framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,128,0.007013760209083557,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,128,0.019170559644699096,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,256,0.006800959706306457,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,256,0.019581760168075564,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,512,0.006632000207901001,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,512,0.01942463994026184,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,1024,0.00689311981201172,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,1024,0.0194652795791626,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,2048,0.00700543999671936,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,2048,0.019569920301437376,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,4096,0.007033280134201051,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,4096,0.019466559886932373,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,8192,0.007407360076904297,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,8192,0.01900320053100586,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,16384,0.00738752007484436,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,16384,0.019113600254058838,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,32768,0.00755840003490448,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,32768,0.019038720130920412,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,65536,0.007771199941635132,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,65536,0.01898687958717346,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,131072,0.007784960269927979,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,131072,0.019159679412841794,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,262144,0.009767040014266967,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,262144,0.020740480422973634,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,524288,0.013642239570617675,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,524288,0.023895680904388428,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,1048576,0.02130687952041626,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,1048576,0.024384639263153075,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,2097152,0.040008959770202634,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,2097152,0.04133632183074951,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,4194304,0.045265278816223144,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,4194304,0.046650238037109375,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,8388608,0.057487998008728024,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,8388608,0.060143041610717776,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,16777216,0.08673407554626464,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,16777216,0.0970032024383545,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,33554432,0.14986944198608398,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,33554432,0.1768339157104492,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,67108864,0.28159423828125,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,67108864,0.3233520126342774,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,134217728,0.5246710586547851,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,134217728,0.6057980728149415,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,268435456,1.0069545745849608,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,268435456,1.1721971130371094,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,2,536870912,1.9088835144042968,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,2,536870912,2.234997100830078,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,128,0.0073363202810287475,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,128,0.020908479690551755,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,256,0.007025600075721741,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,256,0.02081568002700806,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,512,0.007041280269622803,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,512,0.020843520164489746,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,1024,0.007021759748458862,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,1024,0.02091808080673218,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,2048,0.007391999959945679,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,2048,0.02072927951812744,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,4096,0.0073398399353027345,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,4096,0.020483520030975342,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,8192,0.0074559998512268065,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,8192,0.019741760492324827,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,16384,0.007793599963188172,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,16384,0.019696960449218752,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,32768,0.007844480276107787,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,32768,0.019880959987640383,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,65536,0.00819104015827179,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,65536,0.01989567995071411,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,131072,0.008623039722442627,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,131072,0.019972800016403197,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,262144,0.018040000200271605,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,262144,0.02874783992767334,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,524288,0.018853440284729003,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,524288,0.030087680816650392,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,1048576,0.026670079231262207,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,1048576,0.029528319835662842,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,2097152,0.0402950382232666,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,2097152,0.0420527982711792,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,4194304,0.05375167846679687,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,4194304,0.05162176132202149,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,8388608,0.0826255989074707,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,8388608,0.08047616004943847,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,16777216,0.12346816062927246,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,16777216,0.13473376274108886,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,33554432,0.18567968368530274,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,33554432,0.2081376075744629,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,67108864,0.3530521774291992,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,67108864,0.3960700988769531,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,134217728,0.6779618835449218,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,134217728,0.757260513305664,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,268435456,1.2852694702148437,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,268435456,1.452012481689453,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,4,536870912,2.5099673461914063,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,4,536870912,2.835035705566406,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,128,0.007216960191726685,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,128,0.02138175964355469,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,256,0.007213119864463806,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,256,0.020857601165771483,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,512,0.007048640251159669,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,512,0.020320000648498534,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,1024,0.007006080150604247,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,1024,0.021037759780883788,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,2048,0.0073446398973464975,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,2048,0.021240639686584472,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,4096,0.0074259197711944586,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,4096,0.020697600841522217,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,8192,0.007790399789810181,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,8192,0.02000511884689331,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,16384,0.009348160028457642,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,16384,0.01992032051086426,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,32768,0.018261760473251343,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,32768,0.019980159997940065,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,65536,0.009409599900245666,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,65536,0.02016160011291504,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,131072,0.025235838890075683,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,131072,0.033606400489807134,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,262144,0.025121920108795166,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,262144,0.03586944103240967,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,524288,0.03069439888000488,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,524288,0.033206079006195065,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,1048576,0.034475200176239014,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,1048576,0.03747136116027832,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,2097152,0.05691423892974854,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,2097152,0.05789408206939697,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,4194304,0.07906879901885985,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,4194304,0.07521728038787842,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,8388608,0.10941887855529786,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,8388608,0.10652000427246093,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,16777216,0.17018848419189453,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,16777216,0.17664031982421874,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,33554432,0.2767945671081543,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,33554432,0.28910432815551756,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,67108864,0.3968438339233399,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,67108864,0.4234640121459961,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,134217728,0.7140930938720703,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,134217728,0.7850028991699218,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,268435456,1.3381135559082031,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,268435456,1.5002368164062498,vllm_eager
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_graph,float16,8,536870912,2.604142761230469,vllm_graph
vLLM,0.17.0,NVIDIA B200,all_reduce,vLLM_custom_eager,float16,8,536870912,2.931637878417969,vllm_eager
