1854 lines
54 KiB
JSON
1854 lines
54 KiB
JSON
{
|
|
"actual_prefill_vs_decode_pearson": {
|
|
"max": 0.9436904928897797,
|
|
"median": 0.6667880423793499,
|
|
"min": 0.48720283009127985
|
|
},
|
|
"frontier_contract": {
|
|
"allocation": "per-layer fixed Uniform(0.1, 1.0), normalized once and reused for every batch and phase",
|
|
"expert_count": 128,
|
|
"layer_count": 48,
|
|
"mode": "simulation",
|
|
"seed": 42
|
|
},
|
|
"frontier_prefill_vs_decode_pearson": 1.0,
|
|
"phase_summary": {
|
|
"decode": {
|
|
"actual": {
|
|
"expert_utilization": {
|
|
"max": 1.0,
|
|
"median": 0.96875,
|
|
"min": 0.921875
|
|
},
|
|
"load_cv": {
|
|
"max": 1.5122257286912568,
|
|
"median": 1.106820300518243,
|
|
"min": 0.7532423759874015
|
|
},
|
|
"load_gini": {
|
|
"max": 0.647222736853249,
|
|
"median": 0.5427848362430705,
|
|
"min": 0.3623759672389899
|
|
},
|
|
"max_load_ratio": {
|
|
"max": 8.630736064059132,
|
|
"median": 5.642131198028951,
|
|
"min": 3.392670157068063
|
|
}
|
|
},
|
|
"actual_vs_frontier_pearson": {
|
|
"max": 0.17269736915133127,
|
|
"median": -0.024141635934415796,
|
|
"min": -0.20201754939516248
|
|
},
|
|
"frontier_simulation": {
|
|
"expert_utilization": {
|
|
"max": 1.0,
|
|
"median": 1.0,
|
|
"min": 1.0
|
|
},
|
|
"load_cv": {
|
|
"max": 0.5319387026702922,
|
|
"median": 0.4852775824549782,
|
|
"min": 0.4143052194562559
|
|
},
|
|
"load_gini": {
|
|
"max": 0.3057696286957192,
|
|
"median": 0.2796991393497844,
|
|
"min": 0.2373612300777641
|
|
},
|
|
"max_load_ratio": {
|
|
"max": 2.0104712041884816,
|
|
"median": 1.8392362180474282,
|
|
"min": 1.6482907299045273
|
|
}
|
|
},
|
|
"token_count": 6494
|
|
},
|
|
"prefill": {
|
|
"actual": {
|
|
"expert_utilization": {
|
|
"max": 1.0,
|
|
"median": 0.984375,
|
|
"min": 0.9453125
|
|
},
|
|
"load_cv": {
|
|
"max": 1.636444303545179,
|
|
"median": 1.0764214888570005,
|
|
"min": 0.802893981590087
|
|
},
|
|
"load_gini": {
|
|
"max": 0.6967902002938451,
|
|
"median": 0.5262265677660404,
|
|
"min": 0.42613218222079463
|
|
},
|
|
"max_load_ratio": {
|
|
"max": 10.381092808695286,
|
|
"median": 6.121413826515567,
|
|
"min": 3.8360365673842525
|
|
}
|
|
},
|
|
"actual_vs_frontier_pearson": {
|
|
"max": 0.1797880356509229,
|
|
"median": -0.01729201163458405,
|
|
"min": -0.20008636491506096
|
|
},
|
|
"frontier_simulation": {
|
|
"expert_utilization": {
|
|
"max": 1.0,
|
|
"median": 1.0,
|
|
"min": 1.0
|
|
},
|
|
"load_cv": {
|
|
"max": 0.5319950183538887,
|
|
"median": 0.48529878894960354,
|
|
"min": 0.41421985083210333
|
|
},
|
|
"load_gini": {
|
|
"max": 0.3057985522338542,
|
|
"median": 0.2797113699735645,
|
|
"min": 0.23730966555377675
|
|
},
|
|
"max_load_ratio": {
|
|
"max": 2.0093524876774658,
|
|
"median": 1.838817036693769,
|
|
"min": 1.649408097063656
|
|
}
|
|
},
|
|
"token_count": 23737
|
|
}
|
|
},
|
|
"rows": [
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 1.0209765758908849,
|
|
"load_gini": 0.46969517196254795,
|
|
"max_load_ratio": 5.4827484517841345
|
|
},
|
|
"actual_vs_frontier_pearson": -0.09214641606187762,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5013611454809463,
|
|
"load_gini": 0.28879565986118716,
|
|
"max_load_ratio": 1.8509499936807516
|
|
},
|
|
"layer": 0,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.9656310061810012,
|
|
"load_gini": 0.4405106117295783,
|
|
"max_load_ratio": 6.154105405063825
|
|
},
|
|
"actual_vs_frontier_pearson": -0.14455565868271067,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4633214169333404,
|
|
"load_gini": 0.2672016031143363,
|
|
"max_load_ratio": 1.7579306567805535
|
|
},
|
|
"layer": 1,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 1.4100541859296152,
|
|
"load_gini": 0.5967419337426803,
|
|
"max_load_ratio": 8.47284829590934
|
|
},
|
|
"actual_vs_frontier_pearson": 0.10746117235272135,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4915202242607809,
|
|
"load_gini": 0.28343845117853994,
|
|
"max_load_ratio": 1.8556683658423558
|
|
},
|
|
"layer": 2,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 1.1750115312430194,
|
|
"load_gini": 0.49178514160382525,
|
|
"max_load_ratio": 7.965286261953912
|
|
},
|
|
"actual_vs_frontier_pearson": 0.07854011611932744,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5153981688841472,
|
|
"load_gini": 0.29577933447781946,
|
|
"max_load_ratio": 1.899481821628681
|
|
},
|
|
"layer": 3,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 1.2753111721870192,
|
|
"load_gini": 0.5858468253675697,
|
|
"max_load_ratio": 7.488056620465939
|
|
},
|
|
"actual_vs_frontier_pearson": 0.047570378008613835,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5319950183538887,
|
|
"load_gini": 0.3057985522338542,
|
|
"max_load_ratio": 2.0093524876774658
|
|
},
|
|
"layer": 4,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.2689973799039043,
|
|
"load_gini": 0.5545393103330243,
|
|
"max_load_ratio": 9.569532796899356
|
|
},
|
|
"actual_vs_frontier_pearson": -0.06358262657671977,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.46397269621367815,
|
|
"load_gini": 0.2675475161404137,
|
|
"max_load_ratio": 1.7828706239204617
|
|
},
|
|
"layer": 5,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.3617880106985483,
|
|
"load_gini": 0.5929317066973501,
|
|
"max_load_ratio": 10.381092808695286
|
|
},
|
|
"actual_vs_frontier_pearson": -0.01786861982721008,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5065069021962776,
|
|
"load_gini": 0.29121334835910173,
|
|
"max_load_ratio": 1.9075704596200025
|
|
},
|
|
"layer": 6,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.4508622399317908,
|
|
"load_gini": 0.633230738799132,
|
|
"max_load_ratio": 8.563845473311707
|
|
},
|
|
"actual_vs_frontier_pearson": -0.018533706086765106,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4174667228740326,
|
|
"load_gini": 0.23938283981758435,
|
|
"max_load_ratio": 1.649408097063656
|
|
},
|
|
"layer": 7,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0194348783142033,
|
|
"load_gini": 0.5101870727924758,
|
|
"max_load_ratio": 4.746008341407928
|
|
},
|
|
"actual_vs_frontier_pearson": -0.07019410060413582,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.45521406973427264,
|
|
"load_gini": 0.2624579704153853,
|
|
"max_load_ratio": 1.7660192947718751
|
|
},
|
|
"layer": 8,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0499203578258598,
|
|
"load_gini": 0.5348460473101067,
|
|
"max_load_ratio": 4.87475249610313
|
|
},
|
|
"actual_vs_frontier_pearson": 0.055984358629187964,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4981686541927147,
|
|
"load_gini": 0.28707712050280154,
|
|
"max_load_ratio": 1.935880692589628
|
|
},
|
|
"layer": 9,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.9182859260805778,
|
|
"load_gini": 0.4667757285566836,
|
|
"max_load_ratio": 6.065130387159287
|
|
},
|
|
"actual_vs_frontier_pearson": 0.04566572558971987,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4410838523022549,
|
|
"load_gini": 0.2539255039600623,
|
|
"max_load_ratio": 1.7262501579812108
|
|
},
|
|
"layer": 10,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.8938335625889317,
|
|
"load_gini": 0.46192282552028474,
|
|
"max_load_ratio": 4.6934321944643385
|
|
},
|
|
"actual_vs_frontier_pearson": 0.05102058572481217,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4344763177468921,
|
|
"load_gini": 0.250220186312508,
|
|
"max_load_ratio": 1.7592787631124405
|
|
},
|
|
"layer": 11,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.9842127914363633,
|
|
"load_gini": 0.48561951344420096,
|
|
"max_load_ratio": 6.4446223195854575
|
|
},
|
|
"actual_vs_frontier_pearson": -0.117474412760527,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4881943608268652,
|
|
"load_gini": 0.28113027986371497,
|
|
"max_load_ratio": 1.86038673800396
|
|
},
|
|
"layer": 12,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 1.073206640427139,
|
|
"load_gini": 0.5187799341481654,
|
|
"max_load_ratio": 6.354973248514977
|
|
},
|
|
"actual_vs_frontier_pearson": -0.020599525349797518,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4551370538191022,
|
|
"load_gini": 0.2622560506803724,
|
|
"max_load_ratio": 1.7660192947718751
|
|
},
|
|
"layer": 13,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0501051337772571,
|
|
"load_gini": 0.5237924640329443,
|
|
"max_load_ratio": 4.951594557020685
|
|
},
|
|
"actual_vs_frontier_pearson": -0.1404264118915223,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4843196647309149,
|
|
"load_gini": 0.2793572699003666,
|
|
"max_load_ratio": 1.8071365378944264
|
|
},
|
|
"layer": 14,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 1.1589582379096603,
|
|
"load_gini": 0.5428402257024898,
|
|
"max_load_ratio": 6.021990984538905
|
|
},
|
|
"actual_vs_frontier_pearson": 0.010817413166755824,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4523916462775862,
|
|
"load_gini": 0.2604694642067238,
|
|
"max_load_ratio": 1.8084846442263134
|
|
},
|
|
"layer": 15,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.1496287137851553,
|
|
"load_gini": 0.5617946797194253,
|
|
"max_load_ratio": 6.50933142351603
|
|
},
|
|
"actual_vs_frontier_pearson": 0.1797880356509229,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48641030098466176,
|
|
"load_gini": 0.2805311032091249,
|
|
"max_load_ratio": 1.8334246113662216
|
|
},
|
|
"layer": 16,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.2394125145202786,
|
|
"load_gini": 0.5729841103814719,
|
|
"max_load_ratio": 7.101150103214391
|
|
},
|
|
"actual_vs_frontier_pearson": -0.20008636491506096,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.443419246367453,
|
|
"load_gini": 0.25561211794877203,
|
|
"max_load_ratio": 1.7754560390950835
|
|
},
|
|
"layer": 17,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.2168439400279965,
|
|
"load_gini": 0.575970613520032,
|
|
"max_load_ratio": 8.145258457260816
|
|
},
|
|
"actual_vs_frontier_pearson": -0.06530349438725558,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4516257974939683,
|
|
"load_gini": 0.25957785774845177,
|
|
"max_load_ratio": 1.8010700594009352
|
|
},
|
|
"layer": 18,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.2427942152510565,
|
|
"load_gini": 0.5980218283955427,
|
|
"max_load_ratio": 6.7688418924042635
|
|
},
|
|
"actual_vs_frontier_pearson": -0.04180797903095555,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4681222179773095,
|
|
"load_gini": 0.26999935161878086,
|
|
"max_load_ratio": 1.8111808568900871
|
|
},
|
|
"layer": 19,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 0.9706255724368628,
|
|
"load_gini": 0.5072024621371698,
|
|
"max_load_ratio": 4.138012385726924
|
|
},
|
|
"actual_vs_frontier_pearson": 0.007542212699855623,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4876159778421919,
|
|
"load_gini": 0.2809273727461348,
|
|
"max_load_ratio": 1.7734338795972533
|
|
},
|
|
"layer": 20,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.9454143633213303,
|
|
"load_gini": 0.5064155182310317,
|
|
"max_load_ratio": 3.8980494586510512
|
|
},
|
|
"actual_vs_frontier_pearson": -0.005313631721923769,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4407612899555989,
|
|
"load_gini": 0.25342531240784427,
|
|
"max_load_ratio": 1.7006361376753591
|
|
},
|
|
"layer": 21,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.8037705358927761,
|
|
"load_gini": 0.42613218222079463,
|
|
"max_load_ratio": 5.447697687155074
|
|
},
|
|
"actual_vs_frontier_pearson": 0.006893096788933473,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5096984088007344,
|
|
"load_gini": 0.292796698587648,
|
|
"max_load_ratio": 1.9850865737035008
|
|
},
|
|
"layer": 22,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.802893981590087,
|
|
"load_gini": 0.43984914769136796,
|
|
"max_load_ratio": 3.953321818258415
|
|
},
|
|
"actual_vs_frontier_pearson": 0.055518113087740455,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5020829253028674,
|
|
"load_gini": 0.28819319193137294,
|
|
"max_load_ratio": 1.9695833508868013
|
|
},
|
|
"layer": 23,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.8487498379405487,
|
|
"load_gini": 0.44496173234296665,
|
|
"max_load_ratio": 4.682647343809243
|
|
},
|
|
"actual_vs_frontier_pearson": -0.11069406175185324,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4937827979946325,
|
|
"load_gini": 0.28478219657075443,
|
|
"max_load_ratio": 1.8812823861482075
|
|
},
|
|
"layer": 24,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.8842495554569121,
|
|
"load_gini": 0.4609973189272023,
|
|
"max_load_ratio": 5.068205754728904
|
|
},
|
|
"actual_vs_frontier_pearson": 0.04041373985743766,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5094957270929035,
|
|
"load_gini": 0.29290695630766317,
|
|
"max_load_ratio": 1.9264439482664195
|
|
},
|
|
"layer": 25,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.910991140491345,
|
|
"load_gini": 0.47597056415090355,
|
|
"max_load_ratio": 4.189240426338627
|
|
},
|
|
"actual_vs_frontier_pearson": -0.001853343235349625,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5141430671669248,
|
|
"load_gini": 0.2956995210536293,
|
|
"max_load_ratio": 1.870497535493112
|
|
},
|
|
"layer": 26,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.0416423824930179,
|
|
"load_gini": 0.5048694416944011,
|
|
"max_load_ratio": 5.7442810801701985
|
|
},
|
|
"actual_vs_frontier_pearson": 0.021178350576558223,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4563109042451374,
|
|
"load_gini": 0.2625302961884399,
|
|
"max_load_ratio": 1.8846526519779248
|
|
},
|
|
"layer": 27,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.953125,
|
|
"load_cv": 1.0513699281689028,
|
|
"load_gini": 0.5286606714991364,
|
|
"max_load_ratio": 5.484096558116021
|
|
},
|
|
"actual_vs_frontier_pearson": -0.195005081355267,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4931389294946141,
|
|
"load_gini": 0.28378707950667725,
|
|
"max_load_ratio": 1.7983738467371615
|
|
},
|
|
"layer": 28,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.953125,
|
|
"load_cv": 1.079636337286862,
|
|
"load_gini": 0.5304158262943928,
|
|
"max_load_ratio": 6.738509499936807
|
|
},
|
|
"actual_vs_frontier_pearson": -0.016715403441958022,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4500354635096657,
|
|
"load_gini": 0.2587639253854741,
|
|
"max_load_ratio": 1.7316425833087585
|
|
},
|
|
"layer": 29,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.1836591453389935,
|
|
"load_gini": 0.5595659927012682,
|
|
"max_load_ratio": 7.912710115010322
|
|
},
|
|
"actual_vs_frontier_pearson": 0.009080640196552386,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4501844468319078,
|
|
"load_gini": 0.25952453908981754,
|
|
"max_load_ratio": 1.845557568353204
|
|
},
|
|
"layer": 30,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.1837797460931923,
|
|
"load_gini": 0.5779810067089355,
|
|
"max_load_ratio": 6.173652946876185
|
|
},
|
|
"actual_vs_frontier_pearson": -0.11789359921128652,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5029280507742879,
|
|
"load_gini": 0.2893184789305725,
|
|
"max_load_ratio": 1.92846610776425
|
|
},
|
|
"layer": 31,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 0.9485322853469729,
|
|
"load_gini": 0.4766967017472723,
|
|
"max_load_ratio": 5.981547794582298
|
|
},
|
|
"actual_vs_frontier_pearson": -0.11341887598082027,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4862779131682921,
|
|
"load_gini": 0.2800654700467624,
|
|
"max_load_ratio": 1.8307283987024476
|
|
},
|
|
"layer": 32,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.9147261414047101,
|
|
"load_gini": 0.48626378056936437,
|
|
"max_load_ratio": 4.4292033534145006
|
|
},
|
|
"actual_vs_frontier_pearson": -0.0734554681462204,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4826868638135814,
|
|
"load_gini": 0.278262015787589,
|
|
"max_load_ratio": 1.8347727176981083
|
|
},
|
|
"layer": 33,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.822096011945377,
|
|
"load_gini": 0.4335047210051819,
|
|
"max_load_ratio": 5.387032902220162
|
|
},
|
|
"actual_vs_frontier_pearson": 0.041109767501958036,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5205338188216663,
|
|
"load_gini": 0.2990128478219658,
|
|
"max_load_ratio": 1.84286135568943
|
|
},
|
|
"layer": 34,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.8350377321986099,
|
|
"load_gini": 0.4458607441704934,
|
|
"max_load_ratio": 3.8360365673842525
|
|
},
|
|
"actual_vs_frontier_pearson": -0.002498639457316309,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4931172415832444,
|
|
"load_gini": 0.28332819345852456,
|
|
"max_load_ratio": 1.9008299279605678
|
|
},
|
|
"layer": 35,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.936559272961729,
|
|
"load_gini": 0.4615142959830645,
|
|
"max_load_ratio": 5.575767788684333
|
|
},
|
|
"actual_vs_frontier_pearson": -0.04463576992397696,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.47788893428675927,
|
|
"load_gini": 0.27538099801996885,
|
|
"max_load_ratio": 1.7997219530690483
|
|
},
|
|
"layer": 36,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0424990294361898,
|
|
"load_gini": 0.5039767655716814,
|
|
"max_load_ratio": 6.859839069806631
|
|
},
|
|
"actual_vs_frontier_pearson": -0.05047148070882558,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.41421985083210333,
|
|
"load_gini": 0.23730966555377675,
|
|
"max_load_ratio": 1.6965918186796984
|
|
},
|
|
"layer": 37,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 1.081067450691553,
|
|
"load_gini": 0.515087617037747,
|
|
"max_load_ratio": 6.495850360197161
|
|
},
|
|
"actual_vs_frontier_pearson": -0.03189531872213542,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48198382691720465,
|
|
"load_gini": 0.27773105081202343,
|
|
"max_load_ratio": 1.8037662720647092
|
|
},
|
|
"layer": 38,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.140261877881891,
|
|
"load_gini": 0.5301265232021739,
|
|
"max_load_ratio": 6.088722247967309
|
|
},
|
|
"actual_vs_frontier_pearson": -0.1047856089108959,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.49823287732736005,
|
|
"load_gini": 0.28601938692758133,
|
|
"max_load_ratio": 1.9102666722837764
|
|
},
|
|
"layer": 39,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.16210598515629,
|
|
"load_gini": 0.5750575314909214,
|
|
"max_load_ratio": 5.264355226018452
|
|
},
|
|
"actual_vs_frontier_pearson": 0.07101910025349224,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5231673221525147,
|
|
"load_gini": 0.30064416838690655,
|
|
"max_load_ratio": 1.9163331507772676
|
|
},
|
|
"layer": 40,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.335821482050506,
|
|
"load_gini": 0.6172422174506045,
|
|
"max_load_ratio": 8.129755234444117
|
|
},
|
|
"actual_vs_frontier_pearson": 0.04271119117719951,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4592680576699305,
|
|
"load_gini": 0.26492856616253113,
|
|
"max_load_ratio": 1.8145511227198046
|
|
},
|
|
"layer": 41,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.5340944192835082,
|
|
"load_gini": 0.6802103190693853,
|
|
"max_load_ratio": 9.211610565783376
|
|
},
|
|
"actual_vs_frontier_pearson": 0.012121305370270835,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.493325400417425,
|
|
"load_gini": 0.28234122230062764,
|
|
"max_load_ratio": 1.9055483001221722
|
|
},
|
|
"layer": 42,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.5076855954950434,
|
|
"load_gini": 0.6597892728124868,
|
|
"max_load_ratio": 7.8324977882630495
|
|
},
|
|
"actual_vs_frontier_pearson": 0.1787454639272607,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4651216460072469,
|
|
"load_gini": 0.267678755739984,
|
|
"max_load_ratio": 1.870497535493112
|
|
},
|
|
"layer": 43,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 1.4676145461991041,
|
|
"load_gini": 0.666736825683532,
|
|
"max_load_ratio": 6.086700088469478
|
|
},
|
|
"actual_vs_frontier_pearson": -0.1318454052741815,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.47511235321400247,
|
|
"load_gini": 0.2737368414816532,
|
|
"max_load_ratio": 1.84286135568943
|
|
},
|
|
"layer": 44,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.5817687300685335,
|
|
"load_gini": 0.6967902002938451,
|
|
"max_load_ratio": 6.617853983232927
|
|
},
|
|
"actual_vs_frontier_pearson": -0.042931525341602716,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.47921880192242294,
|
|
"load_gini": 0.2763617157549396,
|
|
"max_load_ratio": 1.7754560390950835
|
|
},
|
|
"layer": 45,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.5934457162349587,
|
|
"load_gini": 0.6958883908823776,
|
|
"max_load_ratio": 7.270337447866201
|
|
},
|
|
"actual_vs_frontier_pearson": 0.02211603536084719,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5297987184237168,
|
|
"load_gini": 0.30463451046888834,
|
|
"max_load_ratio": 1.8799342798163206
|
|
},
|
|
"layer": 46,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.636444303545179,
|
|
"load_gini": 0.6918075387317268,
|
|
"max_load_ratio": 8.385895437502633
|
|
},
|
|
"actual_vs_frontier_pearson": -0.01870726023327083,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4952385440074709,
|
|
"load_gini": 0.28391033776382857,
|
|
"max_load_ratio": 1.9170072039432111
|
|
},
|
|
"layer": 47,
|
|
"phase": "prefill",
|
|
"total_routed_tokens": 189896
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.7564860612847418,
|
|
"load_gini": 0.3623759672389899,
|
|
"max_load_ratio": 4.043116723129042
|
|
},
|
|
"actual_vs_frontier_pearson": -0.11076278775511662,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5013193007157182,
|
|
"load_gini": 0.288770162842624,
|
|
"max_load_ratio": 1.8503233754234678
|
|
},
|
|
"layer": 0,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.7532423759874015,
|
|
"load_gini": 0.36718870303356943,
|
|
"max_load_ratio": 4.479211579919926
|
|
},
|
|
"actual_vs_frontier_pearson": -0.20201754939516248,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4633429809491391,
|
|
"load_gini": 0.2672133050700647,
|
|
"max_load_ratio": 1.7566984909146905
|
|
},
|
|
"layer": 1,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.9982932903554329,
|
|
"load_gini": 0.48608902207807203,
|
|
"max_load_ratio": 5.430243301509085
|
|
},
|
|
"actual_vs_frontier_pearson": 0.12507453392144424,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4914341503043837,
|
|
"load_gini": 0.2833871891361257,
|
|
"max_load_ratio": 1.8552510009239298
|
|
},
|
|
"layer": 2,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.887995113903465,
|
|
"load_gini": 0.4407307466507546,
|
|
"max_load_ratio": 4.843855866954112
|
|
},
|
|
"actual_vs_frontier_pearson": 0.09916207755901395,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5153762623410879,
|
|
"load_gini": 0.2957582840891593,
|
|
"max_load_ratio": 1.8995996304280875
|
|
},
|
|
"layer": 3,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0374295217881528,
|
|
"load_gini": 0.5126796730635972,
|
|
"max_load_ratio": 5.198644902987373
|
|
},
|
|
"actual_vs_frontier_pearson": 0.1087630450097359,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5319387026702922,
|
|
"load_gini": 0.3057696286957192,
|
|
"max_load_ratio": 2.0104712041884816
|
|
},
|
|
"layer": 4,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.9864025871023987,
|
|
"load_gini": 0.5000595501616878,
|
|
"max_load_ratio": 6.716353557129658
|
|
},
|
|
"actual_vs_frontier_pearson": -0.060788024591760324,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.46389740539367647,
|
|
"load_gini": 0.26750173236833996,
|
|
"max_load_ratio": 1.7838004311672313
|
|
},
|
|
"layer": 5,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.2953358972433757,
|
|
"load_gini": 0.629114675565907,
|
|
"max_load_ratio": 5.74314752078842
|
|
},
|
|
"actual_vs_frontier_pearson": -0.07165433962410654,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5065225414304516,
|
|
"load_gini": 0.2912228475323375,
|
|
"max_load_ratio": 1.9069910686787803
|
|
},
|
|
"layer": 6,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.2699902455381382,
|
|
"load_gini": 0.6262021312942716,
|
|
"max_load_ratio": 6.66954111487527
|
|
},
|
|
"actual_vs_frontier_pearson": 0.01205569571775735,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.41747315017256753,
|
|
"load_gini": 0.2393889431590699,
|
|
"max_load_ratio": 1.6482907299045273
|
|
},
|
|
"layer": 7,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.0541527297772746,
|
|
"load_gini": 0.5379860857137357,
|
|
"max_load_ratio": 5.171542962734832
|
|
},
|
|
"actual_vs_frontier_pearson": -0.04958458083380716,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4553076424255581,
|
|
"load_gini": 0.2625118498806591,
|
|
"max_load_ratio": 1.7665537419156143
|
|
},
|
|
"layer": 8,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.192957622384299,
|
|
"load_gini": 0.5506200435016939,
|
|
"max_load_ratio": 8.320295657530028
|
|
},
|
|
"actual_vs_frontier_pearson": -0.05509740986705497,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4981311718832551,
|
|
"load_gini": 0.28705553924776717,
|
|
"max_load_ratio": 1.9365568216815523
|
|
},
|
|
"layer": 9,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.8076732412246237,
|
|
"load_gini": 0.4410110534724361,
|
|
"max_load_ratio": 3.392670157068063
|
|
},
|
|
"actual_vs_frontier_pearson": -0.09000721184282558,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4410207617483597,
|
|
"load_gini": 0.2538882044964583,
|
|
"max_load_ratio": 1.7271327379119188
|
|
},
|
|
"layer": 10,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.946297012928892,
|
|
"load_gini": 0.4790344332268248,
|
|
"max_load_ratio": 4.728056667693255
|
|
},
|
|
"actual_vs_frontier_pearson": 0.034097483022671266,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4344893792424141,
|
|
"load_gini": 0.25022677182784103,
|
|
"max_load_ratio": 1.7591623036649215
|
|
},
|
|
"layer": 11,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.9479122820728094,
|
|
"load_gini": 0.48800966517169697,
|
|
"max_load_ratio": 5.097628580227902
|
|
},
|
|
"actual_vs_frontier_pearson": -0.02153501737741947,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48821731504828314,
|
|
"load_gini": 0.28114623835463504,
|
|
"max_load_ratio": 1.8601786264243918
|
|
},
|
|
"layer": 12,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.9047451307311554,
|
|
"load_gini": 0.4740120688327687,
|
|
"max_load_ratio": 4.555589775177086
|
|
},
|
|
"actual_vs_frontier_pearson": -0.05894159287248561,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4550660315489719,
|
|
"load_gini": 0.26222101651524476,
|
|
"max_load_ratio": 1.7665537419156143
|
|
},
|
|
"layer": 13,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.1102662993215426,
|
|
"load_gini": 0.550685007314444,
|
|
"max_load_ratio": 5.5214043732676314
|
|
},
|
|
"actual_vs_frontier_pearson": -0.07473125180903177,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4843214689622731,
|
|
"load_gini": 0.2793576281952572,
|
|
"max_load_ratio": 1.80597474591931
|
|
},
|
|
"layer": 14,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.0424321211996845,
|
|
"load_gini": 0.5389009927433015,
|
|
"max_load_ratio": 4.885740683708038
|
|
},
|
|
"actual_vs_frontier_pearson": -0.122334501258848,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.45239738453493195,
|
|
"load_gini": 0.26047060267169697,
|
|
"max_load_ratio": 1.808438558669541
|
|
},
|
|
"layer": 15,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.1000158272634826,
|
|
"load_gini": 0.5458891139898367,
|
|
"max_load_ratio": 5.321835540498922
|
|
},
|
|
"actual_vs_frontier_pearson": 0.06285259077675373,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48638660674407463,
|
|
"load_gini": 0.2805170517978133,
|
|
"max_load_ratio": 1.833076686171851
|
|
},
|
|
"layer": 16,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.1000359607690584,
|
|
"load_gini": 0.5519229288574068,
|
|
"max_load_ratio": 5.201108715737604
|
|
},
|
|
"actual_vs_frontier_pearson": -0.14790969026732273,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.44351655698688397,
|
|
"load_gini": 0.25566538583692644,
|
|
"max_load_ratio": 1.7764089929165383
|
|
},
|
|
"layer": 17,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.921875,
|
|
"load_cv": 1.275279013013624,
|
|
"load_gini": 0.6363662611641516,
|
|
"max_load_ratio": 5.637203572528488
|
|
},
|
|
"actual_vs_frontier_pearson": -0.017099613299867974,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4514828672543027,
|
|
"load_gini": 0.259494942446874,
|
|
"max_load_ratio": 1.801047120418848
|
|
},
|
|
"layer": 18,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.921875,
|
|
"load_cv": 1.2773056983164457,
|
|
"load_gini": 0.6308308029527256,
|
|
"max_load_ratio": 5.979673544810594
|
|
},
|
|
"actual_vs_frontier_pearson": -0.08479638082455722,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4680898752103892,
|
|
"load_gini": 0.2699814853133662,
|
|
"max_load_ratio": 1.810902371419772
|
|
},
|
|
"layer": 19,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.142557882037356,
|
|
"load_gini": 0.573428958461657,
|
|
"max_load_ratio": 5.856482907299045
|
|
},
|
|
"actual_vs_frontier_pearson": -0.16520423265299047,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48760009420143696,
|
|
"load_gini": 0.2809182634932246,
|
|
"max_load_ratio": 1.7739451801663073
|
|
},
|
|
"layer": 20,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.2250892374067721,
|
|
"load_gini": 0.5839970068524791,
|
|
"max_load_ratio": 6.6547582383738835
|
|
},
|
|
"actual_vs_frontier_pearson": 0.030713168827124045,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.44079057860501963,
|
|
"load_gini": 0.25344789421004,
|
|
"max_load_ratio": 1.7000307976593778
|
|
},
|
|
"layer": 21,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.9505437008130998,
|
|
"load_gini": 0.49268164603865117,
|
|
"max_load_ratio": 5.647058823529412
|
|
},
|
|
"actual_vs_frontier_pearson": -0.06756357177648568,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5096150033498628,
|
|
"load_gini": 0.2927473918232215,
|
|
"max_load_ratio": 1.9858330766861718
|
|
},
|
|
"layer": 22,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.0080029689110093,
|
|
"load_gini": 0.521078351170311,
|
|
"max_load_ratio": 5.109947643979058
|
|
},
|
|
"actual_vs_frontier_pearson": 0.051969177479756205,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.502148823760943,
|
|
"load_gini": 0.288229700011549,
|
|
"max_load_ratio": 1.968586387434555
|
|
},
|
|
"layer": 23,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.0336730047933789,
|
|
"load_gini": 0.5219367156221126,
|
|
"max_load_ratio": 6.2309824453341545
|
|
},
|
|
"actual_vs_frontier_pearson": 0.046624159557719196,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4937452666192137,
|
|
"load_gini": 0.2847580458885124,
|
|
"max_load_ratio": 1.8798891284262396
|
|
},
|
|
"layer": 24,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.9370799623414752,
|
|
"load_gini": 0.49417972359100704,
|
|
"max_load_ratio": 4.348629504157684
|
|
},
|
|
"actual_vs_frontier_pearson": 0.003630591230348502,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5094100439322153,
|
|
"load_gini": 0.29286198077071135,
|
|
"max_load_ratio": 1.9267015706806283
|
|
},
|
|
"layer": 25,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.1307693127116234,
|
|
"load_gini": 0.562033222975054,
|
|
"max_load_ratio": 5.900831536803203
|
|
},
|
|
"actual_vs_frontier_pearson": 0.12889829096423783,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5140701353957097,
|
|
"load_gini": 0.2956587330612874,
|
|
"max_load_ratio": 1.8700338774253156
|
|
},
|
|
"layer": 26,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.1066234149756957,
|
|
"load_gini": 0.5586361569333231,
|
|
"max_load_ratio": 5.615029257776409
|
|
},
|
|
"actual_vs_frontier_pearson": 0.027435906393056993,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4562891247455973,
|
|
"load_gini": 0.26251997035725294,
|
|
"max_load_ratio": 1.8848167539267016
|
|
},
|
|
"layer": 27,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.15194374689878,
|
|
"load_gini": 0.5561218167731752,
|
|
"max_load_ratio": 6.691715429627348
|
|
},
|
|
"actual_vs_frontier_pearson": -0.0851905104644269,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4931885285350055,
|
|
"load_gini": 0.28381426605327986,
|
|
"max_load_ratio": 1.7985833076686173
|
|
},
|
|
"layer": 28,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9296875,
|
|
"load_cv": 1.1356431750011955,
|
|
"load_gini": 0.5496513007198953,
|
|
"max_load_ratio": 5.863874345549738
|
|
},
|
|
"actual_vs_frontier_pearson": -0.016979392910092525,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.45012219999295283,
|
|
"load_gini": 0.25880680724514926,
|
|
"max_load_ratio": 1.7320603634123806
|
|
},
|
|
"layer": 29,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9296875,
|
|
"load_cv": 1.3264510778008716,
|
|
"load_gini": 0.647222736853249,
|
|
"max_load_ratio": 5.84416384354789
|
|
},
|
|
"actual_vs_frontier_pearson": -0.08251025635608537,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.45023597421285283,
|
|
"load_gini": 0.25955389109177696,
|
|
"max_load_ratio": 1.8453957499230058
|
|
},
|
|
"layer": 30,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.921875,
|
|
"load_cv": 1.2804663472936682,
|
|
"load_gini": 0.6305014725130891,
|
|
"max_load_ratio": 5.723437018786572
|
|
},
|
|
"actual_vs_frontier_pearson": -0.07964706640411469,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5029240802777508,
|
|
"load_gini": 0.2893154378079765,
|
|
"max_load_ratio": 1.9291653834308593
|
|
},
|
|
"layer": 31,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.1507881607782275,
|
|
"load_gini": 0.5729281957383738,
|
|
"max_load_ratio": 6.021558361564521
|
|
},
|
|
"actual_vs_frontier_pearson": -0.10452258144829583,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48623369594768334,
|
|
"load_gini": 0.2800406505043116,
|
|
"max_load_ratio": 1.83061287342162
|
|
},
|
|
"layer": 32,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.2686811219716572,
|
|
"load_gini": 0.583570832210502,
|
|
"max_load_ratio": 7.844779796735448
|
|
},
|
|
"actual_vs_frontier_pearson": -0.01888683352026748,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48261072157815704,
|
|
"load_gini": 0.2782189569217739,
|
|
"max_load_ratio": 1.8355404989220818
|
|
},
|
|
"layer": 33,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9921875,
|
|
"load_cv": 0.8839459404636472,
|
|
"load_gini": 0.4794891799160763,
|
|
"max_load_ratio": 3.9618109023714196
|
|
},
|
|
"actual_vs_frontier_pearson": -0.028493286082986517,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5205697308164966,
|
|
"load_gini": 0.2990305353980598,
|
|
"max_load_ratio": 1.8429319371727748
|
|
},
|
|
"layer": 34,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 0.9362312679691519,
|
|
"load_gini": 0.4916611728133662,
|
|
"max_load_ratio": 4.747767169695103
|
|
},
|
|
"actual_vs_frontier_pearson": -0.13290690148552678,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.49297481569205803,
|
|
"load_gini": 0.2832476372420696,
|
|
"max_load_ratio": 1.8995996304280875
|
|
},
|
|
"layer": 35,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 0.9574253286736439,
|
|
"load_gini": 0.49605465140899296,
|
|
"max_load_ratio": 4.767477671696951
|
|
},
|
|
"actual_vs_frontier_pearson": -0.02212359104446765,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.47796935463002527,
|
|
"load_gini": 0.27542581373190633,
|
|
"max_load_ratio": 1.7985833076686173
|
|
},
|
|
"layer": 36,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 0.897443530050085,
|
|
"load_gini": 0.46608167154296276,
|
|
"max_load_ratio": 4.484139205420388
|
|
},
|
|
"actual_vs_frontier_pearson": -0.019587227325278997,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4143052194562559,
|
|
"load_gini": 0.2373612300777641,
|
|
"max_load_ratio": 1.6975669849091468
|
|
},
|
|
"layer": 37,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.0030386322057563,
|
|
"load_gini": 0.5019200415768401,
|
|
"max_load_ratio": 5.198644902987373
|
|
},
|
|
"actual_vs_frontier_pearson": 0.028921565758140203,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.48203207236683104,
|
|
"load_gini": 0.27775819506467503,
|
|
"max_load_ratio": 1.803510933169079
|
|
},
|
|
"layer": 38,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9765625,
|
|
"load_cv": 1.1070171860607907,
|
|
"load_gini": 0.5396805584963043,
|
|
"max_load_ratio": 5.164151524484139
|
|
},
|
|
"actual_vs_frontier_pearson": -0.042234189776402496,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.49820199968589846,
|
|
"load_gini": 0.2860022833577147,
|
|
"max_load_ratio": 1.9094548814290113
|
|
},
|
|
"layer": 39,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.1143666092181908,
|
|
"load_gini": 0.5215875351285801,
|
|
"max_load_ratio": 6.679396365876193
|
|
},
|
|
"actual_vs_frontier_pearson": -0.0013066971352570613,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5231859746946843,
|
|
"load_gini": 0.3006501193409301,
|
|
"max_load_ratio": 1.9168463196797043
|
|
},
|
|
"layer": 40,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9375,
|
|
"load_cv": 1.3163699264356221,
|
|
"load_gini": 0.576891890591315,
|
|
"max_load_ratio": 6.450261780104712
|
|
},
|
|
"actual_vs_frontier_pearson": 0.038349159249901725,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.45930246027155186,
|
|
"load_gini": 0.2649497974091468,
|
|
"max_load_ratio": 1.815829996920234
|
|
},
|
|
"layer": 41,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9609375,
|
|
"load_cv": 1.4367616763803446,
|
|
"load_gini": 0.6064525307014168,
|
|
"max_load_ratio": 8.630736064059132
|
|
},
|
|
"actual_vs_frontier_pearson": -0.005385768830336229,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.49323689442136087,
|
|
"load_gini": 0.2822840073529411,
|
|
"max_load_ratio": 1.9069910686787803
|
|
},
|
|
"layer": 42,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.9453125,
|
|
"load_cv": 1.5122257286912568,
|
|
"load_gini": 0.6255329438712658,
|
|
"max_load_ratio": 8.19956883276871
|
|
},
|
|
"actual_vs_frontier_pearson": 0.17269736915133127,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4650946236339636,
|
|
"load_gini": 0.2676644426586079,
|
|
"max_load_ratio": 1.8700338774253156
|
|
},
|
|
"layer": 43,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.4453751038205662,
|
|
"load_gini": 0.6368104813096704,
|
|
"max_load_ratio": 6.94548814290114
|
|
},
|
|
"actual_vs_frontier_pearson": -0.18470982231222025,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4751830338518044,
|
|
"load_gini": 0.27377735698336925,
|
|
"max_load_ratio": 1.8429319371727748
|
|
},
|
|
"layer": 44,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.96875,
|
|
"load_cv": 1.464195386895173,
|
|
"load_gini": 0.6367434121881737,
|
|
"max_load_ratio": 7.145056975669849
|
|
},
|
|
"actual_vs_frontier_pearson": -0.026159680824363943,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4792291952239595,
|
|
"load_gini": 0.2763707966007083,
|
|
"max_load_ratio": 1.7764089929165383
|
|
},
|
|
"layer": 45,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.3411915327005308,
|
|
"load_gini": 0.5801415850207885,
|
|
"max_load_ratio": 6.0437326763165995
|
|
},
|
|
"actual_vs_frontier_pearson": 0.02225870145809893,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.5297396161604814,
|
|
"load_gini": 0.3045975732406838,
|
|
"max_load_ratio": 1.8798891284262396
|
|
},
|
|
"layer": 46,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
},
|
|
{
|
|
"actual": {
|
|
"expert_utilization": 0.984375,
|
|
"load_cv": 1.351786104177472,
|
|
"load_gini": 0.536063938828149,
|
|
"max_load_ratio": 6.908530951647675
|
|
},
|
|
"actual_vs_frontier_pearson": -0.04368832283746608,
|
|
"frontier_simulation": {
|
|
"expert_utilization": 1.0,
|
|
"load_cv": 0.4952080715601842,
|
|
"load_gini": 0.2838945685440406,
|
|
"max_load_ratio": 1.9168463196797043
|
|
},
|
|
"layer": 47,
|
|
"phase": "decode",
|
|
"total_routed_tokens": 51952
|
|
}
|
|
],
|
|
"schema": "frontier-routing-mismatch.v1",
|
|
"source": "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-trace-routing-20260716-v1-20260716T151552356215Z/artifacts/artifacts/trace-routing-v1/raw/routing.json"
|
|
}
|