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,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.08142079710960388
TRTLLM,1.2.0rc6,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,0.07909759879112244
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.11564160585403442
TRTLLM,1.2.0rc6,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,0.04910080134868622
TRTLLM,1.2.0rc6,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,0.09662719964981079
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.13619199991226197
TRTLLM,1.2.0rc6,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,0.058259201049804685
TRTLLM,1.2.0rc6,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,0.11100800037384033
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,4,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.16223360300064088
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.08197759985923767
TRTLLM,1.2.0rc6,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,0.07804800271987915
TRTLLM,1.2.0rc6,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,0.1440127968788147
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.15247360467910767
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.11535359621047973
TRTLLM,1.2.0rc6,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,0.11401599645614624
TRTLLM,1.2.0rc6,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,0.17423360347747802
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,13,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.17265280485153198
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.11947519779205322
TRTLLM,1.2.0rc6,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,0.13482880592346191
TRTLLM,1.2.0rc6,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,0.20300159454345704
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,17,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.23319039344787598
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,4,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.1382464051246643
TRTLLM,1.2.0rc6,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,0.18010239601135253
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,15,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.2463360071182251
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,24,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.22160000801086427
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,7,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.18367999792099
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,17,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.2156991958618164
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,29,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.23546879291534423
TRTLLM,1.2.0rc6,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,0.26382079124450686
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,16,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.2211967945098877
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,21,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.2342144012451172
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,40,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.24798719882965087
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,25,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.26936318874359133
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,19,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.2182528018951416
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,51,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.2434688091278076
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,31,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.2381376028060913
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,26,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.23991680145263672
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,32,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.28016641139984133
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,64,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.2591104030609131
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,31,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.28149120807647704
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,34,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.2420799970626831
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,39,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.28105599880218507
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,43,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.2520384073257446
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,77,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.24603519439697266
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,46,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.24444160461425782
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,52,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.28375680446624757
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,61,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.2615488052368164
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,117,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.3018239974975586
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,62,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.24578559398651123
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,62,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.28580479621887206
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,76,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.29678080081939695
TRTLLM,1.2.0rc6,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,0.256876802444458
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,141,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.26698880195617675
TRTLLM,1.2.0rc6,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,0.30329599380493166
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,98,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.28854401111602784
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,103,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.26415998935699464
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,182,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.28375680446624757
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,100,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.2951872110366821
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,125,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.26686720848083495
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,127,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.2694591999053955
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,242,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.2911168098449707
TRTLLM,1.2.0rc6,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,0.3006079912185669
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,148,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.2844928026199341
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,151,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.26998400688171387
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,175,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.3081792116165161
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,382,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.32344319820404055
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,190,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.26552319526672363
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,206,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.28362240791320803
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,248,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.33184640407562255
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,450,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.3171328067779541
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,248,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.29931519031524656
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,317,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.3087424039840698
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,331,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.3508863925933838
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,683,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.37796480655670167
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,425,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.33898880481719973
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,420,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.32328319549560547
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,518,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.390118408203125
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,634,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.3686975955963135
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1064,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.461740779876709
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,679,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.36602880954742434
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,1000,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.4456511974334717
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,708,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.4405248165130615
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1374,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.5559040069580078
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,919,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.4202112197875977
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1138,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.451859188079834
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1034,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.5272511959075927
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2000,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.6664000034332276
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1487,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.516435194015503
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,2029,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.5352000236511231
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1355,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.6147583961486817
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2892,7168,2048,8,256,256,1,4,power_law_1.01,accurate,0.7042431831359863
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1882,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.5954368114471436
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2667,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.6482111930847168
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2144,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.7434944152832031
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,4460,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.0028160095214844
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2941,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.6667967796325683
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,4829,7168,2048,8,256,256,1,4,power_law_1.2,accurate,0.8324159622192383
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,2876,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,0.8929599761962891
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3808,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,0.7918015956878662
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,5562,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.0944128036499023
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,5967,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.125715160369873
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4163,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.0868672370910644
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,8867,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.2488256454467774
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,5938,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.0740927696228026
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8102,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.4725055694580078
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,5750,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.3441151618957519
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,12358,7168,2048,8,256,256,1,4,power_law_1.2,accurate,1.727257537841797
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8098,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.3264960289001464
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,10555,7168,2048,8,256,256,1,4,power_law_1.01,accurate,1.7845760345458985
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7044,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,1.6363840103149414
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,15792,7168,2048,8,256,256,1,4,power_law_1.2,accurate,2.696710395812988
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,9847,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,1.5876223564147949
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,19251,7168,2048,8,256,256,1,4,power_law_1.01,accurate,3.1969087600708006
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,11475,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,2.4211456298828127
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,28055,7168,2048,8,256,256,1,4,power_law_1.2,accurate,4.316659164428711
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,14949,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,2.4046592712402344
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,34170,7168,2048,8,256,256,1,4,power_law_1.01,accurate,5.190879821777344
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,23348,7168,2048,8,256,288,1,4,power_law_1.01_eplb,accurate,4.515865707397461
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,55174,7168,2048,8,256,256,1,4,power_law_1.2,accurate,7.675148773193359
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,34823,7168,2048,8,256,256,1,4,power_law_1.01_eplb,accurate,4.7613567352294925
TRTLLM,1.2.0rc6,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,0.049983999133110045
TRTLLM,1.2.0rc6,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,0.058297598361968996
TRTLLM,1.2.0rc6,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,0.07866880297660828
TRTLLM,1.2.0rc6,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,0.10556800365447998
TRTLLM,1.2.0rc6,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,0.07950080037117005
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,7,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.12332799434661865
TRTLLM,1.2.0rc6,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,0.09797760248184204
TRTLLM,1.2.0rc6,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,0.16215039491653443
TRTLLM,1.2.0rc6,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,0.11045119762420655
TRTLLM,1.2.0rc6,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,0.14150400161743165
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,18,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.21854720115661622
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,4,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.17252479791641234
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,24,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.22461440563201904
TRTLLM,1.2.0rc6,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,0.046393600106239316
TRTLLM,1.2.0rc6,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,0.2008064031600952
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,28,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.23160960674285888
TRTLLM,1.2.0rc6,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,0.05064319968223572
TRTLLM,1.2.0rc6,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,0.25031681060791017
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,37,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.24599039554595947
TRTLLM,1.2.0rc6,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,0.06087039709091187
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,20,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.27393279075622556
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.08664320111274719
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,52,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.26099200248718263
TRTLLM,1.2.0rc6,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,0.0868224024772644
TRTLLM,1.2.0rc6,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,0.270470404624939
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.11264640092849731
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,63,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.24571518898010253
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,4,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.09650560021400452
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,3,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.13379839658737183
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,30,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.2752320051193237
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,84,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.25541119575500487
TRTLLM,1.2.0rc6,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,0.11151360273361206
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,1,3,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.13596800565719605
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,44,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.27871360778808596
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,110,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.2700160026550293
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,9,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1329792022705078
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,2,7,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.1712831974029541
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,54,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.28373119831085203
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,137,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.26150400638580323
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,15,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1360576033592224
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,4,10,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.16580480337142944
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,62,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.3042880058288574
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,163,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.2818687915802002
TRTLLM,1.2.0rc6,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,0.13917440176010132
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,6,11,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.13297280073165893
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,85,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.30563199520111084
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,22,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.14507520198822021
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,8,16,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.15326080322265626
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,244,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.2881664037704468
TRTLLM,1.2.0rc6,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,0.30332159996032715
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,10,21,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.1474303960800171
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,32,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.14894720315933227
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,384,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.3198335886001587
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,130,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.32627201080322266
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,12,27,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.1726591944694519
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,42,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1535871982574463
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,16,33,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.17247999906539918
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,502,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.33941121101379396
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,50,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1541632056236267
TRTLLM,1.2.0rc6,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,0.32704639434814453
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,20,46,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.17438080310821533
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,77,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.1583359956741333
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,266,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.3506367921829224
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,881,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.39162240028381345
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,24,52,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.15537279844284058
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,91,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.160588800907135
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,351,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.3717439889907837
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,32,82,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.1593727946281433
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,120,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.16435840129852294
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1105,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.43607678413391116
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,40,99,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.15859839916229249
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,554,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.42003841400146485
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,138,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.17003519535064698
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,48,120,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.17155200242996216
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,202,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.17322880029678345
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1726,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5214784145355225
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,728,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.45267839431762696
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,64,158,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.17683839797973633
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,349,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.19937920570373535
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,96,252,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.18458880186080934
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1099,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.5434944152832031
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,3029,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.5689216136932373
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,488,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.21178240776062013
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,128,370,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.22210559844970704
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,192,519,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.23398399353027344
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1449,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.6294976234436035
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,597,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.23221120834350586
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,3719,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.6879424095153809
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,256,729,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.25065600872039795
TRTLLM,1.2.0rc6,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,0.28117120265960693
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2202,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.7673791885375977
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1173,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.31651840209960935
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1386,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3245120048522949
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,4936,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,0.8228032112121582
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1629,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.4057919979095459
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3023,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,0.9107392311096192
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,2388,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.3839871883392334
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,768,2061,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.4369984149932861
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,2880,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.4559040069580078
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,8872,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.1241151809692382
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4593,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.100985622406006
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,3560,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.5453760147094726
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,4902,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.5865087985992432
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,10648,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.3912575721740723
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,6232,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.3808639526367188
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,4183,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.6866687774658203
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,5678,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.7068096160888672
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,6507,7168,2048,8,256,256,1,8,power_law_1.01,accurate,0.9038335800170898
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7720,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,1.6308351516723634
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,12832,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,1.6675071716308594
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,8308,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,0.8613759994506835
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,8795,7168,2048,8,256,256,1,8,power_law_1.01,accurate,1.0900608062744142
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,12620,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,2.4679616928100585
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,24292,7168,2048,8,256,256,1,4,power_law_1.2_eplb,accurate,2.595327949523926
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,14259,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,1.2770432472229003
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,13172,7168,2048,8,256,256,1,8,power_law_1.01,accurate,1.9327167510986327
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,31063,7168,2048,8,256,256,1,8,power_law_1.01_eplb,accurate,2.4888256072998045
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,26047,7168,2048,8,256,288,1,4,power_law_1.2_eplb,accurate,4.590131378173828
TRTLLM,1.2.0rc6,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,5.083404922485352
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,34672,7168,2048,8,256,256,1,8,power_law_1.01,accurate,3.85203857421875
TRTLLM,1.2.0rc6,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,0.12394239902496337
TRTLLM,1.2.0rc6,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,0.1382143974304199
TRTLLM,1.2.0rc6,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,0.16497279405593873
TRTLLM,1.2.0rc6,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,0.12595839500427247
TRTLLM,1.2.0rc6,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,0.2219520092010498
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.08566399812698364
TRTLLM,1.2.0rc6,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,0.276691198348999
TRTLLM,1.2.0rc6,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,0.13827840089797974
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.11285120248794556
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,7,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.22486400604248047
TRTLLM,1.2.0rc6,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,0.16506880521774292
TRTLLM,1.2.0rc6,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,0.049907198548316954
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.11253119707107544
TRTLLM,1.2.0rc6,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,0.25880959033966067
TRTLLM,1.2.0rc6,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,0.22335360050201417
TRTLLM,1.2.0rc6,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,0.05406079888343811
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,1,3,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.12110719680786133
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,4,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.17701120376586915
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,11,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2517695903778076
TRTLLM,1.2.0rc6,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,0.0633408010005951
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,2,6,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.13982080221176146
TRTLLM,1.2.0rc6,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,0.20076160430908202
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,15,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2725951910018921
TRTLLM,1.2.0rc6,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,0.08092799782752991
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,4,12,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.11459200382232666
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,6,10,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.26538240909576416
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,21,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.26261119842529296
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,2,6,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.09255679845809936
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,6,12,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.16433279514312743
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,14,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2538815975189209
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,24,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2776384115219116
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,4,10,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.10286719799041748
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,8,17,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.17119359970092773
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,16,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.25832960605621336
TRTLLM,1.2.0rc6,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,0.2717375993728638
TRTLLM,1.2.0rc6,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,0.13528319597244262
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,10,23,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.15434880256652833
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,17,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2814143896102905
TRTLLM,1.2.0rc6,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,0.29971840381622317
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,8,15,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.14568959474563598
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,12,26,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.1407871961593628
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,25,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2828351974487305
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,10,20,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.14325120449066162
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,54,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2984639883041382
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,16,44,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.16045440435409547
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,20,33,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.28510079383850095
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,12,29,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.14692480564117433
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,20,55,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.17414400577545167
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,66,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2932735919952393
TRTLLM,1.2.0rc6,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,0.2931519985198975
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,16,40,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.15313279628753662
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,24,70,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.16530560255050658
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,78,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.29173119068145753
TRTLLM,1.2.0rc6,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,0.15317120552062988
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,53,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3217472076416016
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,32,85,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.1664512038230896
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,108,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.2975615978240967
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,24,54,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.1513216018676758
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,40,127,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.1688704013824463
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,71,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.2980544090270996
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,163,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.31084160804748534
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,32,88,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.16508159637451172
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,48,151,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.18705919981002808
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,84,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.30456318855285647
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,214,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.30769920349121094
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,40,108,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.1603775978088379
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,64,205,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.19004160165786743
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,111,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.32622718811035156
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,48,126,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.1706431984901428
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,329,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.44605441093444825
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,96,280,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.20618879795074463
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,163,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.30424959659576417
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,64,193,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.17196799516677858
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,128,540,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.2333375930786133
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,234,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.3368256092071533
TRTLLM,1.2.0rc6,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,0.46513919830322265
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,96,334,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.1920575976371765
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,192,694,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.2541183948516846
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,344,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4489280223846436
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,128,454,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.20304639339447023
TRTLLM,1.2.0rc6,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,0.38815999031066895
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,256,1118,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.30357120037078855
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,489,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4872896194458008
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,192,613,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.22316160202026367
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1421,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.34535040855407717
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,880,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.4170815944671631
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,256,1142,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.2511744022369385
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,702,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.4062016010284424
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,512,2643,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.42351999282836916
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,384,1574,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.29955201148986815
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,1317,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.5071296215057373
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,955,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.43208961486816405
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,512,1722,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.33564159870147703
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,768,3844,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.6064256191253662
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,1476,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.5386112213134766
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,1879,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.5876927852630616
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,768,3121,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.40730881690979004
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,5283,7168,2048,8,256,256,1,8,power_law_1.2,accurate,0.7131455898284912
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,2014,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.5972032070159912
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,2757,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.66212477684021
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,1024,4811,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.47151999473571776
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,7512,7168,2048,8,256,256,1,8,power_law_1.2,accurate,1.0113216400146485
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,3018,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.6804543972015381
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,3517,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.8186880111694336
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,1536,9074,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.6298240184783935
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,11409,7168,2048,8,256,256,1,8,power_law_1.2,accurate,1.2641728401184082
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,4118,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.7637887954711914
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,4675,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,0.8826047897338867
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,2048,8993,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.7532800197601318
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,12878,7168,2048,8,256,256,1,8,power_law_1.2,accurate,1.3257344245910645
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,5069,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,0.8946751594543457
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,2560,14008,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,0.9266816139221191
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,7423,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,1.2892160415649414
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,27432,7168,2048,8,256,256,1,8,power_law_1.2,accurate,2.1635007858276367
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,8160,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,1.3180992126464843
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,4096,21753,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,1.3921088218688964
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,15538,7168,2048,8,256,288,1,8,power_law_1.01_eplb,accurate,2.274220848083496
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,64856,7168,2048,8,256,256,1,8,power_law_1.2,accurate,5.279084777832031
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,16084,7168,2048,8,256,288,1,8,power_law_1.2_eplb,accurate,2.279737663269043
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,8192,63476,7168,2048,8,256,256,1,8,power_law_1.2_eplb,accurate,2.870937538146973
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.07651200294494628
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12083200216293336
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12128640413284301
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12517759799957276
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,1,3,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12371200323104858
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,2,7,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11169919967651368
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,3,10,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11178239583969116
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,4,12,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11175680160522461
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,5,14,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.10586240291595458
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,6,15,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.10062079429626465
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,8,28,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11346559524536133
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,10,35,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11169279813766479
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,12,40,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.10928000211715698
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,16,58,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11381759643554687
TRTLLM,1.2.0rc6,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,0.04872319996356964
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,20,77,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11219199895858764
TRTLLM,1.2.0rc6,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,0.05061759948730469
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,24,98,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.11605119705200195
TRTLLM,1.2.0rc6,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,0.0548799991607666
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,32,146,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12296960353851319
TRTLLM,1.2.0rc6,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,0.06540799736976624
TRTLLM,1.2.0rc6,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,0.07904639840126038
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,48,209,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.12590080499649048
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.07812479734420777
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,8,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.08466560244560242
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.12144639492034912
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,64,240,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.13406720161437988
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,8,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.0963648021221161
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,2,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.12472959756851196
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,12,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.09933440089225769
TRTLLM,1.2.0rc6,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,0.061939197778701785
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,96,428,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.17208319902420044
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,3,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.1261952042579651
TRTLLM,1.2.0rc6,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,0.09978240132331848
TRTLLM,1.2.0rc6,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,0.0624064028263092
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,1,5,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11144319772720337
TRTLLM,1.2.0rc6,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,0.06610559821128845
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,15,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.10035840272903443
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,128,677,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.1893952012062073
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,2,9,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.09050880074501037
TRTLLM,1.2.0rc6,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,0.09073920249938965
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,23,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.10268160104751586
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,3,12,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11236480474472046
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,2,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.13144960403442382
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,32,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.1016767978668213
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,4,14,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.1001855969429016
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,192,815,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.20288000106811524
TRTLLM,1.2.0rc6,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,0.18378880023956298
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,40,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.10520319938659668
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,5,17,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11294080018997192
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,7,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.22383999824523926
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,6,20,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.10366719961166382
TRTLLM,1.2.0rc6,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,0.10264320373535156
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,256,1592,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.28209919929504396
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,8,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.14632320404052734
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,8,36,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.10414079427719117
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,64,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.1063423991203308
TRTLLM,1.2.0rc6,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,0.1564352035522461
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,10,43,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.09619200229644775
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,90,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.11111040115356445
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,12,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1543295979499817
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,12,52,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.10467840433120727
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,102,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.11301120519638061
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,384,2389,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.3145087957382202
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,18,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.14927999973297118
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,16,68,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11233919858932495
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,181,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.1175104022026062
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,18,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.15667200088500977
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,20,66,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11217279434204101
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,28,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1531775951385498
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,24,136,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.11719039678573609
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,233,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.12124799489974976
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,35,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1612223982810974
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,32,200,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.12825599908828736
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,512,2753,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.3951936006546021
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,390,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.141593599319458
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,44,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.16800639629364014
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,48,372,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.147871994972229
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,56,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.16286079883575438
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,559,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.15045119524002076
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,64,391,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.14307839870452882
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,70,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.1693567991256714
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,96,591,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.17953920364379883
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,868,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.17274240255355836
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,115,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.17347840070724488
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,768,4211,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.49560961723327634
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,153,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.17107839584350587
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,128,1111,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.20794239044189453
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,1126,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.20169599056243898
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,232,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.18714879751205443
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,192,1376,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.28368000984191893
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,293,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.19345279932022094
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,5341,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.6233791828155517
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,2004,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.25335040092468264
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,256,2597,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.2921472072601318
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,413,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.21392641067504883
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,384,2885,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.3625472068786621
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,2923,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.2648832082748413
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,643,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.23978879451751708
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,6373,7168,2048,8,256,256,1,16,power_law_1.01,accurate,0.8668607711791992
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,901,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.27942399978637694
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,512,5070,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.5200704097747803
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,4010,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.34364800453186034
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,1345,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.34070401191711425
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,768,6503,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.5716864109039307
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,6135,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.42595200538635253
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,11914,7168,2048,8,256,256,1,16,power_law_1.01,accurate,1.0534720420837402
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,1914,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.40692481994628904
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,8047,7168,2048,8,256,256,1,16,power_law_1.2,accurate,0.679532814025879
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,6556,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.5014656066894532
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,2776,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.4613503932952881
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,12097,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.0642815589904786
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,3437,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.5381951808929444
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,11082,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,0.7088255882263184
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,25801,7168,2048,8,256,256,1,16,power_law_1.01,accurate,2.4655424118041993
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,25605,7168,2048,8,256,256,1,16,power_law_1.2,accurate,1.5246335983276367
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,5256,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,0.7458687782287597
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,17638,7168,2048,8,256,256,1,16,power_law_1.01_eplb,accurate,1.261958408355713
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,42259,7168,2048,8,256,256,1,16,power_law_1.2,accurate,2.432153511047363
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,10458,7168,2048,8,256,288,1,16,power_law_1.01_eplb,accurate,1.3311424255371094
TRTLLM,1.2.0rc6,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,0.04679679870605469
TRTLLM,1.2.0rc6,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,0.0468095988035202
TRTLLM,1.2.0rc6,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,0.05225600004196167
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.06455680131912231
TRTLLM,1.2.0rc6,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,0.057760000228881836
TRTLLM,1.2.0rc6,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,0.07388799786567687
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,10,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.08981119990348815
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,12,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.09731839895248413
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,16,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.09374719858169556
TRTLLM,1.2.0rc6,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,0.09603840112686157
TRTLLM,1.2.0rc6,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,0.0990015983581543
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,40,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.09713919758796692
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,46,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.09902080297470092
TRTLLM,1.2.0rc6,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,0.1016319990158081
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,79,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.10684159994125367
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,113,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.10510079860687256
TRTLLM,1.2.0rc6,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,0.06170880198478699
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,179,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.11221760511398315
TRTLLM,1.2.0rc6,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,0.06376960277557372
TRTLLM,1.2.0rc6,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,0.06616960167884826
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,233,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.11710079908370971
TRTLLM,1.2.0rc6,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,0.09096320271492005
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,338,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.1313024044036865
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,1,3,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.13207039833068848
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.0797760009765625
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08136320114135742
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,2,4,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.18086400032043456
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,531,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.141484797000885
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08092799782752991
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,3,6,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.20663039684295653
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08521599769592285
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,4,7,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.21308159828186035
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,725,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.15358079671859742
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07398399710655212
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,5,11,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.14867839813232422
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,1,6,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07310720086097718
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,6,13,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1537343978881836
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,1,8,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07575680017471313
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,8,18,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.15082240104675293
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,192,1550,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.197324800491333
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,2,11,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07609599828720093
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,10,24,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1620736002922058
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,2,9,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08121600151062011
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,12,27,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1615679979324341
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,3,16,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07531520128250122
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,2265,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.18547199964523314
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,16,38,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.166975998878479
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,4,25,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.0775488018989563
TRTLLM,1.2.0rc6,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,0.04752640128135681
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,20,47,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1593791961669922
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,5,34,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.0802623987197876
TRTLLM,1.2.0rc6,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,0.04909439980983734
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,24,56,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.16524159908294678
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,6,33,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08241919875144958
TRTLLM,1.2.0rc6,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,0.04893440008163452
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,3290,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.2556096076965332
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,8,46,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08112639784812928
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,32,80,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1717311978340149
TRTLLM,1.2.0rc6,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,0.055238401889801024
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,10,65,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08081279993057251
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,48,115,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1789952039718628
TRTLLM,1.2.0rc6,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,0.05203199982643127
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,12,68,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.07916799783706666
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,6,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0546176016330719
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,64,169,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.17914880514144899
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,7,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0701632022857666
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,16,112,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.0890496015548706
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,3739,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.2790463924407959
TRTLLM,1.2.0rc6,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,0.07728639841079712
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,96,240,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.1932736039161682
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,24,154,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.08399999737739564
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,13,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07685120105743408
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,32,248,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.09983999729156494
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,16,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07365120053291321
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,128,325,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.20558719635009765
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,21,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0749888002872467
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,48,392,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.12041599750518799
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,24,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07725440263748169
TRTLLM,1.2.0rc6,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,0.2216639995574951
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,34,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.0716543972492218
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,64,476,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.14106240272521972
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,5188,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.36357760429382324
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,45,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07294719815254211
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,57,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07633280158042907
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,256,704,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.24697599411010743
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,96,832,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.17256959676742553
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,67,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07602559924125671
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,108,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07683200240135193
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,128,1045,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.19455360174179076
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,156,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.07965440154075623
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,384,999,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.2927808046340942
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,224,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.08017280101776122
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,10810,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.5543680191040039
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,331,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.09702399969100953
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,192,1723,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.24288640022277833
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,512,1349,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.3359935998916626
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,534,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.08254079818725586
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,735,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.1155135989189148
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,256,2453,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.28236799240112304
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,995,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.12538880109786987
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,12727,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,0.6299520015716553
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,768,2495,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.4015103816986084
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,384,3925,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.3423743963241577
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,1560,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.13547519445419312
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,1439,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.16253440380096434
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,1024,3093,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.4675136089324951
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,512,4467,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.39343359470367434
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,29872,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,1.2919808387756349
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,3123,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.18659199476242067
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,1280,3811,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.5532159805297852
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,640,5411,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.5980607986450195
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,4583,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.24887681007385254
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,6170,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.3084480047225952
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,10187,7168,2048,8,256,256,1,32,power_law_1.01,accurate,0.8264255523681641
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,2048,6329,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,0.7775872230529786
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,57316,7168,2048,8,256,256,1,16,power_law_1.2_eplb,accurate,2.333555221557617
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,10059,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.4677696228027344
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,17398,7168,2048,8,256,256,1,32,power_law_1.01,accurate,1.4009663581848144
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,4096,12528,7168,2048,8,256,288,1,16,power_law_1.2_eplb,accurate,1.373523235321045
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,18859,7168,2048,8,256,256,1,32,power_law_1.01_eplb,accurate,0.8305855751037597
TRTLLM,1.2.0rc6,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,0.05023360252380371
TRTLLM,1.2.0rc6,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,0.051545602083206174
TRTLLM,1.2.0rc6,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,0.05159040093421936
TRTLLM,1.2.0rc6,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,0.05922560095787048
TRTLLM,1.2.0rc6,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,0.05921279788017273
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,1,5,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.07538560032844543
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,4,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.07761279940605163
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,8,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.0794048011302948
TRTLLM,1.2.0rc6,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,0.07874559760093688
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,8,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08216320276260376
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,11,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08170239925384522
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,14,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08298879861831665
TRTLLM,1.2.0rc6,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,0.08196480274200439
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,25,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08650879859924317
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,39,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.08584319949150085
TRTLLM,1.2.0rc6,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,0.08243839740753174
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07939199805259704
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,53,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.0868287980556488
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08056960105895997
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,95,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.09228799939155578
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08132479786872863
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,3,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08072320222854615
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,113,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.09429759979248047
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,4,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07258880138397217
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,1,8,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07237759828567505
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,173,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.09818879961967468
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,1,7,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07599999904632568
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,2,12,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.0796671986579895
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,236,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.10645760297775268
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,2,18,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08258559703826904
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,3,18,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08278399705886841
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,381,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.11816320419311524
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,4,23,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08243200182914734
TRTLLM,1.2.0rc6,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,0.04736000001430511
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,5,32,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08200960159301758
TRTLLM,1.2.0rc6,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,0.04901759922504425
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,6,52,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07904000282287597
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,475,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.13267840147018434
TRTLLM,1.2.0rc6,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,0.04911360144615173
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,8,71,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.07930240035057068
TRTLLM,1.2.0rc6,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,0.05144320130348205
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,10,97,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08967040181159973
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,5,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.051046401262283325
TRTLLM,1.2.0rc6,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,0.051020801067352295
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,12,89,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.08104320168495179
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,724,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.1541759967803955
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,8,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0732159972190857
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,16,133,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.09303039908409119
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,2,9,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0732479989528656
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,24,259,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.10540159940719604
TRTLLM,1.2.0rc6,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,0.04992640018463135
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,14,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07365120053291321
TRTLLM,1.2.0rc6,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,0.05070719718933105
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,15,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07300480008125305
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,32,389,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.11025279760360718
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,996,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.17669119834899902
TRTLLM,1.2.0rc6,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,0.051052802801132204
TRTLLM,1.2.0rc6,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,0.0780672013759613
TRTLLM,1.2.0rc6,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,0.059596800804138185
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,48,646,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.14039679765701293
TRTLLM,1.2.0rc6,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,0.07206400036811829
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,2,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.060249602794647215
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,41,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07151359915733338
TRTLLM,1.2.0rc6,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,0.06958720088005066
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,64,809,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.15249279737472535
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,59,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0732800006866455
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,1,5,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.07919999957084656
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,10,74,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0673471987247467
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,1602,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.22557439804077148
TRTLLM,1.2.0rc6,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,0.08019199967384338
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,96,1458,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.18829439878463744
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,91,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06908159852027893
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,2,7,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08234239816665649
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,137,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.05873280167579651
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,3,10,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.0819263994693756
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,128,1840,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.21191680431365967
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,4,13,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.07959679961204529
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,220,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.06111360192298889
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,5,18,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08202239871025085
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,344,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.07546240091323853
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,6,20,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08241279721260071
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,192,2212,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.22329599857330323
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,2157,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.23256320953369142
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,615,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.0880128026008606
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,8,33,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.0822655975818634
TRTLLM,1.2.0rc6,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,0.08750079870223999
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,708,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.08822399973869324
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,12,43,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08227840065956116
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,256,2932,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.269651198387146
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,16,57,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.08813440203666686
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,1222,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.1422271966934204
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,24,98,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.09173120260238647
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,32,129,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.09534080028533935
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,2582,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.25770881175994875
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,2144,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.16090879440307618
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,48,172,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.10112639665603637
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,384,8387,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.5741568088531495
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,64,289,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.10881919860839843
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,3037,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.21210238933563233
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,96,367,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.11419520378112794
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,512,16384,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.804633617401123
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,4424,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.26439039707183837
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,128,576,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.13287680149078368
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,4263,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.3835135936737061
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,192,908,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.14083839654922486
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,6404,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.31857919692993164
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,640,13392,7168,2048,8,256,256,1,32,power_law_1.2,accurate,0.8515135765075683
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,256,1145,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.17885440587997437
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,11857,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.5591936111450195
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,384,1676,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.2184448003768921
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,19716,7168,2048,8,256,256,1,32,power_law_1.2,accurate,1.5647232055664062
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,12364,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,0.58122878074646
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,512,2299,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.24841599464416503
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,8647,7168,2048,8,256,288,1,32,power_law_1.01_eplb,accurate,0.6869056224822998
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,640,3011,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.284934401512146
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,30250,7168,2048,8,256,256,1,32,power_law_1.2_eplb,accurate,1.3128191947937011
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,55666,7168,2048,8,256,256,1,32,power_law_1.2,accurate,2.8023231506347654
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,1024,4732,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.3918207883834839
TRTLLM,1.2.0rc6,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,2.771308708190918
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,2048,9675,7168,2048,8,256,288,1,32,power_law_1.2_eplb,accurate,0.7156735897064209
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08017280101776122
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.080595201253891
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08101119995117187
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,1,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.08350719809532166
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,3,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.058387202024459836
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,7,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.055238401889801024
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,5,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05800960063934326
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,1,7,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05581439733505249
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,1,11,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05374079942703247
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,1,18,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.056992000341415404
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,2,18,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05384960174560547
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,2,21,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.057868802547454835
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,3,26,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05842559933662415
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,4,41,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05807999968528747
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,5,60,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.05617920160293579
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,6,71,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.0593280017375946
TRTLLM,1.2.0rc6,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,0.05691519975662231
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,8,92,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.06711680293083191
TRTLLM,1.2.0rc6,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,0.0578495979309082
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,12,166,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.06799359917640686
TRTLLM,1.2.0rc6,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,0.05809280276298523
TRTLLM,1.2.0rc6,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,0.05990399718284607
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,16,224,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.07936000227928161
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,16,2,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.04955520033836365
TRTLLM,1.2.0rc6,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,0.05009920001029968
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,24,370,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.09976959824562073
TRTLLM,1.2.0rc6,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,0.04986880123615265
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,32,471,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.11159039735794067
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,1,8,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.04801279902458191
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,1,11,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.04847359955310822
TRTLLM,1.2.0rc6,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,0.04850560128688812
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,48,760,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.12444800138473511
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,2,18,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.049081599712371825
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,2,24,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.04890879988670349
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,3,27,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.049491199851036075
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1168,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.1646399974822998
TRTLLM,1.2.0rc6,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,0.04953599870204926
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,5,53,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.0493120014667511
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,96,2128,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.20008320808410646
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,6,81,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05072000026702881
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08165760040283203
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,8,102,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05087360143661499
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.084307199716568
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,12,158,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.056467199325561525
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,4,1,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08362879753112792
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,8,2,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.053363198041915895
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,128,2874,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.2679935932159424
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,16,205,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.05847679972648621
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,16,4,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05139200091361999
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,24,339,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.07278720140457154
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,32,8,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.058931201696395874
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,48,8,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05975040197372437
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,1,11,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05739520192146301
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,32,549,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.0812991976737976
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,1,9,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05602560043334961
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,192,2834,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.28158719539642335
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,1,18,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.0574720025062561
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,48,741,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.0871999979019165
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,2,25,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05790719985961914
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,2,36,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.05966079831123352
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,3,45,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.061222398281097413
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,64,928,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.10151040554046631
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,4,70,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.062041598558425906
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,5,82,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.06710399985313416
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,6,106,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.06933760046958923
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,256,4400,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.363155198097229
TRTLLM,1.2.0rc6,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,0.05991680026054382
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,8,133,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.07219840288162231
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,96,1599,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.1354688048362732
TRTLLM,1.2.0rc6,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,0.061324799060821535
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,12,281,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.09444479942321778
TRTLLM,1.2.0rc6,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,0.060812801122665405
TRTLLM,1.2.0rc6,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,0.05111039876937866
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,16,337,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.08933759927749634
TRTLLM,1.2.0rc6,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,0.05121279954910278
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,128,2175,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.14485759735107423
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,24,572,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.11229439973831176
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,32,7,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05314559936523437
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,48,6,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05242239832878113
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,320,6144,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.33990399837493895
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,32,904,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.14151040315628052
TRTLLM,1.2.0rc6,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,0.050918400287628174
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,1,13,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05107839703559876
TRTLLM,1.2.0rc6,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,0.05089920163154602
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,48,1385,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.1697343945503235
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,2,23,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05177599787712097
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,192,4660,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.23853440284729005
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,2,29,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05187199711799621
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1944,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.23697280883789062
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,3,42,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05002880096435547
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,4,54,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.050886398553848265
TRTLLM,1.2.0rc6,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,0.05284479856491089
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,96,2875,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.2297215938568115
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,6,95,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.052876800298690796
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,256,4602,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.24876160621643068
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,8,168,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05824000239372253
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,12,243,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.05902079939842224
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,512,9798,7168,2048,8,256,256,1,64,power_law_1.01,accurate,0.7337088108062744
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,16,354,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.07240960001945496
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,128,4922,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.37246079444885255
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,24,558,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.08020480275154114
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,32,936,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.11501439809799194
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,320,6997,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.32344961166381836
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,192,9007,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.4925055980682373
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,48,1100,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.1094208002090454
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,64,1724,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.13987840414047242
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,256,11815,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.7049983978271485
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,96,3256,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.18428800106048585
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,512,9869,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.44520959854125974
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,128,4993,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.252512001991272
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,320,11582,7168,2048,8,256,256,1,64,power_law_1.2,accurate,0.78438401222229
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,19821,7168,2048,8,256,256,1,64,power_law_1.01,accurate,1.193395233154297
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,192,5639,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.2895679950714111
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,256,8474,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.39957759380340574
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,512,30234,7168,2048,8,256,256,1,64,power_law_1.2,accurate,1.3817279815673829
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,19410,7168,2048,8,256,256,1,64,power_law_1.01_eplb,accurate,0.8308927536010742
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,320,12508,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,0.5462912082672119
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,51893,7168,2048,8,256,256,1,64,power_law_1.2,accurate,3.2437759399414063
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,512,26573,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,1.1266688346862792
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,1024,55092,7168,2048,8,256,256,1,64,power_law_1.2_eplb,accurate,2.302931213378906
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.0407104
TRTLLM,1.2.0rc6,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,0.0395488
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.0578208
TRTLLM,1.2.0rc6,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,0.0245504
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.0483136
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.068096
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.0291296
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,2,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.055504
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,4,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.0811168
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1,1,1,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.0409888
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,1,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.039024
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.0720064
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.0762368
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.0576768
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,2,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.057008
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,5,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.0871168
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,13,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.0863264
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4,1,3,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.0597376
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,6,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.0674144
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,11,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.101501
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,17,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.116595
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8,2,4,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.0691232
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,12,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.0900512
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,15,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.123168
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,24,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.1108
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16,4,7,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.09184
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,17,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.10785
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,29,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.117734
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,21,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.13191
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32,8,16,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.110598
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,21,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.117107
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,40,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.123994
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,25,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.134682
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,48,12,19,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.109126
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,51,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.121734
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,31,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.119069
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,64,16,26,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.119958
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,32,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.140083
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,64,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.129555
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,80,20,31,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.140746
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,34,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.12104
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,39,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.140528
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,96,24,43,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.126019
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,77,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.123018
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,46,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.122221
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,52,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.141878
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,128,32,61,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.130774
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,117,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.150912
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,62,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.122893
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,62,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.142902
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,160,40,76,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.14839
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,74,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.128438
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,141,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.133494
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,85,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.151648
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,192,48,98,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.144272
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,103,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.13208
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,182,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.141878
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,100,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.147594
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,256,64,125,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.133434
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,127,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.13473
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,242,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.145558
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,124,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.150304
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,320,80,148,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.142246
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,151,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.134992
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,175,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.15409
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,382,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.161722
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,384,96,190,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.132762
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,206,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.141811
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,248,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.165923
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,450,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.158566
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,512,128,248,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.149658
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,317,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.154371
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,331,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.175443
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,683,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.188982
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,768,192,425,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.169494
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,420,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.161642
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,518,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.195059
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1024,256,634,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.184349
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1064,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.23087
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,679,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.183014
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,1536,384,1000,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.222826
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,708,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.220262
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1374,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.277952
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,919,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.210106
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1138,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.22593
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1034,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.263626
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2000,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.3332
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1487,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.258218
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,3072,768,2029,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.2676
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1355,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.307379
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2892,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.352122
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1882,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.297718
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,2667,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.324106
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2144,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.371747
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,4460,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.501408
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2941,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.333398
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,4829,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.416208
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,2876,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.44648
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3808,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.395901
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,5562,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.547206
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,5967,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.562858
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4163,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.543434
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,8867,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.624413
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,5938,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.537046
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8102,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.736253
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,5750,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.672058
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,12358,7168,2048,8,256,256,1,2,power_law_1.2,accurate,0.863629
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,8098,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.663248
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,10555,7168,2048,8,256,256,1,2,power_law_1.01,accurate,0.892288
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7044,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,0.818192
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,15792,7168,2048,8,256,256,1,2,power_law_1.2,accurate,1.34836
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,9847,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,0.793811
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,19251,7168,2048,8,256,256,1,2,power_law_1.01,accurate,1.59845
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,11475,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,1.21057
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,28055,7168,2048,8,256,256,1,2,power_law_1.2,accurate,2.15833
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,14949,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,1.20233
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,34170,7168,2048,8,256,256,1,2,power_law_1.01,accurate,2.59544
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,23348,7168,2048,8,256,288,1,2,power_law_1.01_eplb,accurate,2.25793
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,55174,7168,2048,8,256,256,1,2,power_law_1.2,accurate,3.83757
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,34823,7168,2048,8,256,256,1,2,power_law_1.01_eplb,accurate,2.38068
TRTLLM,1.2.0rc6,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,0.024992
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,1,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.0291488
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,1,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.0393344
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.052784
TRTLLM,1.2.0rc6,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,0.0397504
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,7,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.061664
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2,2,2,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.0489888
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,14,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.0810752
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4,1,2,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.0552256
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8,2,3,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.070752
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,18,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.109274
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16,4,4,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.0862624
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,24,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.112307
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32,8,11,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.100403
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,28,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.115805
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,48,12,16,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.125158
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,37,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.122995
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,64,16,20,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.136966
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,52,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.130496
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,80,20,25,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.135235
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,63,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.122858
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,96,24,30,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.137616
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,84,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.127706
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,128,32,44,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.139357
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,110,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.135008
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,160,40,54,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.141866
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,137,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.130752
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,192,48,62,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.152144
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,163,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.140934
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,256,64,85,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.152816
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,244,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.144083
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,320,80,109,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.151661
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,384,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.159917
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,384,96,130,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.163136
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,502,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.169706
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,512,128,178,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.163523
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,768,192,266,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.175318
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,881,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.195811
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1024,256,351,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.185872
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,1105,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.218038
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,1536,384,554,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.210019
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1726,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.260739
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,2048,512,728,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.226339
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,3072,768,1099,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.271747
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,3029,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.284461
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,4096,1024,1449,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.314749
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,3719,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.343971
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,6144,1536,2202,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.38369
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,4936,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.411402
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,8192,2048,3023,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.45537
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,8872,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.562058
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,12288,3072,4593,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.550493
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,10648,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.695629
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,16384,4096,6232,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.690432
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,7720,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,0.815418
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,20480,5120,12832,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,0.833754
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,12620,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,1.23398
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,32768,8192,24292,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,1.29766
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,26047,7168,2048,8,256,288,1,2,power_law_1.2_eplb,accurate,2.29507
TRTLLM,1.2.0rc6,NVIDIA GB200,wideep_moe_eplb,wideep_compute_cutlass,nvfp4,cutlass,65536,16384,65536,7168,2048,8,256,256,1,2,power_law_1.2_eplb,accurate,2.5417
