ring, sampling, token quality, admission/cancellation, complete-request timing and sustained power/thermal gates remain open. AI Hub is the test facility, not a deployment dependency.