framework,version,device,op_name,kernel_source,moe_dtype,moe_kernel,num_tokens,dp_num_tokens,rank0_num_tokens,hidden_size,inter_size,topk,num_experts,num_slots,moe_tp_size,moe_ep_size,distribution,simulation_mode,latency
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.10181119441986083
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.15455360412597657
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.1669119954109192
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,1,1,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.15088640451431273
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,1,2,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.2559360027313232
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.17125760316848754
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,1,2,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.25596160888671876
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,1,2,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.2528831958770752
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,2,2,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.2350464105606079
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,2,4,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.38176000118255615
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,2,4,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.34835200309753417
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,2,3,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.3033152103424072
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,4,4,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.37867519855499265
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,4,6,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.5435776233673095
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,4,6,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.5614719867706299
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,4,6,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.49232001304626466
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,8,9,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.5204800128936767
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,8,8,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.6231296062469482
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,8,11,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.5515200138092041
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,8,11,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.5747392177581787
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,16,23,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.6770688056945801
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,16,19,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.6380159854888916
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,16,22,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.6341824054718017
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,16,19,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.7670015811920166
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,24,28,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.7837247848510742
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,24,30,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9257216453552246
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,24,29,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.8468864440917969
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,24,27,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.9452672004699707
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,32,41,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.8707967758178711
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,32,39,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.8565119743347168
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,32,38,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.8738368034362793
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,32,36,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.9492927551269531
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,40,54,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.989408016204834
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,40,48,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9957375526428223
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,40,47,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.8772543907165528
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,40,45,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.9867520332336426
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,48,63,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.8106495857238769
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,64,87,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.9150079727172852
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,48,60,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9603903770446778
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,48,58,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.8836607933044434
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,48,54,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.977286434173584
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,80,100,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.8955648422241211
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,64,81,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.8552063941955567
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,64,78,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9023551940917969
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,64,73,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.9872896194458007
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,96,125,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.9675071716308594
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,80,100,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.8824000358581543
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,80,98,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9625087738037109
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,80,92,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.0025535583496095
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,128,180,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.9870207786560059
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,96,137,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9727295875549317
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,96,118,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9249855995178222
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,160,221,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.9546367645263671
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,96,109,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.9946623802185058
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,128,164,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.900268840789795
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,128,165,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9537343978881836
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,192,283,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.9920512199401855
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,128,148,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.0174400329589843
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,160,217,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9886079788208008
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,160,201,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9814720153808594
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,256,378,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.05033597946167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,160,184,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.0438719749450684
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,192,248,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.9471551895141601
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,192,244,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9294400215148926
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,384,583,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.117843246459961
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,192,217,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.0339136123657227
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,256,351,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.035814380645752
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,256,337,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.9524224281311036
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,512,718,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.074073600769043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,256,297,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.071622371673584
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,384,497,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.032102394104004
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,384,524,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.104588794708252
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,768,1173,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.258732795715332
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,384,442,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.103769588470459
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,512,678,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.0503999710083007
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,512,705,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.0577280044555664
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1660,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.3347904205322265
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,512,588,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.0876031875610352
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,768,1011,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.1653632164001464
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,768,1044,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.2081472396850585
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,2414,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.5612288475036622
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,768,909,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.1569024085998536
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1421,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.2134336471557616
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1407,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.3457535743713378
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1205,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.3835200309753417
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,3492,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.90645751953125
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,2024,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.4021759986877442
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,2225,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.6050687789916993
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,1816,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.491648006439209
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,2771,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.6232128143310547
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,3160,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.8695104598999024
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,5597,7168,2048,8,256,256,1,2,power_law_1.2,accurate,2.7219776153564452
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,2416,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.6889087677001953
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,4173,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.9832895278930665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,4293,7168,2048,8,256,256,1,2,power_law_1.01,accurate,2.2570623397827148
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,7334,7168,2048,8,256,256,1,2,power_law_1.2,accurate,2.7644159317016603
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,3619,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,2.10449275970459
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,5658,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,2.374380874633789
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,6058,7168,2048,8,256,256,1,2,power_law_1.01,accurate,2.665081596374512
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,10482,7168,2048,8,256,256,1,2,power_law_1.2,accurate,3.7021247863769533
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,4789,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,2.428390312194824
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,8298,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,3.0976383209228517
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,8609,7168,2048,8,256,256,1,2,power_law_1.01,accurate,3.5312320709228517
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,15037,7168,2048,8,256,256,1,2,power_law_1.2,accurate,5.098195266723633
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,7329,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,3.1665727615356447
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,11457,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,3.7781246185302733
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,12848,7168,2048,8,256,256,1,2,power_law_1.01,accurate,4.603731155395508
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,19324,7168,2048,8,256,256,1,2,power_law_1.2,accurate,5.92149772644043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,9657,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,3.900844955444336
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,14267,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,4.589587020874023
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,14990,7168,2048,8,256,256,1,2,power_law_1.01,accurate,5.12603530883789
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,12270,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,4.62097282409668
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,30844,7168,2048,8,256,256,1,2,power_law_1.2,accurate,8.803167724609375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,23461,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,6.690470123291016
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,23880,7168,2048,8,256,256,1,2,power_law_1.01,accurate,7.504422760009765
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,19940,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,6.920185852050781
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,60589,7168,2048,8,256,256,1,2,power_law_1.2,accurate,15.865951538085938
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,48655,7168,2048,8,256,256,1,2,power_law_1.01,accurate,14.427647399902344
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,46452,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,13.203340148925781
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,38651,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,13.064813232421875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.10487680435180664
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,1,1,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.1516095995903015
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,2,2,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.23425920009613038
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,4,5,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.34593279361724855
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,8,10,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.446560001373291
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,16,22,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.5511040210723877
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,24,29,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.786195182800293
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,32,41,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.800160026550293
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.15466879606246947
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,40,49,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.8114751815795899
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,1,2,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.2572096109390259
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.14922239780426025
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,2,4,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.38161280155181887
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,48,58,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.8960639953613281
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.23600640296936035
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,4,6,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.5100031852722168
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,64,81,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.8816639900207519
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.2411263942718506
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,8,9,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.6250944137573242
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,4,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.3806976079940796
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,80,105,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9068351745605469
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,8,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.434335994720459
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,16,19,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.7142399787902832
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,96,120,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9571968078613281
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,14,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.4633279800415039
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,24,28,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.899500846862793
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,19,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5601984024047851
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,128,162,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9714816093444825
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,25,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5082111835479737
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,32,36,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.0033791542053223
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,31,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.564844799041748
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,160,220,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9598976135253906
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,40,45,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.9663104057312012
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.07061759829521179
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,40,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5069695949554444
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,48,55,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.0014335632324218
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.1100160002708435
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,192,255,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9843392372131348
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,54,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.6120255947113037
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,1,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.14966399669647218
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,64,75,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.0428223609924316
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,72,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5425087928771972
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,256,350,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.9541055679321289
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,2,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.24164481163024903
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,80,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.527020788192749
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,80,92,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.9965760231018066
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.3063551902770996
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,107,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5617919921875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,384,537,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.0062015533447266
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,96,113,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.077023983001709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,134,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.6421504020690918
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,12,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.37908480167388914
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,128,147,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.1096447944641112
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,161,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5828288078308106
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,16,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.4602367877960205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,512,755,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.1006912231445312
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,160,188,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.1081215858459472
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,24,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5384640216827392
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,233,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.6077631950378418
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,29,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.4929152011871338
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,192,224,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.1350784301757812
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,371,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.565715217590332
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,768,1139,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.2416959762573243
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,35,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.48509440422058103
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,256,303,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.0372672080993652
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,460,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.6548992156982422
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,47,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.4981696128845215
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1530,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.235910415649414
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,384,454,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.0767744064331055
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,775,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.7094272136688232
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,61,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.526694393157959
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,74,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5352000236511231
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,512,613,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.142745590209961
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,969,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.7252096176147461
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,101,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5462399959564209
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,2309,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.4455103874206543
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,121,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5471104145050049
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,768,921,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.3100735664367675
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1589,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.8908096313476562
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,150,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5405312061309815
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,1024,1219,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.3727295875549317
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,210,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5339968204498291
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,3164,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.6314815521240233
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1961,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.0259200096130372
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,322,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5728640079498291
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,1536,1876,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.5539456367492677
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,428,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5997824192047119
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,4968,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,2.0393535614013674
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,637,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.6309120178222656
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,3162,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.1890239715576172
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,2048,2467,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.7327936172485352
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,867,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.6818304061889648
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,7075,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,2.356582450866699
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1338,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.7657599925994873
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,4492,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.565452766418457
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,3072,3783,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,2.1383743286132812
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2015,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.89486083984375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,4096,4930,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,2.453843116760254
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,10205,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,3.0665407180786133
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2803,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.0415552139282227
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,6728,7168,2048,8,256,256,1,4,power_law_1.01,accurate,2.1292287826538088
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3794,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.2203519821166993
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,6144,7640,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,3.2367488861083986
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,12817,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,3.8490047454833984
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8730,7168,2048,8,256,256,1,4,power_law_1.01,accurate,2.6158655166625975
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,6100,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.6014144897460938
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,8192,10267,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,3.9956417083740234
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,17599,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,4.569343948364258
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,10868,7168,2048,8,256,256,1,4,power_law_1.01,accurate,2.9018815994262694
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8018,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.9938560485839845
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,10240,12760,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,4.720377731323242
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,9716,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,2.31976318359375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,28015,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,6.710816192626953
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,16984,7168,2048,8,256,256,1,4,power_law_1.01,accurate,4.089708709716797
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,16384,20807,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,6.91884765625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,16780,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,3.3769790649414064
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,60473,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,13.148895263671875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,34441,7168,2048,8,256,256,1,4,power_law_1.01,accurate,8.050579071044922
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,32768,41151,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,13.299871826171875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,33178,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,6.300851058959961
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.14991359710693358
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.1485759973526001
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.19775359630584716
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.23593599796295167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,2,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.2310271978378296
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.27486720085144045
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.29509758949279785
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,5,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.3628607988357544
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,5,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.39271039962768556
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,8,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.40836482048034667
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,11,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.41185917854309084
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,15,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.45354881286621096
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,18,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.46419200897216795
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,14,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5570943832397461
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,26,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5016511917114258
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,21,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5276224136352539
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,34,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.6402560234069824
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,24,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5398975849151612
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,38,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5667327880859375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,33,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5637824058532714
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.06973440051078797
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,54,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5029632091522217
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,43,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5758975982666016
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.10944000482559205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,70,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5089471817016602
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,54,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5794432163238525
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.14752639532089235
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,1,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.1494271993637085
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,92,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5868480205535889
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,63,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5777664184570312
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.19646079540252687
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.21427841186523439
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,121,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5277056217193603
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,85,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5955327987670899
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,2,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.23021440505981444
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.28171520233154296
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.281497597694397
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,159,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5535232067108155
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,107,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5655871868133545
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,14,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.3350719928741455
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.3548799991607666
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,222,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5730112075805665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,124,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.6216576099395752
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,19,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.45395841598510744
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,11,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.41509761810302737
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,23,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.4786496162414551
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,286,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.6221248149871826
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,169,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5706495761871337
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,16,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5427519798278808
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,30,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.518720006942749
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,389,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.6287487983703614
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,249,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.6036543846130371
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,19,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5344895839691162
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,36,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.4885248184204102
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,598,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.6771647930145264
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,25,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5258368015289306
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,342,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5991936206817627
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,53,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5037375926971436
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,31,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.558080005645752
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,68,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.547430419921875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,936,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.7718719959259033
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,512,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.6491136074066162
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,43,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5931327819824219
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,79,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5110464096069336
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,51,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6388735771179199
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,108,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5550335884094239
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1236,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.7583807945251465
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,699,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.7008319854736328
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,63,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5993535995483399
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,144,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5496767997741699
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,89,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6130752086639404
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,170,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5086080074310303
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1930,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.8681983947753906
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1074,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.8079487800598144
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,109,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6242496013641358
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,272,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.6060800075531005
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,127,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6676735877990723
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,389,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5967743873596192
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1394,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.9181504249572754
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2852,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.2019904136657715
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,178,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.625984001159668
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,517,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.6356671810150146
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,269,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6646783828735352
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,767,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.6605887889862061
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2143,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.0887167930603028
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,3737,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.2764927864074707
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,369,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6407680034637451
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1145,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.7048448085784912
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,534,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.7089983940124511
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1761,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.7926591873168946
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,719,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.7335040092468261
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,2819,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.261235237121582
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,6444,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.8527616500854491
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,3128,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.8893440246582032
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1082,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.8416192054748535
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1431,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.9365568161010742
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,3855,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.054041576385498
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4337,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.6018815994262696
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,8352,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.962188720703125
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2209,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.1353919982910157
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,6696,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.253593635559082
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,5760,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,2.050521659851074
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3126,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.3067263603210448
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,11755,7168,2048,8,256,256,1,4,power_law_1.2,accurate,2.789164733886719
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,7891,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.6587839126586914
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4534,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.684716796875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7329,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,2.335750389099121
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,14364,7168,2048,8,256,256,1,4,power_law_1.2,accurate,3.3323009490966795
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,10413,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,2.004409599304199
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,6101,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,2.0308351516723633
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,16824,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,2.3705535888671876
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,11410,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,3.4577537536621095
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,32768,7168,2048,8,256,256,1,4,power_law_1.2,accurate,4.495251083374024
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7874,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,2.3936704635620116
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,26191,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,3.48223991394043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,12395,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,3.442700958251953
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,22875,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,6.359193420410156
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,56597,7168,2048,8,256,256,1,4,power_law_1.2,accurate,9.619776153564453
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,65536,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,6.677964782714843
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,25668,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,6.393683242797851
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.13864959478378297
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.2200063943862915
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,3,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.24046080112457274
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,1,4,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.35866880416870117
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,2,6,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3114943981170654
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,4,9,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.2332223892211914
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,6,10,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.35381760597229006
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.058348798751831056
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,8,15,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.29338879585266114
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.07086079716682434
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,10,21,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.28434560298919676
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.0927295982837677
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,12,23,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.2982271909713745
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,1,1,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.14364800453186036
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,16,40,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.36112639904022215
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,5,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1835263967514038
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,20,46,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.2853760004043579
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,8,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.2248255968093872
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,24,55,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.30574080944061277
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,10,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.2928960084915161
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,32,84,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3426111936569214
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.10661120414733886
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,14,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.28392319679260253
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,40,105,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3280191898345947
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.1243008017539978
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,18,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.31278719902038576
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,48,111,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3110016107559204
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.14657919406890868
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,21,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.2844223976135254
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,64,178,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3371968030929565
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,1,2,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.19467519521713256
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,33,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.30392959117889407
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,3,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.24272000789642334
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,96,248,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3452159881591797
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,41,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3019968032836914
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,6,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2680896043777466
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,44,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3194175958633423
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,128,333,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3775808095932007
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,8,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3129856109619141
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,73,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3089024066925049
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,12,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.33660159111022947
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,192,520,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.3716928005218506
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,84,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3305599927902222
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,16,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.35238399505615237
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.13747199773788452
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,19,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.35101439952850344
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,105,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.33575680255889895
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.22076799869537353
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,256,797,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.43927679061889646
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,25,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3247936010360718
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.1686079978942871
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,151,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.32099199295043945
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,33,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.36135680675506593
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,1,4,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.26709120273590087
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1170,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.5379007816314697
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,226,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.37292160987854006
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,2,6,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3069119930267334
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,39,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3584575891494751
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,4,13,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3606911897659302
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,331,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3644095897674561
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,51,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.35439999103546144
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,6,11,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.322873592376709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1780,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.6173888206481933
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,65,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.35934081077575686
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,521,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.402675199508667
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,8,16,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.29173760414123534
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,76,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3548863887786865
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,10,21,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.26255359649658205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,103,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3552896022796631
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,621,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.39720959663391114
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,12,27,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3234112024307251
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,768,2415,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.683462381362915
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,16,41,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3304960012435913
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,161,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.3690239906311035
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,20,47,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3382400035858154
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1039,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.44330878257751466
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,220,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.34454400539398194
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,24,65,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3312511920928955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,32,92,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.31439359188079835
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,328,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.37882239818573
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,3224,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.7995967864990234
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1406,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.49786877632141113
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,40,123,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3487423896789551
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,447,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.41472640037536623
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,48,137,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3493056058883667
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,64,213,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3668800115585327
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,661,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.48749442100524903
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,2016,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.5939712047576904
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,96,343,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.37358078956604
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,4951,7168,2048,8,256,256,1,8,power_law_1.01,accurate,1.0409600257873535
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,128,467,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.3967103958129883
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,903,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.5090943813323975
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,192,668,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.4555647850036621
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,2816,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.6735231876373291
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,256,987,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.4662208080291748
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,1361,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.6030784130096436
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1448,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.5120831966400147
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,6032,7168,2048,8,256,256,1,8,power_law_1.01,accurate,1.257158374786377
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,4263,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.8770432472229004
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,1779,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.6930880069732666
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1969,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.6417344093322754
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,768,3434,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.790502405166626
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,2585,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.8919103622436524
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,8020,7168,2048,8,256,256,1,8,power_law_1.01,accurate,1.6693759918212892
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,5635,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,1.0296704292297363
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,4437,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.9499008178710937
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,3829,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,1.052883243560791
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,8217,7168,2048,8,256,256,1,8,power_law_1.2,accurate,1.2842816352844237
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,9020,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,1.2540927886962892
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,14359,7168,2048,8,256,256,1,8,power_law_1.01,accurate,2.7402111053466798
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,4555,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,1.2085375785827637
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,11711,7168,2048,8,256,256,1,8,power_law_1.2,accurate,1.7978368759155274
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,17461,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,1.8117952346801758
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,14764,7168,2048,8,256,256,1,8,power_law_1.2,accurate,2.0113088607788088
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,7395,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,1.7703039169311523
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,38470,7168,2048,8,256,256,1,8,power_law_1.01,accurate,5.995558547973633
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,26311,7168,2048,8,256,256,1,8,power_law_1.2,accurate,3.2614784240722656
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,23968,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,3.197881507873535
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,14804,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,3.171379280090332
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,53779,7168,2048,8,256,256,1,8,power_law_1.2,accurate,6.944281768798828
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.06069120168685913
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.07301759719848633
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.09478399753570557
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,1,2,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.13384319543838502
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,4,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.170905601978302
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,9,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.19178880453109742
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,12,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.26905601024627684
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,16,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.29240319728851316
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,22,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.3035968065261841
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,26,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.29922559261322024
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,33,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.3098560094833374
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,51,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.31617920398712157
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,56,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.33649280071258547
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,85,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.31916160583496095
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,98,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.30663039684295657
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.10785919427871704
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,139,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.3228928089141846
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.1510208010673523
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.12627840042114258
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.23500800132751465
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,192,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.3351615905761719
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.1470080018043518
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.23548800945281984
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,262,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.3398848056793213
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,1,2,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.1914687991142273
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.24004480838775635
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,1,5,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.23400321006774902
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,3,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2129663944244385
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,441,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.35976319313049315
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,2,8,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.16441600322723388
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,6,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2511552095413208
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,3,8,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.2125823974609375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,702,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.39502720832824706
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,8,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.32356479167938235
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.06032000184059143
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,4,13,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.20008959770202636
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,12,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3213696002960205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,1035,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.41407361030578616
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.06182399988174438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,5,16,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.18338559865951537
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,15,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3554431915283203
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.07185279726982116
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,6,19,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.18684799671173097
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,19,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3373823881149292
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.10925439596176148
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,8,32,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.21857919692993164
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1635,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.47620477676391604
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,26,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3450432062149048
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,3,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.128985595703125
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,10,33,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.20069758892059325
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,34,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.34358398914337157
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,6,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.14729599952697753
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,12,41,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.15548800230026244
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,38,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.34945919513702395
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,2167,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.4961088180541992
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,9,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.18831360340118408
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,16,61,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.21852800846099854
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,54,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3372096061706543
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,11,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.1759168028831482
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,20,73,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.20918400287628175
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,65,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.37703039646148684
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,14,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.19032319784164428
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,24,97,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.18932479619979858
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,80,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.36978559494018554
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,17,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2043839931488037
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,4051,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.6011392116546631
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,32,138,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.21887359619140626
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,24,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.20013439655303955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,114,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3757695913314819
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,48,228,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.22695679664611818
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,33,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.191212797164917
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,173,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3801919937133789
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,41,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.18943359851837158
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,64,285,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.2362368106842041
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,226,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.38017280101776124
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,54,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.17779200077056884
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,5449,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.7107583999633789
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,96,442,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.24272639751434327
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,69,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.1823040008544922
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,366,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4065216064453125
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,93,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.21241600513458253
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,128,544,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.2629823923110962
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,460,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4493055820465088
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,107,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.19275519847869874
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,177,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.20013439655303955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,192,869,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.3349440097808838
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,7656,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.8741120338439942
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,676,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4958911895751953
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,242,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2222656011581421
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,256,1476,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.395961594581604
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,352,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2374272108078003
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,954,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.5500607967376709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,502,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2378175973892212
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,384,1812,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.4481152057647705
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,1495,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.6572095870971679
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,832,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2651648044586182
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,10842,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,1.0742719650268555
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,1109,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2823359966278076
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,512,2818,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.5579584121704102
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,1942,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.7161407947540284
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,1461,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.33518080711364745
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,16232,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,1.3147135734558106
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,768,3611,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.6829887866973877
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,2978,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.8724991798400878
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,2647,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.3918272018432617
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,6354,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.8427328109741211
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,3772,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.4850111961364746
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,4098,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,1.1027008056640626
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,32281,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,2.0038143157958985
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,4957,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.5655424118041992
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,7720,7168,2048,8,256,256,1,16,power_law_1.01,accurate,1.0163904190063477
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,5107,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,1.218348789215088
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,6988,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.675494384765625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,11567,7168,2048,8,256,256,1,16,power_law_1.01,accurate,1.7629695892333985
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,7716,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,1.7607616424560546
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,51476,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,3.5512447357177734
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,9414,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.9586432456970215
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,24184,7168,2048,8,256,256,1,16,power_law_1.01,accurate,3.500543975830078
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,16571,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,3.1900415420532227
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,23495,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,1.7632127761840821
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.08194559812545776
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.08314239978790283
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.09342719912528992
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1241536021232605
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,1,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1411520004272461
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,5,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.16272000074386597
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,6,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1931007981300354
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,9,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.19503359794616698
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,10,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.22485120296478273
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,11,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.21063039302825928
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,16,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.206496000289917
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,22,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.22470400333404542
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,24,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.21162879467010498
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,37,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.21743359565734863
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.05800319910049438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.15160319805145264
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,46,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.22033920288085937
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.059334397315979004
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.23664000034332275
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,53,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.21774721145629883
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.07094399929046631
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.21969919204711913
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,76,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.23457279205322265
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.09118080139160156
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,3,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.24194560050964356
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,5,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.11473280191421509
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,108,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.2358464002609253
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,1,5,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.24270079135894776
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,10,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.1339840054512024
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,2,9,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.23632640838623048
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,150,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.2236095905303955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,9,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.18677760362625123
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,3,12,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.219980788230896
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,14,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.1819264054298401
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,4,12,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.17653119564056396
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.08373119831085205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,228,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.22929279804229735
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,17,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.1727679967880249
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,5,21,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.21529600620269776
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.08467199802398681
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,292,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.25392000675201415
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,19,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.19015040397644042
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,6,28,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.202726411819458
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.0942463994026184
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,27,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.19791359901428224
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,8,39,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.21480319499969483
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.12450560331344604
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,2,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.14358400106430053
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,10,40,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.20124800205230714
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,39,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.19552639722824097
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,469,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.2563136100769043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,5,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.16547839641571044
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,12,52,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.19100799560546874
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,44,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.19447040557861328
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,5,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.19827200174331666
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,16,76,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.1858240008354187
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,63,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.18214399814605714
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,635,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.3033663988113403
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,9,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.19397759437561035
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,20,96,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.21904640197753905
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,98,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.20189440250396729
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,12,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.21675519943237304
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,24,149,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.20570240020751954
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,120,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.2002687931060791
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,14,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.21867520809173585
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,32,172,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.23978240489959718
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,156,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.1974527955055237
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,949,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.3580735921859741
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,20,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.2230463981628418
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,48,241,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.22954239845275878
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,246,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.20979199409484864
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,23,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.23054080009460448
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,28,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.22787199020385743
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,64,356,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.24664320945739746
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,377,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.23284480571746827
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,40,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.22160000801086427
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,1372,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.3868160009384155
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,96,611,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.29176959991455076
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,565,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.2196863889694214
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,45,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.23133440017700196
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,59,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.2365056037902832
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,128,1006,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.3027456045150757
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,791,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.2458048105239868
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,72,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.23584001064300536
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,192,1414,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.3765887975692749
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,122,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.22204160690307617
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,1388,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.26540799140930177
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,1776,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.4959871768951416
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,156,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.23570559024810792
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,256,1865,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.44990081787109376
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,2026,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.26836481094360354
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,254,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.25684480667114257
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,332,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.27567999362945556
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,384,3408,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.508678388595581
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,3099,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.30240640640258787
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,497,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.2758656024932861
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,2659,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.5734591960906983
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,717,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.31371519565582273
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,512,3336,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.543558406829834
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,4768,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.49582719802856445
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,1093,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.3490303993225098
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,3359,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.673740816116333
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,768,6575,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.1742143630981445
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,6903,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.5814591884613037
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,1410,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.3888511896133423
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,16274,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.3188608169555665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,11155,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.8105855941772461
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,2045,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.5047039985656738
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,5408,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.9078144073486328
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,13147,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.3039679527282715
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,2936,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.5763008117675781
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,10802,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.7295551776885987
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,3786,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.6758528232574463
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,21695,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.7815296173095703
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,18464,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,1.160204792022705
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,10583,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,1.6262144088745116
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,5922,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.9108672142028809
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,35503,7168,2048,8,256,256,1,16,power_law_1.2,accurate,3.4421886444091796
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,65536,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,3.5202880859375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,11615,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,1.627084732055664
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.1399359941482544
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.1546880006790161
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.14083199501037597
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.15886080265045166
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.12961280345916748
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,1,5,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.14175360202789306
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,1,7,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.125164794921875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,2,12,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.13023359775543214
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,2,12,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.13164160251617432
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,3,13,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.1302016019821167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,4,20,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.14369280338287355
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,5,29,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.132697594165802
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,6,29,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.13645440340042114
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,8,47,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.1411903977394104
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,10,59,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.12598400115966796
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.05742080211639404
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.059564799070358276
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,12,75,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.11822079420089722
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0585919976234436
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,16,118,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.1343808054924011
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0754751980304718
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,24,151,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.12688000202178956
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07390720248222352
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,5,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.08302720189094544
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,32,232,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.15372159481048583
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,6,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12040959596633911
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,48,369,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.17402880191802977
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,9,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.1276800036430359
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.07046399712562561
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,10,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.13219200372695922
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,64,500,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.18448640108108522
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.07035520076751708
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,15,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12445440292358398
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.07258880138397217
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,22,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12188160419464111
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08699520230293274
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,96,779,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.19634560346603394
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,26,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12237440347671509
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,2,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.10007679462432861
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,33,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12769919633865356
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,4,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.127948796749115
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,39,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12911360263824462
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,128,1122,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.23059840202331544
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,5,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.13452160358428955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,58,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.11898239850997924
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,7,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.13082239627838135
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,69,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.11944960355758667
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,8,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.125600004196167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.13903360366821288
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,100,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.11671040058135987
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,192,1804,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.3297343969345093
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,9,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.14257279634475709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.15873279571533203
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,167,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.14706560373306274
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.13969919681549073
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,13,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.1463744044303894
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.14239360094070436
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,17,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.14243199825286865
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,270,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.08034560084342957
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,5,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.14197759628295897
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,20,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.14746240377426148
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,256,2752,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.4174015998840332
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,1,9,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08634880185127258
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,360,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.13676799535751344
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,31,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.14198399782180787
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,1,7,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.14023040533065795
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,34,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.15700479745864868
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,2,9,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.14260480403900147
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,396,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12437759637832642
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,2,15,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.13471360206604005
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,41,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.14651520252227784
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,3,17,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.12730239629745482
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,57,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.15641599893569946
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,804,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12472319602966309
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,4,26,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.15112320184707642
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,384,2881,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.47166080474853517
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,90,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.15364480018615723
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,5,33,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.143449604511261
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,6,35,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.12755839824676513
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,122,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.15218559503555298
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,958,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.19467519521713256
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,8,79,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.1375615954399109
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,185,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.1457919955253601
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,10,77,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.13585920333862306
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,12,126,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.1379968047142029
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,1351,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.2253887891769409
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,238,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.15263359546661376
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,16,125,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.12789119482040406
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,512,4581,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.6147007942199707
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,24,260,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.15358079671859742
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,367,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.11665920019149781
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,32,459,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.18583680391311647
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,2380,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.1824447989463806
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,48,604,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.1939520001411438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,512,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.20349440574645997
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,64,682,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.19566080570220948
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,96,1250,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.21850879192352296
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,811,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.2243583917617798
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,640,7118,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.7133183956146241
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,3032,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.2821376085281372
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,128,1441,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.2560895919799805
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,900,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.24830079078674316
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,192,3823,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.5913343906402588
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,4769,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.3028543949127197
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,256,5096,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.5243584156036377
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,1511,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.28414719104766845
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,10056,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.8905856132507324
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,6109,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.3845887899398804
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,384,5003,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.5620800018310547
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,1981,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.3262847900390625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,512,11188,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.8718527793884278
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,2471,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.3223104000091553
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,8345,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.48761601448059083
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,640,13893,7168,2048,8,256,256,1,32,power_law_1.2,accurate,1.3223360061645508
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,21696,7168,2048,8,256,256,1,32,power_law_1.01,accurate,2.052947235107422
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,4521,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.5225728034973145
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,22227,7168,2048,8,256,256,1,32,power_law_1.2,accurate,1.9132863998413085
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,30986,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,1.6161727905273438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,8912,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.8623807907104493
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,50715,7168,2048,8,256,256,1,32,power_law_1.2,accurate,3.785548782348633
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.060678398609161376
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0607807993888855
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06222720146179199
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06575999855995178
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,4,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06806399822235107
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,8,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07741439938545228
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,7,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.1215872049331665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,12,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.12049920558929443
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,13,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.12318719625473022
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,17,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.11931519508361817
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,25,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.11508480310440064
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,34,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.11641600131988525
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.07203199863433837
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,39,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.11374720335006713
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.07129600048065185
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.07233920097351074
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,75,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06056960225105286
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08615040183067321
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,70,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.10857599973678589
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.10316159725189208
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,99,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.059628802537918094
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,3,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.11743999719619751
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,6,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13729920387268066
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,130,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07200000286102295
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,6,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13545600175857545
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,9,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13262720108032228
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,228,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06599040031433105
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,9,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13935999870300292
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,15,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.1443519949913025
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,327,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0905344009399414
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,17,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13965439796447754
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,26,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.1500223994255066
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,573,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0910975992679596
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,31,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.1492095947265625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,34,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.14540159702301025
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,659,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.09884799718856811
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,48,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.15118720531463622
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05623040199279785
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09520639777183533
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05891839861869812
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,61,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.15448960065841674
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09426559805870056
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05743359923362732
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,1734,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.19078400135040283
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09543039798736572
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,93,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.1602112054824829
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06147199869155884
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09880959987640381
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,2,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09574400186538697
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.061343997716903687
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,127,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.15534720420837403
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,5,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.10088319778442383
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,5,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06398720145225525
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,2088,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.26483840942382814
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,7,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08957440257072449
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,184,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.1532863974571228
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,6,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06414719820022582
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,1,8,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09105920195579528
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,1,7,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06869120001792908
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,1,13,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09293439984321594
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,268,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.16499199867248535
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,1,10,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.059059202671051025
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,1,13,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09107840061187744
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,1,13,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.0605184018611908
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,3663,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.30940799713134765
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,2,17,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08315520286560059
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,2,20,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.0597760021686554
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,416,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.11129599809646606
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,2,27,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09425920248031616
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,2,23,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06259840130805969
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,3,38,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.10344959497451782
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,3,35,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06125440001487732
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,570,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13859200477600098
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,4,43,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08965759873390197
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,4,41,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06000000238418579
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,4396,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.3716223955154419
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,5,61,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09278079867362976
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,5,56,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05901439785957337
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,6,69,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09121919870376587
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,6,60,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.059359997510910034
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,933,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.23946878910064698
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,8,81,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09477120041847228
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,8,79,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.056953597068786624
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,12,153,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08955519795417785
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,12,161,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06746879816055298
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,16,228,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.1144320011138916
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,16,250,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.06101120114326477
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,7313,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.4658944129943848
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,1309,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.20227200984954835
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,24,324,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.11666560173034668
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,24,328,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.07323520183563233
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,32,492,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.12106239795684814
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,32,442,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.07873280048370361
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,48,572,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.14016000032424927
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,48,630,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.08580480217933655
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,1870,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.28961279392242434
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,10303,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.6685760021209717
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1048,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.19704960584640502
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1042,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.11719679832458496
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,96,1928,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.2647936105728149
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,96,1824,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.1578559994697571
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,2424,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.307590389251709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,12441,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.7637631893157959
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,128,2335,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.3113152027130127
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,128,2071,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.17144320011138917
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,3080,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.4041088104248047
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,192,4010,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.36878719329833987
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,192,3527,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.23653759956359863
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,21755,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,1.231180763244629
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,256,5260,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.4219007968902588
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,256,4810,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.2955967903137207
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,4726,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.528326416015625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,320,7304,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.7007487773895263
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,320,5823,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.34840319156646726
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,512,10569,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.9025471687316895
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,512,14270,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.7725887775421143
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,9724,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.9217472076416016
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,65536,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,3.3902591705322265
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,19966,7168,2048,8,256,256,1,64,power_law_1.01,accurate,1.5291775703430175
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,19040,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,1.0225728034973145
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05753600001335144
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09834240078926086
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06039040088653565
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09445120096206665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.059699201583862306
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0995136022567749
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06225280165672302
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09669119715690613
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,4,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06265599727630615
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08631680011749268
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,7,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0902783989906311
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,9,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06229119896888733
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,8,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08686720132827759
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,7,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.0639360010623932
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,1,10,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08526719808578491
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,1,10,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05982080101966858
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,1,14,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08814079761505127
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,1,14,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.061990398168563846
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,1,17,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0961408019065857
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,1,17,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05999360084533691
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,2,24,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09844480156898498
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,2,24,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06214399933815003
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,2,33,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0979200005531311
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,2,32,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.060063999891281125
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,3,50,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0880832016468048
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,3,36,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06268159747123718
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,4,73,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09784960150718688
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,4,51,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.059724801778793336
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.06912639737129211
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,5,79,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09281920194625855
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,5,81,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.057171201705932616
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07174400091171265
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,6,92,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09168000221252441
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07034879922866821
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,6,96,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05759360194206238
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07420799732208253
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,8,94,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08844159841537476
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,8,143,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.06912639737129211
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.0729856014251709
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,12,236,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.11384320259094238
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,12,289,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.07481600046157837
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,6,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07254400253295898
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,16,341,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.11185280084609986
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,5,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07667840123176575
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,16,327,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.07439360022544861
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,64,9,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07331839799880982
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,24,649,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.14620800018310548
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.05613440275192261
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,80,11,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07108479738235474
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,24,519,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.08945279717445373
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.0588096022605896
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,96,13,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.0755840003490448
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,32,666,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.1454848051071167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.05726720094680786
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,1,19,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07041919827461243
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,32,615,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.08629760146141052
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.061286401748657224
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,1,29,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.06830719709396363
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,48,1071,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.17932800054550171
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06213759779930115
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,1,30,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.06793599724769592
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,48,1296,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.1309056043624878
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,5,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06264320015907288
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,2,44,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.0687936007976532
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,5,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06417919993400574
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,2,55,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07233920097351074
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1752,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.2212735891342163
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,64,8,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.0641215980052948
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1725,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.1909119963645935
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,3,74,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.06784639954566955
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,80,11,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06232960224151611
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,4,93,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.06330879926681518
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,96,13,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06553599834442139
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,1,20,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06064000129699707
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,96,1871,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.2494976043701172
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,6,155,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07617920041084289
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,96,3054,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.225382399559021
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,1,22,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06092159748077393
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,8,194,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.07720959782600403
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,1,30,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.05991680026054382
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,2,45,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.060224002599716185
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,12,300,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.09572479724884034
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,2,59,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.058182400465011594
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,128,2643,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.30561280250549316
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,128,3578,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.2321727991104126
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,16,443,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.1027135968208313
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,3,64,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06167680025100708
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,4,94,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.056460797786712646
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,24,810,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.12837120294570922
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,6,153,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06691840291023254
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,8,212,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.06276479959487916
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,192,6620,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.5403327941894531
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,32,1013,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.14574079513549804
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,192,7318,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.4352128028869629
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,12,275,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.07713919878005981
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,16,498,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.07579519748687744
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,48,1688,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.20300159454345704
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,24,1009,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.10497920513153076
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,256,8772,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.6754047870635986
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,256,8725,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.49166078567504884
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,32,1215,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.11985280513763427
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,64,2098,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.19363199472427367
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,48,1568,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.14291839599609374
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,96,3233,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.3436032056808472
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,320,13355,7168,2048,8,256,256,1,64,power_law_1.2,accurate,1.0188863754272461
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,64,2522,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.1837823987007141
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,320,10897,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.5987711906433105
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,128,3867,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.33165440559387205
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,96,3114,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.21683199405670167
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,512,18421,7168,2048,8,256,256,1,64,power_law_1.2,accurate,1.4303232192993165
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,512,22541,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,1.1911935806274414
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,128,3880,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.2515455961227417
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,160,6124,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.4623551845550537
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,160,5337,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.3203775882720947
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,256,11306,7168,2048,8,256,256,1,128,power_law_1.01,accurate,0.625593614578247
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,62098,7168,2048,8,256,256,1,64,power_law_1.2,accurate,3.1244480133056642
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,256,10741,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,0.5840703964233398
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,65536,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,3.2788158416748048
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,512,28013,7168,2048,8,256,256,1,128,power_law_1.01,accurate,1.4608832359313966
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,512,25429,7168,2048,8,256,256,1,128,power_law_1.01_eplb,accurate,1.3036288261413573
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07215999960899352
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07227519750595093
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07324159741401673
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.06525440216064453
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07292159795761108
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,7,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07201279997825623
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,6,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07531520128250122
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,64,9,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07359359860420227
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.05626239776611328
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,80,12,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07397119998931885
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.05961599946022034
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,96,14,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07274240255355835
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.057625597715377806
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,1,27,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07496960163116455
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.0637503981590271
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,1,32,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07363200187683105
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,4,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06335999965667724
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,1,34,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07649279832839966
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,8,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06532480120658875
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,2,48,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.06803200244903565
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,7,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06220800280570984
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,2,78,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07342720031738281
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,64,10,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06183040142059326
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,3,105,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07043840289115906
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.05727360248565674
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,80,14,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06452479958534241
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.05822719931602478
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,4,127,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.06389120221138
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,96,19,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06451839804649354
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.05826560258865356
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,1,22,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06182399988174438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,6,202,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.07340800166130065
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06285439729690552
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,1,35,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.061459201574325564
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,8,331,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.09070720076560974
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06225280165672302
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,1,43,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.059059202671051025
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,5,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06536319851875305
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,12,769,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.1096384048461914
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,2,54,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.060159999132156375
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,5,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06599680185317994
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,2,92,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.0554751992225647
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,64,8,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06440960168838501
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,16,806,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.11349760293960572
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.05634559988975525
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,3,119,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.05537279844284058
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,80,10,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06283519864082336
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.057081598043441775
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,96,14,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06574079990386963
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,4,136,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06714879870414733
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,24,1497,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.1527616024017334
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.057606399059295654
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,128,18,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06480640172958374
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.05998079776763916
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,6,211,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.06211199760437012
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,160,23,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.0667136013507843
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06079360246658325
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,32,1449,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.1670912027359009
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,8,324,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.07271680235862732
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,192,26,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06656640172004699
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,6,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06409599781036376
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,1,47,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.057760000228881836
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,5,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06421120166778564
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,12,679,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.09075199961662292
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,64,8,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.063155198097229
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,1,57,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.05942400097846985
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,48,3525,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.24956159591674804
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,80,10,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06180480122566223
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,1,81,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.05588480234146118
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,16,998,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.10421119928359986
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,96,13,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06526079773902893
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,2,95,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.055980801582336426
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,128,22,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.0663424015045166
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,24,1117,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.11115520000457764
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,3,136,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06694399714469909
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,160,22,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06565759778022766
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,64,4029,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.27256319522857664
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,192,30,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.06372479796409607
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,4,214,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.06280959844589233
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,1,44,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.05920000076293945
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,32,2163,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.16488319635391235
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,6,347,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.0713599979877472
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,1,52,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.05934720039367676
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,1,86,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.05459840297698974
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,8,520,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.08920320272445678
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,2,94,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.054841601848602296
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,48,3246,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.23662080764770507
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,96,6588,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.5713535785675049
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,3,204,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.0642304003238678
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,12,682,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.09102720022201538
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,4,252,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.058719998598098753
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,16,1146,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.10868480205535888
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,64,5520,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.35007998943328855
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,6,355,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.07031040191650391
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,8,424,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.07659519910812378
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,128,14235,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.786521577835083
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,24,1551,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.14229120016098024
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,12,759,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.08874239921569824
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,96,5657,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.34204161167144775
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,16,944,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.10558079481124878
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,32,2816,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.19326720237731934
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,160,15066,7168,2048,8,256,256,1,128,power_law_1.2,accurate,0.8109951972961426
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,24,1425,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.13562239408493043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,128,11247,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.6212800025939942
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,48,3443,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.2221695899963379
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,32,1893,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.1506175994873047
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,64,4619,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.28652799129486084
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,160,14198,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,0.7662335872650147
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,48,3397,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.224236798286438
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,256,24743,7168,2048,8,256,256,1,128,power_law_1.2,accurate,1.8495616912841797
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,64,4914,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.3006975889205933
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,80,7040,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.39590399265289306
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,256,23024,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,1.202943992614746
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,80,7421,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.41365761756896974
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,128,8277,7168,2048,8,256,256,1,256,power_law_1.01,accurate,0.45435519218444825
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,512,52685,7168,2048,8,256,256,1,128,power_law_1.2,accurate,3.1270336151123046
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,128,10679,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,0.5750144004821778
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,512,57108,7168,2048,8,256,256,1,128,power_law_1.2_eplb,accurate,2.8601728439331056
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,256,22146,7168,2048,8,256,256,1,256,power_law_1.01,accurate,1.1236864089965821
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,256,19851,7168,2048,8,256,256,1,256,power_law_1.01_eplb,accurate,1.0167872428894043
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.058259201049804685
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.059359997510910034
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.059545600414276124
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.0640064001083374
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06334720253944397
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,9,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06389120221138
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,7,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06572800278663635
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05717759728431702
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,64,9,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06568959951400757
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05672320127487183
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,80,14,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06678400039672852
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05829120278358459
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,96,17,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.0651968002319336
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06144639849662781
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,128,25,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06755840182304382
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.061337602138519284
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,160,32,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.0676800012588501
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,7,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.0626688003540039
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,192,40,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06544640064239501
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,6,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06557440161705017
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,1,66,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06124160289764404
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,64,9,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06386560201644897
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,1,82,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.05864319801330566
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,80,14,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06581760048866273
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,1,108,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.057574397325515746
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,96,18,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06262400150299072
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,2,141,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.0695743978023529
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,128,26,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06629760265350342
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,3,218,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.06258559823036194
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,160,35,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06527360081672669
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,4,311,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.07573760151863099
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,192,43,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06295679807662964
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,1,55,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05906559824943543
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,6,531,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.08941439986228943
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,1,69,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05626239776611328
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,1,82,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.05583360195159912
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,8,655,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.09411839842796325
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,2,143,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.07002879977226258
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,12,1201,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.12150399684906006
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,3,222,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.06083199977874756
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,4,327,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.07157760262489318
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,16,1762,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.14424959421157837
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,6,636,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.08276479840278625
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,8,683,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.09026560187339783
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,24,3996,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.25499520301818845
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,12,1244,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.1184831976890564
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,16,1548,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.1388800024986267
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,32,4246,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.27003519535064696
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,24,3637,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.23713281154632568
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,48,6552,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.3833791971206665
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,32,3936,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.24425599575042725
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,64,11144,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.6011263847351074
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,48,6294,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.36837120056152345
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,64,9384,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.5154623985290527
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,80,15776,7168,2048,8,256,256,1,256,power_law_1.2,accurate,0.8273280143737793
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,80,13279,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,0.7011263847351075
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,128,30067,7168,2048,8,256,256,1,256,power_law_1.2,accurate,1.497606372833252
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,128,32768,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,1.6275840759277345
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,256,59849,7168,2048,8,256,256,1,256,power_law_1.2,accurate,2.932908821105957
TRTLLM,1.1.0,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,256,65536,7168,2048,8,256,256,1,256,power_law_1.2_eplb,accurate,3.198303985595703
