vh
a6c1d3c454
A/B of the live STT seat (fv-ml1 GPU 0, sherpa-onnx int8 v3) against nvidia/parakeet-unified-en-0.6b, measured on GPU 3 with the seat's own image, k2-fsa's published unified int8 export, fp32/fp16 exports made with k2-fsa's recipe, v2 int8, and NeMo 3.0.0 (fp32, bf16 autocast, bf16 weights). - Seat int8 graph runs on one CPU thread (cpu/wall 1.00, GPU 2-9%). - unified-en under NeMo: -121/-234/-530 ms vs the seat at 1-3/3-8/8-20 s (paired, n=120/bin; floor <=6 ms; +50 ms positive control reads +52-54). - unified-en WER lower in every runtime: -0.7 pp clean, -1.5 pp other, -3.2 to -4.4 pp AMI (paired CIs exclude 0). - Seat defects found: hard 400 s input ceiling (HTTP 500), truncation after a quiet 1.5 s pause, and severe long-window dropouts (int8 v3 only). - B-bf16w needs +0.8 to +1.5 GB over the seat's 1,690 MiB on GPU 0. Raw requests, hypotheses, manifests and the full harness under services/parakeet-ab-2026-09-30/. No deploy; live seat untouched apart from 240 light test requests.
49 lines
4.2 KiB
JSON
49 lines
4.2 KiB
JSON
{"arm": "ab-b32", "phase": "rest-after-warmup", "t0": 1790816217.9522808, "t1": 1790816220.952287}
|
|
{"arm": "ab-b32c", "phase": "rest-after-warmup", "t0": 1790816238.7243755, "t1": 1790816241.7243812}
|
|
{"arm": "ab-b16w", "phase": "rest-after-warmup", "t0": 1790816260.0628917, "t1": 1790816263.0628967}
|
|
{"arm": "ab-b16", "phase": "rest-after-warmup", "t0": 1790816284.5180502, "t1": 1790816287.5180552}
|
|
{"arm": "ab-cf32", "phase": "rest-after-warmup", "t0": 1790816348.5764616, "t1": 1790816351.5764678}
|
|
{"arm": "ab-af32", "phase": "rest-after-warmup", "t0": 1790816409.543435, "t1": 1790816412.543441}
|
|
{"arm": "ab-af32", "phase": "lat-r0", "t0": 1790816412.9853637, "t1": 1790816419.8702936}
|
|
{"arm": "ab-b32", "phase": "lat-r0", "t0": 1790816419.8705037, "t1": 1790816423.7948604}
|
|
{"arm": "ab-b16", "phase": "lat-r0", "t0": 1790816423.795097, "t1": 1790816427.9718645}
|
|
{"arm": "ab-b16w", "phase": "lat-r0", "t0": 1790816427.9720802, "t1": 1790816431.769075}
|
|
{"arm": "ab-b32c", "phase": "lat-r0", "t0": 1790816431.7693086, "t1": 1790816435.7430315}
|
|
{"arm": "ab-cf32", "phase": "lat-r0", "t0": 1790816435.7432907, "t1": 1790816443.4020047}
|
|
{"arm": "ab-b16w", "phase": "lat-r1", "t0": 1790816443.4022298, "t1": 1790816446.8991585}
|
|
{"arm": "ab-af32", "phase": "lat-r1", "t0": 1790816446.8993726, "t1": 1790816453.7491817}
|
|
{"arm": "ab-b16", "phase": "lat-r1", "t0": 1790816453.7493386, "t1": 1790816457.4571297}
|
|
{"arm": "ab-b32", "phase": "lat-r1", "t0": 1790816457.457347, "t1": 1790816461.1599212}
|
|
{"arm": "ab-b32c", "phase": "lat-r1", "t0": 1790816461.1601648, "t1": 1790816464.6924193}
|
|
{"arm": "ab-cf32", "phase": "lat-r1", "t0": 1790816464.6926153, "t1": 1790816472.223418}
|
|
{"arm": "ab-b16", "phase": "lat-r2", "t0": 1790816472.2237296, "t1": 1790816475.8168836}
|
|
{"arm": "ab-b32", "phase": "lat-r2", "t0": 1790816475.8171268, "t1": 1790816479.7100127}
|
|
{"arm": "ab-b16w", "phase": "lat-r2", "t0": 1790816479.7102478, "t1": 1790816482.9155467}
|
|
{"arm": "ab-b32c", "phase": "lat-r2", "t0": 1790816482.9157612, "t1": 1790816486.4219022}
|
|
{"arm": "ab-af32", "phase": "lat-r2", "t0": 1790816486.4221354, "t1": 1790816493.3614554}
|
|
{"arm": "ab-cf32", "phase": "lat-r2", "t0": 1790816493.3616917, "t1": 1790816500.8425043}
|
|
{"arm": "ab-b32", "phase": "lat-r3", "t0": 1790816500.8428423, "t1": 1790816504.3418753}
|
|
{"arm": "ab-cf32", "phase": "lat-r3", "t0": 1790816504.3420897, "t1": 1790816511.8687708}
|
|
{"arm": "ab-b16w", "phase": "lat-r3", "t0": 1790816511.8689802, "t1": 1790816515.484867}
|
|
{"arm": "ab-b16", "phase": "lat-r3", "t0": 1790816515.485102, "t1": 1790816519.2602904}
|
|
{"arm": "ab-b32c", "phase": "lat-r3", "t0": 1790816519.2605364, "t1": 1790816522.768376}
|
|
{"arm": "ab-af32", "phase": "lat-r3", "t0": 1790816522.76861, "t1": 1790816529.5814555}
|
|
{"arm": "ab-b16w", "phase": "lat-r4", "t0": 1790816529.5817509, "t1": 1790816532.7475722}
|
|
{"arm": "ab-cf32", "phase": "lat-r4", "t0": 1790816532.7478075, "t1": 1790816540.445296}
|
|
{"arm": "ab-b16", "phase": "lat-r4", "t0": 1790816540.4454498, "t1": 1790816543.9971998}
|
|
{"arm": "ab-b32", "phase": "lat-r4", "t0": 1790816543.9974234, "t1": 1790816547.4305873}
|
|
{"arm": "ab-af32", "phase": "lat-r4", "t0": 1790816547.4308534, "t1": 1790816554.350192}
|
|
{"arm": "ab-b32c", "phase": "lat-r4", "t0": 1790816554.350425, "t1": 1790816558.0701263}
|
|
{"arm": "ab-af32", "phase": "lat-r5", "t0": 1790816558.0704808, "t1": 1790816564.9396}
|
|
{"arm": "ab-cf32", "phase": "lat-r5", "t0": 1790816564.9398334, "t1": 1790816572.5606947}
|
|
{"arm": "ab-b16", "phase": "lat-r5", "t0": 1790816572.5609298, "t1": 1790816576.3371167}
|
|
{"arm": "ab-b32", "phase": "lat-r5", "t0": 1790816576.3373232, "t1": 1790816579.774904}
|
|
{"arm": "ab-b16w", "phase": "lat-r5", "t0": 1790816579.7751667, "t1": 1790816582.9084818}
|
|
{"arm": "ab-b32c", "phase": "lat-r5", "t0": 1790816582.9087172, "t1": 1790816586.3651376}
|
|
{"arm": "ab-b32", "phase": "conc4", "t0": 1790816586.3654053, "t1": 1790816593.2398071}
|
|
{"arm": "ab-b32c", "phase": "conc4", "t0": 1790816593.2400835, "t1": 1790816600.0950725}
|
|
{"arm": "ab-b16w", "phase": "conc4", "t0": 1790816600.095266, "t1": 1790816606.2720919}
|
|
{"arm": "ab-b16", "phase": "conc4", "t0": 1790816606.272367, "t1": 1790816613.296962}
|
|
{"arm": "ab-cf32", "phase": "conc4", "t0": 1790816613.297279, "t1": 1790816628.326475}
|
|
{"arm": "ab-af32", "phase": "conc4", "t0": 1790816628.3267124, "t1": 1790816642.1572175}
|