ruv 575ee4d2eb feat(benchmarks): static PTQ int8 (calibrated) results + overnight capture script ...

Conv-only static QDQ beats dynamic int8 on accuracy (PCK@20 96.61-96.63%
vs 96.52%, MPJPE +10% vs +18% over fp32) at ~equal size/latency; all-ops
QDQ strictly worse (int8 activations through attention glue). Entropy
calibration verified bit-identical to MinMax on this data. Deployment:
ONNX fp32 for speed (3.2ms), static conv-only QDQ for smallest (2.53MB).

Also: scripts/overnight-empty-capture.py — segmented UDP CSI recorder for
empty-room baselines (no glob collisions, detach-safe).

Co-Authored-By: claude-flow <ruv@ruv.net>

2026-06-10 23:04:38 -04:00

..

remote

feat(benchmarks): WiFlow-STD reproduction harness + measurement (a) results (ADR-152 §2.2)

2026-06-10 18:54:03 -04:00

results

feat(benchmarks): static PTQ int8 (calibrated) results + overnight capture script

2026-06-10 23:04:38 -04:00

.gitignore

feat(benchmarks): edge optimization measured + measurement (b) blocked + 92.9% retraction

2026-06-10 22:05:39 -04:00

eval_ort_accuracy.py

feat(benchmarks): edge optimization measured + measurement (b) blocked + 92.9% retraction

2026-06-10 22:05:39 -04:00

eval_repro.py

feat(benchmarks): WiFlow-STD reproduction harness + measurement (a) results (ADR-152 §2.2)

2026-06-10 18:54:03 -04:00

export_to_safetensors.py

feat(train): WiFlow-STD PyTorch->tch weight import + numerical parity proof

2026-06-10 20:23:38 -04:00

onnx_bench.py

feat(benchmarks): edge optimization measured + measurement (b) blocked + 92.9% retraction

2026-06-10 22:05:39 -04:00

quantize_bench.py

feat(benchmarks): edge optimization measured + measurement (b) blocked + 92.9% retraction

2026-06-10 22:05:39 -04:00

RESULTS.md

feat(benchmarks): static PTQ int8 (calibrated) results + overnight capture script

2026-06-10 23:04:38 -04:00

static_ptq_bench.py

feat(benchmarks): static PTQ int8 (calibrated) results + overnight capture script

2026-06-10 23:04:38 -04:00