framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,128,0.006637120246887207,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,128,0.048793921470642085,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,256,0.00701312005519867,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,256,0.04898015975952148,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,512,0.007014080286026,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,512,0.05074719905853271,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,1024,0.007207679748535156,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,1024,0.04902048110961914,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,2048,0.006996799707412719,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,2048,0.04868031978607178,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,4096,0.007205759882926941,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,4096,0.04659711837768555,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,8192,0.007268480062484742,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,8192,0.046251840591430664,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,16384,0.007041599750518798,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,16384,0.04570432186126709,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,32768,0.006997119784355163,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,32768,0.04560736179351806,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,65536,0.007433279752731323,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,65536,0.04702911853790283,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,131072,0.007947520017623902,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,131072,0.04707808017730713,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,262144,0.009927679896354676,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,262144,0.0416534423828125,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,524288,0.01347391963005066,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,524288,0.03997056007385254,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,1048576,0.018888640403747558,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,1048576,0.05857664108276367,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,2097152,0.037976961135864254,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,2097152,0.05925951957702637,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,4194304,0.04393439769744873,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,4194304,0.06106592178344726,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,8388608,0.055706877708435056,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,8388608,0.06194464206695557,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,16777216,0.08460351943969727,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,16777216,0.09781439781188965,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,33554432,0.14841440200805664,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,33554432,0.17303007125854492,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,67108864,0.2705881690979004,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,67108864,0.3132691192626953,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,134217728,0.5226252746582032,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,134217728,0.6049465560913087,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,268435456,1.0058076477050781,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,268435456,1.1682252502441406,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,2,536870912,1.9053184509277343,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,2,536870912,2.223113555908203,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,128,0.007138239741325378,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,128,0.04279488086700439,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,256,0.007489280104637145,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,256,0.04287744045257568,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,512,0.007581440210342408,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,512,0.04183328151702881,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,1024,0.007634239792823791,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,1024,0.04359295845031738,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,2048,0.00745631992816925,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,2048,0.04283487796783447,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,4096,0.00779583990573883,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,4096,0.04277503967285156,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,8192,0.007888320088386535,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,8192,0.04135776042938232,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,16384,0.007647039890289306,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,16384,0.04289087772369385,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,32768,0.007751039862632751,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,32768,0.043157119750976566,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,65536,0.008074560165405273,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,65536,0.043067522048950195,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,131072,0.008767359852790832,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,131072,0.04480544090270996,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,262144,0.018601280450820924,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,262144,0.04255008220672608,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,524288,0.020959360599517824,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,524288,0.03945823907852173,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,1048576,0.02465440034866333,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,1048576,0.06505184173583985,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,2097152,0.03783584117889404,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,2097152,0.06339488029479981,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,4194304,0.051752319335937505,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,4194304,0.06282976150512695,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,8388608,0.08002047538757324,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,8388608,0.08047167778015137,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,16777216,0.1248265552520752,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,16777216,0.13834303855895996,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,33554432,0.1911795234680176,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,33554432,0.214898567199707,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,67108864,0.3641904067993164,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,67108864,0.41032161712646487,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,134217728,0.7059049224853515,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,134217728,0.7897500610351562,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,268435456,1.3534892272949217,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,268435456,1.5185218811035157,vllm_eager
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_graph,float16,4,536870912,2.555708465576172,vllm_graph
vLLM,0.14.0,NVIDIA GB200,all_reduce,vLLM_custom_eager,float16,4,536870912,2.9099075317382814,vllm_eager
