nirnay / code /bench /results /nirnay_latency.json
nandukmelath's picture
Nirnay v2 (bge-base, default) + v1 (fast/): weights, package, code, benchmarks
00ee773 verified
Raw History Blame Contribute Delete
6.49 kB
{
"meta": {
"timestamp": "2026-10-24 13:26:46",
"cpu": "2x Xeon E5-2630 v3 (HP Z840)",
"model": "nirnay-v1",
"model_dir": "D:\\Nirnay\\models\\nirnay-v1"
},
"cold_load_ms": 3172.2,
"rss_delta_mb": 239.8,
"disk_mb": 155.0,
"by_threads": {
"1": {
"1_questions": {
"warm": {
"p50_ms": 155.79,
"p95_ms": 157.33,
"mean_ms": 155.9,
"min_ms": 154.0
},
"cold_schema": {
"p50_ms": 208.13,
"p95_ms": 221.99,
"mean_ms": 210.47,
"min_ms": 205.82
},
"ms_per_question_warm": 155.79
},
"5_questions": {
"warm": {
"p50_ms": 227.52,
"p95_ms": 282.88,
"mean_ms": 239.52,
"min_ms": 223.26
},
"cold_schema": {
"p50_ms": 422.09,
"p95_ms": 456.87,
"mean_ms": 427.43,
"min_ms": 414.39
},
"ms_per_question_warm": 45.504
},
"10_questions": {
"warm": {
"p50_ms": 294.01,
"p95_ms": 335.97,
"mean_ms": 301.52,
"min_ms": 291.33
},
"cold_schema": {
"p50_ms": 627.96,
"p95_ms": 694.3,
"mean_ms": 639.69,
"min_ms": 619.22
},
"ms_per_question_warm": 29.401
},
"50_questions": {
"warm": {
"p50_ms": 978.39,
"p95_ms": 1319.05,
"mean_ms": 1044.65,
"min_ms": 962.45
},
"cold_schema": {
"p50_ms": 2501.77,
"p95_ms": 2679.58,
"mean_ms": 2539.42,
"min_ms": 2486.15
},
"ms_per_question_warm": 19.568
}
},
"4": {
"1_questions": {
"warm": {
"p50_ms": 59.8,
"p95_ms": 66.91,
"mean_ms": 61.15,
"min_ms": 59.27
},
"cold_schema": {
"p50_ms": 84.99,
"p95_ms": 91.05,
"mean_ms": 86.08,
"min_ms": 84.18
},
"ms_per_question_warm": 59.8
},
"5_questions": {
"warm": {
"p50_ms": 87.08,
"p95_ms": 89.14,
"mean_ms": 87.32,
"min_ms": 85.9
},
"cold_schema": {
"p50_ms": 161.54,
"p95_ms": 163.33,
"mean_ms": 159.93,
"min_ms": 153.22
},
"ms_per_question_warm": 17.416
},
"10_questions": {
"warm": {
"p50_ms": 111.85,
"p95_ms": 122.66,
"mean_ms": 114.12,
"min_ms": 109.99
},
"cold_schema": {
"p50_ms": 228.91,
"p95_ms": 252.13,
"mean_ms": 231.37,
"min_ms": 216.58
},
"ms_per_question_warm": 11.185
},
"50_questions": {
"warm": {
"p50_ms": 344.2,
"p95_ms": 371.43,
"mean_ms": 344.91,
"min_ms": 329.21
},
"cold_schema": {
"p50_ms": 808.22,
"p95_ms": 820.51,
"mean_ms": 809.48,
"min_ms": 797.08
},
"ms_per_question_warm": 6.884
}
},
"8": {
"1_questions": {
"warm": {
"p50_ms": 54.23,
"p95_ms": 58.44,
"mean_ms": 54.75,
"min_ms": 52.42
},
"cold_schema": {
"p50_ms": 80.75,
"p95_ms": 88.41,
"mean_ms": 81.55,
"min_ms": 75.4
},
"ms_per_question_warm": 54.23
},
"5_questions": {
"warm": {
"p50_ms": 76.11,
"p95_ms": 84.53,
"mean_ms": 76.35,
"min_ms": 70.96
},
"cold_schema": {
"p50_ms": 151.79,
"p95_ms": 159.15,
"mean_ms": 152.76,
"min_ms": 147.3
},
"ms_per_question_warm": 15.222
},
"10_questions": {
"warm": {
"p50_ms": 94.79,
"p95_ms": 109.49,
"mean_ms": 96.91,
"min_ms": 88.94
},
"cold_schema": {
"p50_ms": 194.29,
"p95_ms": 207.88,
"mean_ms": 194.25,
"min_ms": 182.81
},
"ms_per_question_warm": 9.479
},
"50_questions": {
"warm": {
"p50_ms": 258.01,
"p95_ms": 283.36,
"mean_ms": 260.65,
"min_ms": 236.94
},
"cold_schema": {
"p50_ms": 572.9,
"p95_ms": 629.95,
"mean_ms": 578.51,
"min_ms": 544.46
},
"ms_per_question_warm": 5.16
}
},
"16": {
"1_questions": {
"warm": {
"p50_ms": 58.69,
"p95_ms": 59.86,
"mean_ms": 58.08,
"min_ms": 53.37
},
"cold_schema": {
"p50_ms": 79.2,
"p95_ms": 93.85,
"mean_ms": 81.18,
"min_ms": 73.67
},
"ms_per_question_warm": 58.69
},
"5_questions": {
"warm": {
"p50_ms": 81.79,
"p95_ms": 90.32,
"mean_ms": 83.24,
"min_ms": 79.45
},
"cold_schema": {
"p50_ms": 158.53,
"p95_ms": 167.37,
"mean_ms": 156.65,
"min_ms": 145.52
},
"ms_per_question_warm": 16.358
},
"10_questions": {
"warm": {
"p50_ms": 106.15,
"p95_ms": 116.47,
"mean_ms": 108.42,
"min_ms": 102.73
},
"cold_schema": {
"p50_ms": 221.71,
"p95_ms": 232.53,
"mean_ms": 219.72,
"min_ms": 197.73
},
"ms_per_question_warm": 10.615
},
"50_questions": {
"warm": {
"p50_ms": 278.8,
"p95_ms": 309.64,
"mean_ms": 272.85,
"min_ms": 229.42
},
"cold_schema": {
"p50_ms": 600.19,
"p95_ms": 660.85,
"mean_ms": 608.41,
"min_ms": 572.76
},
"ms_per_question_warm": 5.576
}
}
},
"backend": "torch",
"throughput_10q": {
"states": 64,
"seconds": 4.467,
"decisions_per_s": 143.3,
"states_per_s": 14.3
},
"peak_rss_mb": 564.9
}