diff --git a/runs/frontier-fidelity-envelope-v1/.gitignore b/runs/frontier-fidelity-envelope-v1/.gitignore index 71cf748..acd9559 100644 --- a/runs/frontier-fidelity-envelope-v1/.gitignore +++ b/runs/frontier-fidelity-envelope-v1/.gitignore @@ -4,6 +4,9 @@ a2-measured-fix-smoke/ simulator-a1/ simulator-a2/ simulator-a3/ +simulator-f1/ +simulator-f2/ +simulator-t1/ fleet-state/ fleet-artifacts/ profiles/profile-v3-batch/ diff --git a/runs/frontier-fidelity-envelope-v1/results/a3/capacity.csv b/runs/frontier-fidelity-envelope-v1/results/a3/capacity.csv new file mode 100644 index 0000000..da354c9 --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/results/a3/capacity.csv @@ -0,0 +1,13 @@ +config,tp,mns,real,simulator +tp1_mns8,1,8,7.0,8.0 +tp1_mns16,1,16,7.0,8.0 +tp1_mns32,1,32,7.0,8.0 +tp1_mns64,1,64,7.0,8.0 +tp2_mns8,2,8,7.0,8.0 +tp2_mns16,2,16,7.0,8.0 +tp2_mns32,2,32,7.0,8.0 +tp2_mns64,2,64,7.0,8.0 +tp4_mns8,4,8,8.0,8.0 +tp4_mns16,4,16,8.0,8.0 +tp4_mns32,4,32,8.0,8.0 +tp4_mns64,4,64,8.0,8.0 diff --git a/runs/frontier-fidelity-envelope-v1/results/a3/comparison.json b/runs/frontier-fidelity-envelope-v1/results/a3/comparison.json new file mode 100644 index 0000000..b011206 --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/results/a3/comparison.json @@ -0,0 +1,5638 @@ +{ + "contract": { + "input_tokens": 2048, + "model": "Qwen3-30B-A3B", + "output_tokens": 1, + "prefix_caching": false, + "real_anchor_merge": "both_fresh_server_rounds_must_pass", + "target_pass_rate": 0.95, + "ttft_slo_ms": 1256.0 + }, + "metrics": { + "anchor_confusion": { + "real_fail_sim_fail": 24, + "real_fail_sim_pass": 8, + "real_pass_sim_fail": 0, + "real_pass_sim_pass": 28 + }, + "anchor_grid_coverage": { + "real_only": 36, + "shared": 60, + "simulator_only": 0 + }, + "config_order": [ + "tp1_mns8", + "tp1_mns16", + "tp1_mns32", + "tp1_mns64", + "tp2_mns8", + "tp2_mns16", + "tp2_mns32", + "tp2_mns64", + "tp4_mns8", + "tp4_mns16", + "tp4_mns32", + "tp4_mns64" + ], + "kendall": { + "both_ties": 34, + "concordant": 0, + "discordant": 0, + "kendall_tau_b": null, + "real_only_ties": 0, + "simulator_only_ties": 32 + }, + "pairwise_non_tied": { + "all": { + "accuracy": 0.0, + "comparable": 32, + "correct": 0 + }, + "within_tp": {} + }, + "real_best_capacity_per_gpu": 8.0, + "real_top_set": [ + "tp4_mns8", + "tp4_mns16", + "tp4_mns32", + "tp4_mns64" + ], + "simulator_best_capacity_per_gpu": 8.0, + "simulator_top_set": [ + "tp1_mns8", + "tp1_mns16", + "tp1_mns32", + "tp1_mns64", + "tp2_mns8", + "tp2_mns16", + "tp2_mns32", + "tp2_mns64", + "tp4_mns8", + "tp4_mns16", + "tp4_mns32", + "tp4_mns64" + ], + "top1_regret_best": 0.0, + "top1_regret_worst": 0.125, + "top_set_exact_match": false, + "top_set_overlap": [ + "tp4_mns16", + "tp4_mns32", + "tp4_mns64", + "tp4_mns8" + ] + }, + "objective": "maximum_tested_slo_feasible_offered_request_rate_per_gpu", + "real": { + "tp1_mns16": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 153.15452707000077, + 152.64549700077623 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r4p00.json", + "sha256": "d82e386ec3effdf57c148171311928298405da0e4f896584bf05db2485bb6bd4", + "summary": { + "admission_lag_max_ms": 0.11588609777390957, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 158.6234939750284, + "ttft_p50_ms": 151.05855697765946, + "ttft_p95_ms": 153.15452707000077 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r4p00.json", + "sha256": "1565be89131e3c8a9c352c6f29694ef2793734b0623a2db1cbea7e0b0e8a2bf6", + "summary": { + "admission_lag_max_ms": 0.11451996397227049, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.18561100866646, + "ttft_p50_ms": 151.18564700242132, + "ttft_p95_ms": 152.64549700077623 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 5.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 152.8308589477092, + 152.43642206769437 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r5p00.json", + "sha256": "e4447d94a326c541412b40157cfe283a6350680e3285b5d10fed6b8de5c0cd93", + "summary": { + "admission_lag_max_ms": 0.17109804321080446, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 154.35138496104628, + "ttft_p50_ms": 150.91424097772688, + "ttft_p95_ms": 152.8308589477092 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r5p00.json", + "sha256": "572c17e309c6c9f23c490979e7acb885475f142bc530ecfee88916b048259ea2", + "summary": { + "admission_lag_max_ms": 0.1321269664913416, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 157.11930498946458, + "ttft_p50_ms": 150.50161199178547, + "ttft_p95_ms": 152.43642206769437 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 6.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 153.56597595382482, + 152.16859406791627 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r6p00.json", + "sha256": "68c8b39154d46f1ceda17f47e8cbd84a8c8a9f088c5e82d8f0ac25a7180dc14c", + "summary": { + "admission_lag_max_ms": 0.11641765013337135, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 156.50876995641738, + "ttft_p50_ms": 150.41159105021507, + "ttft_p95_ms": 153.56597595382482 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r6p00.json", + "sha256": "a970a2cdf7034026a56af3ea4da59856fd588da35d66ef4f9e15481a892edcc9", + "summary": { + "admission_lag_max_ms": 0.11551892384886742, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 168.65782206878066, + "ttft_p50_ms": 150.03317291848361, + "ttft_p95_ms": 152.16859406791627 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 7.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 300.90478900820017, + 293.0445579113439 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r7p00.json", + "sha256": "1aa74c18adbff53b36ae8a6f07b5b0865f6b68fb3e7862a70b36f67c18f0bba0", + "summary": { + "admission_lag_max_ms": 0.10505830869078636, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 304.15459000505507, + "ttft_p50_ms": 263.44375393819064, + "ttft_p95_ms": 300.90478900820017 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r7p00.json", + "sha256": "e14bc9dba35d453862839eeed77262a1ed9b8394404ca90c60ec8cb26a1aa73b", + "summary": { + "admission_lag_max_ms": 0.10963028762489557, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 295.1545650139451, + "ttft_p50_ms": 243.97043196950108, + "ttft_p95_ms": 293.0445579113439 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 8.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1323.4397460473701, + 1315.5151740647852 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r8p00.json", + "sha256": "9aeee7453ffe5098d203e02b84f5009ed347c478b736e89297bd3695dd81d477", + "summary": { + "admission_lag_max_ms": 0.1251589274033904, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.90625, + "passed": 58, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1477.699184906669, + "ttft_p50_ms": 776.2576789828017, + "ttft_p95_ms": 1323.4397460473701 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r8p00.json", + "sha256": "7630f4901cf4ec4ea955738b119a1c32aba53f9ca1edb8819556fc60ae1fda09", + "summary": { + "admission_lag_max_ms": 0.11316593736410141, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.921875, + "passed": 59, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1404.566578916274, + "ttft_p50_ms": 761.7035870207474, + "ttft_p95_ms": 1315.5151740647852 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4892.253074911423, + 4893.758205929771 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r16p00.json", + "sha256": "d3a2a85cf06a4f080edc9a3809f05493ac55b0aa1d890806ffbc551c8eca14f8", + "summary": { + "admission_lag_max_ms": 0.39830803871154785, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4971.215195022523, + "ttft_p50_ms": 2742.7029330283403, + "ttft_p95_ms": 4892.253074911423 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r16p00.json", + "sha256": "67110284c38a9a60228f00b1d2b0de9477d6dec660f1c91eb4f39e1b79e272a5", + "summary": { + "admission_lag_max_ms": 0.11306500528007746, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4972.818040987477, + "ttft_p50_ms": 2742.512951954268, + "ttft_p95_ms": 4893.758205929771 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 6650.756464921869, + 6751.845097984187 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r32p00.json", + "sha256": "affc62b27382400497bb82e9da31ed9891b60155318fc4f250dca0704ccc5e02", + "summary": { + "admission_lag_max_ms": 0.11010200250893831, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6920.621327939443, + "ttft_p50_ms": 3650.839229929261, + "ttft_p95_ms": 6650.756464921869 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r32p00.json", + "sha256": "df0c8d703b023692814be41f47da4db4ce1b7eb0c3423a742c10884b6447583f", + "summary": { + "admission_lag_max_ms": 0.1288279891014099, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.15625, + "passed": 10, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6897.979240980931, + "ttft_p50_ms": 3730.34580796957, + "ttft_p95_ms": 6751.845097984187 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 7732.2855240199715, + 7744.056030060165 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round1/results/r64p00.json", + "sha256": "fbfad78129c843c85905c3f94f278f468a0a019c0cdf0b62140d0fea8f0aaffc", + "summary": { + "admission_lag_max_ms": 0.11090096086263657, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7836.245508980937, + "ttft_p50_ms": 4328.587644966319, + "ttft_p95_ms": 7732.2855240199715 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z/artifacts/artifacts/real-tp1-mns16-v1/round2/results/r64p00.json", + "sha256": "0f005c642327708e1ceb058f71f806b4d46ea666f45695605552e8f234521608", + "summary": { + "admission_lag_max_ms": 0.10862003546208143, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7847.871358972043, + "ttft_p50_ms": 4332.9262170009315, + "ttft_p95_ms": 7744.056030060165 + } + } + ] + } + ], + "capacity": 7.0, + "capacity_per_gpu": 7.0, + "mns": 16, + "name": "tp1_mns16", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z", + "tp": 1 + }, + "tp1_mns32": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 153.44141807872802, + 152.2580359596759 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r4p00.json", + "sha256": "211fa276e4a99a061126882085b9850549b453520ac3414dd2d332c89e52b431", + "summary": { + "admission_lag_max_ms": 0.09491504170000553, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 169.56262104213238, + "ttft_p50_ms": 150.8563900133595, + "ttft_p95_ms": 153.44141807872802 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r4p00.json", + "sha256": "17428902a5b20e9084f20792963f0620a1af06cdc1afb76ee99fd15dd905b89e", + "summary": { + "admission_lag_max_ms": 0.12033001985400915, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.99700601119548, + "ttft_p50_ms": 150.5822929320857, + "ttft_p95_ms": 152.2580359596759 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 5.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 152.05012692604214, + 152.00478909537196 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r5p00.json", + "sha256": "b0eb19ff7ab8a26e34e91f213a983e58689c63071285e27b3719bda0a765ff32", + "summary": { + "admission_lag_max_ms": 0.11265103239566088, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.48999004345387, + "ttft_p50_ms": 150.21017810795456, + "ttft_p95_ms": 152.05012692604214 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r5p00.json", + "sha256": "3b857dff42dd16ebee95356ceacd6b58ff7bd0a9b91f3d6208d489b84985697f", + "summary": { + "admission_lag_max_ms": 0.11149293277412653, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.33813393954188, + "ttft_p50_ms": 150.45341395307332, + "ttft_p95_ms": 152.00478909537196 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 6.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 152.74227899499238, + 152.82704099081457 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r6p00.json", + "sha256": "7615fa664a99f02f9f6d4f1fb0f2d6cba11842772c086634b939bc08cd0f2f4d", + "summary": { + "admission_lag_max_ms": 0.1096969936043024, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 154.2669020127505, + "ttft_p50_ms": 150.6725640501827, + "ttft_p95_ms": 152.74227899499238 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r6p00.json", + "sha256": "345e3cd3f8cd4d7c7a5e9343f937dd9b1b5928500133830f3046b743da6df64e", + "summary": { + "admission_lag_max_ms": 0.12192933354526758, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.12304301187396, + "ttft_p50_ms": 150.68741200957447, + "ttft_p95_ms": 152.82704099081457 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 7.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 302.5522669777274, + 294.5978350471705 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r7p00.json", + "sha256": "3f54a3e3f2279a5325a173ff714d600fb1a2cdfcfba1e8e5c980f7b21b99be59", + "summary": { + "admission_lag_max_ms": 0.11122168507426977, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 304.24335307907313, + "ttft_p50_ms": 262.4604320153594, + "ttft_p95_ms": 302.5522669777274 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r7p00.json", + "sha256": "0832169cdf1bef49c5006464c6cbcd50f4567e0ab90bec98d722f64f7b476763", + "summary": { + "admission_lag_max_ms": 0.1059951027855277, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 297.1431679325178, + "ttft_p50_ms": 249.51713997870684, + "ttft_p95_ms": 294.5978350471705 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 8.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1321.9368209829554, + 1318.7565059633926 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r8p00.json", + "sha256": "a0f5900ac9d2c47dcc7d7ac08d3f3d23e1d6f529fdd0200989afa819dd3df055", + "summary": { + "admission_lag_max_ms": 0.17414195463061333, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.90625, + "passed": 58, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1475.8616799954325, + "ttft_p50_ms": 775.7647309917957, + "ttft_p95_ms": 1321.9368209829554 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r8p00.json", + "sha256": "c0fc3c85095ddacc66849e6f58be0851abab6d4022ee4eb7120c8cb8ea9e5d2c", + "summary": { + "admission_lag_max_ms": 2.749277977272868, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.921875, + "passed": 59, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1407.0930189918727, + "ttft_p50_ms": 763.8335369993001, + "ttft_p95_ms": 1318.7565059633926 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4893.3289990527555, + 4893.17379205022 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r16p00.json", + "sha256": "4e9ccaf5384f622b1da7c58d3de803c86099dd9deae91e363aabccc14aeabc05", + "summary": { + "admission_lag_max_ms": 0.11893699411302805, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4972.85822511185, + "ttft_p50_ms": 2742.8449960425496, + "ttft_p95_ms": 4893.3289990527555 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r16p00.json", + "sha256": "c207e415186aed1148e16085145bcdfd189314e7b92f74ea72fb2e4972ef4eae", + "summary": { + "admission_lag_max_ms": 0.09541702456772327, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4971.2590809213, + "ttft_p50_ms": 2740.232800017111, + "ttft_p95_ms": 4893.17379205022 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 6653.632597066462, + 6647.155530983582 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r32p00.json", + "sha256": "780f563a2739d892761905031f82ca8bd0ac4f4eb485c454390dcb2bcdfc1b81", + "summary": { + "admission_lag_max_ms": 0.105265062302351, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6922.717701992951, + "ttft_p50_ms": 3652.452490059659, + "ttft_p95_ms": 6653.632597066462 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r32p00.json", + "sha256": "2a5eb8fb37f03e759de149d24f9b23c812a26afe72ffaaffa030688904634e43", + "summary": { + "admission_lag_max_ms": 0.10375294368714094, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6917.111590038985, + "ttft_p50_ms": 3649.0042679943144, + "ttft_p95_ms": 6647.155530983582 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 7747.191094094887, + 7839.257381972857 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round1/results/r64p00.json", + "sha256": "b94a39a91842ab39d545c473fc4f872ec79eb1888943c3e03ab03c18961a9105", + "summary": { + "admission_lag_max_ms": 0.6591080455109477, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7851.370444986969, + "ttft_p50_ms": 4335.578468977474, + "ttft_p95_ms": 7747.191094094887 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z/artifacts/artifacts/real-tp1-mns32-v1/round2/results/r64p00.json", + "sha256": "69ab47305c4bb14f3b4ed5bc0b18a37b8a5e7c1cdbee9232e6d2b9616ed10f84", + "summary": { + "admission_lag_max_ms": 0.10753795504570007, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.125, + "passed": 8, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7885.349316056818, + "ttft_p50_ms": 4009.1016669757664, + "ttft_p95_ms": 7839.257381972857 + } + } + ] + } + ], + "capacity": 7.0, + "capacity_per_gpu": 7.0, + "mns": 32, + "name": "tp1_mns32", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z", + "tp": 1 + }, + "tp1_mns64": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 156.76817402709275, + 152.9719429090619 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r4p00.json", + "sha256": "df798dd7a1cde3cfaf698286f44bbac2b0730a2c806f125268bdf7226e2bb47c", + "summary": { + "admission_lag_max_ms": 0.1147780567407608, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 162.9874319769442, + "ttft_p50_ms": 152.67050196416676, + "ttft_p95_ms": 156.76817402709275 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r4p00.json", + "sha256": "df80057224d69ce2a79f937aa45c2093d66d131765f3c7bfcc2289a9aaff3d6a", + "summary": { + "admission_lag_max_ms": 0.2131829969584942, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 153.917470946908, + "ttft_p50_ms": 150.96506499685347, + "ttft_p95_ms": 152.9719429090619 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 5.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 153.4973110537976, + 155.23677493911237 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r5p00.json", + "sha256": "e614ffaf879859407fc6420847f0140a81bda9a94fcc438fb6782c9019bf5378", + "summary": { + "admission_lag_max_ms": 0.1283780438825488, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 155.08639602921903, + "ttft_p50_ms": 152.1085740532726, + "ttft_p95_ms": 153.4973110537976 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r5p00.json", + "sha256": "13e825eb5ced3d0cb2b5dd68d9243fdf1cee8779b90e9aad77be04053be22be2", + "summary": { + "admission_lag_max_ms": 0.1683390000835061, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 155.6691420264542, + "ttft_p50_ms": 152.42570801638067, + "ttft_p95_ms": 155.23677493911237 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 6.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 152.53963600844145, + 152.85827894695103 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r6p00.json", + "sha256": "ac530f98f57ec452538ebe2311291d0a7c83bbd5554e82f53e78a2311802e411", + "summary": { + "admission_lag_max_ms": 0.13490929268300533, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 157.18225098680705, + "ttft_p50_ms": 150.29372205026448, + "ttft_p95_ms": 152.53963600844145 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r6p00.json", + "sha256": "2f1445ee0890b0972252d10180ef6f00cbad45ee73bd174f21b370cdf950b654", + "summary": { + "admission_lag_max_ms": 0.16514293383806944, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 158.88926689513028, + "ttft_p50_ms": 150.2534201135859, + "ttft_p95_ms": 152.85827894695103 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 7.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 346.96662798523903, + 295.1050230767578 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r7p00.json", + "sha256": "c4765794d442daf7eeb64a2f587792701ebe7659c7702a3bbbfc8900801fff0e", + "summary": { + "admission_lag_max_ms": 0.16206549480557442, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 354.881492909044, + "ttft_p50_ms": 310.15722593292594, + "ttft_p95_ms": 346.96662798523903 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r7p00.json", + "sha256": "30addb463839e23a7356a5cf6d4060ec1eebfa2b9564fa9500be5bcb7aecb537", + "summary": { + "admission_lag_max_ms": 0.10446517262607813, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 296.6241450048983, + "ttft_p50_ms": 246.06937903445214, + "ttft_p95_ms": 295.1050230767578 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 8.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1376.107804942876, + 1321.1636180058122 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r8p00.json", + "sha256": "89e2975f9c7319e477789ea77015541d5235efa61d90fb8bd5b0420a58f7f186", + "summary": { + "admission_lag_max_ms": 0.1416490413248539, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.890625, + "passed": 57, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1501.330335973762, + "ttft_p50_ms": 802.2130889585242, + "ttft_p95_ms": 1376.107804942876 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r8p00.json", + "sha256": "9ca6935f1320033c3ec4b72a35c5d9909ccd7b0fff2e4b2c92c6a26f52887df1", + "summary": { + "admission_lag_max_ms": 0.12422900181263685, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.921875, + "passed": 59, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1410.047525074333, + "ttft_p50_ms": 766.8001820566133, + "ttft_p95_ms": 1321.1636180058122 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4950.155022088438, + 4892.750842031091 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r16p00.json", + "sha256": "4fce2749435bf2e8c77999e11be0a788cd85f459a647760f3db34f1dbc0aee85", + "summary": { + "admission_lag_max_ms": 0.10520406067371368, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.171875, + "passed": 11, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 5031.207721913233, + "ttft_p50_ms": 2775.8131299633533, + "ttft_p95_ms": 4950.155022088438 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r16p00.json", + "sha256": "33ded0192d8986dbcf0a7f5ab6329a440994243d4e528d6c1b278bb6b92720fd", + "summary": { + "admission_lag_max_ms": 0.14005706179887056, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4971.664474112913, + "ttft_p50_ms": 2743.070787983015, + "ttft_p95_ms": 4892.750842031091 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 6750.775643973611, + 6754.7571890754625 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r32p00.json", + "sha256": "44030c4d3374cecebe67b886b428ba8014c87d941c00a178d6049bc57c809f46", + "summary": { + "admission_lag_max_ms": 0.7358170114457607, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.15625, + "passed": 10, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6899.517056066543, + "ttft_p50_ms": 3728.6347220651805, + "ttft_p95_ms": 6750.775643973611 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r32p00.json", + "sha256": "f397bad36552c6df5bf5bcee702cfa968ddf3e20b20e9976554ddfbe838ef3b2", + "summary": { + "admission_lag_max_ms": 0.7476879982277751, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.15625, + "passed": 10, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6901.606322033331, + "ttft_p50_ms": 3729.3128239689395, + "ttft_p95_ms": 6754.7571890754625 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 7800.616211956367, + 7748.703246936202 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round1/results/r64p00.json", + "sha256": "61eadc2af60386d6972d69fe7b083874e705864162073d54d1bde2bc0e6de7e7", + "summary": { + "admission_lag_max_ms": 0.11387700214982033, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7905.77321103774, + "ttft_p50_ms": 4366.223473916762, + "ttft_p95_ms": 7800.616211956367 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z/artifacts/artifacts/real-tp1-mns64-v1/round2/results/r64p00.json", + "sha256": "db02e9a814154a9259cbe3fefdf15700923458f7e1208596e861b69b6a6ddd41", + "summary": { + "admission_lag_max_ms": 0.14900392852723598, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7852.748095989227, + "ttft_p50_ms": 4337.09563605953, + "ttft_p95_ms": 7748.703246936202 + } + } + ] + } + ], + "capacity": 7.0, + "capacity_per_gpu": 7.0, + "mns": 64, + "name": "tp1_mns64", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z", + "tp": 1 + }, + "tp1_mns8": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 154.07429495826364, + 153.62840006127954 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r4p00.json", + "sha256": "23d5ab491b70c02f51c2f06dec0407edc0627819a33b9bcef1349d974a4e20fe", + "summary": { + "admission_lag_max_ms": 0.14362391084432602, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 171.46877304185182, + "ttft_p50_ms": 151.62539994344115, + "ttft_p95_ms": 154.07429495826364 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r4p00.json", + "sha256": "563aa230e0f5e420cf453dfd9b829d7df493919192d638cbe76f85eb77b6d2e7", + "summary": { + "admission_lag_max_ms": 0.11323392391204834, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 155.62100906390697, + "ttft_p50_ms": 151.69239102397114, + "ttft_p95_ms": 153.62840006127954 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 5.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 154.20477103907615, + 153.81942794192582 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r5p00.json", + "sha256": "6f2a04a5c11eed5826e2a6e1ce6e6f8d98504b31d0fec30d9440df2cdb62e56d", + "summary": { + "admission_lag_max_ms": 0.10794296395033598, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 155.37187200970948, + "ttft_p50_ms": 152.59731002151966, + "ttft_p95_ms": 154.20477103907615 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r5p00.json", + "sha256": "27a0bfa65a66e5cc77e10a5cf39a846f0fd1dd266c10926dae9db7a80c581d6e", + "summary": { + "admission_lag_max_ms": 0.1204309519380331, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 155.40366806089878, + "ttft_p50_ms": 152.0398510619998, + "ttft_p95_ms": 153.81942794192582 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 6.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 153.00951700191945, + 153.5524509381503 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r6p00.json", + "sha256": "4562fae05d510cd50f2c9d7479916d056c8ac4d04851d2388b98f8410192c23a", + "summary": { + "admission_lag_max_ms": 0.10347133502364159, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 156.3970319693908, + "ttft_p50_ms": 150.414744974114, + "ttft_p95_ms": 153.00951700191945 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r6p00.json", + "sha256": "bd13283e2807150396b278f46ff630ae9820895e0c054a9efcfd8fc213947d47", + "summary": { + "admission_lag_max_ms": 0.13600767124444246, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 159.63539096992463, + "ttft_p50_ms": 151.83425601571798, + "ttft_p95_ms": 153.5524509381503 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 7.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 336.8800369789824, + 329.65395506471395 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r7p00.json", + "sha256": "fd5afe52238d6ddb72bdf7f7b142fddbeb3f62e160e4a1b3a639b308717964e6", + "summary": { + "admission_lag_max_ms": 0.14346837997436523, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 341.1952300230041, + "ttft_p50_ms": 294.81438896618783, + "ttft_p95_ms": 336.8800369789824 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r7p00.json", + "sha256": "b0fdc1ef617313261c490010adeb2f4ee93bbb053f1db10734ed5a63fd37fa76", + "summary": { + "admission_lag_max_ms": 0.11047185398638248, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 334.06115300022066, + "ttft_p50_ms": 281.1476809438318, + "ttft_p95_ms": 329.65395506471395 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 8.0, + "round_feasible": [ + false, + true + ], + "round_ttft_p95_ms": [ + 1289.2397560644895, + 1254.356418037787 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r8p00.json", + "sha256": "8877ed44b90f421afd22ec110fe6337581c67cd973b44378bcb7528fc8624de3", + "summary": { + "admission_lag_max_ms": 0.09899691212922335, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.9375, + "passed": 60, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1454.3529660440981, + "ttft_p50_ms": 813.0663179326802, + "ttft_p95_ms": 1289.2397560644895 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r8p00.json", + "sha256": "7ab24f5ab33bf166589af7e2f5c2a274ceef6a64f0029a96ec1879b42332123d", + "summary": { + "admission_lag_max_ms": 0.11694896966218948, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 0.953125, + "passed": 61, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1347.0055809011683, + "ttft_p50_ms": 744.0344559727237, + "ttft_p95_ms": 1254.356418037787 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4903.538785991259, + 4900.28300601989 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r16p00.json", + "sha256": "363d9b43164b5451bc3dd668caa0ce141d15114087cf444e7fd79d12b7563fd6", + "summary": { + "admission_lag_max_ms": 0.11420203372836113, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4982.48829995282, + "ttft_p50_ms": 2749.9175920384005, + "ttft_p95_ms": 4903.538785991259 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r16p00.json", + "sha256": "8aec70dfd6b69f4903bccc02c5d5474165ed07d9dd010c860ff9b447429da950", + "summary": { + "admission_lag_max_ms": 0.19414699636399746, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4979.007771937177, + "ttft_p50_ms": 2749.115635990165, + "ttft_p95_ms": 4900.28300601989 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 6760.187636013143, + 6534.733187989332 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r32p00.json", + "sha256": "7684ac0d2d6f851218b88fb37db3a5b54d877a6f1ead3bb604a6cb4d9a61cf3f", + "summary": { + "admission_lag_max_ms": 0.43534801807254553, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.15625, + "passed": 10, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6907.940509961918, + "ttft_p50_ms": 3735.540736000985, + "ttft_p95_ms": 6760.187636013143 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r32p00.json", + "sha256": "947974a9ae04e170b8c1aad32ddbda8c1000f3902a9b572c6338a7d0d4f23146", + "summary": { + "admission_lag_max_ms": 1.221106038428843, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 6798.48151106853, + "ttft_p50_ms": 3595.78589303419, + "ttft_p95_ms": 6534.733187989332 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 7754.704045015387, + 7756.428897031583 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round1/results/r64p00.json", + "sha256": "abfc2a158087f0ad75b560dd420604a9f81359ffdf39f0ce8e2ea25b451e6904", + "summary": { + "admission_lag_max_ms": 0.13654399663209915, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7857.916672946885, + "ttft_p50_ms": 4341.527819051407, + "ttft_p95_ms": 7754.704045015387 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z/artifacts/artifacts/real-tp1-mns8-v1/round2/results/r64p00.json", + "sha256": "192c0c850a52d2b64780bc6321cbfd10faddcc0ff10741ddad283b7d789b9def", + "summary": { + "admission_lag_max_ms": 0.11000398080796003, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.109375, + "passed": 7, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 7859.458995983005, + "ttft_p50_ms": 4341.894763987511, + "ttft_p95_ms": 7756.428897031583 + } + } + ] + } + ], + "capacity": 7.0, + "capacity_per_gpu": 7.0, + "mns": 8, + "name": "tp1_mns8", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z", + "tp": 1 + }, + "tp2_mns16": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 99.1211449727416, + 101.08128399588168 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r4p00.json", + "sha256": "f149882da390114f2a2a140a3bd2ca6f15b6f7761591ea7b49b5a961c9d18df5", + "summary": { + "admission_lag_max_ms": 0.10085199028253555, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 101.28417506348342, + "ttft_p50_ms": 95.90165293775499, + "ttft_p95_ms": 99.1211449727416 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r4p00.json", + "sha256": "154dfaf2729001a8ef5b7430290f4ca906df1d4758e867d5b350489410f36d5c", + "summary": { + "admission_lag_max_ms": 0.11863000690937042, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 105.76263803523034, + "ttft_p50_ms": 96.27402503974736, + "ttft_p95_ms": 101.08128399588168 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 97.1850419882685, + 97.03529300168157 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r8p00.json", + "sha256": "9338d5cb994b0c12a6420f030ae2a9335a04986b4a81e31f4781ac838aeef2d4", + "summary": { + "admission_lag_max_ms": 0.12352492194622755, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 102.26434096693993, + "ttft_p50_ms": 95.07436107378453, + "ttft_p95_ms": 97.1850419882685 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r8p00.json", + "sha256": "a35b67970948976482ee835e595ee5631fa921607276d5ea7928460514f0d06b", + "summary": { + "admission_lag_max_ms": 0.11170096695423126, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 101.72453604172915, + "ttft_p50_ms": 94.03157397173345, + "ttft_p95_ms": 97.03529300168157 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 10.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 96.10817302018404, + 94.37024802900851 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r10p00.json", + "sha256": "f67e0f4f6b41861fcbcbee04b4efbf0dee46e273239ec94ef22d650ba6afa22a", + "summary": { + "admission_lag_max_ms": 0.10012194979935884, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 102.40214702207595, + "ttft_p50_ms": 94.17163301259279, + "ttft_p95_ms": 96.10817302018404 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r10p00.json", + "sha256": "5a64b52f0298e9ecda2fddb8230cc27ab4b9c1db01575f68b5331816b324d1d7", + "summary": { + "admission_lag_max_ms": 0.10465399827808142, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 101.07159405015409, + "ttft_p50_ms": 93.24954205658287, + "ttft_p95_ms": 94.37024802900851 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 12.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 447.0475659472868, + 454.8899739747867 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r12p00.json", + "sha256": "c6e27b9078281a54feeb1fab09616385a70de819791ce14f46b001bf540a80ab", + "summary": { + "admission_lag_max_ms": 0.12051803059875965, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 455.56937297806144, + "ttft_p50_ms": 280.09015799034387, + "ttft_p95_ms": 447.0475659472868 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r12p00.json", + "sha256": "861f9c4186c33e9ee6365d19c53e98c10615c5ee529a84e69f84dc292f53ac4f", + "summary": { + "admission_lag_max_ms": 0.11322635691612959, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 457.4526828946546, + "ttft_p50_ms": 290.1579710887745, + "ttft_p95_ms": 454.8899739747867 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 14.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1000.5082850111648, + 1004.6314439969137 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r14p00.json", + "sha256": "2dcaf1087edd1f12fbfe0618a6ab636b22484756a5729919bc0e51de153f8f4a", + "summary": { + "admission_lag_max_ms": 0.1881921198219061, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1089.2393708927557, + "ttft_p50_ms": 611.2175739835948, + "ttft_p95_ms": 1000.5082850111648 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r14p00.json", + "sha256": "16937b31f72298c7a50ec3c7b9c1b9c798f305d83469b3d3c058efc4d53135da", + "summary": { + "admission_lag_max_ms": 0.10659568943083286, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1096.6680150013417, + "ttft_p50_ms": 617.3943240428343, + "ttft_p95_ms": 1004.6314439969137 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1458.4544260287657, + 1471.3564389385283 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r16p00.json", + "sha256": "5576a7fd06977f51f1c43bb73c3c474cc21a93cd288b422ec3d09db671b3afe7", + "summary": { + "admission_lag_max_ms": 0.135264010168612, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1583.2410319708288, + "ttft_p50_ms": 904.5663280412555, + "ttft_p95_ms": 1458.4544260287657 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r16p00.json", + "sha256": "2ccf16426046c412829442f72932268d5b27ce051a4312d135d32b240b48ec7c", + "summary": { + "admission_lag_max_ms": 0.11741300113499165, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.78125, + "passed": 50, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1537.9131329245865, + "ttft_p50_ms": 896.1611289996654, + "ttft_p95_ms": 1471.3564389385283 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 3348.027400090359, + 3348.575382027775 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r32p00.json", + "sha256": "d258760d15f1fdf2adefca72f7a3505c0b839036799dcb074450f7f526f76718", + "summary": { + "admission_lag_max_ms": 0.18610397819429636, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3442.3487790627405, + "ttft_p50_ms": 1801.6119640087709, + "ttft_p95_ms": 3348.027400090359 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r32p00.json", + "sha256": "edf5cccc1400639a94b5430897a04f20d2214bb1ffee7f4bfa62514150179a6a", + "summary": { + "admission_lag_max_ms": 0.1144439447671175, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3443.0311269825324, + "ttft_p50_ms": 1801.435066969134, + "ttft_p95_ms": 3348.575382027775 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4286.911493982188, + 4287.898648995906 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round1/results/r64p00.json", + "sha256": "2a6303efb826637c3c6448d6f766f7de4fc4d7b63db163a48f72c20df68394f4", + "summary": { + "admission_lag_max_ms": 0.10997895151376724, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4340.311317006126, + "ttft_p50_ms": 2396.4726959820837, + "ttft_p95_ms": 4286.911493982188 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z/artifacts/artifacts/real-tp2-mns16-v1/round2/results/r64p00.json", + "sha256": "6fafe3bf206583b8689be1b2b57d6b92cbd1ba54f7fb230ca05f6c68b3ab9757", + "summary": { + "admission_lag_max_ms": 0.12669397983700037, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4340.538130956702, + "ttft_p50_ms": 2394.112486974336, + "ttft_p95_ms": 4287.898648995906 + } + } + ] + } + ], + "capacity": 14.0, + "capacity_per_gpu": 7.0, + "mns": 16, + "name": "tp2_mns16", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z", + "tp": 2 + }, + "tp2_mns32": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 97.2925309324637, + 96.13083000294864 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r4p00.json", + "sha256": "812f8a108a9390fb759b46e5bd41e3ddad66bc144d47bf4024d834e42b47c47e", + "summary": { + "admission_lag_max_ms": 0.11386198457330465, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.76773096807301, + "ttft_p50_ms": 95.72782705072314, + "ttft_p95_ms": 97.2925309324637 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r4p00.json", + "sha256": "70f922eb7651d46d81d7d394b8453c5b778f921069cf48fb2ae2677afa94a842", + "summary": { + "admission_lag_max_ms": 0.30313804745674133, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 129.06646099872887, + "ttft_p50_ms": 94.53632694203407, + "ttft_p95_ms": 96.13083000294864 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 95.42673197574914, + 98.94769196398556 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r8p00.json", + "sha256": "24097d38b0ffb45f6cdbd93f0aad04dce3b7da41ff45e948eb3a794902b6547f", + "summary": { + "admission_lag_max_ms": 0.11571904178708792, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.04934400785714, + "ttft_p50_ms": 94.34181300457567, + "ttft_p95_ms": 95.42673197574914 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r8p00.json", + "sha256": "7251f076d34ba9484b08fa3fc4b898d2ba8d1199cb1eaeaaffc19930f8c889b9", + "summary": { + "admission_lag_max_ms": 0.1177559606730938, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.93720292206854, + "ttft_p50_ms": 94.77824496570975, + "ttft_p95_ms": 98.94769196398556 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 10.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 96.28185909241438, + 95.9081610199064 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r10p00.json", + "sha256": "85796457715a25c15fc909f6def32776fbd49956eb7890df1c4abcef8c5afe78", + "summary": { + "admission_lag_max_ms": 0.14786410611122847, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 102.46025596279651, + "ttft_p50_ms": 93.54631893802434, + "ttft_p95_ms": 96.28185909241438 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r10p00.json", + "sha256": "0bf4d4d006bf0b5bd8b989ceeeb4bacb0e3ee1b8ea66f30b3deb120a14871cf0", + "summary": { + "admission_lag_max_ms": 0.11162098962813616, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.54603894241154, + "ttft_p50_ms": 93.77408598084003, + "ttft_p95_ms": 95.9081610199064 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 12.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 439.0145930228755, + 429.9719400005415 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r12p00.json", + "sha256": "68de6e36bd7c5624cab107fe5cb21bb817711691e545439a958c926a27e3fe29", + "summary": { + "admission_lag_max_ms": 0.10475562885403633, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 445.90794993564487, + "ttft_p50_ms": 270.3119309153408, + "ttft_p95_ms": 439.0145930228755 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r12p00.json", + "sha256": "d8d3a44a07b46f1673817ece2948fd97c30f41bd451d93a2fa02edf074536c63", + "summary": { + "admission_lag_max_ms": 0.4323029424995184, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 434.6071509644389, + "ttft_p50_ms": 265.2625450864434, + "ttft_p95_ms": 429.9719400005415 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 14.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1014.2896320903674, + 999.8802930349484 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r14p00.json", + "sha256": "10196c2f9a9b69c2e47f64d06bf5ee6863af4cd2b612301622f0e304949a942a", + "summary": { + "admission_lag_max_ms": 0.09005970787256956, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1105.1907669752836, + "ttft_p50_ms": 621.6445650206879, + "ttft_p95_ms": 1014.2896320903674 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r14p00.json", + "sha256": "c10cab85588a3563bed36fc4480462635d34d240938beaa0e5034f13186578a6", + "summary": { + "admission_lag_max_ms": 0.11327012907713652, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1090.6395060010254, + "ttft_p50_ms": 623.5586369875818, + "ttft_p95_ms": 999.8802930349484 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1488.3801550604403, + 1484.6198010491207 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r16p00.json", + "sha256": "ef5046c49c32178855bed6a0488ea21e8da22f786ddaabc503bea2d1025bb2f3", + "summary": { + "admission_lag_max_ms": 0.7892940193414688, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1613.616122980602, + "ttft_p50_ms": 905.2404230460525, + "ttft_p95_ms": 1488.3801550604403 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r16p00.json", + "sha256": "de02c1d49414cffe7e20023abf4d502a4efc6d2e01fcb3e49e928081aaef77b6", + "summary": { + "admission_lag_max_ms": 0.10548799764364958, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1550.9842830942944, + "ttft_p50_ms": 904.9503420246765, + "ttft_p95_ms": 1484.6198010491207 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 3391.811708919704, + 3362.9186518955976 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r32p00.json", + "sha256": "876c5444d17c4ce81eca907ae17d5cf7062ffa67fa266b491580e52f712614ec", + "summary": { + "admission_lag_max_ms": 0.12557301670312881, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3486.074637970887, + "ttft_p50_ms": 1840.204154024832, + "ttft_p95_ms": 3391.811708919704 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r32p00.json", + "sha256": "02b4cd93ce1f9816e822022fb608ba8d4c89265f673f6d27c8fa3d4f1523ffe0", + "summary": { + "admission_lag_max_ms": 0.11014204937964678, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3456.965696066618, + "ttft_p50_ms": 1812.343467026949, + "ttft_p95_ms": 3362.9186518955976 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4306.351027917117, + 4302.602957002819 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round1/results/r64p00.json", + "sha256": "13ba282f179b2d6ce13a0aa2fd04eb29f82e579b7995e3406cb0afd69a2d5ee2", + "summary": { + "admission_lag_max_ms": 0.3526840591803193, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4359.505696920678, + "ttft_p50_ms": 2411.1130589153618, + "ttft_p95_ms": 4306.351027917117 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z/artifacts/artifacts/real-tp2-mns32-v1/round2/results/r64p00.json", + "sha256": "767622d62f0757ef2d8c7651ea470bfa2e62edad0ed4e75f66f84a0b722e35da", + "summary": { + "admission_lag_max_ms": 0.10626309085637331, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4355.409953044727, + "ttft_p50_ms": 2409.5974090741947, + "ttft_p95_ms": 4302.602957002819 + } + } + ] + } + ], + "capacity": 14.0, + "capacity_per_gpu": 7.0, + "mns": 32, + "name": "tp2_mns32", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z", + "tp": 2 + }, + "tp2_mns64": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 96.54812794178724, + 95.08875198662281 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r4p00.json", + "sha256": "6e6a45d0c1010094c3395969815fd8cf9c35b919db202d772fb514744f981ebc", + "summary": { + "admission_lag_max_ms": 0.11306197848170996, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 99.27749610505998, + "ttft_p50_ms": 93.01548008807003, + "ttft_p95_ms": 96.54812794178724 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r4p00.json", + "sha256": "58aa71d694704cfea15c37cb29420dbac1dc505045c29a731e8be904846cca5f", + "summary": { + "admission_lag_max_ms": 0.11853000614792109, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 98.74943597242236, + "ttft_p50_ms": 93.57558900956064, + "ttft_p95_ms": 95.08875198662281 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 94.24941800534725, + 96.21043398510665 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r8p00.json", + "sha256": "311350af236cb0306aa635f05f8999b66bf5080bad7cc1f6703fffe0be31aee6", + "summary": { + "admission_lag_max_ms": 0.19771000370383263, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.52211803849787, + "ttft_p50_ms": 92.4453770276159, + "ttft_p95_ms": 94.24941800534725 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r8p00.json", + "sha256": "60fe989a36a526e76f6f5c98e217b9feaf081bc680b16b9d53650bf085047c06", + "summary": { + "admission_lag_max_ms": 0.3933720290660858, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 102.97230398282409, + "ttft_p50_ms": 93.26327103190124, + "ttft_p95_ms": 96.21043398510665 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 10.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 97.98661107197404, + 92.9473100695759 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r10p00.json", + "sha256": "d0a04444a0e41bf76f71ac6ea3815bfbe76f7a51f8ddf669976bc3a4896dd0c1", + "summary": { + "admission_lag_max_ms": 0.10270101483911276, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 103.49742788821459, + "ttft_p50_ms": 94.55623896792531, + "ttft_p95_ms": 97.98661107197404 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r10p00.json", + "sha256": "f330f26f45f6039c48150a0b6b146ca3dbb9ea8b2dff086a9efd5938ab981eb7", + "summary": { + "admission_lag_max_ms": 0.11239596642553806, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 97.4133670097217, + "ttft_p50_ms": 91.2688790122047, + "ttft_p95_ms": 92.9473100695759 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 12.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 260.68170997314155, + 319.04888805001974 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r12p00.json", + "sha256": "fed5eacf18018402b0c2675b8c5595f92693a9bddec30f354d989098f9fe4965", + "summary": { + "admission_lag_max_ms": 0.11784560047090054, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 345.19017208367586, + "ttft_p50_ms": 211.41935500781983, + "ttft_p95_ms": 260.68170997314155 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r12p00.json", + "sha256": "f2be27cbeff3e5c1a95050418f2dc3c712a4648cada8cdb3d36382cfffc9593a", + "summary": { + "admission_lag_max_ms": 0.12930459342896938, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 326.86293800361454, + "ttft_p50_ms": 229.8808820778504, + "ttft_p95_ms": 319.04888805001974 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 14.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 895.7287209341303, + 912.1733580250293 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r14p00.json", + "sha256": "c2e157808e0308cc443f3abdf5114898cef66c435688d24321f80bef2d966208", + "summary": { + "admission_lag_max_ms": 0.14902697876095772, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 962.3544370988384, + "ttft_p50_ms": 576.2168710352853, + "ttft_p95_ms": 895.7287209341303 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r14p00.json", + "sha256": "749ef7d31bf8782d42719b75ee1a2e4a443f097510af721d02cf53df191bc66e", + "summary": { + "admission_lag_max_ms": 0.42896345257759094, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 986.0872740391642, + "ttft_p50_ms": 569.7114520007744, + "ttft_p95_ms": 912.1733580250293 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1364.6171110449359, + 1366.6857269126922 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r16p00.json", + "sha256": "7545ea9210348019de88ede5d2cba7527913534274deee034fa6811c173a6e72", + "summary": { + "admission_lag_max_ms": 0.12612901628017426, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.859375, + "passed": 55, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1435.8308439841494, + "ttft_p50_ms": 848.9949750946835, + "ttft_p95_ms": 1364.6171110449359 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r16p00.json", + "sha256": "d75e27137f6a7811126d8c957bb5a081944fc1bb7643852cdd785111773f6010", + "summary": { + "admission_lag_max_ms": 0.1258660340681672, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.859375, + "passed": 55, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1436.2916270038113, + "ttft_p50_ms": 846.2129919789732, + "ttft_p95_ms": 1366.6857269126922 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 3243.9196659252048, + 3214.524411014281 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r32p00.json", + "sha256": "366ebddc167338ed096f373de5667740721422c13d2db1f887ed09186afafd19", + "summary": { + "admission_lag_max_ms": 0.09669095743447542, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3337.068362045102, + "ttft_p50_ms": 1751.3202880509198, + "ttft_p95_ms": 3243.9196659252048 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r32p00.json", + "sha256": "01b8f5081e36a64d5776285c8c5f423b53b1ad684c4b8f95a8caa5d997c6bf09", + "summary": { + "admission_lag_max_ms": 0.8310690755024552, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.328125, + "passed": 21, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3277.297410997562, + "ttft_p50_ms": 1790.5721209244803, + "ttft_p95_ms": 3214.524411014281 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4319.725521025248, + 4367.031275993213 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round1/results/r64p00.json", + "sha256": "92202709fd6bbd558ff6f1986df213848254110c4274447a7bd5879153290d46", + "summary": { + "admission_lag_max_ms": 0.10392803233116865, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4360.145599930547, + "ttft_p50_ms": 2403.5812759539112, + "ttft_p95_ms": 4319.725521025248 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z/artifacts/artifacts/real-tp2-mns64-v1/round2/results/r64p00.json", + "sha256": "9660d15d31bc70c5550ed29477f563313d4d81baf0e561126aebd3b7d097cfa3", + "summary": { + "admission_lag_max_ms": 0.17276196740567684, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.25, + "passed": 16, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4414.442001027055, + "ttft_p50_ms": 2260.2788450894877, + "ttft_p95_ms": 4367.031275993213 + } + } + ] + } + ], + "capacity": 14.0, + "capacity_per_gpu": 7.0, + "mns": 64, + "name": "tp2_mns64", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z", + "tp": 2 + }, + "tp2_mns8": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 97.17482794076204, + 98.03147299680859 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r4p00.json", + "sha256": "0e315f6d9e0087ad553e874b0c16c2e6de79eb6eca22eb118dc5deca52e6efe3", + "summary": { + "admission_lag_max_ms": 0.11708191595971584, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 101.92840604577214, + "ttft_p50_ms": 95.13642196543515, + "ttft_p95_ms": 97.17482794076204 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r4p00.json", + "sha256": "baa394a24d2a3db40a3a7ae6dae0c5ce49ff05499a1e17eaa0bca4b479ae33bf", + "summary": { + "admission_lag_max_ms": 0.10946101974695921, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 100.52333807107061, + "ttft_p50_ms": 93.19468098692596, + "ttft_p95_ms": 98.03147299680859 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 97.67980605829507, + 93.80627400241792 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r8p00.json", + "sha256": "303df27c6d4604d5af65f9056acacb7c4373fb6eb7edb6e5d42689eeaa384aaa", + "summary": { + "admission_lag_max_ms": 0.11111306957900524, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 113.2329449756071, + "ttft_p50_ms": 93.05417106952518, + "ttft_p95_ms": 97.67980605829507 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r8p00.json", + "sha256": "cdbb1bca17b24b67e18ef3dd4997494d6bd2a887b3b5d1264fbf411e688c53bc", + "summary": { + "admission_lag_max_ms": 0.12463307939469814, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 101.12245299387723, + "ttft_p50_ms": 92.26030297577381, + "ttft_p95_ms": 93.80627400241792 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 10.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 94.45723297540098, + 93.57301203999668 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r10p00.json", + "sha256": "fe17ef2417319307d651bc10b01a618234d61bab797ad87a7332cab20d78ecb2", + "summary": { + "admission_lag_max_ms": 0.3267280990257859, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 99.28593400400132, + "ttft_p50_ms": 92.22474799025804, + "ttft_p95_ms": 94.45723297540098 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r10p00.json", + "sha256": "550f693708a34624a82d77afd4da0ab4bc01e09f42242a3a79e4532897681a51", + "summary": { + "admission_lag_max_ms": 0.11238199658691883, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 97.77756198309362, + "ttft_p50_ms": 92.07128698471934, + "ttft_p95_ms": 93.57301203999668 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 12.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 320.49866404850036, + 336.9630330707878 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r12p00.json", + "sha256": "edab8169453151de732923c076b8f7dca4082ee47ccd2597e65e7ba68516bc2e", + "summary": { + "admission_lag_max_ms": 0.09914662223309278, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 327.930542989634, + "ttft_p50_ms": 231.4880370395258, + "ttft_p95_ms": 320.49866404850036 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r12p00.json", + "sha256": "a6dde4363d362ca10254409e3dc734d0f0e640e63156426f3f4fd7115db49e16", + "summary": { + "admission_lag_max_ms": 0.17016706988215446, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 428.6657409975305, + "ttft_p50_ms": 237.09688894450665, + "ttft_p95_ms": 336.9630330707878 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 14.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 911.7195709841326, + 910.7652489328757 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r14p00.json", + "sha256": "7eba81ee5c85c457a056f9aef28abd94d375fdaa116929cfe0a0c282a0738bd2", + "summary": { + "admission_lag_max_ms": 0.13163150288164616, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 999.2953530745581, + "ttft_p50_ms": 580.023527960293, + "ttft_p95_ms": 911.7195709841326 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r14p00.json", + "sha256": "6a113d3d876ff01a9c1101e8b75efa2fc71f96683d6bb430e26d5976426b5d23", + "summary": { + "admission_lag_max_ms": 0.11634768452495337, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 988.8870939612389, + "ttft_p50_ms": 578.9327350212261, + "ttft_p95_ms": 910.7652489328757 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 16.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 1389.8832038976252, + 1365.0342189939693 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r16p00.json", + "sha256": "d677251ee1a941c52469052a239c2e13be0a837285069aa8182a16e196445e48", + "summary": { + "admission_lag_max_ms": 0.11070503387600183, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.828125, + "passed": 53, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1460.5093500576913, + "ttft_p50_ms": 864.3530959961936, + "ttft_p95_ms": 1389.8832038976252 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r16p00.json", + "sha256": "f5177f53aa0f87a4e7291f470978d86ad8bc46bd1a8362b79f48cd128646dc10", + "summary": { + "admission_lag_max_ms": 0.3672370221465826, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.828125, + "passed": 53, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1491.050804965198, + "ttft_p50_ms": 841.7225609300658, + "ttft_p95_ms": 1365.0342189939693 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 32.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 3258.010295103304, + 3258.664960041642 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r32p00.json", + "sha256": "02a6ac2b9c2b9ab97854d6fcc13b64bbb36a079e4afdae84231e4cb187fac4ef", + "summary": { + "admission_lag_max_ms": 0.1296499976888299, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3352.278529899195, + "ttft_p50_ms": 1755.0864539807662, + "ttft_p95_ms": 3258.010295103304 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r32p00.json", + "sha256": "9ca2f44fc7e3cb3d38ada548af858903b31bbee5cd50f1723c29255ebd4224e9", + "summary": { + "admission_lag_max_ms": 1.1343059595674276, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 3352.7101600775495, + "ttft_p50_ms": 1755.3453800501302, + "ttft_p95_ms": 3258.664960041642 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 4190.255103982054, + 4189.1050330596045 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round1/results/r64p00.json", + "sha256": "53908d3a254e19691839831b4337014b09591ced9189c2a9b29e0d92e80fcb56", + "summary": { + "admission_lag_max_ms": 0.11591101065278053, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4242.029857938178, + "ttft_p50_ms": 2340.5529640149325, + "ttft_p95_ms": 4190.255103982054 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z/artifacts/artifacts/real-tp2-mns8-v1/round2/results/r64p00.json", + "sha256": "1c8c85c873aa96c01dc0200f848474a53c1173ef72f3828e8c23aeae70e92b11", + "summary": { + "admission_lag_max_ms": 0.284243025816977, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.234375, + "passed": 15, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 4240.770195960067, + "ttft_p50_ms": 2339.3583690049127, + "ttft_p95_ms": 4189.1050330596045 + } + } + ] + } + ], + "capacity": 14.0, + "capacity_per_gpu": 7.0, + "mns": 8, + "name": "tp2_mns8", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z", + "tp": 2 + }, + "tp4_mns16": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 74.66826005838811, + 66.5705680148676 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r4p00.json", + "sha256": "fb354fc8a097caa21844d4ea5515b98b03459f400816e5e5590a6cda81cf4815", + "summary": { + "admission_lag_max_ms": 0.1035250024870038, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 93.38432096410543, + "ttft_p50_ms": 65.75745099689811, + "ttft_p95_ms": 74.66826005838811 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r4p00.json", + "sha256": "ac8e8412f79a8a32ebf085445a94aa548dd2d646fce0b9f56b4378081dcc8960", + "summary": { + "admission_lag_max_ms": 0.12108997907489538, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 79.89856600761414, + "ttft_p50_ms": 62.386377016082406, + "ttft_p95_ms": 66.5705680148676 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 70.44799195136875, + 72.29184894822538 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r8p00.json", + "sha256": "05ebbd68bbf24e9c8b25684474411c7db122b92701cc1617f9d786cf19592751", + "summary": { + "admission_lag_max_ms": 0.1825150102376938, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 96.92095802165568, + "ttft_p50_ms": 60.598687967285514, + "ttft_p95_ms": 70.44799195136875 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r8p00.json", + "sha256": "d24f7daaf41073cc8aa030b74c06203783684c3e556770366883c37b826406f2", + "summary": { + "admission_lag_max_ms": 0.10628998279571533, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 75.13270596973598, + "ttft_p50_ms": 60.161596979014575, + "ttft_p95_ms": 72.29184894822538 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 16.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 66.02151494007558, + 107.58775402791798 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r16p00.json", + "sha256": "a29b2128983af630e1da8f970cbda802849036bc522eb506821bf2fe1a34172f", + "summary": { + "admission_lag_max_ms": 0.10468997061252594, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 72.93136999942362, + "ttft_p50_ms": 59.5597909996286, + "ttft_p95_ms": 66.02151494007558 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r16p00.json", + "sha256": "275ae35651cbe2200d16c9b06a68568efda527564065255a10dee6fe93346010", + "summary": { + "admission_lag_max_ms": 0.3144120564684272, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 119.78317995090038, + "ttft_p50_ms": 59.12985489703715, + "ttft_p95_ms": 107.58775402791798 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 20.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 193.8455159543082, + 187.52167106140405 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r20p00.json", + "sha256": "4cb485f30b81c004bab379e6522d94b811c0bdc0be6d8e42c9d8ae842d9600dc", + "summary": { + "admission_lag_max_ms": 0.10407401714473963, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 197.7908629924059, + "ttft_p50_ms": 142.73560303263366, + "ttft_p95_ms": 193.8455159543082 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r20p00.json", + "sha256": "63aa9dc03a08b781c5d6e8805f2b61698c75e71673608ecd38bd1132c62e6987", + "summary": { + "admission_lag_max_ms": 0.09386602323502302, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 190.29632199089974, + "ttft_p50_ms": 136.1723990412429, + "ttft_p95_ms": 187.52167106140405 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 24.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 568.2198900030926, + 573.4539540717378 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r24p00.json", + "sha256": "5e4d5b44850090d07fe1d3abc44eddea3bfdf2b682f59b12a45a866621e38f98", + "summary": { + "admission_lag_max_ms": 0.7511250441893935, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 613.2891150191426, + "ttft_p50_ms": 375.27791305910796, + "ttft_p95_ms": 568.2198900030926 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r24p00.json", + "sha256": "239e7789aed2e9b2bcf1d127215a6b122071eb856be2db64ccc184fda153aa48", + "summary": { + "admission_lag_max_ms": 0.11180061846971512, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 614.9205249967054, + "ttft_p50_ms": 375.5133739905432, + "ttft_p95_ms": 573.4539540717378 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 28.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 860.0242879474536, + 886.1482660286129 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r28p00.json", + "sha256": "8b249b52779e5b3cd9028aaee30ee73ef34b0b0a64dcd75b9aad57266c29fbc2", + "summary": { + "admission_lag_max_ms": 0.11425174307078123, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 930.8854410191998, + "ttft_p50_ms": 534.7980660153553, + "ttft_p95_ms": 860.0242879474536 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r28p00.json", + "sha256": "c9b28c8f3187542a343ecf1067783569ec2abfa64631d6827a94ba6876aa7b63", + "summary": { + "admission_lag_max_ms": 0.312068616040051, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 922.4315020255744, + "ttft_p50_ms": 549.7413279954344, + "ttft_p95_ms": 886.1482660286129 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 32.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1147.1082780044526, + 1137.236590962857 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r32p00.json", + "sha256": "d27ff9f940a6cd7fba339d9a176de39cf4b5bd164dcc500f2af36bcb07cdef30", + "summary": { + "admission_lag_max_ms": 0.16146607231348753, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1201.626160996966, + "ttft_p50_ms": 696.4708579471335, + "ttft_p95_ms": 1147.1082780044526 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r32p00.json", + "sha256": "093044e87fffdaa222e5bad3807cbf4900aa64891d5b4ba27f187c7a411252d7", + "summary": { + "admission_lag_max_ms": 0.10430300608277321, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1201.045956928283, + "ttft_p50_ms": 684.3201500596479, + "ttft_p95_ms": 1137.236590962857 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 2064.763988018967, + 2039.6780440350994 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round1/results/r64p00.json", + "sha256": "fad7bdd80d5ca603790a4e5d19dadd4f494540b106ce583e3e8c1c05d2374a58", + "summary": { + "admission_lag_max_ms": 6.394341005943716, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.53125, + "passed": 34, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2144.80376499705, + "ttft_p50_ms": 1158.5425049997866, + "ttft_p95_ms": 2064.763988018967 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z/artifacts/artifacts/real-tp4-mns16-v1/round2/results/r64p00.json", + "sha256": "24fdb945706dbbdd29916e0f08e0eff807334db212fded94c729da2e1b46efd3", + "summary": { + "admission_lag_max_ms": 0.10294502135366201, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.546875, + "passed": 35, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2125.4275270039216, + "ttft_p50_ms": 1131.1213460285217, + "ttft_p95_ms": 2039.6780440350994 + } + } + ] + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 16, + "name": "tp4_mns16", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z", + "tp": 4 + }, + "tp4_mns32": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 67.80304794665426, + 64.59337309934199 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r4p00.json", + "sha256": "2e82a30f02c2702668e59661930431cf82323746a111d42bf38b7fdc0d083b3d", + "summary": { + "admission_lag_max_ms": 0.1332589890807867, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 72.74050498381257, + "ttft_p50_ms": 64.10286005120724, + "ttft_p95_ms": 67.80304794665426 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r4p00.json", + "sha256": "8294196c7dc52765a6820f5e1d8439c480490daed9505263f707810f73c95783", + "summary": { + "admission_lag_max_ms": 0.1283179735764861, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 70.6796320155263, + "ttft_p50_ms": 60.962425894103944, + "ttft_p95_ms": 64.59337309934199 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 70.46496798284352, + 68.22343391831964 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r8p00.json", + "sha256": "38bbfeab47e5eb6296fcccf20f04ef25dc24db658be92300c492734bbfe333cf", + "summary": { + "admission_lag_max_ms": 0.10555307380855083, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 87.7126749837771, + "ttft_p50_ms": 62.65173899009824, + "ttft_p95_ms": 70.46496798284352 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r8p00.json", + "sha256": "716f90177d6f203717c80a8e9b28697e53141f204af470493c299bc73a1ea3ee", + "summary": { + "admission_lag_max_ms": 0.24518708232790232, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 80.40577603969723, + "ttft_p50_ms": 60.29119901359081, + "ttft_p95_ms": 68.22343391831964 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 16.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 89.53388908412308, + 76.93439698778093 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r16p00.json", + "sha256": "063016efe822c31c2425e7d963eda4c054f3f7c0341df0a3cd73ab86a942e873", + "summary": { + "admission_lag_max_ms": 0.2996840048581362, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 97.41822991054505, + "ttft_p50_ms": 60.97789097111672, + "ttft_p95_ms": 89.53388908412308 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r16p00.json", + "sha256": "eecc6bdeff90a8bf195dfc5387c6bc2f4db20ed7c9aeb9b4ebe050efccc6051a", + "summary": { + "admission_lag_max_ms": 0.36267400719225407, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 79.42048599943519, + "ttft_p50_ms": 60.29245804529637, + "ttft_p95_ms": 76.93439698778093 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 20.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 206.2494510319084, + 201.13301498349756 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r20p00.json", + "sha256": "22a62729358fea7c38ca166a1a49f1235e98c7f07ecfe4948322110cda4b7178", + "summary": { + "admission_lag_max_ms": 0.11367700062692165, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 255.14698098413646, + "ttft_p50_ms": 155.46583198010921, + "ttft_p95_ms": 206.2494510319084 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r20p00.json", + "sha256": "be103a759100bd2ab6dd31ee4ce41b6f0da40d232a60ebe137e3d933194a487c", + "summary": { + "admission_lag_max_ms": 0.17853605095297098, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 209.22932599205524, + "ttft_p50_ms": 155.12292610947043, + "ttft_p95_ms": 201.13301498349756 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 24.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 580.5047769099474, + 577.3971510352567 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r24p00.json", + "sha256": "890a733a97f399deb4c5efcca661d9a442ed76080cd5d8b6a2c09368aef701b7", + "summary": { + "admission_lag_max_ms": 0.1084543764591217, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 621.5670789824799, + "ttft_p50_ms": 374.6346529806033, + "ttft_p95_ms": 580.5047769099474 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r24p00.json", + "sha256": "87e8543652b1743b1b9b8c3de0fc680401b0bf9a33d61fa6132120c944b2a6eb", + "summary": { + "admission_lag_max_ms": 0.11801638174802065, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 623.2662099646404, + "ttft_p50_ms": 382.11427396163344, + "ttft_p95_ms": 577.3971510352567 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 28.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 873.0683700414374, + 889.4930009264499 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r28p00.json", + "sha256": "f877781730a5437b4cbdb76948915f1c871574b19a6df27076d21b65bb18e678", + "summary": { + "admission_lag_max_ms": 0.6807943573221564, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 943.9258739585057, + "ttft_p50_ms": 542.9852310335264, + "ttft_p95_ms": 873.0683700414374 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r28p00.json", + "sha256": "82151dc539b35302489c17976cfe89b8f4c64944298954607fad8f5c9cb9e7d8", + "summary": { + "admission_lag_max_ms": 6.750226486474276, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 936.5667289821431, + "ttft_p50_ms": 546.0105559322983, + "ttft_p95_ms": 889.4930009264499 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 32.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1135.3467140579596, + 1146.743991994299 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r32p00.json", + "sha256": "082cc4c47b2eec6e49f0edd14cff5a51b880bd998d98da8b9830bbf7615483ef", + "summary": { + "admission_lag_max_ms": 0.10443700011819601, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1199.8225760180503, + "ttft_p50_ms": 680.4794340860099, + "ttft_p95_ms": 1135.3467140579596 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r32p00.json", + "sha256": "517d3cbc1104b5ea1bdb0e381c033cf978181a93449b6498d30b52cd3dd3e770", + "summary": { + "admission_lag_max_ms": 0.4415180301293731, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1210.0602959981188, + "ttft_p50_ms": 693.0603379150853, + "ttft_p95_ms": 1146.743991994299 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 2084.9693010095507, + 2041.9401369290426 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round1/results/r64p00.json", + "sha256": "7e11771ed598826fb3c5594c7e05d0989e01030be2d8b36fa21813aff462496d", + "summary": { + "admission_lag_max_ms": 0.10768393985927105, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.5625, + "passed": 36, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2132.128676981665, + "ttft_p50_ms": 1113.0280290963128, + "ttft_p95_ms": 2084.9693010095507 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z/artifacts/artifacts/real-tp4-mns32-v1/round2/results/r64p00.json", + "sha256": "cf550423832a92cba942b7e1dc14a42000e8c9507403de8cd49aa56af9db6706", + "summary": { + "admission_lag_max_ms": 0.11262902989983559, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.546875, + "passed": 35, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2122.8132860269397, + "ttft_p50_ms": 1131.2310450011864, + "ttft_p95_ms": 2041.9401369290426 + } + } + ] + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 32, + "name": "tp4_mns32", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z", + "tp": 4 + }, + "tp4_mns64": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 66.89354998525232, + 83.23234610725194 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r4p00.json", + "sha256": "266cba4088f8a9483ceb41fdc02d04a2181cb7ffb6b52372b6b0687e62f2863e", + "summary": { + "admission_lag_max_ms": 0.10568101424723864, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 72.09669100120664, + "ttft_p50_ms": 62.014322029426694, + "ttft_p95_ms": 66.89354998525232 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r4p00.json", + "sha256": "26f3926c40504b96a37cb7b9dae1af5040c085c1a730e20feaf1d3e9ddbf8e54", + "summary": { + "admission_lag_max_ms": 0.108166947029531, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 96.42452490516007, + "ttft_p50_ms": 61.261708033271134, + "ttft_p95_ms": 83.23234610725194 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 67.26283801253885, + 67.34374002553523 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r8p00.json", + "sha256": "e84e6ba2c8822e012aceea7cc6f376ea1b6265fe4a3bb68b7badfc36b8813545", + "summary": { + "admission_lag_max_ms": 0.15694997273385525, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 86.54619695153087, + "ttft_p50_ms": 59.63437003083527, + "ttft_p95_ms": 67.26283801253885 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r8p00.json", + "sha256": "dae50073c9f79919181e8dcf274ca8929ba41c512f64363e40187de68839b87e", + "summary": { + "admission_lag_max_ms": 0.09962497279047966, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 69.00161597877741, + "ttft_p50_ms": 60.2562150452286, + "ttft_p95_ms": 67.34374002553523 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 16.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 71.37343694921583, + 64.9534990079701 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r16p00.json", + "sha256": "776c9956f65acf7c9b6c8df517a4f8c4d462289893a19e625a2a0ad7dcbf5506", + "summary": { + "admission_lag_max_ms": 0.1066719414666295, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 82.25136005785316, + "ttft_p50_ms": 58.52357891853899, + "ttft_p95_ms": 71.37343694921583 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r16p00.json", + "sha256": "90057204a5f6343ee4ee9cdd08323d87d83660d5bf8f02480882eb479d2e587e", + "summary": { + "admission_lag_max_ms": 0.10102998930960894, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 65.3582609957084, + "ttft_p50_ms": 58.98014397826046, + "ttft_p95_ms": 64.9534990079701 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 20.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 192.98715889453888, + 258.1990590551868 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r20p00.json", + "sha256": "6ce029388b206d0663acd41f4da1d67a6562389e1c84cf1826ce77c05d176d9a", + "summary": { + "admission_lag_max_ms": 0.12691703159362078, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 198.67739104665816, + "ttft_p50_ms": 142.75032398290932, + "ttft_p95_ms": 192.98715889453888 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r20p00.json", + "sha256": "0dbfb3f012574988ddebfb0af1532c735ac62d6d88863d24cd820aed50a92a59", + "summary": { + "admission_lag_max_ms": 0.3961039474233985, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 259.55891504418105, + "ttft_p50_ms": 162.6282810466364, + "ttft_p95_ms": 258.1990590551868 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 24.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 583.3089610096067, + 571.3486529421061 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r24p00.json", + "sha256": "0774ba2f285d81447c63343b32551d6fc33369b54010629381efa9a37df23a9c", + "summary": { + "admission_lag_max_ms": 0.09822729043662548, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 628.714332007803, + "ttft_p50_ms": 366.9840869260952, + "ttft_p95_ms": 583.3089610096067 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r24p00.json", + "sha256": "1986a05c366ebe73af0f89d2cd0e4efd3e577c800cf1e80997cd26ea0aaaec09", + "summary": { + "admission_lag_max_ms": 0.11414836626499891, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 617.9165309295058, + "ttft_p50_ms": 376.65714998729527, + "ttft_p95_ms": 571.3486529421061 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 28.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 869.708820944652, + 866.4831441128626 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r28p00.json", + "sha256": "ab987729fc833a59817eabaa18c320b0d0a2ecd402fd02bdb534b6c773c029d8", + "summary": { + "admission_lag_max_ms": 0.578614417463541, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 941.3277570856735, + "ttft_p50_ms": 540.4732850147411, + "ttft_p95_ms": 869.708820944652 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r28p00.json", + "sha256": "38f9c82c9ee30927cc0ba857e5e3fce2e79b8a24de03efaec77ce27258633a7e", + "summary": { + "admission_lag_max_ms": 0.12532551772892475, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 938.2033230504021, + "ttft_p50_ms": 541.139594046399, + "ttft_p95_ms": 866.4831441128626 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 32.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1132.6635719742626, + 1144.196804962121 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r32p00.json", + "sha256": "358f5b8e1b72bff5752c076f2966b966d98f6f5767ce52d5a255b7f68f061e54", + "summary": { + "admission_lag_max_ms": 0.32093306072056293, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1197.7118930080906, + "ttft_p50_ms": 678.4568330040202, + "ttft_p95_ms": 1132.6635719742626 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r32p00.json", + "sha256": "3f6dbb760582fdd7470422d8edf6b62949d158ad0d437f04015441a0a4509003", + "summary": { + "admission_lag_max_ms": 1.1304179206490517, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1193.1529619032517, + "ttft_p50_ms": 692.5535890040919, + "ttft_p95_ms": 1144.196804962121 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 2042.27216495201, + 2043.943466967903 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round1/results/r64p00.json", + "sha256": "a24eae7db6eb4cd105ed903917988e8fe6eef5d4cc2549d4e05e9a03dfafb107", + "summary": { + "admission_lag_max_ms": 0.134881935082376, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.546875, + "passed": 35, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2132.23939598538, + "ttft_p50_ms": 1137.3268909519538, + "ttft_p95_ms": 2042.27216495201 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z/artifacts/artifacts/real-tp4-mns64-v1/round2/results/r64p00.json", + "sha256": "e85efaada0b64bb9dd686f59282171ccb256a1d8483c6f037589a68e3e1a104a", + "summary": { + "admission_lag_max_ms": 0.11892011389136314, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.546875, + "passed": 35, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2125.3299129894003, + "ttft_p50_ms": 1134.724210947752, + "ttft_p95_ms": 2043.943466967903 + } + } + ] + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 64, + "name": "tp4_mns64", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z", + "tp": 4 + }, + "tp4_mns8": { + "anchors": [ + { + "conservative_feasible": true, + "rate": 4.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 82.33969006687403, + 69.57727600820363 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r4p00.json", + "sha256": "e69b89ad2e07e6560ebd870773532b33e4415e9ca1515e585cbe465616678569", + "summary": { + "admission_lag_max_ms": 0.14186999760568142, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 86.54833806212991, + "ttft_p50_ms": 62.749656033702195, + "ttft_p95_ms": 82.33969006687403 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r4p00.json", + "sha256": "2c76dcf112a828a6720e079d29bb54a6a65d37d26abaaafdc6d5580751dbad65", + "summary": { + "admission_lag_max_ms": 0.11478399392217398, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 81.3983449479565, + "ttft_p50_ms": 62.220132909715176, + "ttft_p95_ms": 69.57727600820363 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 8.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 63.71171900536865, + 68.14328604377806 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r8p00.json", + "sha256": "0e5d687dfcdb4b6a0d1d060ceefe8a74faab94a5b241580d45b53c452f893b23", + "summary": { + "admission_lag_max_ms": 0.09818794205784798, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 69.63772000744939, + "ttft_p50_ms": 60.89008797425777, + "ttft_p95_ms": 63.71171900536865 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r8p00.json", + "sha256": "1bf016b603473d46fda4bc6d4daeaee5e1091ff4cda1cab7224e0c5edea512f8", + "summary": { + "admission_lag_max_ms": 0.11399004142731428, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 76.76690397784114, + "ttft_p50_ms": 60.9520060243085, + "ttft_p95_ms": 68.14328604377806 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 16.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 133.86176899075508, + 61.95738597307354 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r16p00.json", + "sha256": "197808b1b5fa8082e603a255f358826b5f8f48707223c7fd63abfe029674b817", + "summary": { + "admission_lag_max_ms": 0.10530801955610514, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 191.68931001331657, + "ttft_p50_ms": 61.92394997924566, + "ttft_p95_ms": 133.86176899075508 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r16p00.json", + "sha256": "53e809470daf9efcdabb254e85e8cc3b45d0ab8593c6c2f3aa3396e60d1656da", + "summary": { + "admission_lag_max_ms": 0.1288399798795581, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 70.40882098954171, + "ttft_p50_ms": 59.417470009066164, + "ttft_p95_ms": 61.95738597307354 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 20.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 201.29013701807708, + 200.0734859611839 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r20p00.json", + "sha256": "05b977824fb301517cdbe2d077b360b6053d57fee1c20acad3537321827d027a", + "summary": { + "admission_lag_max_ms": 0.3848400665447116, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 204.5107699232176, + "ttft_p50_ms": 150.0862929970026, + "ttft_p95_ms": 201.29013701807708 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r20p00.json", + "sha256": "baf422b7a3c95b0d3d376a31b0a30e075f69216c794ebc2ef51afafe1ec55229", + "summary": { + "admission_lag_max_ms": 0.24754798505455256, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 204.8192109214142, + "ttft_p50_ms": 148.99871195666492, + "ttft_p95_ms": 200.0734859611839 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 24.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 564.2021730309352, + 560.4996709153056 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r24p00.json", + "sha256": "029b2a9695322d536539c3dd7665725f933b21a8bb48fa94c0448c9f7f8bf0eb", + "summary": { + "admission_lag_max_ms": 0.14956865925341845, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 610.4218009859324, + "ttft_p50_ms": 368.3909500250593, + "ttft_p95_ms": 564.2021730309352 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r24p00.json", + "sha256": "94b46583af3a7a4bbb9f7256c4c4a5ce804b3092459abe3984ea0ce67f591f3c", + "summary": { + "admission_lag_max_ms": 0.11110503692179918, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 606.6705230623484, + "ttft_p50_ms": 367.50321893487126, + "ttft_p95_ms": 560.4996709153056 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 28.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 870.7952990662307, + 894.6108610834926 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r28p00.json", + "sha256": "07bfe263107bb3d2acaa356257343f90045e4295c799178bc0a0b5f1c1994284", + "summary": { + "admission_lag_max_ms": 0.6798935355618596, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 942.792383953929, + "ttft_p50_ms": 543.340316042304, + "ttft_p95_ms": 870.7952990662307 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r28p00.json", + "sha256": "861dd55076ef40a78c34824345c5279dff86262b604db120e9e38efa27b5be55", + "summary": { + "admission_lag_max_ms": 0.09912776295095682, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 930.2741150604561, + "ttft_p50_ms": 549.9696529004723, + "ttft_p95_ms": 894.6108610834926 + } + } + ] + }, + { + "conservative_feasible": true, + "rate": 32.0, + "round_feasible": [ + true, + true + ], + "round_ttft_p95_ms": [ + 1136.603599996306, + 1134.0074760373682 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r32p00.json", + "sha256": "332b805322b571087b018ee459dd91c443dfe0b2bb539b5bdfb4de902bba6da2", + "summary": { + "admission_lag_max_ms": 0.19067595712840557, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1201.8779760692269, + "ttft_p50_ms": 683.3873559953645, + "ttft_p95_ms": 1136.603599996306 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r32p00.json", + "sha256": "996a12a5ed962218dfbe6f30be688119030f10f6539647b6e4a9f25f032b22cb", + "summary": { + "admission_lag_max_ms": 0.11958007235080004, + "completed": 64, + "failed": 0, + "slo": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 1198.4891929896548, + "ttft_p50_ms": 682.7383439522237, + "ttft_p95_ms": 1134.0074760373682 + } + } + ] + }, + { + "conservative_feasible": false, + "rate": 64.0, + "round_feasible": [ + false, + false + ], + "round_ttft_p95_ms": [ + 2040.8174369949847, + 2082.3236129945144 + ], + "rounds": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round1/results/r64p00.json", + "sha256": "c5f17f625071e905b5f1bf8045cd233b2fea0fe6ca00044935dee9c663d1526b", + "summary": { + "admission_lag_max_ms": 0.7995429914444685, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.546875, + "passed": 35, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2131.160691031255, + "ttft_p50_ms": 1134.8516750149429, + "ttft_p95_ms": 2040.8174369949847 + } + }, + { + "path": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z/artifacts/artifacts/real-tp4-mns8-v1/round2/results/r64p00.json", + "sha256": "775db5b941347d9fd5d6df32d6ee037f8168bd47b9eae9a7bdb3dbd8c40aa990", + "summary": { + "admission_lag_max_ms": 0.17620902508497238, + "completed": 64, + "failed": 0, + "slo": { + "feasible": false, + "pass_rate": 0.5625, + "passed": 36, + "ttft_threshold_ms": 1256.0 + }, + "ttft_max_ms": 2129.518053960055, + "ttft_p50_ms": 1112.3740070033818, + "ttft_p95_ms": 2082.3236129945144 + } + } + ] + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 8, + "name": "tp4_mns8", + "source_run": "/home/gahow/phd/aituner/runs/frontier-phase-factorial-v0/fleet-artifacts-exclusive/qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z", + "tp": 4 + } + }, + "real_campaign_resources": { + "fresh_server_anchors": 192, + "gpu_hours": 12.074444444444445, + "measured_requests": 12288, + "runs": [ + { + "duration_seconds": 908.0, + "gpu_hours": 0.25222222222222224, + "run": "qwen30-prefill-real-tp1-mns16-20260717-v2-exclusive-20260716T164730291172Z", + "tp": 1 + }, + { + "duration_seconds": 540.0, + "gpu_hours": 0.15, + "run": "qwen30-prefill-real-tp1-mns16-20260717-v3-refine-20260716T181031481585Z", + "tp": 1 + }, + { + "duration_seconds": 937.0, + "gpu_hours": 0.2602777777777778, + "run": "qwen30-prefill-real-tp1-mns32-20260717-v2-exclusive-20260716T164731376588Z", + "tp": 1 + }, + { + "duration_seconds": 558.0, + "gpu_hours": 0.155, + "run": "qwen30-prefill-real-tp1-mns32-20260717-v3-refine-20260716T181032543571Z", + "tp": 1 + }, + { + "duration_seconds": 972.0, + "gpu_hours": 0.27, + "run": "qwen30-prefill-real-tp1-mns64-20260717-v2-exclusive-20260716T164732482666Z", + "tp": 1 + }, + { + "duration_seconds": 572.0, + "gpu_hours": 0.15888888888888889, + "run": "qwen30-prefill-real-tp1-mns64-20260717-v3-refine-20260716T181033562318Z", + "tp": 1 + }, + { + "duration_seconds": 886.0, + "gpu_hours": 0.2461111111111111, + "run": "qwen30-prefill-real-tp1-mns8-20260717-v2-exclusive-20260716T164729267199Z", + "tp": 1 + }, + { + "duration_seconds": 525.0, + "gpu_hours": 0.14583333333333334, + "run": "qwen30-prefill-real-tp1-mns8-20260717-v3-refine-20260716T181030504792Z", + "tp": 1 + }, + { + "duration_seconds": 983.0, + "gpu_hours": 0.5461111111111111, + "run": "qwen30-prefill-real-tp2-mns16-20260717-v2-exclusive-20260716T164734745039Z", + "tp": 2 + }, + { + "duration_seconds": 569.0, + "gpu_hours": 0.3161111111111111, + "run": "qwen30-prefill-real-tp2-mns16-20260717-v3-refine-20260716T181035686437Z", + "tp": 2 + }, + { + "duration_seconds": 1033.0, + "gpu_hours": 0.5738888888888889, + "run": "qwen30-prefill-real-tp2-mns32-20260717-v2-exclusive-20260716T170909930153Z", + "tp": 2 + }, + { + "duration_seconds": 584.0, + "gpu_hours": 0.3244444444444444, + "run": "qwen30-prefill-real-tp2-mns32-20260717-v3-refine-20260716T182308386197Z", + "tp": 2 + }, + { + "duration_seconds": 1063.0, + "gpu_hours": 0.5905555555555555, + "run": "qwen30-prefill-real-tp2-mns64-20260717-v2-exclusive-20260716T170910954319Z", + "tp": 2 + }, + { + "duration_seconds": 597.0, + "gpu_hours": 0.33166666666666667, + "run": "qwen30-prefill-real-tp2-mns64-20260717-v3-refine-20260716T182309391513Z", + "tp": 2 + }, + { + "duration_seconds": 957.0, + "gpu_hours": 0.5316666666666666, + "run": "qwen30-prefill-real-tp2-mns8-20260717-v2-exclusive-20260716T164733581881Z", + "tp": 2 + }, + { + "duration_seconds": 563.0, + "gpu_hours": 0.31277777777777777, + "run": "qwen30-prefill-real-tp2-mns8-20260717-v3-refine-20260716T181034592841Z", + "tp": 2 + }, + { + "duration_seconds": 1008.0, + "gpu_hours": 1.12, + "run": "qwen30-prefill-real-tp4-mns16-20260717-v2-exclusive-20260716T172922998830Z", + "tp": 4 + }, + { + "duration_seconds": 559.0, + "gpu_hours": 0.6211111111111111, + "run": "qwen30-prefill-real-tp4-mns16-20260717-v4-refine-20260716T183829231278Z", + "tp": 4 + }, + { + "duration_seconds": 1024.0, + "gpu_hours": 1.1377777777777778, + "run": "qwen30-prefill-real-tp4-mns32-20260717-v2-exclusive-20260716T172923955862Z", + "tp": 4 + }, + { + "duration_seconds": 556.0, + "gpu_hours": 0.6177777777777778, + "run": "qwen30-prefill-real-tp4-mns32-20260717-v4-refine-20260716T183830222875Z", + "tp": 4 + }, + { + "duration_seconds": 959.0, + "gpu_hours": 1.0655555555555556, + "run": "qwen30-prefill-real-tp4-mns64-20260717-v2b-exclusive-20260716T175248149074Z", + "tp": 4 + }, + { + "duration_seconds": 567.0, + "gpu_hours": 0.63, + "run": "qwen30-prefill-real-tp4-mns64-20260717-v4-refine-20260716T184826349294Z", + "tp": 4 + }, + { + "duration_seconds": 995.0, + "gpu_hours": 1.1055555555555556, + "run": "qwen30-prefill-real-tp4-mns8-20260717-v2-exclusive-20260716T170911977980Z", + "tp": 4 + }, + { + "duration_seconds": 550.0, + "gpu_hours": 0.6111111111111112, + "run": "qwen30-prefill-real-tp4-mns8-20260717-v3-refine-20260716T182310414444Z", + "tp": 4 + } + ], + "successful_fleet_jobs": 24, + "warmup_requests": 4512 + }, + "schema": "qwen30-prefill-fidelity-comparison-v1", + "simulator": { + "tp1_mns16": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 166.6461504479848 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 575.6631856791392 + }, + { + "feasible": false, + "pass_rate": 0.265625, + "rate": 16.0, + "ttft_p95_ms": 3529.1762077816857 + }, + { + "feasible": false, + "pass_rate": 0.1875, + "rate": 32.0, + "ttft_p95_ms": 5290.0301465092725 + }, + { + "feasible": false, + "pass_rate": 0.140625, + "rate": 64.0, + "ttft_p95_ms": 6180.6551465092725 + } + ], + "capacity": 8.0, + "capacity_per_gpu": 8.0, + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "tp1_mns32": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 166.6461504479848 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 575.6631856791392 + }, + { + "feasible": false, + "pass_rate": 0.265625, + "rate": 16.0, + "ttft_p95_ms": 3529.1762077816857 + }, + { + "feasible": false, + "pass_rate": 0.1875, + "rate": 32.0, + "ttft_p95_ms": 5290.0301465092725 + }, + { + "feasible": false, + "pass_rate": 0.140625, + "rate": 64.0, + "ttft_p95_ms": 6180.6551465092725 + } + ], + "capacity": 8.0, + "capacity_per_gpu": 8.0, + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "tp1_mns64": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 166.6461504479848 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 575.6631856791392 + }, + { + "feasible": false, + "pass_rate": 0.265625, + "rate": 16.0, + "ttft_p95_ms": 3529.1762077816857 + }, + { + "feasible": false, + "pass_rate": 0.1875, + "rate": 32.0, + "ttft_p95_ms": 5290.0301465092725 + }, + { + "feasible": false, + "pass_rate": 0.140625, + "rate": 64.0, + "ttft_p95_ms": 6180.6551465092725 + } + ], + "capacity": 8.0, + "capacity_per_gpu": 8.0, + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "tp1_mns8": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 166.6461504479848 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 575.6631856791392 + }, + { + "feasible": false, + "pass_rate": 0.265625, + "rate": 16.0, + "ttft_p95_ms": 3529.1762077816857 + }, + { + "feasible": false, + "pass_rate": 0.1875, + "rate": 32.0, + "ttft_p95_ms": 5290.0301465092725 + }, + { + "feasible": false, + "pass_rate": 0.140625, + "rate": 64.0, + "ttft_p95_ms": 6180.6551465092725 + } + ], + "capacity": 8.0, + "capacity_per_gpu": 8.0, + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "tp2_mns16": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 118.55152318709017 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 118.55152318708929 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 786.9875986853465 + }, + { + "feasible": false, + "pass_rate": 0.421875, + "rate": 32.0, + "ttft_p95_ms": 2566.908354521784 + }, + { + "feasible": false, + "pass_rate": 0.3125, + "rate": 64.0, + "ttft_p95_ms": 3439.824414996849 + } + ], + "capacity": 16.0, + "capacity_per_gpu": 8.0, + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "tp2_mns32": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 118.55152318709017 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 118.55152318708929 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 786.9875986853465 + }, + { + "feasible": false, + "pass_rate": 0.421875, + "rate": 32.0, + "ttft_p95_ms": 2566.908354521784 + }, + { + "feasible": false, + "pass_rate": 0.3125, + "rate": 64.0, + "ttft_p95_ms": 3439.824414996849 + } + ], + "capacity": 16.0, + "capacity_per_gpu": 8.0, + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "tp2_mns64": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 118.55152318709017 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 118.55152318708929 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 786.9875986853465 + }, + { + "feasible": false, + "pass_rate": 0.421875, + "rate": 32.0, + "ttft_p95_ms": 2566.908354521784 + }, + { + "feasible": false, + "pass_rate": 0.3125, + "rate": 64.0, + "ttft_p95_ms": 3439.824414996849 + } + ], + "capacity": 16.0, + "capacity_per_gpu": 8.0, + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "tp2_mns8": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 118.55152318709017 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 118.55152318708929 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 786.9875986853465 + }, + { + "feasible": false, + "pass_rate": 0.421875, + "rate": 32.0, + "ttft_p95_ms": 2566.908354521784 + }, + { + "feasible": false, + "pass_rate": 0.3125, + "rate": 64.0, + "ttft_p95_ms": 3439.824414996849 + } + ], + "capacity": 16.0, + "capacity_per_gpu": 8.0, + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "tp4_mns16": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 86.51189290936045 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 86.51189290936001 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 252.16908885537626 + }, + { + "feasible": true, + "pass_rate": 0.984375, + "rate": 32.0, + "ttft_p95_ms": 1219.2127732631457 + }, + { + "feasible": false, + "pass_rate": 0.515625, + "rate": 64.0, + "ttft_p95_ms": 2106.0636248827373 + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "tp4_mns32": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 86.51189290936045 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 86.51189290936001 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 252.16908885537626 + }, + { + "feasible": true, + "pass_rate": 0.984375, + "rate": 32.0, + "ttft_p95_ms": 1219.2127732631457 + }, + { + "feasible": false, + "pass_rate": 0.515625, + "rate": 64.0, + "ttft_p95_ms": 2106.0636248827373 + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "tp4_mns64": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 86.51189290936045 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 86.51189290936001 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 252.16908885537626 + }, + { + "feasible": true, + "pass_rate": 0.984375, + "rate": 32.0, + "ttft_p95_ms": 1219.2127732631457 + }, + { + "feasible": false, + "pass_rate": 0.515625, + "rate": 64.0, + "ttft_p95_ms": 2106.0636248827373 + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "tp4_mns8": { + "anchors": [ + { + "feasible": true, + "pass_rate": 1.0, + "rate": 4.0, + "ttft_p95_ms": 86.51189290936045 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 8.0, + "ttft_p95_ms": 86.51189290936001 + }, + { + "feasible": true, + "pass_rate": 1.0, + "rate": 16.0, + "ttft_p95_ms": 252.16908885537626 + }, + { + "feasible": true, + "pass_rate": 0.984375, + "rate": 32.0, + "ttft_p95_ms": 1219.2127732631457 + }, + { + "feasible": false, + "pass_rate": 0.515625, + "rate": 64.0, + "ttft_p95_ms": 2106.0636248827373 + } + ], + "capacity": 32.0, + "capacity_per_gpu": 8.0, + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + } + }, + "simulator_sources": [ + { + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/simulator-a3/frontier_surface_frozen.json", + "sha256": "fe7392fbcd20f6a96562773b3644df9dac23c8562462356acc86585f7f1d40f0" + } + ] +} diff --git a/runs/frontier-fidelity-envelope-v1/results/a3/frontier_surface_frozen.json b/runs/frontier-fidelity-envelope-v1/results/a3/frontier_surface_frozen.json new file mode 100644 index 0000000..34009eb --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/results/a3/frontier_surface_frozen.json @@ -0,0 +1,1736 @@ +{ + "capacity": [ + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 8.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 8.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 8.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 8.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 16.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 16.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 16.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 16.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 8.0, + "upper_censored": false + } + ], + "collective": { + "allreduce_csv": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/profiles/measured-allreduce.csv", + "allreduce_csv_sha256": "9d693fd406616b599e57bcde66c980c7fc2831b3acf37d3eb633cec80ea0070d", + "backend": "vidur" + }, + "config_results": [ + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 157.35246539115906, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "705075942dea05ebacb4c83c2587a865e40e1c7af243d2894fa7a05409f8fe59", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.020947591286289, + "ttft_max_ms": 166.6461504479848, + "ttft_p50_ms": 166.64615044798398, + "ttft_p95_ms": 166.6461504479848 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.141019344329834, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "5dbdc7f1672445ef892ed104c0ea1bfb02baba3f87e8a3e707c523707861b747", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 7.75274759930899, + "ttft_max_ms": 586.937565208359, + "ttft_p50_ms": 436.98066306434936, + "ttft_p95_ms": 575.6631856791392 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.098863363265991, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "58835b434e8f85b91a8d14681bf4a6a0df5f38c69ead31bbc6e02c1c50f8a9c1", + "score": { + "feasible": false, + "pass_rate": 0.265625, + "passed": 17, + "throughput_requests_per_second": 8.523874834682788, + "ttft_max_ms": 3654.1762077816857, + "ttft_p50_ms": 1994.5136762864174, + "ttft_p95_ms": 3529.1762077816857 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.089977025985718, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a83336424fd7682e60e1a02a62d5c1edbcded34f14919b6011c8bc76a3f24de1", + "score": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 5483.402803563629, + "ttft_p50_ms": 2911.6176150140045, + "ttft_p95_ms": 5290.0301465092725 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.088061094284058, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "000767c0b12d8ec4c9e83f978b3552c8c6b99a2b42414100fc1a7b3e8e5c721e", + "score": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 6436.527803563629, + "ttft_p50_ms": 3380.3676150140045, + "ttft_p95_ms": 6180.6551465092725 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 8.947673320770264, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "705075942dea05ebacb4c83c2587a865e40e1c7af243d2894fa7a05409f8fe59", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.020947591286289, + "ttft_max_ms": 166.6461504479848, + "ttft_p50_ms": 166.64615044798398, + "ttft_p95_ms": 166.6461504479848 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.042973518371582, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "5dbdc7f1672445ef892ed104c0ea1bfb02baba3f87e8a3e707c523707861b747", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 7.75274759930899, + "ttft_max_ms": 586.937565208359, + "ttft_p50_ms": 436.98066306434936, + "ttft_p95_ms": 575.6631856791392 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.092506170272827, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "58835b434e8f85b91a8d14681bf4a6a0df5f38c69ead31bbc6e02c1c50f8a9c1", + "score": { + "feasible": false, + "pass_rate": 0.265625, + "passed": 17, + "throughput_requests_per_second": 8.523874834682788, + "ttft_max_ms": 3654.1762077816857, + "ttft_p50_ms": 1994.5136762864174, + "ttft_p95_ms": 3529.1762077816857 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.044708013534546, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a83336424fd7682e60e1a02a62d5c1edbcded34f14919b6011c8bc76a3f24de1", + "score": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 5483.402803563629, + "ttft_p50_ms": 2911.6176150140045, + "ttft_p95_ms": 5290.0301465092725 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.29175853729248, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "000767c0b12d8ec4c9e83f978b3552c8c6b99a2b42414100fc1a7b3e8e5c721e", + "score": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 6436.527803563629, + "ttft_p50_ms": 3380.3676150140045, + "ttft_p95_ms": 6180.6551465092725 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 8.991463661193848, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "705075942dea05ebacb4c83c2587a865e40e1c7af243d2894fa7a05409f8fe59", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.020947591286289, + "ttft_max_ms": 166.6461504479848, + "ttft_p50_ms": 166.64615044798398, + "ttft_p95_ms": 166.6461504479848 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.138724327087402, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "5dbdc7f1672445ef892ed104c0ea1bfb02baba3f87e8a3e707c523707861b747", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 7.75274759930899, + "ttft_max_ms": 586.937565208359, + "ttft_p50_ms": 436.98066306434936, + "ttft_p95_ms": 575.6631856791392 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.188970565795898, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "58835b434e8f85b91a8d14681bf4a6a0df5f38c69ead31bbc6e02c1c50f8a9c1", + "score": { + "feasible": false, + "pass_rate": 0.265625, + "passed": 17, + "throughput_requests_per_second": 8.523874834682788, + "ttft_max_ms": 3654.1762077816857, + "ttft_p50_ms": 1994.5136762864174, + "ttft_p95_ms": 3529.1762077816857 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.189127683639526, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a83336424fd7682e60e1a02a62d5c1edbcded34f14919b6011c8bc76a3f24de1", + "score": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 5483.402803563629, + "ttft_p50_ms": 2911.6176150140045, + "ttft_p95_ms": 5290.0301465092725 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.189879894256592, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "000767c0b12d8ec4c9e83f978b3552c8c6b99a2b42414100fc1a7b3e8e5c721e", + "score": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 6436.527803563629, + "ttft_p50_ms": 3380.3676150140045, + "ttft_p95_ms": 6180.6551465092725 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 8.999529123306274, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "705075942dea05ebacb4c83c2587a865e40e1c7af243d2894fa7a05409f8fe59", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.020947591286289, + "ttft_max_ms": 166.6461504479848, + "ttft_p50_ms": 166.64615044798398, + "ttft_p95_ms": 166.6461504479848 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.192278385162354, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "5dbdc7f1672445ef892ed104c0ea1bfb02baba3f87e8a3e707c523707861b747", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 7.75274759930899, + "ttft_max_ms": 586.937565208359, + "ttft_p50_ms": 436.98066306434936, + "ttft_p95_ms": 575.6631856791392 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.040328979492188, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "58835b434e8f85b91a8d14681bf4a6a0df5f38c69ead31bbc6e02c1c50f8a9c1", + "score": { + "feasible": false, + "pass_rate": 0.265625, + "passed": 17, + "throughput_requests_per_second": 8.523874834682788, + "ttft_max_ms": 3654.1762077816857, + "ttft_p50_ms": 1994.5136762864174, + "ttft_p95_ms": 3529.1762077816857 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 8.990798473358154, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a83336424fd7682e60e1a02a62d5c1edbcded34f14919b6011c8bc76a3f24de1", + "score": { + "feasible": false, + "pass_rate": 0.1875, + "passed": 12, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 5483.402803563629, + "ttft_p50_ms": 2911.6176150140045, + "ttft_p95_ms": 5290.0301465092725 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.192291736602783, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "000767c0b12d8ec4c9e83f978b3552c8c6b99a2b42414100fc1a7b3e8e5c721e", + "score": { + "feasible": false, + "pass_rate": 0.140625, + "passed": 9, + "throughput_requests_per_second": 8.660758725922316, + "ttft_max_ms": 6436.527803563629, + "ttft_p50_ms": 3380.3676150140045, + "ttft_p95_ms": 6180.6551465092725 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.093063116073608, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "195c2750784a03930b1ae369648ed6361fb8d1ea506f0828bc7c9818a4f55961", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.033134335322499, + "ttft_max_ms": 118.55152318709017, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318709017 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 160.23531913757324, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "733c6a6bca1efc56f9a160dd89a9156446581d5a1c4c072b90f0a7659acc637e", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.006453678862723, + "ttft_max_ms": 118.55152318708929, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318708929 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.289193630218506, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "37ba1e79ce560f7449049a3ef35429ee1601b35f249800f5e2e0b4d96121f0a4", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 13.760059242174107, + "ttft_max_ms": 849.4875986853465, + "ttft_p50_ms": 539.7284419614418, + "ttft_p95_ms": 786.9875986853465 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.239543676376343, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "1fb27df916c2d9002f5584c9a3ac3ca5fceed2f5fece4f40e69748dbad7aee3b", + "score": { + "feasible": false, + "pass_rate": 0.421875, + "passed": 27, + "throughput_requests_per_second": 14.110410219984, + "ttft_max_ms": 2660.658354521784, + "ttft_p50_ms": 1414.3128122232463, + "ttft_p95_ms": 2566.908354521784 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.141466856002808, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "ccefacf7e9b214037083892f6fa84c5e0fd288f041ebfe21880b9709c8b25727", + "score": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "throughput_requests_per_second": 14.110410219983999, + "ttft_max_ms": 3582.533354521785, + "ttft_p50_ms": 1896.1470654856275, + "ttft_p95_ms": 3439.824414996849 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.042891025543213, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "195c2750784a03930b1ae369648ed6361fb8d1ea506f0828bc7c9818a4f55961", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.033134335322499, + "ttft_max_ms": 118.55152318709017, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318709017 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.136998653411865, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "733c6a6bca1efc56f9a160dd89a9156446581d5a1c4c072b90f0a7659acc637e", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.006453678862723, + "ttft_max_ms": 118.55152318708929, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318708929 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.13957166671753, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "37ba1e79ce560f7449049a3ef35429ee1601b35f249800f5e2e0b4d96121f0a4", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 13.760059242174107, + "ttft_max_ms": 849.4875986853465, + "ttft_p50_ms": 539.7284419614418, + "ttft_p95_ms": 786.9875986853465 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.098120927810669, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "1fb27df916c2d9002f5584c9a3ac3ca5fceed2f5fece4f40e69748dbad7aee3b", + "score": { + "feasible": false, + "pass_rate": 0.421875, + "passed": 27, + "throughput_requests_per_second": 14.110410219984, + "ttft_max_ms": 2660.658354521784, + "ttft_p50_ms": 1414.3128122232463, + "ttft_p95_ms": 2566.908354521784 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.140004634857178, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "ccefacf7e9b214037083892f6fa84c5e0fd288f041ebfe21880b9709c8b25727", + "score": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "throughput_requests_per_second": 14.110410219983999, + "ttft_max_ms": 3582.533354521785, + "ttft_p50_ms": 1896.1470654856275, + "ttft_p95_ms": 3439.824414996849 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.090512752532959, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "195c2750784a03930b1ae369648ed6361fb8d1ea506f0828bc7c9818a4f55961", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.033134335322499, + "ttft_max_ms": 118.55152318709017, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318709017 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.102084636688232, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "733c6a6bca1efc56f9a160dd89a9156446581d5a1c4c072b90f0a7659acc637e", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.006453678862723, + "ttft_max_ms": 118.55152318708929, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318708929 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.09265947341919, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "37ba1e79ce560f7449049a3ef35429ee1601b35f249800f5e2e0b4d96121f0a4", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 13.760059242174107, + "ttft_max_ms": 849.4875986853465, + "ttft_p50_ms": 539.7284419614418, + "ttft_p95_ms": 786.9875986853465 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.14569616317749, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "1fb27df916c2d9002f5584c9a3ac3ca5fceed2f5fece4f40e69748dbad7aee3b", + "score": { + "feasible": false, + "pass_rate": 0.421875, + "passed": 27, + "throughput_requests_per_second": 14.110410219984, + "ttft_max_ms": 2660.658354521784, + "ttft_p50_ms": 1414.3128122232463, + "ttft_p95_ms": 2566.908354521784 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.191733598709106, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "ccefacf7e9b214037083892f6fa84c5e0fd288f041ebfe21880b9709c8b25727", + "score": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "throughput_requests_per_second": 14.110410219983999, + "ttft_max_ms": 3582.533354521785, + "ttft_p50_ms": 1896.1470654856275, + "ttft_p95_ms": 3439.824414996849 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.146432876586914, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "195c2750784a03930b1ae369648ed6361fb8d1ea506f0828bc7c9818a4f55961", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.033134335322499, + "ttft_max_ms": 118.55152318709017, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318709017 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.068153619766235, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "733c6a6bca1efc56f9a160dd89a9156446581d5a1c4c072b90f0a7659acc637e", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.006453678862723, + "ttft_max_ms": 118.55152318708929, + "ttft_p50_ms": 118.55152318708929, + "ttft_p95_ms": 118.55152318708929 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.139062404632568, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "37ba1e79ce560f7449049a3ef35429ee1601b35f249800f5e2e0b4d96121f0a4", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 13.760059242174107, + "ttft_max_ms": 849.4875986853465, + "ttft_p50_ms": 539.7284419614418, + "ttft_p95_ms": 786.9875986853465 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.143300771713257, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "1fb27df916c2d9002f5584c9a3ac3ca5fceed2f5fece4f40e69748dbad7aee3b", + "score": { + "feasible": false, + "pass_rate": 0.421875, + "passed": 27, + "throughput_requests_per_second": 14.110410219984, + "ttft_max_ms": 2660.658354521784, + "ttft_p50_ms": 1414.3128122232463, + "ttft_p95_ms": 2566.908354521784 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.092010021209717, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "ccefacf7e9b214037083892f6fa84c5e0fd288f041ebfe21880b9709c8b25727", + "score": { + "feasible": false, + "pass_rate": 0.3125, + "passed": 20, + "throughput_requests_per_second": 14.110410219983999, + "ttft_max_ms": 3582.533354521785, + "ttft_p50_ms": 1896.1470654856275, + "ttft_p95_ms": 3439.824414996849 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 161.351567029953, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 1.0, + "request_count": 64, + "request_metrics_sha256": "d1d9086461911338e6c87a8303aed090f25e2bbbfd184341e2df531f64c23355", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.041293968822475, + "ttft_max_ms": 86.51189290936045, + "ttft_p50_ms": 86.51189290936001, + "ttft_p95_ms": 86.51189290936045 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.191051244735718, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "ee23707d601c6b4559c7f019b09020acdb589844eeffb79f6b46f2365805ac99", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.038674169035575, + "ttft_max_ms": 86.51189290936001, + "ttft_p50_ms": 86.51189290935957, + "ttft_p95_ms": 86.51189290936001 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.050537109375, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "63cce9144c69236a07bb94cdef18c7d09dddeaa1c0ba549a36ff29742c780c2a", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 15.535832140198398, + "ttft_max_ms": 276.9276050512883, + "ttft_p50_ms": 189.66908885537626, + "ttft_p95_ms": 252.16908885537626 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.04306149482727, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "568476977590c981958cc0acc4e07547c079a2ae00520b59f8ed7761acb67bfb", + "score": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63, + "throughput_requests_per_second": 19.925437356037587, + "ttft_max_ms": 1281.7127732631457, + "ttft_p50_ms": 729.6185162106785, + "ttft_p95_ms": 1219.2127732631457 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.14296841621399, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "f789fb967a981edf5c112d0e1a5d0a034456e2cf51f49d6e2e8597afe9df0f2f", + "score": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33, + "throughput_requests_per_second": 20.34996647854275, + "ttft_max_ms": 2191.8433254998167, + "ttft_p50_ms": 1169.8561499618286, + "ttft_p95_ms": 2106.0636248827373 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.14107608795166, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 1.0, + "request_count": 64, + "request_metrics_sha256": "d1d9086461911338e6c87a8303aed090f25e2bbbfd184341e2df531f64c23355", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.041293968822475, + "ttft_max_ms": 86.51189290936045, + "ttft_p50_ms": 86.51189290936001, + "ttft_p95_ms": 86.51189290936045 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.139281272888184, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "ee23707d601c6b4559c7f019b09020acdb589844eeffb79f6b46f2365805ac99", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.038674169035575, + "ttft_max_ms": 86.51189290936001, + "ttft_p50_ms": 86.51189290935957, + "ttft_p95_ms": 86.51189290936001 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.240875244140625, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "63cce9144c69236a07bb94cdef18c7d09dddeaa1c0ba549a36ff29742c780c2a", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 15.535832140198398, + "ttft_max_ms": 276.9276050512883, + "ttft_p50_ms": 189.66908885537626, + "ttft_p95_ms": 252.16908885537626 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.090924739837646, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "568476977590c981958cc0acc4e07547c079a2ae00520b59f8ed7761acb67bfb", + "score": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63, + "throughput_requests_per_second": 19.925437356037587, + "ttft_max_ms": 1281.7127732631457, + "ttft_p50_ms": 729.6185162106785, + "ttft_p95_ms": 1219.2127732631457 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.041213274002075, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "f789fb967a981edf5c112d0e1a5d0a034456e2cf51f49d6e2e8597afe9df0f2f", + "score": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33, + "throughput_requests_per_second": 20.34996647854275, + "ttft_max_ms": 2191.8433254998167, + "ttft_p50_ms": 1169.8561499618286, + "ttft_p95_ms": 2106.0636248827373 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.092233896255493, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 1.0, + "request_count": 64, + "request_metrics_sha256": "d1d9086461911338e6c87a8303aed090f25e2bbbfd184341e2df531f64c23355", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.041293968822475, + "ttft_max_ms": 86.51189290936045, + "ttft_p50_ms": 86.51189290936001, + "ttft_p95_ms": 86.51189290936045 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.093050956726074, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "ee23707d601c6b4559c7f019b09020acdb589844eeffb79f6b46f2365805ac99", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.038674169035575, + "ttft_max_ms": 86.51189290936001, + "ttft_p50_ms": 86.51189290935957, + "ttft_p95_ms": 86.51189290936001 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.242159605026245, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "63cce9144c69236a07bb94cdef18c7d09dddeaa1c0ba549a36ff29742c780c2a", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 15.535832140198398, + "ttft_max_ms": 276.9276050512883, + "ttft_p50_ms": 189.66908885537626, + "ttft_p95_ms": 252.16908885537626 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.19024658203125, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "568476977590c981958cc0acc4e07547c079a2ae00520b59f8ed7761acb67bfb", + "score": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63, + "throughput_requests_per_second": 19.925437356037587, + "ttft_max_ms": 1281.7127732631457, + "ttft_p50_ms": 729.6185162106785, + "ttft_p95_ms": 1219.2127732631457 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.18896198272705, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "f789fb967a981edf5c112d0e1a5d0a034456e2cf51f49d6e2e8597afe9df0f2f", + "score": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33, + "throughput_requests_per_second": 20.34996647854275, + "ttft_max_ms": 2191.8433254998167, + "ttft_p50_ms": 1169.8561499618286, + "ttft_p95_ms": 2106.0636248827373 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.088153839111328, + "offered_request_rate": 4.0, + "offered_request_rate_per_gpu": 1.0, + "request_count": 64, + "request_metrics_sha256": "d1d9086461911338e6c87a8303aed090f25e2bbbfd184341e2df531f64c23355", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 4.041293968822475, + "ttft_max_ms": 86.51189290936045, + "ttft_p50_ms": 86.51189290936001, + "ttft_p95_ms": 86.51189290936045 + }, + "status": "completed", + "trace_sha256": "0cef631f351f5afd880172d0467d62d9813de5ba36fd3795f91afe3256450c05" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.048948764801025, + "offered_request_rate": 8.0, + "offered_request_rate_per_gpu": 2.0, + "request_count": 64, + "request_metrics_sha256": "ee23707d601c6b4559c7f019b09020acdb589844eeffb79f6b46f2365805ac99", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 8.038674169035575, + "ttft_max_ms": 86.51189290936001, + "ttft_p50_ms": 86.51189290935957, + "ttft_p95_ms": 86.51189290936001 + }, + "status": "completed", + "trace_sha256": "c49314cbe4e4b07a9335b57e861ec5b6541ea4cc061ca9abf923453276c419fd" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.04839539527893, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "63cce9144c69236a07bb94cdef18c7d09dddeaa1c0ba549a36ff29742c780c2a", + "score": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64, + "throughput_requests_per_second": 15.535832140198398, + "ttft_max_ms": 276.9276050512883, + "ttft_p50_ms": 189.66908885537626, + "ttft_p95_ms": 252.16908885537626 + }, + "status": "completed", + "trace_sha256": "2504b6340f764e726110e48251f8017a2f9ba67b44b8af7d3a838bf674823d33" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.139516353607178, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "568476977590c981958cc0acc4e07547c079a2ae00520b59f8ed7761acb67bfb", + "score": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63, + "throughput_requests_per_second": 19.925437356037587, + "ttft_max_ms": 1281.7127732631457, + "ttft_p50_ms": 729.6185162106785, + "ttft_p95_ms": 1219.2127732631457 + }, + "status": "completed", + "trace_sha256": "5508a3ce590f7d1b0b3c211a2d4f2131886a7ae2d85cafa0c436a8e3029675f1" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.145836114883423, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "f789fb967a981edf5c112d0e1a5d0a034456e2cf51f49d6e2e8597afe9df0f2f", + "score": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33, + "throughput_requests_per_second": 20.34996647854275, + "ttft_max_ms": 2191.8433254998167, + "ttft_p50_ms": 1169.8561499618286, + "ttft_p95_ms": 2106.0636248827373 + }, + "status": "completed", + "trace_sha256": "8987c3ebd4a7db983dc0e99e44710cd44a9d5d21cdb8e6b58878b160918593d9" + } + ] + } + ], + "contract": { + "arrival": "open_loop_uniform", + "input_tokens": 2048, + "output_tokens": 1, + "prefix_caching": false, + "rates": [ + 4.0, + 8.0, + 16.0, + 32.0, + 64.0 + ], + "requests_per_anchor": 64, + "target_pass_rate": 0.95, + "ttft_slo_ms": 1256.0 + }, + "frontier": { + "git_head": "d9cfeb6d8791fbf2f295dd9744c56a666171776e", + "git_status_short": " M frontier/cc_backend/backends/vidur_cc_backend.py\n M frontier/config/config.py\n M frontier/entities/request.py\n M frontier/events/cluster_schedule_event.py\n M frontier/execution_time_predictor/sklearn_execution_time_predictor.py\n M frontier/metrics/constants.py\n M frontier/metrics/metrics_store.py\n M frontier/profiling/common/layers/rotary_embedding.py\n M frontier/profiling/moe/moe_impl.py\n M frontier/profiling/moe/moe_vllm_kernel.py\n M frontier/scheduler/cluster_scheduler/__init__.py\n M frontier/scheduler/cluster_scheduler/base_cluster_scheduler.py\n M frontier/scheduler/cluster_scheduler/cluster_scheduler_registry.py\n M frontier/scheduler/cluster_scheduler/sticky_lor_cluster_scheduler.py\n M frontier/scheduler/replica_scheduler/base_replica_scheduler.py\n M frontier/scheduler/replica_scheduler/vllm_v1_engine_replica_scheduler.py\n M frontier/scheduler/replica_stage_scheduler/replica_stage_schduler.py\n M frontier/simulator.py\n M frontier/types/cluster_scheduler_type.py\n?? data/profiling/compute/h20/\n?? frontier/scheduler/cluster_scheduler/prefix_lor_cluster_scheduler.py\n?? runs/\n?? tests/unit/test_attn_prefill_prediction_fallback.py\n?? tests/unit/test_vidur_cc_large_payload.py\n", + "source": "/tmp/replayserve-frontier-rs1b" + }, + "profiles": { + "coverage": { + "attention": { + "1": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + }, + "2": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + }, + "4": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + } + }, + "manifest": { + "attention_tp_coverage": [ + 1, + 2, + 4 + ], + "environment_contract": { + "dtype": "bfloat16", + "frontier_commit": "d9cfeb6d8791fbf2f295dd9744c56a666171776e", + "hardware": "NVIDIA H20", + "model": "Qwen3-30B-A3B", + "tensor_parallel_sizes": [ + 1, + 2, + 4 + ], + "vllm_source_commit": "88d34c6409e9fb3c7b8ca0c04756f061d2099eb1", + "vllm_version": "0.20.0" + }, + "inputs": { + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp1-20260717-v1-20260717T014550833983Z/artifacts/artifacts/attention-composition-tp1-v1/raw/flashattn-composition-tp1.json": "ce461be92a5059e2d168b2cbf21b55d9a55ee71c3b0586a3e63cd8288938168d", + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp2-20260717-v1-20260717T014551831545Z/artifacts/artifacts/attention-composition-tp2-v1/raw/flashattn-composition-tp2.json": "997d67fb3a28b488c3aa4945ae7c6d4ecf8a08726ee3cb4652d16c221ff395ad", + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp4-20260717-v1-20260717T014552897206Z/artifacts/artifacts/attention-composition-tp4-v1/raw/flashattn-composition-tp4.json": "046cffd4c11d8a36513167ec25d3e1ca016324715679ea688a58d4a71dfe62b2", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-allreduce-full-tp2-20260716-v1-dispatch-aware-20260716T140743025781Z/artifacts/artifacts/allreduce-full-tp2-v1/raw/allreduce-tp2.json": "97c3c76b5a04e95bd9192423c2b891667c668f39cc0dfecbd097d749939f2d0a", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-allreduce-full-tp4-20260716-v1-dispatch-aware-20260716T141106009788Z/artifacts/artifacts/allreduce-full-tp4-v1/raw/allreduce-tp4.json": "809df9baa6f468cf12bf0c99827475acc67894dd9f3f948976590b665fac0e76", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp1-20260716-v2-20260716T135132587012Z/artifacts/artifacts/flashattn-kv-full-v2-tp1/raw/flashattn-tp1.json": "dcb4c1bf7e76b9c765f78ddd2b8a734f2d7ba2adac13ce017689a8a77fe69a27", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp2-20260716-v2-20260716T135134194295Z/artifacts/artifacts/flashattn-kv-full-v2-tp2/raw/flashattn-tp2.json": "43ce042556ba887c8860614b43ccf0f564e5cebc1a0cffbce299d0acb9fa8d07", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp4-20260716-v2-20260716T135135197200Z/artifacts/artifacts/flashattn-kv-full-v2-tp4/raw/flashattn-tp4.json": "84eef31bcad0f556907a093318a420959d14fdc94474823d11f659704bdfec73", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-frontier-linear-full-20260716-v2-max-tokens-20260716T144444676943Z/artifacts/artifacts/frontier-linear-full-v2/profiles/compute/h20/qwen3-a3b-30b-moe/linear_op.csv": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-moe-full-20260716-v1-local-shard-20260716T141334565164Z/artifacts/artifacts/moe-full-v1/raw/moe-full.json": "588f6ad0d69c9636d1b852e3df0a12d13cfe731f050ea7ec7aea457cceefbde8", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-router-full-20260716-v3-tp-context-20260716T145446098505Z/artifacts/artifacts/router-full-v3/raw/router.json": "1962972e983bff3e06a721ef4ae4ec65728ff669681497a4a7e7f769b88b4931" + }, + "outputs": { + "allreduce.json": "b38d14f990578d668523d25b107aceed433da5020d8ada3b6e44d3562261a3b3", + "attention.csv": "03180eec28087a301c9d8c185874c50bdb28b31e5e9af35fcce5a6a5a9e62147", + "attention_true_mixed_fused.csv": "cccf76e6cb6a9359f188b07d4dc2942442568a6e8cf3ff3d77ab3d35a052cecb", + "linear_op.csv": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "moe.csv": "dd698113975e5ed77f51dfd8a69aebf764ed651c3f6f34b71914d25d6e001415" + }, + "profile_id": "qwen3-30b-a3b-bf16-vllm020-h20-tp1-2-4-fused-mixed-total-conserving-pure-prefill-batch-composition", + "projection_contract": { + "allreduce": "Frozen exact runtime measurements; base profile-only comparison keeps the historical Frontier CC backend fixed to isolate compute profile fidelity", + "attention": "Pure prefill/extend/decode FA3 core plus separately measured KV update; input/output reshape assumed zero; exported mean is used as median target; true mixed rows use a total-conserving compatibility projection", + "attention_pure_prefill_batch_composition": "Direct FA3 measurements for 2/4 requests at query length 2048 and 2/4/8/16 requests at query length 512 for each TP; included only when batch-composition attention inputs are supplied", + "attention_true_mixed": "The directly measured fused total is preserved in diagnostics. Frontier's two targets are projected by the same-TP pure prefill/decode reference ratio, with projected prefill + decode exactly equal to the fused total; the split is a schema compatibility attribution, not an observation", + "linear": "Frontier profiler using vLLM 0.20 CUDA operators", + "moe": "Replicated gate and fused top-k plus TP-local modular expert kernel; expert measurement already includes prepare/finalize so shuffling is zero" + }, + "row_counts": { + "allreduce": 24, + "attention_frontier_compatible": 150, + "attention_true_mixed_fused_diagnostic": 30, + "linear": 36, + "moe": 72 + }, + "schema_version": "frontier_qwen30_vllm020_frozen_profile.v3" + } + }, + "root": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/profiles/profile-v3-batch-final", + "sha256": { + "attention": "03180eec28087a301c9d8c185874c50bdb28b31e5e9af35fcce5a6a5a9e62147", + "linear": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "manifest": "70aecb6d70f9e509e9816b1a0929163b9505760c2ac30bbd4cd8e51badfb7635", + "moe": "dd698113975e5ed77f51dfd8a69aebf764ed651c3f6f34b71914d25d6e001415" + } + }, + "schema": "frontier-qwen30-prefill-surface-v1", + "status": "frozen_before_real" +} diff --git a/runs/frontier-fidelity-envelope-v1/results/a3/qwen30-prefill-ranking.png b/runs/frontier-fidelity-envelope-v1/results/a3/qwen30-prefill-ranking.png new file mode 100644 index 0000000..ed8a221 Binary files /dev/null and b/runs/frontier-fidelity-envelope-v1/results/a3/qwen30-prefill-ranking.png differ diff --git a/runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py b/runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py index 3d8aebc..f461645 100644 --- a/runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py +++ b/runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py @@ -148,6 +148,14 @@ def find_request_metrics(run_dir: Path) -> Path: return matches[0] +def trace_manifest_entry(trace: dict[str, Any]) -> dict[str, Any]: + return { + key: str(value) if isinstance(value, Path) else value + for key, value in trace.items() + if key != "shapes" + } + + def score(path: Path, expected_shapes: list[tuple[int, int]]) -> dict[str, Any]: with path.open(newline="") as source: rows = list(csv.DictReader(source)) @@ -409,10 +417,7 @@ def main() -> None: "allreduce_csv": str(args.allreduce_csv) if args.allreduce_csv else None, "allreduce_csv_sha256": BASE.sha256(args.allreduce_csv) if args.allreduce_csv else None, }, - "traces": [ - {key: value for key, value in trace.items() if key != "shapes"} - for trace in traces - ], + "traces": [trace_manifest_entry(trace) for trace in traces], "config_results": config_results, "rankings": rankings, } diff --git a/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py b/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py index 3f9813f..20f4e32 100644 --- a/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py +++ b/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py @@ -143,6 +143,7 @@ class FidelityEnvelopeTest(unittest.TestCase): parsed = module.parse_trace(f"u0p01={trace}") self.assertEqual(parsed["requests"], 2) self.assertEqual(parsed["shapes"], [(512, 1), (2048, 2)]) + json.dumps(module.trace_manifest_entry(parsed)) uniform = module.parse_trace( f"r1={trace}", rate_contract="uniform-spacing" )