framework,version,device,op_name,kernel_source,batch_size,isl,num_heads,num_key_value_heads,head_dim,beam_width,attn_dtype,kv_cache_dtype,step,latency
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,8,128,1,float16,fp8,0,22.973108927408855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,4,128,1,float16,float16,0,30.607200622558594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,1,128,1,float16,fp8,0,22.565589904785156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,1,128,1,float16,float16,0,30.48017120361328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,2,128,1,float16,float16,0,29.990806579589844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,1,128,1,float16,fp8,0,11.40298080444336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,8,128,1,float16,float16,0,30.79649607340495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,64,128,1,float16,fp8,0,12.576160430908203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,64,128,1,float16,float16,0,16.73804219563802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,2,128,1,float16,float16,0,15.568740844726562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,1,128,1,float16,float16,0,15.840896606445312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,2,128,1,float16,fp8,0,22.9979248046875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,64,4,128,1,float16,fp8,0,22.916030883789062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,2,128,1,float16,fp8,0,11.613455454508463
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,4,128,1,float16,fp8,0,11.569466908772787
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,4,128,1,float16,float16,0,16.059429168701172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,8,128,1,float16,float16,0,15.560128529866537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,1,128,1,float16,fp8,0,5.569264094034831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,1,128,1,float16,float16,0,7.2917226155598955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,64,128,1,float16,float16,0,7.77238400777181
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,64,8,128,1,float16,fp8,0,11.633466084798178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,64,128,1,float16,fp8,0,6.267050425211589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,2,128,1,float16,float16,0,7.536687850952148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,2,128,1,float16,fp8,0,5.879178365071614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,4,128,1,float16,fp8,0,5.884112040201823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,1,128,1,float16,float16,0,3.8767786026000977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,4,128,1,float16,float16,0,7.62070910135905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,64,128,1,float16,fp8,0,3.2349440256754556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,64,128,1,float16,float16,0,3.7176478703816733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,8,128,1,float16,fp8,0,5.83787218729655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,1,128,1,float16,fp8,0,2.8474133809407554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,64,8,128,1,float16,float16,0,7.977104187011719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,2,128,1,float16,fp8,0,3.133461316426595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,4,128,1,float16,float16,0,3.584869384765625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,8,128,1,float16,fp8,0,2.9349387486775718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,8,128,1,float16,float16,0,3.490917205810547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,2,128,1,float16,float16,0,3.529621442159017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,64,4,128,1,float16,fp8,0,2.8758986790974936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,1,128,1,float16,fp8,0,12.943098704020182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,1,128,1,float16,float16,0,17.715323130289715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,2,128,1,float16,fp8,0,13.16213353474935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,4,128,1,float16,fp8,0,13.47104008992513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,2,128,1,float16,float16,0,17.778981526692707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,4,128,1,float16,float16,0,18.117589314778645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,8,128,1,float16,fp8,0,13.478127797444662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,1,128,1,float16,fp8,0,6.443136215209961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,1,128,1,float16,float16,0,8.861279805501303
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,64,8,128,1,float16,float16,0,17.783109029134113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,64,128,1,float16,float16,0,9.165248235066732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,64,128,1,float16,fp8,0,7.484762827555339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,2,128,1,float16,float16,0,8.728869120279947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,2,128,1,float16,fp8,0,6.697349548339844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,4,128,1,float16,fp8,0,6.693946838378906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,1,128,1,float16,float16,0,4.3258771896362305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,1,128,1,float16,fp8,0,3.238389333089193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,4,128,1,float16,float16,0,8.852218627929688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,64,128,1,float16,fp8,0,3.7907307942708335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,8,128,1,float16,fp8,0,6.748832066853841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,64,8,128,1,float16,float16,0,9.170389175415039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,64,128,1,float16,float16,0,4.3840532302856445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,2,128,1,float16,float16,0,3.9298505783081055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,2,128,1,float16,fp8,0,3.6059414545694985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,4,128,1,float16,float16,0,4.148581186930339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,4,128,1,float16,fp8,0,3.9489545822143555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,8,128,1,float16,float16,0,4.000906626383464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,64,8,128,1,float16,fp8,0,3.429391860961914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,1,128,1,float16,float16,0,2.1889440218607583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,1,128,1,float16,fp8,0,1.7243626912434895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,64,128,1,float16,fp8,0,2.0419947306315103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,2,128,1,float16,fp8,0,1.7060906092325847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,64,128,1,float16,float16,0,2.227663993835449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,2,128,1,float16,float16,0,2.3908692995707193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,4,128,1,float16,fp8,0,1.7002719243367512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,4,128,1,float16,float16,0,2.1319252649943032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,8,128,1,float16,float16,0,2.163952032725016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,64,8,128,1,float16,fp8,0,1.7179412841796875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,1,128,1,float16,fp8,0,9.261807759602865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,2,128,1,float16,fp8,0,9.084810892740885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,1,128,1,float16,float16,0,12.343755086263021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,4,128,1,float16,fp8,0,9.488992055257162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,8,128,1,float16,fp8,0,9.483311971028646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,2,128,1,float16,float16,0,12.42722193400065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,4,128,1,float16,float16,0,13.005855560302734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,64,8,128,1,float16,float16,0,12.80075200398763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,1,128,1,float16,fp8,0,4.556645393371582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,1,128,1,float16,float16,0,5.896965026855469
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,2,128,1,float16,float16,0,5.781040191650391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,64,128,1,float16,float16,0,6.446298599243164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,64,128,1,float16,fp8,0,5.431994756062825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,4,128,1,float16,float16,0,5.9785919189453125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,4,128,1,float16,fp8,0,4.864378611246745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,8,128,1,float16,float16,0,6.338015874226888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,1,128,1,float16,float16,0,2.791136105855306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,1,128,1,float16,fp8,0,2.6770401000976562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,2,128,1,float16,float16,0,2.7920799255371094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,8,128,1,float16,fp8,0,4.978384017944336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,64,128,1,float16,float16,0,3.1461973190307617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,64,128,1,float16,fp8,0,2.886101404825846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,4,128,1,float16,fp8,0,2.4282933870951333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,4,128,1,float16,float16,0,2.8638881047566733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,64,2,128,1,float16,fp8,0,4.753061294555664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,8,128,1,float16,fp8,0,2.402538617451986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,8,128,1,float16,float16,0,2.9021708170572915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,1,128,1,float16,float16,0,1.431615988413493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,64,128,1,float16,fp8,0,1.4482986132303874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,1,128,1,float16,fp8,0,1.1743786334991455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,2,128,1,float16,float16,0,1.4447520573933919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,2,128,1,float16,fp8,0,1.404538631439209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,64,128,1,float16,float16,0,1.6057599385579426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,4,128,1,float16,float16,0,1.4739413261413574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,4,128,1,float16,fp8,0,1.218506654103597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,8,128,1,float16,fp8,0,1.225221316019694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,64,8,128,1,float16,float16,0,1.4582187334696453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,64,2,128,1,float16,fp8,0,3.0350774129231772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,1,128,1,float16,fp8,0,11.923802693684896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,2,128,1,float16,fp8,0,12.190170288085938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,1,128,1,float16,float16,0,16.14674123128255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,4,128,1,float16,fp8,0,12.427898406982422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,2,128,1,float16,float16,0,16.3712158203125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,4,128,1,float16,float16,0,16.627066294352215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,8,128,1,float16,float16,0,16.56395721435547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,1,128,1,float16,fp8,0,5.788981119791667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,64,8,128,1,float16,fp8,0,12.466932932535807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,1,128,1,float16,float16,0,7.907205581665039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,64,128,1,float16,float16,0,8.808479944864908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,64,128,1,float16,fp8,0,7.437466939290364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,2,128,1,float16,fp8,0,6.15666135152181
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,2,128,1,float16,float16,0,7.969962437947591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,4,128,1,float16,fp8,0,6.1822509765625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,1,128,1,float16,float16,0,3.6823787689208984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,4,128,1,float16,float16,0,7.777605056762695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,1,128,1,float16,fp8,0,3.4440320332845054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,64,128,1,float16,fp8,0,3.717482566833496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,8,128,1,float16,float16,0,8.080656051635742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,64,8,128,1,float16,fp8,0,6.270293553670247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,64,128,1,float16,float16,0,4.177157402038574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,2,128,1,float16,float16,0,3.7412052154541016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,2,128,1,float16,fp8,0,3.4408801396687827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,4,128,1,float16,fp8,0,3.2597386042277017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,4,128,1,float16,float16,0,3.73582394917806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,8,128,1,float16,float16,0,3.7412586212158203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,1,128,1,float16,fp8,0,1.6017066637674968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,1,128,1,float16,float16,0,1.9759467442830403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,64,8,128,1,float16,fp8,0,3.2679465611775718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,64,128,1,float16,float16,0,2.0905332565307617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,2,128,1,float16,float16,0,1.864730676015218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,64,128,1,float16,fp8,0,1.9352693557739258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,4,128,1,float16,fp8,0,1.5935413042704265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,4,128,1,float16,float16,0,2.1745173136393228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,8,128,1,float16,float16,0,1.9490399360656738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,8,128,1,float16,fp8,0,1.5923360188802083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,1,128,1,float16,fp8,0,0.7987146377563477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,2,128,1,float16,float16,0,1.0050559838612874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,2,128,1,float16,fp8,0,0.8491893609364828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,4,128,1,float16,float16,0,0.9695573647816976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,64,128,1,float16,fp8,0,0.9839146931966146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,64,128,1,float16,float16,0,1.0750986735026042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,8,128,1,float16,fp8,0,0.8739093144734701
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,8,128,1,float16,float16,0,0.9815200169881185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,64,2,128,1,float16,fp8,0,1.7522400220235188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,1,128,1,float16,float16,0,0.9453012943267822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,1,128,1,float16,fp8,0,6.829792022705078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,64,4,128,1,float16,fp8,0,0.8214080333709717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,1,128,1,float16,float16,0,9.517199834187826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,2,128,1,float16,fp8,0,6.871349334716797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,2,128,1,float16,float16,0,9.107690811157227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,4,128,1,float16,float16,0,9.286287943522135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,4,128,1,float16,fp8,0,7.221743901570638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,1,128,1,float16,float16,0,4.375712076822917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,1,128,1,float16,fp8,0,3.612127939860026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,2,128,1,float16,float16,0,4.273578643798828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,8,128,1,float16,fp8,0,7.323594411214192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,64,128,1,float16,float16,0,5.046677271525065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,2,128,1,float16,fp8,0,3.4811573028564453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,64,8,128,1,float16,float16,0,9.588250478108725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,64,128,1,float16,fp8,0,4.6888853708903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,4,128,1,float16,float16,0,4.958762804667155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,4,128,1,float16,fp8,0,3.635557174682617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,1,128,1,float16,float16,0,2.4040959676106772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,1,128,1,float16,fp8,0,1.7320747375488281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,8,128,1,float16,fp8,0,3.6715094248453775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,64,128,1,float16,fp8,0,2.320730686187744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,64,8,128,1,float16,float16,0,4.523514747619629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,2,128,1,float16,fp8,0,1.7705814043680828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,2,128,1,float16,float16,0,2.119061311086019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,64,128,1,float16,float16,0,2.5383946100870767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,4,128,1,float16,fp8,0,1.8332746823628743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,4,128,1,float16,float16,0,2.194218635559082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,8,128,1,float16,float16,0,2.2184267044067383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,1,128,1,float16,float16,0,1.0713919798533122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,2,128,1,float16,float16,0,1.0858240127563477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,64,8,128,1,float16,fp8,0,1.8416639963785808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,2,128,1,float16,fp8,0,1.0792853037516277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,4,128,1,float16,fp8,0,0.9631253083546957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,4,128,1,float16,float16,0,1.1563146909077961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,64,128,1,float16,fp8,0,1.1885120073954265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,64,128,1,float16,float16,0,1.3887839317321777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,8,128,1,float16,float16,0,1.11844801902771
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,8,128,1,float16,fp8,0,0.9513866901397705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,1,128,1,float16,float16,0,0.5630079905192057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,1,128,1,float16,fp8,0,0.4875839948654175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,4,128,1,float16,float16,0,0.5737653175989786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,64,128,1,float16,fp8,0,0.6244639952977499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,2,128,1,float16,fp8,0,0.48606932163238525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,2,128,1,float16,float16,0,0.5653813282648722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,64,128,1,float16,float16,0,0.6628640095392863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,4,128,1,float16,fp8,0,0.508629322052002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,8,128,1,float16,float16,0,0.5751093228658041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,64,1,128,1,float16,fp8,0,0.9008533159891764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,1,128,1,float16,fp8,0,6.396634419759114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,2,128,1,float16,fp8,0,6.537888209025065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,64,8,128,1,float16,fp8,0,0.5048853158950806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,1,128,1,float16,float16,0,8.568437576293945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,2,128,1,float16,float16,0,8.79363759358724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,4,128,1,float16,float16,0,8.886005401611328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,4,128,1,float16,fp8,0,6.922234853108724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,1,128,1,float16,fp8,0,3.623413403828939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,1,128,1,float16,float16,0,4.180549303690593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,8,128,1,float16,float16,0,9.222965240478516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,2,128,1,float16,float16,0,4.078271865844727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,64,128,1,float16,fp8,0,4.771562576293945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,64,8,128,1,float16,fp8,0,7.05793571472168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,64,128,1,float16,float16,0,5.027946790059407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,4,128,1,float16,float16,0,4.181557337443034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,4,128,1,float16,fp8,0,3.6874825159708657
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,8,128,1,float16,fp8,0,3.7600425084431968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,1,128,1,float16,float16,0,1.9475679397583008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,8,128,1,float16,float16,0,4.196506818135579
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,1,128,1,float16,fp8,0,1.8327946662902832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,2,128,1,float16,float16,0,2.014613310496012
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,64,128,1,float16,fp8,0,2.42740805943807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,64,128,1,float16,float16,0,2.523541291554769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,2,128,1,float16,fp8,0,1.6993227005004883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,64,2,128,1,float16,fp8,0,3.299861272176107
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,4,128,1,float16,float16,0,2.0684693654378257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,4,128,1,float16,fp8,0,2.075594743092855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,1,128,1,float16,float16,0,0.9960693518320719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,8,128,1,float16,fp8,0,1.7871947288513184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,64,8,128,1,float16,float16,0,2.097503980000814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,1,128,1,float16,fp8,0,0.8391466935475668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,2,128,1,float16,float16,0,1.0100053151448567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,2,128,1,float16,fp8,0,0.9571093718210856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,64,128,1,float16,fp8,0,1.2277759710947673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,4,128,1,float16,fp8,0,0.9066987037658691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,4,128,1,float16,float16,0,1.051146666208903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,8,128,1,float16,float16,0,1.0653759638468425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,64,128,1,float16,float16,0,1.275712013244629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,64,8,128,1,float16,fp8,0,0.913978656133016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,2,128,1,float16,fp8,0,0.4648053248723348
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,1,128,1,float16,float16,0,0.5348853270212809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,2,128,1,float16,float16,0,0.5160319805145264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,4,128,1,float16,fp8,0,0.4757759968439738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,1,128,1,float16,fp8,0,0.4453493356704712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,4,128,1,float16,float16,0,0.5412160158157349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,64,128,1,float16,fp8,0,0.6432799895604452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,1,128,1,float16,float16,0,0.2768373290697734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,8,128,1,float16,fp8,0,0.4805599848429362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,8,128,1,float16,float16,0,0.5485386848449707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,64,64,128,1,float16,float16,0,0.6501173178354899
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,1,128,1,float16,fp8,0,0.2456106742223104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,2,128,1,float16,float16,0,0.2815626660982768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,2,128,1,float16,fp8,0,0.24925865729649863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,4,128,1,float16,float16,0,0.28674666086832684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,4,128,1,float16,fp8,0,0.2634879946708679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,8,128,1,float16,fp8,0,0.2640533248583476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,64,128,1,float16,fp8,0,0.3433973391850789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,2,128,1,float16,fp8,0,3.942655881245931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,1,128,1,float16,fp8,0,3.843621253967285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,2,128,1,float16,float16,0,4.727909406026204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,4,128,1,float16,float16,0,5.132543881734212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,4,128,1,float16,fp8,0,4.235439936319987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,8,128,1,float16,float16,0,0.28945066531499225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,64,64,128,1,float16,float16,0,0.3760319948196411
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,1,128,1,float16,float16,0,2.35424534479777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,1,128,1,float16,fp8,0,1.9390133221944172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,8,128,1,float16,float16,0,5.121040026346843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,8,128,1,float16,fp8,0,4.301194508870442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,2,128,1,float16,float16,0,2.368325392405192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,2,128,1,float16,fp8,0,2.08951473236084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,64,128,1,float16,float16,0,3.1725546518961587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,64,128,1,float16,fp8,0,3.2041333516438804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,4,128,1,float16,float16,0,2.506506601969401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,4,128,1,float16,fp8,0,2.5808960596720376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,8,128,1,float16,float16,0,2.5260960261027017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,64,8,128,1,float16,fp8,0,2.286736011505127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,1,128,1,float16,float16,0,1.1817386945088704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,1,128,1,float16,fp8,0,1.0825599829355876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,2,128,1,float16,float16,0,1.1901439825693767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,2,128,1,float16,fp8,0,1.0539147059122722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,4,128,1,float16,float16,0,1.2550666332244873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,64,128,1,float16,fp8,0,1.5935626029968262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,4,128,1,float16,fp8,0,1.1348640124003093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,8,128,1,float16,float16,0,1.2703200181325276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,8,128,1,float16,fp8,0,1.1552800337473552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,1,128,1,float16,fp8,0,0.525269349416097
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,2,128,1,float16,float16,0,0.6138399839401245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,1,128,1,float16,float16,0,0.6208853324254354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,4,128,1,float16,fp8,0,0.5948533217112223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,4,128,1,float16,float16,0,0.6420160134633383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,64,128,1,float16,fp8,0,0.8189173539479574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,64,1,128,1,float16,float16,0,4.6886240641276045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,8,128,1,float16,float16,0,0.6402133305867513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,8,128,1,float16,fp8,0,0.5754293203353882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,1,128,1,float16,float16,0,0.3105173309644063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,2,128,1,float16,float16,0,0.31626667579015094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,1,128,1,float16,fp8,0,0.28123199939727783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,64,128,1,float16,float16,0,0.421125332514445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,2,128,1,float16,fp8,0,0.2853386600812276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,4,128,1,float16,float16,0,0.3328426678975423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,64,128,1,float16,fp8,0,0.42180267969767254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,64,64,128,1,float16,float16,0,1.6025439898173015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,4,128,1,float16,fp8,0,0.30879465738932294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,8,128,1,float16,float16,0,0.33508265018463135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,64,8,128,1,float16,fp8,0,0.30585600932439166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,1,128,1,float16,float16,0,0.17193599541982016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,1,128,1,float16,fp8,0,0.1564959983030955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,2,128,1,float16,fp8,0,0.5319199959437052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,2,128,1,float16,fp8,0,0.16033066312472025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,4,128,1,float16,float16,0,0.1820746660232544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,64,64,128,1,float16,float16,0,0.8423199653625488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,64,128,1,float16,float16,0,0.24943466981252035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,8,128,1,float16,fp8,0,0.16894932587941489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,8,128,1,float16,float16,0,0.1853813330332438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,1,128,1,float16,fp8,0,3.8022025426228843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,1,128,1,float16,float16,0,4.718496004740397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,2,128,1,float16,fp8,0,3.9507360458374023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,4,128,1,float16,fp8,0,0.1683680017789205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,2,128,1,float16,float16,0,4.708805402119954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,4,128,1,float16,float16,0,5.03985055287679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,4,128,1,float16,fp8,0,4.320810635884603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,8,128,1,float16,fp8,0,4.405104001363118
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,2,128,1,float16,float16,0,0.17449599504470825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,64,8,128,1,float16,float16,0,5.190863927205403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,64,128,1,float16,float16,0,3.390080134073893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,1,128,1,float16,float16,0,2.2769920031229653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,1,128,1,float16,fp8,0,1.9177974065144856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,64,128,1,float16,fp8,0,3.5050878524780273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,2,128,1,float16,float16,0,2.349952061971029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,4,128,1,float16,float16,0,2.5322559674580893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,2,128,1,float16,fp8,0,1.9875307083129883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,4,128,1,float16,fp8,0,2.3695680300394693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,8,128,1,float16,float16,0,2.522101402282715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,64,8,128,1,float16,fp8,0,2.2180959383646646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,1,128,1,float16,float16,0,1.16102933883667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,1,128,1,float16,fp8,0,0.9738453229268392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,4,128,1,float16,float16,0,1.2710506916046143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,2,128,1,float16,float16,0,1.172053337097168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,64,128,1,float16,float16,0,1.7028746604919434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,2,128,1,float16,fp8,0,1.015455961227417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,4,128,1,float16,fp8,0,1.158357302347819
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,64,64,128,1,float16,fp8,0,0.23106666405995688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,64,128,1,float16,fp8,0,1.767562707265218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,8,128,1,float16,float16,0,1.271456003189087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,1,128,1,float16,float16,0,0.5801440080006918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,1,128,1,float16,fp8,0,0.5190133253733317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,2,128,1,float16,float16,0,0.5965280135472616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,2,128,1,float16,fp8,0,0.5226186513900757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,4,128,1,float16,float16,0,0.6349440018335978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,4,128,1,float16,fp8,0,0.566490650177002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,8,128,1,float16,float16,0,0.6445279916127523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,64,128,1,float16,fp8,0,0.8882719675699869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,8,128,1,float16,fp8,0,0.5761760075887045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,64,64,128,1,float16,float16,0,0.8530399799346924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,1,128,1,float16,float16,0,0.2983199954032898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,2,128,1,float16,fp8,0,0.27848533789316815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,1,128,1,float16,fp8,0,0.26983465751012164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,2,128,1,float16,float16,0,0.3086293339729309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,4,128,1,float16,fp8,0,0.3013919989267985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,64,128,1,float16,fp8,0,0.4625920057296753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,8,128,1,float16,float16,0,0.33345067501068115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,8,128,1,float16,fp8,0,0.3081120053927104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,64,128,1,float16,float16,0,0.43792001406351727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,2,128,1,float16,float16,0,0.1704746683438619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,64,8,128,1,float16,fp8,0,1.1389973163604736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,2,128,1,float16,fp8,0,0.15338666240374246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,4,128,1,float16,fp8,0,0.16426133116086325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,64,128,1,float16,fp8,0,0.24633600314458212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,8,128,1,float16,float16,0,0.18041066328684488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,8,128,1,float16,fp8,0,0.16772266228993735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,1,128,1,float16,float16,0,0.09463999668757121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,1,128,1,float16,fp8,0,0.08714666962623596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,2,128,1,float16,float16,0,0.09530666470527649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,2,128,1,float16,fp8,0,0.0897653301556905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,64,128,1,float16,float16,0,0.14248533050219217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,4,128,1,float16,fp8,0,0.09314133723576863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,8,128,1,float16,float16,0,0.09942400455474854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,8,128,1,float16,fp8,0,0.09459732969601949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,4,128,1,float16,float16,0,0.17983466386795044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,64,128,1,float16,float16,0,0.23361066977183023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,64,4,128,1,float16,float16,0,0.3381706476211548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,64,128,1,float16,fp8,0,0.14283733566602072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,64,4,128,1,float16,float16,0,0.10053867101669312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,1,128,1,float16,fp8,0,0.1499786674976349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,1,128,1,float16,float16,0,2.8089866638183594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,1,128,1,float16,fp8,0,2.3728052775065103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,2,128,1,float16,float16,0,2.898751894632975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,4,128,1,float16,float16,0,3.1601813634236655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,4,128,1,float16,fp8,0,2.7613652547200522
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,8,128,1,float16,fp8,0,2.9150721232096353
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,1,128,1,float16,float16,0,1.4418400128682454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,1,128,1,float16,fp8,0,1.1956693331400554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,2,128,1,float16,float16,0,1.4503092765808105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,64,1,128,1,float16,float16,0,0.16741333405176798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,64,128,1,float16,fp8,0,2.3904693921407065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,64,128,1,float16,float16,0,2.2542452812194824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,2,128,1,float16,fp8,0,1.2528800169626872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,2,128,1,float16,fp8,0,2.4776479403177896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,4,128,1,float16,float16,0,1.5736427307128906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,4,128,1,float16,fp8,0,1.3959627151489258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,1,128,1,float16,float16,0,0.7096906503041586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,8,128,1,float16,fp8,0,1.4230720202128093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,64,8,128,1,float16,float16,0,1.5907306671142578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,1,128,1,float16,fp8,0,0.616592009862264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,64,8,128,1,float16,float16,0,3.2002080281575522
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,2,128,1,float16,float16,0,0.7267733414967855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,4,128,1,float16,float16,0,0.7887360254923502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,2,128,1,float16,fp8,0,0.6412586768468221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,64,128,1,float16,float16,0,1.1295466423034668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,4,128,1,float16,fp8,0,0.7105813026428223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,64,128,1,float16,fp8,0,1.1910773118336995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,1,128,1,float16,fp8,0,0.32391466697057086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,8,128,1,float16,fp8,0,0.7216426531473795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,2,128,1,float16,float16,0,0.37398401896158856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,1,128,1,float16,float16,0,0.36483200391133624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,64,8,128,1,float16,float16,0,0.803546667098999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,4,128,1,float16,fp8,0,0.3692213296890259
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,2,128,1,float16,fp8,0,0.33982932567596436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,64,128,1,float16,float16,0,0.5671626726786295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,1,128,1,float16,float16,0,0.19359999895095825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,8,128,1,float16,float16,0,0.4087573289871216
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,64,128,1,float16,fp8,0,0.609221339225769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,8,128,1,float16,fp8,0,0.37675201892852783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,2,128,1,float16,fp8,0,0.18402665853500366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,1,128,1,float16,fp8,0,0.1770346760749817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,4,128,1,float16,fp8,0,0.20107734203338623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,8,128,1,float16,fp8,0,0.20246932903925577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,8,128,1,float16,float16,0,0.2208319902420044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,64,128,1,float16,float16,0,0.3412426710128784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,1,128,1,float16,float16,0,0.10984533031781514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,2,128,1,float16,float16,0,0.11195733149846394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,1,128,1,float16,fp8,0,0.09995733698209126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,2,128,1,float16,fp8,0,0.10148266951243083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,64,128,1,float16,float16,0,0.18311999241511026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,64,128,1,float16,fp8,0,0.18058133125305176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,4,128,1,float16,float16,0,0.12055466572443645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,4,128,1,float16,fp8,0,0.1072106659412384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,8,128,1,float16,fp8,0,0.11001066366831462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,1,128,1,float16,float16,0,0.0660693347454071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,64,4,128,1,float16,float16,0,0.40486399332682294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,1,128,1,float16,fp8,0,0.05991466840108236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,64,128,1,float16,fp8,0,0.10041600465774536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,64,128,1,float16,float16,0,0.10480533043543498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,2,128,1,float16,float16,0,0.20081067085266113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,2,128,1,float16,fp8,0,0.06126399834950765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,2,128,1,float16,float16,0,0.06649066507816315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,64,128,1,float16,fp8,0,0.3187733292579651
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,4,128,1,float16,float16,0,0.06894400219122569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,4,128,1,float16,fp8,0,0.06559999783833821
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,8,128,1,float16,fp8,0,0.06617600222428639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,64,8,128,1,float16,float16,0,0.06946666538715363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,64,8,128,1,float16,float16,0,0.12156266967455547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,1,128,1,float16,fp8,0,2.495141347249349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,2,128,1,float16,fp8,0,2.6485652923583984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,2,128,1,float16,float16,0,3.0376691818237305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,1,128,1,float16,float16,0,2.9400691986083984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,64,4,128,1,float16,float16,0,0.21876800060272217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,4,128,1,float16,fp8,0,3.020234743754069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,4,128,1,float16,float16,0,3.3944854736328125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,8,128,1,float16,float16,0,3.466597239176432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,1,128,1,float16,fp8,0,1.2650026480356853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,1,128,1,float16,float16,0,1.4986027081807454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,64,8,128,1,float16,fp8,0,3.1039838790893555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,8,128,1,float16,fp8,0,1.5569173494974773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,8,128,1,float16,float16,0,1.719919999440511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,4,128,1,float16,fp8,0,1.526261329650879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,4,128,1,float16,float16,0,1.7218400637308757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,2,128,1,float16,fp8,0,1.3281066417694092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,64,128,1,float16,float16,0,2.5959946314493814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,64,128,1,float16,fp8,0,2.7961654663085938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,2,128,1,float16,float16,0,0.7656640211741129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,1,128,1,float16,float16,0,0.7411466439565023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,1,128,1,float16,fp8,0,0.6506453355153402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,2,128,1,float16,fp8,0,0.6835306485493978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,64,128,1,float16,float16,0,1.290170669555664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,4,128,1,float16,float16,0,0.8474720319112142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,8,128,1,float16,fp8,0,0.7902666727701823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,64,128,1,float16,fp8,0,1.4097867012023926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,8,128,1,float16,float16,0,0.8613279660542806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,64,4,128,1,float16,fp8,0,0.7725706895192465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,1,128,1,float16,float16,0,0.3770506779352824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,1,128,1,float16,fp8,0,0.33177600304285687
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,2,128,1,float16,fp8,0,0.35148266951243085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,4,128,1,float16,float16,0,0.4310506582260132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,64,128,1,float16,fp8,0,0.7156906922658285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,4,128,1,float16,fp8,0,0.403653343518575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,2,128,1,float16,float16,0,0.3914773464202881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,64,2,128,1,float16,float16,0,1.5321332613627117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,64,128,1,float16,float16,0,0.6512266794840494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,8,128,1,float16,fp8,0,0.4091626803080241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,2,128,1,float16,fp8,0,0.1939199964205424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,1,128,1,float16,float16,0,0.20544000466664633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,1,128,1,float16,fp8,0,0.18341867129007974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,2,128,1,float16,float16,0,0.20893865823745728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,64,128,1,float16,fp8,0,0.37094398339589435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,4,128,1,float16,fp8,0,0.21198399861653647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,8,128,1,float16,fp8,0,0.2174826661745707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,8,128,1,float16,float16,0,0.23242666323979697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,64,128,1,float16,float16,0,0.33349867661794025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,1,128,1,float16,float16,0,0.1123466690381368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,1,128,1,float16,fp8,0,0.10021332899729411
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,2,128,1,float16,fp8,0,0.10291199882825215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,2,128,1,float16,float16,0,0.11645866433779399
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,64,128,1,float16,float16,0,0.21011734008789062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,4,128,1,float16,fp8,0,0.11686933040618896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,4,128,1,float16,float16,0,0.12665599584579468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,64,128,1,float16,fp8,0,0.1979466676712036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,8,128,1,float16,fp8,0,0.1218826671441396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,64,8,128,1,float16,float16,0,0.1274133324623108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,1,128,1,float16,float16,0,0.06451733410358429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,64,8,128,1,float16,float16,0,0.43883200486501056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,1,128,1,float16,fp8,0,0.05876799921194712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,2,128,1,float16,float16,0,0.06632000207901001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,4,128,1,float16,float16,0,0.06806399921576183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,4,128,1,float16,fp8,0,0.06406933565934499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,64,4,128,1,float16,float16,0,0.22740799188613892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,8,128,1,float16,fp8,0,0.06596266726652782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,64,128,1,float16,fp8,0,0.11045866211255391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,64,128,1,float16,float16,0,0.10193600257237752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,1,128,1,float16,float16,0,0.04133866727352142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,2,128,1,float16,float16,0,0.04275199770927429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,2,128,1,float16,fp8,0,0.03931200007597605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,4,128,1,float16,float16,0,0.04410133262475332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,64,128,1,float16,fp8,0,0.06046933432420095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,4,128,1,float16,fp8,0,0.04145599901676178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,64,128,1,float16,float16,0,0.06588799754778545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,8,128,1,float16,float16,0,0.04366933306058248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,8,128,1,float16,fp8,0,0.041802664597829185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,2,128,1,float16,fp8,0,0.06001600126425425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,64,8,128,1,float16,float16,0,0.07027733325958252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,1,128,1,float16,float16,0,2.136768023173014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,64,1,128,1,float16,fp8,0,0.03758399933576584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,2,128,1,float16,float16,0,2.2451465924580893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,2,128,1,float16,fp8,0,1.996448040008545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,1,128,1,float16,fp8,0,1.8517600695292156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,4,128,1,float16,float16,0,2.6174400647481284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,4,128,1,float16,fp8,0,2.3681653340657554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,1,128,1,float16,float16,0,1.0895733038584392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,1,128,1,float16,fp8,0,0.9330080350240072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,8,128,1,float16,float16,0,2.6478880246480307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,2,128,1,float16,float16,0,1.1380373636881511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,64,128,1,float16,fp8,0,2.4563306172688804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,2,128,1,float16,fp8,0,1.0046933492024739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,4,128,1,float16,float16,0,1.322538693745931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,4,128,1,float16,fp8,0,1.1906452973683674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,64,128,1,float16,float16,0,2.181920051574707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,64,8,128,1,float16,fp8,0,2.448736031850179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,8,128,1,float16,float16,0,1.3292213280995686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,1,128,1,float16,float16,0,0.5552800099054972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,64,8,128,1,float16,fp8,0,1.2243893146514893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,64,128,1,float16,float16,0,1.08732803662618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,1,128,1,float16,fp8,0,0.4787786801656087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,4,128,1,float16,fp8,0,0.6078879833221436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,64,128,1,float16,fp8,0,1.2393439610799153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,1,128,1,float16,float16,0,0.2960853377978007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,8,128,1,float16,fp8,0,0.6341386636098226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,2,128,1,float16,float16,0,0.5828959941864014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,8,128,1,float16,float16,0,0.6702880064646403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,1,128,1,float16,fp8,0,0.25406932830810547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,64,128,1,float16,float16,0,0.5529333353042603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,4,128,1,float16,float16,0,0.6679999828338623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,64,128,1,float16,fp8,0,0.6296799977620443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,2,128,1,float16,float16,0,0.2982986569404602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,4,128,1,float16,fp8,0,0.3168693383534749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,8,128,1,float16,fp8,0,0.32474666833877563
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,2,128,1,float16,fp8,0,0.2686400016148885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,1,128,1,float16,float16,0,0.15988266468048096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,4,128,1,float16,float16,0,0.3454293409983317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,64,8,128,1,float16,float16,0,0.3468533356984456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,2,128,1,float16,float16,0,0.16500266393025717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,1,128,1,float16,fp8,0,0.1376426617304484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,64,2,128,1,float16,fp8,0,0.5170666774113973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,2,128,1,float16,fp8,0,0.14401066303253174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,64,128,1,float16,float16,0,0.3306826750437419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,64,128,1,float16,fp8,0,0.32466665903727215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,8,128,1,float16,float16,0,0.18574933211008707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,1,128,1,float16,fp8,0,0.0751039981842041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,1,128,1,float16,float16,0,0.08880533774693807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,8,128,1,float16,fp8,0,0.1714400053024292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,2,128,1,float16,fp8,0,0.07958933214346568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,64,128,1,float16,fp8,0,0.17314666509628296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,4,128,1,float16,float16,0,0.10236799716949463
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,2,128,1,float16,float16,0,0.09222400188446045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,4,128,1,float16,fp8,0,0.16942399740219116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,4,128,1,float16,fp8,0,0.09227733810742696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,8,128,1,float16,float16,0,0.101583997408549
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,8,128,1,float16,fp8,0,0.09493866562843323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,1,128,1,float16,fp8,0,0.044490665197372437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,1,128,1,float16,float16,0,0.0506879985332489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,2,128,1,float16,float16,0,0.051551997661590576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,2,128,1,float16,fp8,0,0.0458186666170756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,4,128,1,float16,float16,0,0.05508266886075338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,8,128,1,float16,float16,0,0.055733333031336464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,8,128,1,float16,fp8,0,0.0517546683549881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,64,128,1,float16,fp8,0,0.09522133072217305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,64,128,1,float16,float16,0,0.08657067020734151
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,1,128,1,float16,float16,0,0.031583999594052635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,64,4,128,1,float16,float16,0,0.18545067310333252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,1,128,1,float16,fp8,0,0.029045333464940388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,2,128,1,float16,float16,0,0.032485333581765495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,2,128,1,float16,fp8,0,0.02956266701221466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,64,128,1,float16,fp8,0,0.0517546683549881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,4,128,1,float16,fp8,0,0.031285333136717476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,4,128,1,float16,float16,0,0.033733333150545754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,64,128,1,float16,float16,0,0.04971200227737427
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,8,128,1,float16,fp8,0,0.033471999069054924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,1,128,1,float16,float16,0,0.019146667172511418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,64,8,128,1,float16,float16,0,0.03429866582155228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,2,128,1,float16,float16,0,0.019098666807015736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,64,64,128,1,float16,float16,0,0.15337066849072775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,2,128,1,float16,fp8,0,0.019082666685183842
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,1,128,1,float16,fp8,0,0.019199999670187633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,4,128,1,float16,float16,0,0.020997333029905956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,64,4,128,1,float16,fp8,0,0.050288001696268715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,8,128,1,float16,float16,0,0.02142400046189626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,64,128,1,float16,fp8,0,0.03286399940649668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,64,128,1,float16,float16,0,0.025477332373460133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,1,128,1,float16,float16,0,0.8171306451161703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,1,128,1,float16,fp8,0,0.7435466448465983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,2,128,1,float16,float16,0,0.8767146269480387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,2,128,1,float16,fp8,0,0.8117600282033285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,4,128,1,float16,float16,0,1.053818702697754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,1,128,1,float16,float16,0,0.42209601402282715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,4,128,1,float16,fp8,0,1.0017759799957275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,4,128,1,float16,fp8,0,0.021354667842388153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,64,8,128,1,float16,fp8,0,0.02126399924357732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,8,128,1,float16,float16,0,1.0768427054087322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,1,128,1,float16,fp8,0,0.38257598876953125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,64,128,1,float16,fp8,0,1.1373813152313232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,64,128,1,float16,float16,0,0.9668160279591879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,2,128,1,float16,float16,0,0.4496906598409017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,2,128,1,float16,fp8,0,0.4204853375752767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,4,128,1,float16,float16,0,0.5338559945424398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,1,128,1,float16,float16,0,0.21950932343800864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,4,128,1,float16,fp8,0,0.5091679890950521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,1,128,1,float16,fp8,0,0.2002133329709371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,8,128,1,float16,float16,0,0.539354681968689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,64,8,128,1,float16,fp8,0,0.5268160104751587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,64,128,1,float16,float16,0,0.48972801367441815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,2,128,1,float16,fp8,0,0.22054932514826456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,64,128,1,float16,fp8,0,0.5764586528142294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,4,128,1,float16,float16,0,0.27482134103775024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,4,128,1,float16,fp8,0,0.2622986634572347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,8,128,1,float16,float16,0,0.28197866678237915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,1,128,1,float16,float16,0,0.11959999799728394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,1,128,1,float16,fp8,0,0.10905599594116211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,2,128,1,float16,float16,0,0.1267413298288981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,64,128,1,float16,fp8,0,0.2975200017293294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,64,8,128,1,float16,fp8,0,1.0374773343404133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,2,128,1,float16,fp8,0,0.1167733371257782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,4,128,1,float16,float16,0,0.14857600132624307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,4,128,1,float16,fp8,0,0.1392586628595988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,64,128,1,float16,float16,0,0.2526400089263916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,8,128,1,float16,float16,0,0.16040533781051636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,1,128,1,float16,fp8,0,0.06007466713587443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,1,128,1,float16,float16,0,0.06714666883150737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,2,128,1,float16,float16,0,0.07222400108973186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,2,128,1,float16,fp8,0,0.06292266647020976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,4,128,1,float16,float16,0,0.08043733239173889
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,4,128,1,float16,fp8,0,0.07296533385912578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,64,128,1,float16,fp8,0,0.15651200215021768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,2,128,1,float16,float16,0,0.23275200525919595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,64,8,128,1,float16,fp8,0,0.14378133416175842
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,8,128,1,float16,fp8,0,0.07682133217652638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,8,128,1,float16,float16,0,0.08266133566697438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,1,128,1,float16,float16,0,0.03933866570393244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,1,128,1,float16,fp8,0,0.03640533238649368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,2,128,1,float16,float16,0,0.040720000863075256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,2,128,1,float16,fp8,0,0.037632000943024956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,4,128,1,float16,fp8,0,0.04185600082079569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,4,128,1,float16,float16,0,0.043466667334238686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,8,128,1,float16,fp8,0,0.04355733096599579
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,64,8,128,1,float16,fp8,0,0.28700800736745197
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,8,128,1,float16,float16,0,0.045466666420300804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,1,128,1,float16,float16,0,0.02409599969784419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,2,128,1,float16,float16,0,0.02531733363866806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,64,128,1,float16,float16,0,0.07684800028800964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,1,128,1,float16,fp8,0,0.022272000710169475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,2,128,1,float16,fp8,0,0.023754666248957317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,4,128,1,float16,float16,0,0.025631998976071674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,8,128,1,float16,float16,0,0.02703999976317088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,4,128,1,float16,fp8,0,0.025333332518736523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,8,128,1,float16,fp8,0,0.02717866748571396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,1,128,1,float16,float16,0,0.016458666572968166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,1,128,1,float16,fp8,0,0.015040000279744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,2,128,1,float16,float16,0,0.016757333030303318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,2,128,1,float16,fp8,0,0.01523200049996376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,64,64,128,1,float16,float16,0,0.13514133294423422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,4,128,1,float16,float16,0,0.0170666662355264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,64,128,1,float16,fp8,0,0.029498666524887085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,64,128,1,float16,float16,0,0.02182399978240331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,8,128,1,float16,fp8,0,0.019253333409627277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,8,128,1,float16,float16,0,0.017423999806245167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,1,128,1,float16,float16,0,0.014757333944241205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,1,128,1,float16,fp8,0,0.014794666320085526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,64,64,128,1,float16,fp8,0,0.08795733253161113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,2,128,1,float16,float16,0,0.01504533365368843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,2,128,1,float16,fp8,0,0.015184000134468079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,4,128,1,float16,float16,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,4,128,1,float16,fp8,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,8,128,1,float16,float16,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,8,128,1,float16,fp8,0,0.015135999768972397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,64,128,1,float16,float16,0,0.04252799848715464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,64,4,128,1,float16,fp8,0,0.01718933383623759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,1,128,1,float16,float16,0,0.512656013170878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,1,128,1,float16,fp8,0,0.4633493423461914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,2,128,1,float16,float16,0,0.5560373465220133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,64,128,1,float16,float16,0,0.016927999754746754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,2,128,1,float16,fp8,0,0.49904000759124756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,4,128,1,float16,float16,0,0.6307573318481445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,64,64,128,1,float16,fp8,0,0.04567466676235199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,1,128,1,float16,float16,0,0.26711465915044147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,8,128,1,float16,fp8,0,0.6092106501261393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,1,128,1,float16,fp8,0,0.24470933278401694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,64,128,1,float16,float16,0,0.5504586696624756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,2,128,1,float16,float16,0,0.2792960007985433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,2,128,1,float16,fp8,0,0.25945599873860675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,4,128,1,float16,float16,0,0.3256053328514099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,4,128,1,float16,fp8,0,0.31278934081395465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,8,128,1,float16,float16,0,0.3299893339474996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,64,64,128,1,float16,fp8,0,0.02117866774400075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,4,128,1,float16,fp8,0,0.5928373336791992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,1,128,1,float16,fp8,0,0.130213330189387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,8,128,1,float16,fp8,0,0.31518399715423584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,1,128,1,float16,float16,0,0.14562132954597473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,2,128,1,float16,float16,0,0.15147200226783752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,4,128,1,float16,fp8,0,0.1588053305943807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,64,8,128,1,float16,float16,0,0.6368639866511027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,64,128,1,float16,fp8,0,0.31833066542943317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,4,128,1,float16,float16,0,0.17065600554148355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,64,128,1,float16,float16,0,0.27349332968393963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,8,128,1,float16,fp8,0,0.16427733500798544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,1,128,1,float16,float16,0,0.07890133559703827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,64,64,128,1,float16,fp8,0,0.6170186599095663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,2,128,1,float16,fp8,0,0.0751039981842041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,64,128,1,float16,fp8,0,0.166703999042511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,4,128,1,float16,float16,0,0.09635733564694722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,4,128,1,float16,fp8,0,0.08640533685684204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,8,128,1,float16,float16,0,0.09915733337402344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,64,128,1,float16,float16,0,0.14509866635004678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,8,128,1,float16,fp8,0,0.08842666943868001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,1,128,1,float16,fp8,0,0.039333333571751915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,1,128,1,float16,float16,0,0.04381333291530609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,2,128,1,float16,float16,0,0.04555733501911163
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,2,128,1,float16,fp8,0,0.13818132877349854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,2,128,1,float16,fp8,0,0.041477332512537636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,64,128,1,float16,float16,0,0.07900266846021016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,64,128,1,float16,fp8,0,0.090938667456309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,4,128,1,float16,float16,0,0.047824000318845115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,4,128,1,float16,fp8,0,0.04533333579699198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,8,128,1,float16,fp8,0,0.04569066564242045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,64,8,128,1,float16,float16,0,0.1734559933344523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,1,128,1,float16,fp8,0,0.07238933444023132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,2,128,1,float16,float16,0,0.029050665597120922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,64,2,128,1,float16,float16,0,0.0843999981880188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,4,128,1,float16,float16,0,0.02959466725587845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,4,128,1,float16,fp8,0,0.028773332635561626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,64,128,1,float16,fp8,0,0.048911998669306435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,8,128,1,float16,float16,0,0.029802667597929638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,64,128,1,float16,float16,0,0.05053333441416422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,1,128,1,float16,float16,0,0.0170666662355264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,1,128,1,float16,fp8,0,0.016864000509182613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,2,128,1,float16,float16,0,0.01704000060757001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,2,128,1,float16,fp8,0,0.017130666722853977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,64,128,1,float16,fp8,0,0.029866665601730347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,4,128,1,float16,float16,0,0.01922133316596349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,4,128,1,float16,fp8,0,0.01907733331123988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,8,128,1,float16,float16,0,0.01926933353145917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,64,8,128,1,float16,float16,0,0.04987200101216634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,8,128,1,float16,fp8,0,0.01916266605257988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,1,128,1,float16,float16,0,0.01302933320403099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,1,128,1,float16,fp8,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,2,128,1,float16,float16,0,0.013082666943470636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,2,128,1,float16,fp8,0,0.027285332481066387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,1,128,1,float16,fp8,0,0.02513066679239273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,2,128,1,float16,fp8,0,0.0129120002190272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,64,128,1,float16,float16,0,0.016447999825080235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,4,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,64,128,1,float16,fp8,0,0.019071999937295914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,8,128,1,float16,float16,0,0.013594667116800943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,8,128,1,float16,fp8,0,0.029530666768550873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,8,128,1,float16,fp8,0,0.01313599944114685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,1,128,1,float16,float16,0,0.01302933320403099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,1,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,2,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,2,128,1,float16,fp8,0,0.012970666090647379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,4,128,1,float16,float16,0,0.012634667257467905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,4,128,1,float16,fp8,0,0.01267733300725619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,8,128,1,float16,float16,0,0.012938667088747025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,64,128,1,float16,fp8,0,0.017029333859682083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,64,128,1,float16,float16,0,0.014896000425020853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,64,8,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,64,64,128,1,float16,float16,0,0.02346133440732956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,64,1,128,1,float16,float16,0,0.027029333015282948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,64,4,128,1,float16,fp8,0,0.013162666310866674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,1,128,1,float16,fp8,0,0.3885759909947713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,1,128,1,float16,float16,0,0.4532053470611572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,2,128,1,float16,fp8,0,0.4052586555480957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,4,128,1,float16,float16,0,0.5073066552480062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,2,128,1,float16,float16,0,0.4668746789296468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,4,128,1,float16,fp8,0,0.4549013376235962
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,8,128,1,float16,float16,0,0.5112266540527344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,1,128,1,float16,float16,0,0.2374080022176107
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,1,128,1,float16,fp8,0,0.20390399297078451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,64,128,1,float16,fp8,0,0.39213867982228595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,64,128,1,float16,float16,0,0.4081386725107829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,64,8,128,1,float16,fp8,0,0.4628640015920003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,2,128,1,float16,fp8,0,0.21303999423980713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,4,128,1,float16,float16,0,0.2635200023651123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,4,128,1,float16,fp8,0,0.23324799537658691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,8,128,1,float16,float16,0,0.26492265860239667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,1,128,1,float16,float16,0,0.1267626682917277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,1,128,1,float16,fp8,0,0.1090880036354065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,2,128,1,float16,float16,0,0.1321386694908142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,4,128,1,float16,float16,0,0.1405226687590281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,2,128,1,float16,fp8,0,0.11505599816640218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,8,128,1,float16,float16,0,0.14245333274205527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,8,128,1,float16,fp8,0,0.12638399998346964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,64,128,1,float16,fp8,0,0.2037386695543925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,8,128,1,float16,fp8,0,0.23895466327667236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,64,128,1,float16,float16,0,0.19183466831843057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,64,2,128,1,float16,float16,0,0.24497600396474203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,64,128,1,float16,float16,0,0.1116373340288798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,1,128,1,float16,float16,0,0.06795733173688252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,64,128,1,float16,fp8,0,0.10919466614723206
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,8,128,1,float16,float16,0,0.07509333391984303
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,4,128,1,float16,fp8,0,0.06423999865849812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,4,128,1,float16,float16,0,0.07353599866231282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,8,128,1,float16,fp8,0,0.0661599983771642
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,64,4,128,1,float16,fp8,0,0.12377066413561504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,2,128,1,float16,float16,0,0.07028266787528992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,1,128,1,float16,fp8,0,0.059477334221204124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,64,2,128,1,float16,fp8,0,0.06085333228111267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,1,128,1,float16,float16,0,0.0390133336186409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,64,128,1,float16,float16,0,0.06264000137646993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,2,128,1,float16,fp8,0,0.0352960005402565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,4,128,1,float16,fp8,0,0.0373333344856898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,2,128,1,float16,float16,0,0.039477333426475525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,8,128,1,float16,float16,0,0.041434665520985924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,1,128,1,float16,float16,0,0.0239680012067159
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,8,128,1,float16,fp8,0,0.03738666574160258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,2,128,1,float16,float16,0,0.023946667710940044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,2,128,1,float16,fp8,0,0.02314666658639908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,4,128,1,float16,float16,0,0.025439999997615814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,64,128,1,float16,fp8,0,0.03586133321126302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,8,128,1,float16,float16,0,0.025663999219735462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,8,128,1,float16,fp8,0,0.025360000630219776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,64,128,1,float16,fp8,0,0.022592000663280487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,64,128,1,float16,float16,0,0.01911466692884763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,1,128,1,float16,float16,0,0.017050666113694508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,4,128,1,float16,float16,0,0.041562666495641075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,1,128,1,float16,fp8,0,0.03356799980004629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,2,128,1,float16,float16,0,0.017136000096797943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,1,128,1,float16,fp8,0,0.01642666632930438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,4,128,1,float16,fp8,0,0.015274666249752045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,1,128,1,float16,fp8,0,0.02309333284695943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,8,128,1,float16,fp8,0,0.016330666840076447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,8,128,1,float16,float16,0,0.016906666258970898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,4,128,1,float16,fp8,0,0.024735999604066212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,1,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,1,128,1,float16,fp8,0,0.012703999876976013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,2,128,1,float16,float16,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,2,128,1,float16,fp8,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,64,64,128,1,float16,float16,0,0.029850666721661884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,4,128,1,float16,fp8,0,0.013605333864688873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,4,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,8,128,1,float16,float16,0,0.012730666746695837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,8,128,1,float16,fp8,0,0.012698666503032049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,1,128,1,float16,float16,0,0.011354666203260422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,1,128,1,float16,fp8,0,0.011855999628702799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,2,128,1,float16,float16,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,2,128,1,float16,fp8,0,0.012400000045696894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,64,128,1,float16,float16,0,0.013365333278973898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,64,64,128,1,float16,fp8,0,0.055999999245007835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,4,128,1,float16,float16,0,0.011338666081428528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,4,128,1,float16,fp8,0,0.012047999848922094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,8,128,1,float16,fp8,0,0.01267733300725619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,2,128,1,float16,fp8,0,0.01664000004529953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,8,128,1,float16,float16,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,1,128,1,float16,float16,0,0.4252693255742391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,64,128,1,float16,float16,0,0.014752000570297241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,1,128,1,float16,fp8,0,0.35731732845306396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,2,128,1,float16,float16,0,0.42902398109436035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,4,128,1,float16,float16,0,0.45070934295654297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,64,64,128,1,float16,fp8,0,0.014965333044528961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,8,128,1,float16,float16,0,0.4511893192927043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,8,128,1,float16,fp8,0,0.3931093215942383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,4,128,1,float16,fp8,0,0.39273067315419513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,64,4,128,1,float16,float16,0,0.016895999511082966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,1,128,1,float16,float16,0,0.22336000204086304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,1,128,1,float16,fp8,0,0.18497065703074136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,2,128,1,float16,float16,0,0.23242133855819702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,2,128,1,float16,fp8,0,0.1888213356335958
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,64,128,1,float16,float16,0,0.3080586592356364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,64,128,1,float16,fp8,0,0.2791360020637512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,4,128,1,float16,float16,0,0.233407994111379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,64,64,128,1,float16,fp8,0,0.01699200024207433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,64,2,128,1,float16,fp8,0,0.365013321240743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,4,128,1,float16,fp8,0,0.19977066914240518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,8,128,1,float16,float16,0,0.2396906614303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,1,128,1,float16,float16,0,0.12045866250991821
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,2,128,1,float16,float16,0,0.12105066577593486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,1,128,1,float16,fp8,0,0.09681600332260132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,64,8,128,1,float16,fp8,0,0.20320000251134238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,4,128,1,float16,float16,0,0.12685333689053854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,4,128,1,float16,fp8,0,0.10177600383758545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,8,128,1,float16,float16,0,0.12592533230781555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,8,128,1,float16,fp8,0,0.1035040020942688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,64,128,1,float16,float16,0,0.15359466274579367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,1,128,1,float16,float16,0,0.06520533561706543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,2,128,1,float16,fp8,0,0.05475200215975443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,4,128,1,float16,float16,0,0.06731200218200684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,64,128,1,float16,float16,0,0.0899786651134491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,1,128,1,float16,fp8,0,0.05343999962011973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,2,128,1,float16,float16,0,0.06621866424878438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,4,128,1,float16,fp8,0,0.05610666672388712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,8,128,1,float16,fp8,0,0.05690666536490122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,8,128,1,float16,float16,0,0.06833600004514058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,1,128,1,float16,float16,0,0.03722666700681051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,1,128,1,float16,fp8,0,0.03160000095764796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,2,128,1,float16,fp8,0,0.031370667119820915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,2,128,1,float16,fp8,0,0.09903466701507568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,4,128,1,float16,fp8,0,0.03333866596221924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,4,128,1,float16,float16,0,0.03748800108830134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,64,64,128,1,float16,fp8,0,0.1470133364200592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,64,128,1,float16,float16,0,0.04541333516438802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,64,128,1,float16,fp8,0,0.045423999428749084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,8,128,1,float16,fp8,0,0.03339733431736628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,2,128,1,float16,float16,0,0.023333333432674408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,1,128,1,float16,fp8,0,0.02126399924357732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,1,128,1,float16,float16,0,0.023034666975339253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,2,128,1,float16,fp8,0,0.021007999777793884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,4,128,1,float16,float16,0,0.023247999449570973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,4,128,1,float16,fp8,0,0.021082667013009388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,64,128,1,float16,float16,0,0.02534399926662445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,64,128,1,float16,fp8,0,0.027141332626342773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,8,128,1,float16,float16,0,0.02332799881696701
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,1,128,1,float16,float16,0,0.015205333630243937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,1,128,1,float16,fp8,0,0.015253332753976187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,2,128,1,float16,float16,0,0.014949332922697067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,2,128,1,float16,fp8,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,64,64,128,1,float16,fp8,0,0.07795199751853943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,4,128,1,float16,float16,0,0.015642666568358738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,4,128,1,float16,fp8,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,64,128,1,float16,fp8,0,0.019189332922299702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,64,128,1,float16,float16,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,8,128,1,float16,float16,0,0.016506666938463848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,2,128,1,float16,float16,0,0.03749333322048187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,1,128,1,float16,fp8,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,64,128,1,float16,fp8,0,0.016762666404247284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,4,128,1,float16,float16,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,4,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,64,8,128,1,float16,float16,0,0.03839999934037527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,8,128,1,float16,float16,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,8,128,1,float16,fp8,0,0.012879999975363413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,64,8,128,1,float16,fp8,0,0.021327999730904896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,1,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,1,128,1,float16,fp8,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,2,128,1,float16,float16,0,0.010762666662534079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,64,128,1,float16,float16,0,0.01332266628742218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,64,128,1,float16,fp8,0,0.015008000036080679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,2,128,1,float16,fp8,0,0.01110400011142095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,4,128,1,float16,float16,0,0.011706666400035223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,64,8,128,1,float16,fp8,0,0.014949332922697067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,8,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,4,128,1,float16,fp8,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,64,8,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,2,128,1,float16,float16,0,0.012703999876976013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,2,128,1,float16,fp8,0,0.012144000579913458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,1,128,1,float16,float16,0,0.41575467586517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,1,128,1,float16,fp8,0,0.33688533306121826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,2,128,1,float16,float16,0,0.42976001898447674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,4,128,1,float16,float16,0,0.4376213153203328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,4,128,1,float16,fp8,0,0.35354665915171307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,8,128,1,float16,float16,0,0.42952001094818115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,8,128,1,float16,fp8,0,0.3534986575444539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,1,128,1,float16,float16,0,0.21833600600560507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,64,128,1,float16,fp8,0,0.22195200125376383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,64,128,1,float16,float16,0,0.2595360080401103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,1,128,1,float16,fp8,0,0.1734293301900228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,2,128,1,float16,float16,0,0.22186134258906046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,1,128,1,float16,float16,0,0.012517333030700684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,2,128,1,float16,fp8,0,0.17422932386398315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,64,64,128,1,float16,float16,0,0.014442666123310724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,8,128,1,float16,fp8,0,0.18437866369883218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,8,128,1,float16,float16,0,0.2258560061454773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,1,128,1,float16,float16,0,0.11538666486740112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,64,2,128,1,float16,fp8,0,0.3463519811630249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,4,128,1,float16,fp8,0,0.18129066626230875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,1,128,1,float16,fp8,0,0.09285333752632141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,2,128,1,float16,float16,0,0.11672533551851909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,2,128,1,float16,fp8,0,0.09289066990216573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,4,128,1,float16,fp8,0,0.09505599737167358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,8,128,1,float16,float16,0,0.11993066469828288
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,64,128,1,float16,fp8,0,0.11988266309102376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,4,128,1,float16,float16,0,0.11572800079981486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,8,128,1,float16,fp8,0,0.09697066744168599
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,64,64,128,1,float16,float16,0,0.13843199610710144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,1,128,1,float16,fp8,0,0.05203733344872793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,2,128,1,float16,float16,0,0.06425066788991292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,1,128,1,float16,float16,0,0.06334400177001953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,64,128,1,float16,float16,0,0.07702933251857758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,4,128,1,float16,fp8,0,0.05407466491063436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,64,128,1,float16,fp8,0,0.06505600114663442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,2,128,1,float16,fp8,0,0.05190933247407278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,8,128,1,float16,float16,0,0.06427733103434245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,1,128,1,float16,float16,0,0.03659199923276901
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,8,128,1,float16,fp8,0,0.055231998364130654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,1,128,1,float16,fp8,0,0.031343999008337654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,2,128,1,float16,float16,0,0.036677333215872444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,4,128,1,float16,float16,0,0.036303999523321785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,4,128,1,float16,fp8,0,0.031290667752424874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,2,128,1,float16,fp8,0,0.03146666785081228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,64,4,128,1,float16,float16,0,0.23101866245269775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,64,128,1,float16,float16,0,0.03939199944337209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,64,128,1,float16,fp8,0,0.03760000069936117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,8,128,1,float16,fp8,0,0.03124266614516576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,64,8,128,1,float16,float16,0,0.03669866671164831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,2,128,1,float16,float16,0,0.023317334552605946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,1,128,1,float16,fp8,0,0.021338666478792827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,64,128,1,float16,fp8,0,0.02535466601451238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,1,128,1,float16,float16,0,0.023024000227451324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,4,128,1,float16,float16,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,2,128,1,float16,fp8,0,0.021349333226680756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,8,128,1,float16,fp8,0,0.021216000119845074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,1,128,1,float16,float16,0,0.015397333850463232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,1,128,1,float16,fp8,0,0.015040000279744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,64,4,128,1,float16,float16,0,0.06351466476917267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,2,128,1,float16,float16,0,0.015034666905800501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,2,128,1,float16,fp8,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,4,128,1,float16,fp8,0,0.01479999969402949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,64,128,1,float16,float16,0,0.017946666727463405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,4,128,1,float16,float16,0,0.015397333850463232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,8,128,1,float16,fp8,0,0.015119999647140503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,8,128,1,float16,float16,0,0.015930666277805965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,1,128,1,float16,float16,0,0.012170666207869848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,1,128,1,float16,fp8,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,2,128,1,float16,float16,0,0.012746666868527731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,2,128,1,float16,fp8,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,4,128,1,float16,float16,0,0.012698666503032049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,4,128,1,float16,fp8,0,0.011994666109482447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,8,128,1,float16,float16,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,64,128,1,float16,fp8,0,0.016165333489576977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,64,128,1,float16,float16,0,0.014874666929244995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,64,8,128,1,float16,fp8,0,0.0124746672809124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,1,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,64,128,1,float16,float16,0,0.025013332565625507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,1,128,1,float16,float16,0,0.011333333949247995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,2,128,1,float16,float16,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,4,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,64,128,1,float16,fp8,0,0.014981333166360855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,2,128,1,float16,fp8,0,0.010693332801262537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,8,128,1,float16,float16,0,0.010842667271693548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,8,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,4,128,1,float16,fp8,0,0.020917333662509918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,64,8,128,1,float16,float16,0,0.02332799881696701
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,64,4,128,1,float16,float16,0,0.02187199890613556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,64,4,128,1,float16,fp8,0,0.019274666905403137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,4,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,64,64,128,1,float16,float16,0,0.013082666943470636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,64,8,128,1,float16,float16,0,0.02917333443959554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,64,64,128,1,float16,fp8,0,0.018858666221300762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,64,8,128,1,float16,fp8,0,0.025429333249727886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,64,128,1,float16,fp8,0,0.08061866462230682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,4,128,1,float16,float16,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,8,128,1,float16,float16,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,4,128,1,float16,fp8,0,0.014885333677132925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,8,128,1,float16,fp8,0,0.016879999389251072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,4,128,1,float16,float16,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,4,128,1,float16,fp8,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,8,128,1,float16,float16,0,0.013141332815090815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,64,128,1,float16,fp8,0,0.0454773356517156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,8,128,1,float16,fp8,0,0.012794667234023413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,64,64,128,1,float16,float16,0,0.06015466650327047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,8,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,64,128,1,float16,fp8,0,0.02938133229811986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,8,128,1,float16,float16,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,64,128,1,float16,float16,0,0.029578665892283123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,64,128,1,float16,float16,0,0.020682666450738907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,4,128,1,float16,float16,0,0.00874133345981439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,4,128,1,float16,fp8,0,0.010378666842977205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,64,64,128,1,float16,float16,0,0.1053013304869334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,64,128,1,float16,fp8,0,0.021183999876181286
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,8,128,1,float16,float16,0,0.009349333122372627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,64,8,128,1,float16,fp8,0,0.009061333412925402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,4,128,1,float16,float16,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,8,128,1,float16,float16,0,0.008805333326260248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,4,128,1,float16,fp8,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,64,128,1,float16,fp8,0,0.016970666746298473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,8,128,1,float16,fp8,0,0.00901333304742972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,64,64,128,1,float16,float16,0,0.015125333021084467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,4,128,1,float16,float16,0,0.008938666433095932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,8,128,1,float16,float16,0,0.00933333362142245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,8,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,4,128,1,float16,float16,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,64,4,128,1,float16,fp8,0,0.01071999967098236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,64,128,1,float16,float16,0,0.013199999928474426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,64,128,1,float16,fp8,0,0.014741333822409311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,4,128,1,float16,float16,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,4,128,1,float16,fp8,0,0.008879999940594038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,64,128,1,float16,float16,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,4,128,1,float16,float16,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,64,128,1,float16,float16,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,64,128,1,float16,fp8,0,0.013189333180586496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,8,128,1,float16,float16,0,0.008896000062425932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,8,128,1,float16,fp8,0,0.008810666700204214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,64,4,128,1,float16,fp8,0,0.009082666908701261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,64,128,1,float16,fp8,0,0.013525333255529404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,64,4,128,1,float16,fp8,0,0.008869333192706108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,8,128,1,float16,fp8,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,1,128,1,float16,fp8,0,17.115887959798176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,2,128,1,float16,fp8,0,17.260772705078125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,64,8,128,1,float16,float16,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,1,128,1,float16,float16,0,22.856608072916668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,2,128,1,float16,float16,0,22.908597310384113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,4,128,1,float16,fp8,0,17.57431411743164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,4,128,1,float16,float16,0,23.139307657877605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,48,128,1,float16,fp8,0,9.490325291951498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,1,128,1,float16,fp8,0,8.562661488850912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,1,128,1,float16,float16,0,11.533402760823568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,48,128,1,float16,float16,0,12.108783721923828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,8,128,1,float16,fp8,0,17.792762756347656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,48,8,128,1,float16,float16,0,23.537071228027344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,2,128,1,float16,fp8,0,8.747461318969727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,2,128,1,float16,float16,0,11.587163289388021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,4,128,1,float16,fp8,0,8.823087692260742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,1,128,1,float16,fp8,0,4.575130780537923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,1,128,1,float16,float16,0,5.730768203735352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,48,128,1,float16,float16,0,5.99514643351237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,4,128,1,float16,float16,0,11.74386723836263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,8,128,1,float16,fp8,0,8.714911778767904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,48,8,128,1,float16,float16,0,12.162879943847656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,2,128,1,float16,fp8,0,4.3367414474487305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,2,128,1,float16,float16,0,5.46012814839681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,4,128,1,float16,float16,0,5.922543843587239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,4,128,1,float16,fp8,0,4.532304128011067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,8,128,1,float16,fp8,0,4.3914079666137695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,8,128,1,float16,float16,0,5.507434844970703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,48,48,128,1,float16,fp8,0,5.013312021891276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,1,128,1,float16,float16,0,3.0980211893717446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,48,128,1,float16,float16,0,2.871664047241211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,48,128,1,float16,fp8,0,2.5308799743652344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,1,128,1,float16,fp8,0,2.5408266385396323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,2,128,1,float16,float16,0,2.6262826919555664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,2,128,1,float16,fp8,0,2.2112852732340493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,4,128,1,float16,float16,0,2.806682586669922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,4,128,1,float16,fp8,0,2.2332372665405273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,8,128,1,float16,float16,0,2.701541264851888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,48,8,128,1,float16,fp8,0,2.3716160456339517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,1,128,1,float16,fp8,0,9.983333587646484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,2,128,1,float16,fp8,0,9.915301640828451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,1,128,1,float16,float16,0,13.266192118326822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,4,128,1,float16,fp8,0,10.186911900838217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,2,128,1,float16,float16,0,13.26214345296224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,8,128,1,float16,fp8,0,10.161823908487955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,4,128,1,float16,float16,0,14.01736577351888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,48,8,128,1,float16,float16,0,13.8647092183431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,1,128,1,float16,fp8,0,4.863797187805176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,48,128,1,float16,float16,0,7.239632288614909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,48,128,1,float16,fp8,0,5.715450922648112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,1,128,1,float16,float16,0,6.658373514811198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,2,128,1,float16,float16,0,6.586938858032227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,2,128,1,float16,fp8,0,5.273152033487956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,4,128,1,float16,float16,0,6.510090510050456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,4,128,1,float16,fp8,0,5.079194704691569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,1,128,1,float16,float16,0,3.00712521870931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,1,128,1,float16,fp8,0,2.466906706492106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,8,128,1,float16,fp8,0,5.033141454060872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,48,128,1,float16,fp8,0,3.034682591756185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,48,128,1,float16,float16,0,3.3497654596964517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,48,8,128,1,float16,float16,0,6.400378545125325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,2,128,1,float16,fp8,0,2.5599306424458823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,2,128,1,float16,float16,0,3.1510238647460938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,4,128,1,float16,fp8,0,2.5651520093282065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,4,128,1,float16,float16,0,3.167477289835612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,8,128,1,float16,float16,0,3.0745118459065757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,48,8,128,1,float16,fp8,0,2.596826712290446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,1,128,1,float16,float16,0,1.5385440190633137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,1,128,1,float16,fp8,0,1.6071467399597168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,2,128,1,float16,float16,0,1.5550026893615723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,48,128,1,float16,float16,0,1.7207999229431152
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,2,128,1,float16,fp8,0,1.579535961151123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,4,128,1,float16,float16,0,1.811338742574056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,4,128,1,float16,fp8,0,1.333679993947347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,8,128,1,float16,float16,0,1.5846452713012695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,8,128,1,float16,fp8,0,1.444101333618164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,1,128,1,float16,fp8,0,6.850122451782227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,48,48,128,1,float16,fp8,0,1.7615466117858887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,2,128,1,float16,fp8,0,7.078053156534831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,1,128,1,float16,float16,0,9.294911702473959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,4,128,1,float16,fp8,0,7.2913869222005205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,2,128,1,float16,float16,0,9.357583999633789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,4,128,1,float16,float16,0,9.501845041910807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,8,128,1,float16,float16,0,9.636330922444662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,1,128,1,float16,fp8,0,3.4811840057373047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,1,128,1,float16,float16,0,4.459653218587239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,2,128,1,float16,float16,0,4.500202814737956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,48,128,1,float16,float16,0,4.7437388102213545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,48,8,128,1,float16,fp8,0,7.2713063557942705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,4,128,1,float16,float16,0,4.606869379679362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,4,128,1,float16,fp8,0,3.6933066050211587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,48,128,1,float16,fp8,0,2.149866739908854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,8,128,1,float16,float16,0,4.598064104715983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,8,128,1,float16,fp8,0,3.8531999588012695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,1,128,1,float16,fp8,0,1.8035893440246582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,48,128,1,float16,float16,0,2.4038240114847818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,2,128,1,float16,float16,0,2.160282611846924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,4,128,1,float16,float16,0,2.2052000363667807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,48,128,1,float16,fp8,0,4.198053359985352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,48,2,128,1,float16,fp8,0,4.3416798909505205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,8,128,1,float16,float16,0,2.2061440149943032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,1,128,1,float16,float16,0,1.0858186880747478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,8,128,1,float16,fp8,0,2.166170597076416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,1,128,1,float16,fp8,0,1.0890239874521892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,48,128,1,float16,fp8,0,1.1117760340372722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,2,128,1,float16,float16,0,1.0958986282348633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,48,128,1,float16,float16,0,1.2549066543579102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,2,128,1,float16,fp8,0,0.9701600074768066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,1,128,1,float16,float16,0,2.24018128712972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,4,128,1,float16,float16,0,1.1352266470591228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,4,128,1,float16,fp8,0,0.9673066933949789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,8,128,1,float16,float16,0,1.1241226991017659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,48,8,128,1,float16,fp8,0,0.9866027037302653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,4,128,1,float16,fp8,0,1.887290636698405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,1,128,1,float16,fp8,0,9.058293024698893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,2,128,1,float16,fp8,0,9.185237248738607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,48,2,128,1,float16,fp8,0,2.0526347160339355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,4,128,1,float16,fp8,0,9.441541035970053
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,1,128,1,float16,float16,0,12.316858927408854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,2,128,1,float16,float16,0,12.36190923055013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,4,128,1,float16,float16,0,12.46844228108724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,8,128,1,float16,float16,0,12.756752014160156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,1,128,1,float16,float16,0,5.708474477132161
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,1,128,1,float16,fp8,0,4.756373405456543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,48,128,1,float16,fp8,0,5.696970621744792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,48,128,1,float16,float16,0,6.464842478434245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,48,8,128,1,float16,fp8,0,9.588213602701822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,4,128,1,float16,float16,0,5.953317642211914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,4,128,1,float16,fp8,0,4.985114733378093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,48,128,1,float16,fp8,0,2.8859306971232095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,8,128,1,float16,fp8,0,4.790234565734863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,8,128,1,float16,float16,0,6.202581405639648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,48,128,1,float16,float16,0,3.1801493962605796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,1,128,1,float16,float16,0,2.9311307271321616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,1,128,1,float16,fp8,0,2.3461333910624185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,2,128,1,float16,float16,0,5.633077621459961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,2,128,1,float16,float16,0,2.7923574447631836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,2,128,1,float16,fp8,0,2.320570627848307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,4,128,1,float16,fp8,0,2.525616010030111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,4,128,1,float16,float16,0,2.9068320592244468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,8,128,1,float16,fp8,0,2.4484000205993652
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,1,128,1,float16,float16,0,1.4040212631225586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,48,8,128,1,float16,float16,0,2.8753814697265625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,48,128,1,float16,float16,0,1.6201866467793782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,1,128,1,float16,fp8,0,1.1847840150197346
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,2,128,1,float16,fp8,0,1.1956266562143962
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,2,128,1,float16,float16,0,1.5827892621358235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,4,128,1,float16,float16,0,1.461008071899414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,4,128,1,float16,fp8,0,1.5366133054097493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,8,128,1,float16,float16,0,1.4660533269246419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,8,128,1,float16,fp8,0,1.2557546297709148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,1,128,1,float16,float16,0,0.7428320248921713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,48,2,128,1,float16,fp8,0,4.862016042073567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,2,128,1,float16,float16,0,0.7381172974904379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,2,128,1,float16,fp8,0,0.6631733179092407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,48,128,1,float16,fp8,0,0.8406506379445394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,48,128,1,float16,float16,0,0.8872906366984049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,4,128,1,float16,float16,0,0.7476747035980225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,4,128,1,float16,fp8,0,0.6584106683731079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,8,128,1,float16,float16,0,0.764784018198649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,8,128,1,float16,fp8,0,0.687338670094808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,48,48,128,1,float16,fp8,0,1.5725280443827312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,1,128,1,float16,fp8,0,5.251887957255046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,48,1,128,1,float16,fp8,0,0.6337226629257202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,2,128,1,float16,fp8,0,5.370106379191081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,1,128,1,float16,float16,0,6.945728302001953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,2,128,1,float16,float16,0,6.9429976145426435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,4,128,1,float16,float16,0,7.04963747660319
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,4,128,1,float16,fp8,0,5.639834721883138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,1,128,1,float16,float16,0,3.5557867685953775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,2,128,1,float16,float16,0,3.220815976460775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,1,128,1,float16,fp8,0,2.9435412089029946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,48,128,1,float16,fp8,0,3.690688133239746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,2,128,1,float16,fp8,0,2.71507199605306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,8,128,1,float16,fp8,0,5.694896062215169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,48,8,128,1,float16,float16,0,7.366959889729817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,4,128,1,float16,float16,0,3.513930638631185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,4,128,1,float16,fp8,0,3.1319093704223633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,1,128,1,float16,float16,0,1.6098559697469075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,8,128,1,float16,float16,0,3.3759946823120117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,48,128,1,float16,float16,0,1.931173324584961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,8,128,1,float16,fp8,0,2.8919785817464194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,48,128,1,float16,fp8,0,1.8081653912862141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,1,128,1,float16,fp8,0,1.4798026084899902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,2,128,1,float16,float16,0,1.6265920003255208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,2,128,1,float16,fp8,0,1.4499893188476562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,4,128,1,float16,float16,0,1.7221066157023113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,4,128,1,float16,fp8,0,1.4579466183980305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,1,128,1,float16,float16,0,0.8205013275146484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,8,128,1,float16,float16,0,1.7190826733907063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,48,8,128,1,float16,fp8,0,1.4705759684244792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,1,128,1,float16,fp8,0,0.7564160029093424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,48,48,128,1,float16,float16,0,3.8424905141194663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,2,128,1,float16,fp8,0,0.7328853607177734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,2,128,1,float16,float16,0,0.8380426565806071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,4,128,1,float16,float16,0,0.8743306795756022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,4,128,1,float16,fp8,0,0.7623840173085531
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,48,128,1,float16,fp8,0,0.9925173123677572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,8,128,1,float16,float16,0,0.8824480374654134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,1,128,1,float16,float16,0,0.4399786790211995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,8,128,1,float16,fp8,0,0.7675573031107584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,2,128,1,float16,fp8,0,0.41866668065388996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,2,128,1,float16,float16,0,0.44539201259613037
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,48,128,1,float16,fp8,0,0.49928534030914307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,4,128,1,float16,float16,0,0.45897066593170166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,4,128,1,float16,fp8,0,0.4147040049235026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,48,128,1,float16,float16,0,0.5156053304672241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,8,128,1,float16,float16,0,0.45981868108113605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,48,48,128,1,float16,float16,0,0.9978773593902588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,1,128,1,float16,fp8,0,5.056090672810872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,2,128,1,float16,fp8,0,5.188549359639485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,1,128,1,float16,fp8,0,0.3896373510360718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,1,128,1,float16,float16,0,6.142101287841797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,2,128,1,float16,float16,0,6.401509602864583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,4,128,1,float16,float16,0,6.801765441894531
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,48,8,128,1,float16,fp8,0,0.41979201634724933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,1,128,1,float16,float16,0,3.0885705947875977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,1,128,1,float16,fp8,0,2.700181325276693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,2,128,1,float16,float16,0,3.0575199127197266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,4,128,1,float16,fp8,0,5.576170603434245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,8,128,1,float16,fp8,0,5.635295867919922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,48,128,1,float16,float16,0,3.8956054051717124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,48,8,128,1,float16,float16,0,6.861696243286133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,48,128,1,float16,fp8,0,3.8168907165527344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,2,128,1,float16,fp8,0,2.903813362121582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,4,128,1,float16,float16,0,3.251498540242513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,4,128,1,float16,fp8,0,3.0472373962402344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,8,128,1,float16,float16,0,3.2773707707722983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,1,128,1,float16,float16,0,1.5139360427856445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,48,8,128,1,float16,fp8,0,2.838864008585612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,1,128,1,float16,fp8,0,1.3512800534566243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,2,128,1,float16,fp8,0,1.381648063659668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,2,128,1,float16,float16,0,1.5585974057515461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,48,128,1,float16,float16,0,1.9628853797912598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,4,128,1,float16,float16,0,1.6525920232137044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,48,128,1,float16,fp8,0,1.9947679837544758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,8,128,1,float16,float16,0,1.6265759468078613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,1,128,1,float16,fp8,0,0.7375679810841879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,1,128,1,float16,float16,0,0.778223991394043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,2,128,1,float16,float16,0,0.7999413013458252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,48,128,1,float16,float16,0,0.9855999946594238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,2,128,1,float16,fp8,0,0.6930560270945231
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,48,128,1,float16,fp8,0,1.0214347044626872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,4,128,1,float16,float16,0,0.8489653269449869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,4,128,1,float16,fp8,0,0.736629327138265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,8,128,1,float16,float16,0,0.8426613012949625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,1,128,1,float16,float16,0,0.41043734550476074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,48,8,128,1,float16,fp8,0,0.7503946622212728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,1,128,1,float16,fp8,0,0.36150399843851727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,2,128,1,float16,float16,0,0.41530664761861164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,48,128,1,float16,float16,0,0.5245120127995809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,2,128,1,float16,fp8,0,0.37004268169403076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,48,128,1,float16,fp8,0,0.5142506758371989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,4,128,1,float16,float16,0,0.4318026701609294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,4,128,1,float16,fp8,0,0.39842132727305096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,8,128,1,float16,float16,0,0.43220798174540204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,48,8,128,1,float16,fp8,0,0.3972800175348918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,4,128,1,float16,fp8,0,1.4285227457682292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,1,128,1,float16,float16,0,0.22156266371409097
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,2,128,1,float16,float16,0,0.22470400730768839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,1,128,1,float16,fp8,0,0.2045066754023234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,2,128,1,float16,fp8,0,0.2047413388888041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,4,128,1,float16,float16,0,0.23581333955128989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,48,8,128,1,float16,fp8,0,1.4515573183695476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,4,128,1,float16,fp8,0,0.21761600176493326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,8,128,1,float16,fp8,0,0.21949867407480875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,48,128,1,float16,float16,0,0.277834673722585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,48,128,1,float16,fp8,0,0.2789813280105591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,48,8,128,1,float16,float16,0,0.23998934030532837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,2,128,1,float16,fp8,0,3.1732266743977866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,1,128,1,float16,fp8,0,3.069845199584961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,1,128,1,float16,float16,0,3.625253359476725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,2,128,1,float16,float16,0,3.671205202738444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,4,128,1,float16,fp8,0,3.4519360860188804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,4,128,1,float16,float16,0,3.96285343170166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,8,128,1,float16,float16,0,3.930389404296875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,1,128,1,float16,float16,0,1.8005493481953938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,1,128,1,float16,fp8,0,1.5522400538126628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,2,128,1,float16,float16,0,1.953312079111735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,2,128,1,float16,fp8,0,1.7246185938517253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,4,128,1,float16,float16,0,1.9769493738810222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,48,8,128,1,float16,fp8,0,3.516458511352539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,48,128,1,float16,float16,0,2.456826686859131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,4,128,1,float16,fp8,0,1.8981067339579265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,8,128,1,float16,float16,0,2.0111145973205566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,1,128,1,float16,float16,0,0.9093226591746012
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,8,128,1,float16,fp8,0,1.8702880541483562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,2,128,1,float16,fp8,0,0.8202986717224121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,2,128,1,float16,float16,0,0.9347093105316162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,48,128,1,float16,float16,0,1.23363200823466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,1,128,1,float16,fp8,0,0.7985386848449707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,4,128,1,float16,fp8,0,0.9196320374806722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,4,128,1,float16,float16,0,0.9941279888153076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,48,128,1,float16,fp8,0,1.2532320022583008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,8,128,1,float16,fp8,0,0.9247679710388184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,1,128,1,float16,fp8,0,0.41894932587941486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,1,128,1,float16,float16,0,0.4726773500442505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,2,128,1,float16,float16,0,0.4818613529205322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,2,128,1,float16,fp8,0,0.43668798605600995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,4,128,1,float16,fp8,0,0.4691093365351359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,4,128,1,float16,float16,0,0.5074559847513834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,48,128,1,float16,float16,0,0.6249173482259115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,48,128,1,float16,fp8,0,0.6403946479161581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,8,128,1,float16,fp8,0,0.4755306641260783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,48,8,128,1,float16,float16,0,0.5138346751530966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,1,128,1,float16,float16,0,0.25494933128356934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,2,128,1,float16,fp8,0,0.23965332905451456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,2,128,1,float16,float16,0,0.2582826614379883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,48,48,128,1,float16,fp8,0,2.5265493392944336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,4,128,1,float16,fp8,0,0.25779199600219727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,48,128,1,float16,fp8,0,0.3413066864013672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,48,128,1,float16,float16,0,0.34301332632700604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,8,128,1,float16,fp8,0,0.25778667132059735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,1,128,1,float16,float16,0,0.1437333325544993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,1,128,1,float16,fp8,0,0.132533331712087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,2,128,1,float16,float16,0,0.145306666692098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,48,128,1,float16,fp8,0,0.19079999128977457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,48,8,128,1,float16,float16,0,1.0100373427073162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,4,128,1,float16,float16,0,0.14948266744613647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,4,128,1,float16,fp8,0,0.14225066701571146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,8,128,1,float16,float16,0,0.15212800105412802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,8,128,1,float16,fp8,0,0.14486933747927347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,1,128,1,float16,fp8,0,0.23081600666046143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,4,128,1,float16,float16,0,0.27558932701746625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,48,8,128,1,float16,float16,0,0.27450132369995117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,1,128,1,float16,float16,0,3.546565373738607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,2,128,1,float16,float16,0,3.687514623006185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,48,128,1,float16,float16,0,0.20224533478418985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,2,128,1,float16,fp8,0,3.2372213999430337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,4,128,1,float16,fp8,0,3.618618647257487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,4,128,1,float16,float16,0,3.9894240697224936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,8,128,1,float16,float16,0,4.167728106180827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,8,128,1,float16,fp8,0,3.7735039393107095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,1,128,1,float16,float16,0,1.7954079310099285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,1,128,1,float16,fp8,0,1.674880027770996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,48,1,128,1,float16,fp8,0,3.094853401184082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,48,128,1,float16,float16,0,2.6444907188415527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,2,128,1,float16,float16,0,1.8414133389790852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,48,128,1,float16,fp8,0,2.756143887837728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,2,128,1,float16,fp8,0,1.6394826571146648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,48,2,128,1,float16,fp8,0,0.13404800494511923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,4,128,1,float16,float16,0,2.013823986053467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,4,128,1,float16,fp8,0,1.866714636484782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,8,128,1,float16,float16,0,2.0469226837158203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,48,8,128,1,float16,fp8,0,1.8606613477071126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,1,128,1,float16,float16,0,0.9051146507263184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,1,128,1,float16,fp8,0,0.7987360159556071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,2,128,1,float16,float16,0,0.9287146727244059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,2,128,1,float16,fp8,0,0.8388853073120117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,4,128,1,float16,float16,0,1.0151786804199219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,48,128,1,float16,fp8,0,1.3916319211324055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,4,128,1,float16,fp8,0,0.930560032526652
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,48,128,1,float16,float16,0,1.3324693044026692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,8,128,1,float16,float16,0,1.0244159698486328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,48,8,128,1,float16,fp8,0,0.9435413678487142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,1,128,1,float16,fp8,0,0.4153226613998413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,2,128,1,float16,fp8,0,0.43231467405955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,2,128,1,float16,float16,0,0.4718559980392456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,4,128,1,float16,fp8,0,0.48478933175404865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,48,128,1,float16,fp8,0,0.7104053497314453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,48,128,1,float16,float16,0,0.6702826817830404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,1,128,1,float16,float16,0,0.250874658425649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,8,128,1,float16,fp8,0,0.49035199483235675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,8,128,1,float16,float16,0,0.5214133262634277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,1,128,1,float16,fp8,0,0.22882666190465292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,48,128,1,float16,fp8,0,0.3720426559448242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,48,128,1,float16,float16,0,0.3808319965998332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,2,128,1,float16,float16,0,0.24839999278386435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,2,128,1,float16,fp8,0,0.23415466149648032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,4,128,1,float16,float16,0,0.27452800671259564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,4,128,1,float16,fp8,0,0.25674132506052655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,8,128,1,float16,float16,0,0.27618666489919025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,48,8,128,1,float16,fp8,0,0.2630560000737508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,1,128,1,float16,float16,0,0.1357439955075582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,2,128,1,float16,fp8,0,0.1301866670449575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,1,128,1,float16,fp8,0,0.12713600198427835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,2,128,1,float16,float16,0,0.14123732844988504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,1,128,1,float16,float16,0,0.4607359965642293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,4,128,1,float16,float16,0,0.1515679955482483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,4,128,1,float16,fp8,0,0.13959999879201254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,48,128,1,float16,float16,0,0.19048533837000528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,8,128,1,float16,float16,0,0.15412267049153647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,8,128,1,float16,fp8,0,0.1434986690680186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,1,128,1,float16,float16,0,0.08228266735871632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,48,4,128,1,float16,float16,0,0.5171626806259155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,1,128,1,float16,fp8,0,0.07827199995517731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,2,128,1,float16,fp8,0,0.0790293316046397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,2,128,1,float16,float16,0,0.08364799618721008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,4,128,1,float16,float16,0,0.08613333106040955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,4,128,1,float16,fp8,0,0.0825439989566803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,48,128,1,float16,fp8,0,0.1195093293984731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,48,128,1,float16,float16,0,0.1220960021018982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,8,128,1,float16,fp8,0,0.08452266454696655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,1,128,1,float16,fp8,0,1.9660479227701824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,1,128,1,float16,float16,0,2.2019306818644204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,4,128,1,float16,float16,0,2.575215975443522
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,4,128,1,float16,fp8,0,2.351679960886637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,2,128,1,float16,float16,0,2.2976853052775064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,48,48,128,1,float16,fp8,0,0.20242132743199667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,1,128,1,float16,float16,0,1.1171786785125732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,1,128,1,float16,fp8,0,0.9947360356648763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,8,128,1,float16,float16,0,2.5926879247029624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,48,8,128,1,float16,float16,0,0.08893332878748576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,8,128,1,float16,fp8,0,2.45853328704834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,48,128,1,float16,float16,0,1.7703359921773274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,48,128,1,float16,fp8,0,1.884127934773763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,2,128,1,float16,float16,0,1.15337069829305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,2,128,1,float16,fp8,0,1.046224037806193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,4,128,1,float16,float16,0,1.288266658782959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,4,128,1,float16,fp8,0,1.1897546450297039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,8,128,1,float16,float16,0,1.290058692296346
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,1,128,1,float16,float16,0,0.5619093179702759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,48,8,128,1,float16,fp8,0,1.2115840117136638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,2,128,1,float16,float16,0,0.5856373310089111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,48,2,128,1,float16,fp8,0,2.064751942952474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,2,128,1,float16,fp8,0,0.5374933481216431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,48,128,1,float16,float16,0,0.8848959604899088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,4,128,1,float16,float16,0,0.6431039969126383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,4,128,1,float16,fp8,0,0.6102613210678101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,48,128,1,float16,fp8,0,0.9547999699910482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,8,128,1,float16,float16,0,0.6559786796569824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,1,128,1,float16,float16,0,0.2959679961204529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,1,128,1,float16,fp8,0,0.27107733488082886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,8,128,1,float16,fp8,0,0.619978666305542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,8,128,1,float16,float16,0,0.33878934383392334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,4,128,1,float16,float16,0,0.3397333224614461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,2,128,1,float16,fp8,0,0.28543466329574585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,8,128,1,float16,fp8,0,0.3282346725463867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,48,128,1,float16,float16,0,0.44971732298533124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,1,128,1,float16,float16,0,0.1621280014514923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,48,128,1,float16,float16,0,0.2718399961789449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,2,128,1,float16,float16,0,0.16701332728068033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,48,128,1,float16,fp8,0,0.2619946599006653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,48,1,128,1,float16,fp8,0,0.5183466672897339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,2,128,1,float16,fp8,0,0.1595253348350525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,4,128,1,float16,float16,0,0.18066666523615518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,8,128,1,float16,float16,0,0.1845866640408834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,8,128,1,float16,fp8,0,0.1770026683807373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,1,128,1,float16,float16,0,0.0926080048084259
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,1,128,1,float16,fp8,0,0.08558400472005208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,48,128,1,float16,fp8,0,0.1464853286743164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,48,128,1,float16,fp8,0,0.49084798494974774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,2,128,1,float16,fp8,0,0.08979733784993489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,4,128,1,float16,float16,0,0.10013866424560547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,2,128,1,float16,float16,0,0.30664533376693726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,4,128,1,float16,fp8,0,0.09583466251691182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,8,128,1,float16,float16,0,0.10276800394058228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,8,128,1,float16,fp8,0,0.09602666894594829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,1,128,1,float16,fp8,0,0.15229333440462747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,1,128,1,float16,float16,0,0.05806399881839752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,48,128,1,float16,float16,0,0.08647466699282329
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,48,4,128,1,float16,fp8,0,0.17695466677347818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,48,128,1,float16,fp8,0,0.08162666857242584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,2,128,1,float16,float16,0,0.059631998340288796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,2,128,1,float16,fp8,0,0.05600533386071523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,4,128,1,float16,float16,0,0.06161599854628245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,4,128,1,float16,fp8,0,0.058970664938290916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,8,128,1,float16,float16,0,0.061610668897628784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,8,128,1,float16,fp8,0,0.05945600072542826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,48,128,1,float16,float16,0,0.1507253348827362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,48,2,128,1,float16,float16,0,0.09511466821034749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,48,4,128,1,float16,fp8,0,0.319541335105896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,1,128,1,float16,float16,0,2.328981399536133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,1,128,1,float16,fp8,0,2.1192906697591147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,2,128,1,float16,float16,0,2.445312023162842
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,48,1,128,1,float16,fp8,0,0.05346133311589559
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,2,128,1,float16,fp8,0,2.2522613207499185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,4,128,1,float16,fp8,0,2.631594657897949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,4,128,1,float16,float16,0,2.7618560791015625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,1,128,1,float16,float16,0,1.1570453643798828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,8,128,1,float16,float16,0,2.815082550048828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,1,128,1,float16,fp8,0,1.0673973560333252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,48,128,1,float16,float16,0,2.0383146603902182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,48,8,128,1,float16,fp8,0,2.7096853256225586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,48,128,1,float16,fp8,0,2.2748053868611655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,4,128,1,float16,float16,0,1.3874613444010417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,2,128,1,float16,fp8,0,1.1588586966196697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,8,128,1,float16,float16,0,1.411237398783366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,8,128,1,float16,fp8,0,1.3583520253499348
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,1,128,1,float16,float16,0,0.5869653224945068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,1,128,1,float16,fp8,0,0.5494879881540934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,48,128,1,float16,float16,0,1.0199146270751953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,2,128,1,float16,fp8,0,0.5801013310750326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,2,128,1,float16,float16,0,0.619925340016683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,4,128,1,float16,float16,0,0.701807975769043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,4,128,1,float16,fp8,0,0.674074649810791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,8,128,1,float16,float16,0,0.7166079680124918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,8,128,1,float16,fp8,0,0.6928479671478271
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,1,128,1,float16,fp8,0,0.2871039907137553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,2,128,1,float16,float16,0,0.32125866413116455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,1,128,1,float16,float16,0,0.31306666135787964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,48,128,1,float16,float16,0,0.5166773398717245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,2,128,1,float16,fp8,0,0.30322132507960003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,48,128,1,float16,fp8,0,0.5784639914830526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,2,128,1,float16,float16,0,1.2341759999593098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,4,128,1,float16,float16,0,0.3612906535466512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,4,128,1,float16,fp8,0,0.3527946472167969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,8,128,1,float16,float16,0,0.368613322575887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,1,128,1,float16,float16,0,0.168938676516215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,48,8,128,1,float16,fp8,0,0.360858678817749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,48,4,128,1,float16,fp8,0,1.3291786511739094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,1,128,1,float16,fp8,0,0.1572213371594747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,2,128,1,float16,float16,0,0.18001067638397217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,2,128,1,float16,fp8,0,0.1653600037097931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,4,128,1,float16,float16,0,0.196234663327535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,48,48,128,1,float16,fp8,0,1.1368213494618733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,48,128,1,float16,float16,0,0.2689653237660726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,8,128,1,float16,float16,0,0.20161600907643637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,48,128,1,float16,fp8,0,0.30212799708048504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,8,128,1,float16,fp8,0,0.1929546594619751
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,1,128,1,float16,float16,0,0.09595200419425964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,2,128,1,float16,fp8,0,0.09165333708127339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,4,128,1,float16,fp8,0,0.09939733147621155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,2,128,1,float16,float16,0,0.10012267033259074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,48,128,1,float16,float16,0,0.14833600322405496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,48,128,1,float16,fp8,0,0.16316266854604086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,8,128,1,float16,float16,0,0.11158933242162068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,8,128,1,float16,fp8,0,0.1034879982471466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,48,4,128,1,float16,fp8,0,0.18920532862345377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,1,128,1,float16,fp8,0,0.053029333551724754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,2,128,1,float16,float16,0,0.0581226646900177
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,4,128,1,float16,float16,0,0.06041066845258077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,2,128,1,float16,fp8,0,0.055311997731526695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,4,128,1,float16,fp8,0,0.06044800082842509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,8,128,1,float16,float16,0,0.061493332187334694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,48,128,1,float16,float16,0,0.09156266848246257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,8,128,1,float16,fp8,0,0.06052800019582113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,1,128,1,float16,float16,0,0.0396373321612676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,1,128,1,float16,fp8,0,0.037685332198937736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,2,128,1,float16,float16,0,0.04026666780312856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,2,128,1,float16,fp8,0,0.03754133234421412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,48,128,1,float16,fp8,0,0.05411200225353241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,48,128,1,float16,float16,0,0.049423997600873314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,8,128,1,float16,float16,0,0.04170133173465729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,8,128,1,float16,fp8,0,0.04173333446184794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,4,128,1,float16,float16,0,0.11007466912269592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,48,1,128,1,float16,fp8,0,0.08714666962623596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,1,128,1,float16,float16,0,0.056874667604764305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,48,48,128,1,float16,fp8,0,0.09225066502888997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,1,128,1,float16,float16,0,1.7371253967285156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,1,128,1,float16,fp8,0,1.6153440475463867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,2,128,1,float16,float16,0,1.865872065226237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,4,128,1,float16,fp8,0,0.03980266551176707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,2,128,1,float16,fp8,0,1.7506613731384277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,4,128,1,float16,float16,0,2.2082932790120444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,4,128,1,float16,fp8,0,2.132319927215576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,1,128,1,float16,float16,0,0.8752586841583252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,48,128,1,float16,float16,0,1.7158239682515461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,48,128,1,float16,fp8,0,1.9887785911560059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,8,128,1,float16,float16,0,2.233242670694987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,48,8,128,1,float16,fp8,0,2.2033279736836753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,2,128,1,float16,float16,0,0.9403839906056722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,48,4,128,1,float16,float16,0,0.04186133543650309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,2,128,1,float16,fp8,0,0.8849759896596273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,1,128,1,float16,fp8,0,0.8165706793467203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,4,128,1,float16,float16,0,1.1091307004292805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,1,128,1,float16,float16,0,0.4522133270899455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,48,128,1,float16,float16,0,0.8640213012695312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,8,128,1,float16,float16,0,1.124224026997884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,2,128,1,float16,float16,0,0.48074134190877277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,4,128,1,float16,fp8,0,1.073466698328654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,48,128,1,float16,fp8,0,1.0027466615041096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,2,128,1,float16,fp8,0,0.4581173261006673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,1,128,1,float16,fp8,0,0.4243893225987752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,4,128,1,float16,fp8,0,0.552186648050944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,8,128,1,float16,float16,0,0.5744479894638062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,4,128,1,float16,float16,0,0.5638346672058105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,1,128,1,float16,float16,0,0.2397600015004476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,48,8,128,1,float16,fp8,0,0.5621386766433716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,48,128,1,float16,fp8,0,0.512234648068746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,2,128,1,float16,fp8,0,0.23668799797693887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,48,128,1,float16,float16,0,0.4429493347803752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,1,128,1,float16,fp8,0,0.22159467140833536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,4,128,1,float16,float16,0,0.2937440077463786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,48,8,128,1,float16,fp8,0,1.1048213640848796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,4,128,1,float16,fp8,0,0.29125332832336426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,1,128,1,float16,float16,0,0.1322719951470693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,8,128,1,float16,fp8,0,0.2933066685994466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,2,128,1,float16,float16,0,0.1393173336982727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,8,128,1,float16,float16,0,0.30182933807373047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,48,128,1,float16,float16,0,0.24506133794784546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,4,128,1,float16,fp8,0,0.1527253290017446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,48,128,1,float16,fp8,0,0.26623467604319256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,4,128,1,float16,float16,0,0.1588053305943807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,8,128,1,float16,float16,0,0.1621280014514923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,1,128,1,float16,fp8,0,0.12231466174125671
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,1,128,1,float16,float16,0,0.07393066585063934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,2,128,1,float16,float16,0,0.07937600215276082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,8,128,1,float16,fp8,0,0.1646399994691213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,48,128,1,float16,fp8,0,0.14414933323860168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,4,128,1,float16,fp8,0,0.08053866525491078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,48,128,1,float16,float16,0,0.12769066294034323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,4,128,1,float16,float16,0,0.08870399991671245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,8,128,1,float16,float16,0,0.08964799841245015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,8,128,1,float16,fp8,0,0.08417066931724548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,1,128,1,float16,fp8,0,0.041797334949175514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,1,128,1,float16,float16,0,0.044810667634010315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,48,2,128,1,float16,float16,0,0.2526400089263916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,2,128,1,float16,float16,0,0.045594667394955955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,48,128,1,float16,float16,0,0.08273066580295563
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,4,128,1,float16,float16,0,0.0484746644894282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,48,128,1,float16,fp8,0,0.08057599763075511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,8,128,1,float16,float16,0,0.049770668148994446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,48,2,128,1,float16,fp8,0,0.12993599971135458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,1,128,1,float16,float16,0,0.02945599953333537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,1,128,1,float16,fp8,0,0.029333333174387615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,48,128,1,float16,fp8,0,0.045834665497144066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,48,128,1,float16,float16,0,0.0397173340121905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,2,128,1,float16,float16,0,0.03147733211517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,4,128,1,float16,fp8,0,0.03268266717592875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,2,128,1,float16,fp8,0,0.03081600119670232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,2,128,1,float16,fp8,0,0.07044266661008199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,48,1,128,1,float16,fp8,0,0.06819200019041698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,8,128,1,float16,float16,0,0.033573334415753685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,8,128,1,float16,fp8,0,0.03276266654332479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,1,128,1,float16,float16,0,0.021194666624069214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,1,128,1,float16,fp8,0,0.020703999946514767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,2,128,1,float16,float16,0,0.020981334149837494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,2,128,1,float16,fp8,0,0.01956266661485036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,4,128,1,float16,float16,0,0.021226666867733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,48,128,1,float16,fp8,0,0.030346666773160298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,4,128,1,float16,fp8,0,0.021226666867733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,48,128,1,float16,float16,0,0.025114665428797405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,8,128,1,float16,float16,0,0.021498667697111767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,48,8,128,1,float16,fp8,0,0.02292266736427943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,2,128,1,float16,fp8,0,0.042912001411120095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,8,128,1,float16,fp8,0,0.04786133269468943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,1,128,1,float16,float16,0,0.7074666817982992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,1,128,1,float16,fp8,0,0.6791893641153971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,2,128,1,float16,fp8,0,0.7449599901835123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,2,128,1,float16,float16,0,0.7721172968546549
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,48,4,128,1,float16,float16,0,0.03281066566705704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,4,128,1,float16,float16,0,0.9425013065338135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,4,128,1,float16,fp8,0,0.9354133605957031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,1,128,1,float16,float16,0,0.363322655359904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,8,128,1,float16,float16,0,0.9525439739227295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,48,4,128,1,float16,fp8,0,0.04763199885686239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,1,128,1,float16,fp8,0,0.3481813271840413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,48,8,128,1,float16,fp8,0,0.9624479611714681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,2,128,1,float16,float16,0,0.397274653116862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,48,128,1,float16,float16,0,0.7742453416188558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,2,128,1,float16,fp8,0,0.38391466935475665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,48,128,1,float16,fp8,0,0.9319252967834473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,1,128,1,float16,float16,0,0.19346133867899576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,1,128,1,float16,fp8,0,0.18294399976730347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,4,128,1,float16,fp8,0,0.475872000058492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,4,128,1,float16,float16,0,0.47680532932281494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,48,128,1,float16,fp8,0,0.4734239975611369
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,8,128,1,float16,fp8,0,0.48901331424713135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,2,128,1,float16,float16,0,0.20588266849517822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,2,128,1,float16,fp8,0,0.19977599382400513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,48,128,1,float16,float16,0,0.397055983543396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,4,128,1,float16,float16,0,0.2509920001029968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,8,128,1,float16,fp8,0,0.254202663898468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,4,128,1,float16,fp8,0,0.25698665777842206
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,1,128,1,float16,float16,0,0.1090826690196991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,48,8,128,1,float16,float16,0,0.2539786696434021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,2,128,1,float16,fp8,0,0.10925867160161336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,48,128,1,float16,fp8,0,0.24811200300852457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,1,128,1,float16,fp8,0,0.10121066371599834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,2,128,1,float16,float16,0,0.11409599582354228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,4,128,1,float16,float16,0,0.134634663661321
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,48,128,1,float16,float16,0,0.20966400702794394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,4,128,1,float16,fp8,0,0.1325706640879313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,8,128,1,float16,float16,0,0.13688533504803976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,48,8,128,1,float16,fp8,0,0.13685866196950278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,1,128,1,float16,fp8,0,0.05821333328882853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,2,128,1,float16,fp8,0,0.06136000156402588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,4,128,1,float16,float16,0,0.07667199770609538
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,2,128,1,float16,float16,0,0.06647466619809468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,4,128,1,float16,fp8,0,0.069882666071256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,48,128,1,float16,float16,0,0.11549866199493408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,48,128,1,float16,fp8,0,0.13387733697891235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,48,8,128,1,float16,float16,0,0.4867466688156128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,8,128,1,float16,fp8,0,0.07406933108965556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,1,128,1,float16,float16,0,0.03571200122435888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,2,128,1,float16,float16,0,0.03730133424202601
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,1,128,1,float16,fp8,0,0.03356266766786575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,2,128,1,float16,fp8,0,0.03571200122435888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,48,128,1,float16,fp8,0,0.07214400172233582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,4,128,1,float16,fp8,0,0.03962666789690653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,4,128,1,float16,float16,0,0.039706667264302574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,48,128,1,float16,float16,0,0.06417066852251689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,8,128,1,float16,fp8,0,0.04144533226887385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,48,8,128,1,float16,float16,0,0.04005333284536997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,1,128,1,float16,float16,0,0.025248001019159954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,2,128,1,float16,fp8,0,0.026698666314284008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,2,128,1,float16,float16,0,0.02717333287000656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,4,128,1,float16,float16,0,0.027797333896160126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,4,128,1,float16,fp8,0,0.02743999908367793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,8,128,1,float16,float16,0,0.027727998793125153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,8,128,1,float16,fp8,0,0.029306667546431225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,1,128,1,float16,float16,0,0.06392000118891399
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,1,128,1,float16,float16,0,0.017045332739750545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,1,128,1,float16,fp8,0,0.017125333348910015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,2,128,1,float16,float16,0,0.017082666357358296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,2,128,1,float16,fp8,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,4,128,1,float16,float16,0,0.018858666221300762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,48,128,1,float16,float16,0,0.021397332350413006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,4,128,1,float16,fp8,0,0.01887999971707662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,8,128,1,float16,float16,0,0.01894933357834816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,48,8,128,1,float16,float16,0,0.07856533428033192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,8,128,1,float16,fp8,0,0.019130667050679524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,1,128,1,float16,float16,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,1,128,1,float16,fp8,0,0.01676799977819125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,2,128,1,float16,float16,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,2,128,1,float16,fp8,0,0.016805333395799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,4,128,1,float16,float16,0,0.016810666769742966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,4,128,1,float16,fp8,0,0.01691199963291486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,8,128,1,float16,fp8,0,0.017050666113694508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,8,128,1,float16,float16,0,0.01691199963291486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,1,128,1,float16,fp8,0,0.024634666740894318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,48,128,1,float16,fp8,0,0.04142399877309799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,48,48,128,1,float16,float16,0,0.03559466699759165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,1,128,1,float16,float16,0,0.43430932362874347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,1,128,1,float16,fp8,0,0.40860267480214435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,2,128,1,float16,float16,0,0.4676213264465332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,48,48,128,1,float16,fp8,0,0.027402666707833607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,2,128,1,float16,fp8,0,0.44838933149973553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,4,128,1,float16,float16,0,0.5467199881871542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,48,128,1,float16,fp8,0,0.02090666691462199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,4,128,1,float16,fp8,0,0.5354506572087606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,48,48,128,1,float16,float16,0,0.018826667219400406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,1,128,1,float16,float16,0,0.22735466559727988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,8,128,1,float16,float16,0,0.5546720027923584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,48,8,128,1,float16,fp8,0,0.5515413284301758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,48,128,1,float16,float16,0,0.45828266938527423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,1,128,1,float16,fp8,0,0.21360532442728677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,48,128,1,float16,fp8,0,0.5027519861857096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,2,128,1,float16,float16,0,0.243231991926829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,2,128,1,float16,fp8,0,0.2291253407796224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,4,128,1,float16,float16,0,0.2858186761538188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,4,128,1,float16,fp8,0,0.27937599023183185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,1,128,1,float16,float16,0,0.12402133146921794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,1,128,1,float16,fp8,0,0.11421333750089009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,2,128,1,float16,float16,0,0.13115732868512472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,8,128,1,float16,fp8,0,0.2993973294893901
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,2,128,1,float16,fp8,0,0.12377599875132243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,48,128,1,float16,fp8,0,0.26020266612370807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,4,128,1,float16,float16,0,0.15076800187428793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,48,128,1,float16,float16,0,0.2225386699040731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,4,128,1,float16,fp8,0,0.1477173368136088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,8,128,1,float16,float16,0,0.15476800004641214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,2,128,1,float16,float16,0,0.07384000221888225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,48,8,128,1,float16,fp8,0,0.15270933508872986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,1,128,1,float16,fp8,0,0.06243200103441874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,2,128,1,float16,fp8,0,0.06597333153088887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,4,128,1,float16,float16,0,0.08233066896597545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,48,128,1,float16,fp8,0,0.13865066568056741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,48,128,1,float16,float16,0,0.11980799833933513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,8,128,1,float16,fp8,0,0.08055999875068665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,1,128,1,float16,float16,0,0.039434666434923805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,1,128,1,float16,fp8,0,0.03749866783618927
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,48,128,1,float16,fp8,0,0.07657599945863088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,2,128,1,float16,float16,0,0.041536000867684685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,48,128,1,float16,float16,0,0.0755626658598582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,2,128,1,float16,fp8,0,0.03975466638803482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,4,128,1,float16,float16,0,0.04349866509437561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,48,8,128,1,float16,float16,0,0.28838932514190674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,4,128,1,float16,fp8,0,0.04404800136884054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,8,128,1,float16,fp8,0,0.04379733403523763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,48,8,128,1,float16,float16,0,0.04486933350563049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,1,128,1,float16,float16,0,0.025114665428797405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,1,128,1,float16,fp8,0,0.023376000424226124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,2,128,1,float16,fp8,0,0.025146665672461193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,1,128,1,float16,float16,0,0.06897066533565521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,4,128,1,float16,fp8,0,0.0271519993742307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,48,128,1,float16,fp8,0,0.040181333820025124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,4,128,1,float16,float16,0,0.02749866743882497
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,48,128,1,float16,float16,0,0.03355200091997782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,8,128,1,float16,float16,0,0.02701866626739502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,8,128,1,float16,fp8,0,0.027136000494162243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,4,128,1,float16,fp8,0,0.07508266468842824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,2,128,1,float16,float16,0,0.01904533306757609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,1,128,1,float16,fp8,0,0.017642666896184284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,48,8,128,1,float16,float16,0,0.08388266960779826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,4,128,1,float16,float16,0,0.019173332800467808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,4,128,1,float16,fp8,0,0.019354666272799175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,48,128,1,float16,float16,0,0.023317334552605946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,48,128,1,float16,fp8,0,0.027632000545660656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,1,128,1,float16,float16,0,0.012975999464591345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,1,128,1,float16,fp8,0,0.012944000462690989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,2,128,1,float16,float16,0,0.013023999830087027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,48,128,1,float16,float16,0,0.014874666929244995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,2,128,1,float16,fp8,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,4,128,1,float16,float16,0,0.013125333935022354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,4,128,1,float16,fp8,0,0.01302933320403099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,8,128,1,float16,float16,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,8,128,1,float16,fp8,0,0.013013333082199097
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,48,2,128,1,float16,float16,0,0.025008000433444977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,1,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,1,128,1,float16,fp8,0,0.0129120002190272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,2,128,1,float16,float16,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,48,128,1,float16,float16,0,0.014933332800865173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,48,128,1,float16,fp8,0,0.015189333508412043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,2,128,1,float16,fp8,0,0.012746666868527731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,2,128,1,float16,fp8,0,0.01893866683046023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,4,128,1,float16,fp8,0,0.012698666503032049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,8,128,1,float16,float16,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,8,128,1,float16,fp8,0,0.01309866706530253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,8,128,1,float16,fp8,0,0.020975999534130096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,8,128,1,float16,float16,0,0.01930133377512296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,48,48,128,1,float16,fp8,0,0.018976000448067982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,1,128,1,float16,float16,0,0.3518079916636149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,2,128,1,float16,float16,0,0.36501868565877277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,1,128,1,float16,fp8,0,0.31623466809590656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,2,128,1,float16,fp8,0,0.333840012550354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,4,128,1,float16,float16,0,0.4078559875488281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,48,1,128,1,float16,float16,0,0.018933333456516266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,48,4,128,1,float16,float16,0,0.013194666554530462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,8,128,1,float16,float16,0,0.41492267449696857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,4,128,1,float16,fp8,0,0.3814133405685425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,48,8,128,1,float16,fp8,0,0.3909813165664673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,1,128,1,float16,float16,0,0.1865440011024475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,1,128,1,float16,fp8,0,0.1678559978802999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,2,128,1,float16,float16,0,0.1949546734491984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,48,128,1,float16,float16,0,0.2861706614494324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,2,128,1,float16,fp8,0,0.17645333210627237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,4,128,1,float16,float16,0,0.21357866128285727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,4,128,1,float16,fp8,0,0.19821866353352866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,8,128,1,float16,float16,0,0.2150613268216451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,8,128,1,float16,fp8,0,0.2023893396059672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,1,128,1,float16,float16,0,0.102783997853597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,1,128,1,float16,fp8,0,0.09026133020718892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,2,128,1,float16,float16,0,0.10550933082898457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,2,128,1,float16,fp8,0,0.09305066863695781
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,48,128,1,float16,float16,0,0.1659999986489614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,8,128,1,float16,fp8,0,0.1070240040620168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,4,128,1,float16,float16,0,0.11474133531252544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,8,128,1,float16,float16,0,0.11711466312408447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,48,128,1,float16,fp8,0,0.16371200482050577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,1,128,1,float16,float16,0,0.05535466472307841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,1,128,1,float16,fp8,0,0.05006400247414907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,2,128,1,float16,float16,0,0.056799997886021934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,4,128,1,float16,float16,0,0.05983999868233999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,48,128,1,float16,float16,0,0.089519997437795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,48,128,1,float16,fp8,0,0.08885332942008972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,8,128,1,float16,float16,0,0.06065600117047628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,8,128,1,float16,fp8,0,0.05801066756248474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,2,128,1,float16,fp8,0,0.052933335304260254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,1,128,1,float16,float16,0,0.0331839993596077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,2,128,1,float16,float16,0,0.03408000121514002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,48,48,128,1,float16,fp8,0,0.3121653397878011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,2,128,1,float16,fp8,0,0.03146133323510488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,4,128,1,float16,float16,0,0.03581333408753077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,48,128,1,float16,fp8,0,0.04764799773693085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,48,128,1,float16,float16,0,0.04185600082079569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,8,128,1,float16,float16,0,0.0355679988861084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,8,128,1,float16,fp8,0,0.03419200082619985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,48,4,128,1,float16,fp8,0,0.10517866412798564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,4,128,1,float16,fp8,0,0.03389866650104523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,48,128,1,float16,float16,0,0.025370667378107708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,1,128,1,float16,fp8,0,0.02067199970285098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,2,128,1,float16,float16,0,0.021130666136741638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,2,128,1,float16,fp8,0,0.020303999384244282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,4,128,1,float16,float16,0,0.022463999688625336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,4,128,1,float16,fp8,0,0.021226666867733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,48,4,128,1,float16,fp8,0,0.0561706672112147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,8,128,1,float16,float16,0,0.024288001159826916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,8,128,1,float16,fp8,0,0.02258666604757309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,1,128,1,float16,float16,0,0.016544000556071598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,1,128,1,float16,fp8,0,0.015066667149464289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,2,128,1,float16,float16,0,0.016544000556071598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,2,128,1,float16,fp8,0,0.015119999647140503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,48,1,128,1,float16,fp8,0,0.03049066662788391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,4,128,1,float16,float16,0,0.01709866647919019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,4,128,1,float16,fp8,0,0.016783999900023144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,8,128,1,float16,float16,0,0.016906666258970898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,48,128,1,float16,fp8,0,0.02103466788927714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,8,128,1,float16,fp8,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,1,128,1,float16,fp8,0,0.01312000056107839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,1,128,1,float16,float16,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,2,128,1,float16,float16,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,2,128,1,float16,fp8,0,0.012879999975363413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,48,128,1,float16,fp8,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,4,128,1,float16,float16,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,4,128,1,float16,fp8,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,8,128,1,float16,float16,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,48,128,1,float16,float16,0,0.0138026662170887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,1,128,1,float16,float16,0,0.011370666325092316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,1,128,1,float16,float16,0,0.0210506667693456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,1,128,1,float16,fp8,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,2,128,1,float16,float16,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,48,128,1,float16,fp8,0,0.015072000523408255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,48,128,1,float16,float16,0,0.013199999928474426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,4,128,1,float16,float16,0,0.01137599969903628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,4,128,1,float16,fp8,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,8,128,1,float16,float16,0,0.011717333147923151
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,48,48,128,1,float16,float16,0,0.018181333939234417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,1,128,1,float16,fp8,0,0.2823893427848816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,2,128,1,float16,float16,0,0.33769599596659344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,48,8,128,1,float16,fp8,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,2,128,1,float16,fp8,0,0.2874720096588135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,4,128,1,float16,float16,0,0.34991999467213947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,48,48,128,1,float16,fp8,0,0.029333333174387615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,4,128,1,float16,fp8,0,0.31141332785288495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,2,128,1,float16,fp8,0,0.012650666137536367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,48,8,128,1,float16,fp8,0,0.01268799975514412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,1,128,1,float16,float16,0,0.17593065897623697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,8,128,1,float16,fp8,0,0.31406400601069134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,8,128,1,float16,float16,0,0.3569493293762207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,1,128,1,float16,fp8,0,0.1451359987258911
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,48,128,1,float16,float16,0,0.22789865732192993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,2,128,1,float16,float16,0,0.1774079998334249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,2,128,1,float16,fp8,0,0.14863999684651694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,4,128,1,float16,float16,0,0.189903994401296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,8,128,1,float16,float16,0,0.18946667512257895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,48,1,128,1,float16,float16,0,0.3309813340504964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,4,128,1,float16,fp8,0,0.15973333517710367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,1,128,1,float16,float16,0,0.09330133597056071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,1,128,1,float16,fp8,0,0.07859733204046886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,48,128,1,float16,float16,0,0.12220799922943115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,48,128,1,float16,fp8,0,0.1165013313293457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,2,128,1,float16,float16,0,0.09486400087674458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,4,128,1,float16,float16,0,0.0960640013217926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,8,128,1,float16,fp8,0,0.16517333189646402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,4,128,1,float16,fp8,0,0.08294933537642162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,1,128,1,float16,fp8,0,0.04353600243727366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,8,128,1,float16,fp8,0,0.08586666981379192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,2,128,1,float16,float16,0,0.05182399849096934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,48,128,1,float16,float16,0,0.06977066894372304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,48,128,1,float16,fp8,0,0.06193066636721293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,2,128,1,float16,fp8,0,0.04606399933497111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,4,128,1,float16,float16,0,0.053904001911481224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,4,128,1,float16,fp8,0,0.04863466819127401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,8,128,1,float16,float16,0,0.05407466491063436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,48,48,128,1,float16,fp8,0,0.22014933824539185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,1,128,1,float16,float16,0,0.031301334500312805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,48,128,1,float16,fp8,0,0.037578667203585304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,48,128,1,float16,float16,0,0.03527999917666117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,2,128,1,float16,float16,0,0.031136001149813335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,1,128,1,float16,fp8,0,0.027386667827765148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,2,128,1,float16,fp8,0,0.02735999971628189
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,4,128,1,float16,fp8,0,0.029152000943819683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,4,128,1,float16,float16,0,0.03207999964555105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,8,128,1,float16,float16,0,0.03292266776164373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,48,8,128,1,float16,fp8,0,0.029530666768550873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,2,128,1,float16,fp8,0,0.08037866652011871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,1,128,1,float16,fp8,0,0.018960000326236088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,1,128,1,float16,float16,0,0.019509332875410717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,2,128,1,float16,float16,0,0.01956266661485036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,48,128,1,float16,fp8,0,0.02314666658639908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,2,128,1,float16,fp8,0,0.018965333700180054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,4,128,1,float16,fp8,0,0.019173332800467808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,48,128,1,float16,float16,0,0.021295999487241108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,8,128,1,float16,float16,0,0.019893333315849304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,8,128,1,float16,fp8,0,0.018885333091020584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,48,8,128,1,float16,float16,0,0.09739200274149577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,1,128,1,float16,float16,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,2,128,1,float16,float16,0,0.015146666516860327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,1,128,1,float16,fp8,0,0.014949332922697067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,48,128,1,float16,fp8,0,0.01871466636657715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,48,128,1,float16,float16,0,0.01693333312869072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,4,128,1,float16,float16,0,0.014943999548753103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,4,128,1,float16,fp8,0,0.015008000036080679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,8,128,1,float16,fp8,0,0.04782933493455251
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,8,128,1,float16,float16,0,0.015237333873907724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,8,128,1,float16,fp8,0,0.015034666905800501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,1,128,1,float16,float16,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,1,128,1,float16,fp8,0,0.012122667084137598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,2,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,2,128,1,float16,fp8,0,0.012074666718641916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,4,128,1,float16,float16,0,0.011359999577204386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,48,128,1,float16,float16,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,48,128,1,float16,fp8,0,0.015013333410024643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,4,128,1,float16,fp8,0,0.011354666203260422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,8,128,1,float16,fp8,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,48,8,128,1,float16,float16,0,0.011727999895811081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,1,128,1,float16,float16,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,1,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,2,128,1,float16,float16,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,48,128,1,float16,fp8,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,2,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,4,128,1,float16,float16,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,48,4,128,1,float16,float16,0,0.020469332734743755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,4,128,1,float16,fp8,0,0.01163200040658315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,48,128,1,float16,float16,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,8,128,1,float16,fp8,0,0.011685332904259363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,48,1,128,1,float16,float16,0,0.0516480008761088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,48,8,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,48,2,128,1,float16,fp8,0,0.014896000425020853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,1,128,1,float16,float16,0,0.31725333134333294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,2,128,1,float16,fp8,0,0.26136000951131183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,4,128,1,float16,fp8,0,0.27484800418217975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,4,128,1,float16,float16,0,0.3306666612625122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,2,128,1,float16,float16,0,0.33078932762145996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,8,128,1,float16,float16,0,0.34098132451375324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,1,128,1,float16,fp8,0,0.1341973344484965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,48,128,1,float16,fp8,0,0.17270932594935098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,8,128,1,float16,fp8,0,0.27484800418217975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,4,128,1,float16,float16,0,0.16940265893936157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,4,128,1,float16,fp8,0,0.14065066973368326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,8,128,1,float16,float16,0,0.17458132902781168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,8,128,1,float16,fp8,0,0.14267733693122864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,1,128,1,float16,float16,0,0.08910399675369263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,48,1,128,1,float16,fp8,0,0.2572159965833028
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,48,128,1,float16,float16,0,0.1093386709690094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,48,128,1,float16,fp8,0,0.09481066465377808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,1,128,1,float16,fp8,0,0.07250666618347168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,2,128,1,float16,fp8,0,0.07489599784215291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,2,128,1,float16,float16,0,0.09085333347320557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,4,128,1,float16,fp8,0,0.07553599774837494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,2,128,1,float16,float16,0,0.17156267166137695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,8,128,1,float16,float16,0,0.09251200159390767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,8,128,1,float16,fp8,0,0.07623999814192454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,48,128,1,float16,float16,0,0.19765333334604898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,1,128,1,float16,float16,0,0.050255998969078064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,1,128,1,float16,fp8,0,0.04188266893227895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,2,128,1,float16,fp8,0,0.13569600383440653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,48,128,1,float16,float16,0,0.055434669057528176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,2,128,1,float16,fp8,0,0.04163199911514918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,4,128,1,float16,float16,0,0.05141866703828176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,48,128,1,float16,fp8,0,0.05159999926884969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,4,128,1,float16,fp8,0,0.04355733096599579
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,2,128,1,float16,float16,0,0.04996799925963084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,8,128,1,float16,float16,0,0.05193600058555603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,1,128,1,float16,float16,0,0.03030933439731598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,1,128,1,float16,fp8,0,0.027061333258946735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,48,8,128,1,float16,fp8,0,0.04383466641108195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,2,128,1,float16,float16,0,0.029440000653266907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,2,128,1,float16,fp8,0,0.027072000006834667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,4,128,1,float16,float16,0,0.030192000170548756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,4,128,1,float16,fp8,0,0.02665599932273229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,8,128,1,float16,float16,0,0.030405332644780476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,48,128,1,float16,float16,0,0.03147733211517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,48,128,1,float16,fp8,0,0.031285333136717476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,1,128,1,float16,float16,0,0.018922666708628338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,48,128,1,float16,fp8,0,0.021312000850836437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,48,4,128,1,float16,float16,0,0.0920906662940979
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,1,128,1,float16,fp8,0,0.017680000513792038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,2,128,1,float16,float16,0,0.019050666441520054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,8,128,1,float16,float16,0,0.019296000401178997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,8,128,1,float16,fp8,0,0.018272000054518383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,4,128,1,float16,fp8,0,0.01899733394384384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,1,128,1,float16,float16,0,0.015157333264748255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,1,128,1,float16,fp8,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,2,128,1,float16,float16,0,0.01498666654030482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,48,128,1,float16,fp8,0,0.017263999829689663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,2,128,1,float16,fp8,0,0.014965333044528961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,4,128,1,float16,float16,0,0.015226667126019796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,8,128,1,float16,float16,0,0.014805333067973455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,8,128,1,float16,fp8,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,48,8,128,1,float16,fp8,0,0.027066667874654133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,2,128,1,float16,fp8,0,0.01775466650724411
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,4,128,1,float16,float16,0,0.019237333287795384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,1,128,1,float16,float16,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,48,128,1,float16,float16,0,0.012874666601419449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,1,128,1,float16,fp8,0,0.011391999820868174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,48,1,128,1,float16,float16,0,0.16698133945465088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,2,128,1,float16,fp8,0,0.01257066677014033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,2,128,1,float16,float16,0,0.011482667177915573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,4,128,1,float16,float16,0,0.01209066684047381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,4,128,1,float16,fp8,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,8,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,8,128,1,float16,fp8,0,0.012671999633312225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,1,128,1,float16,float16,0,0.011349332829316458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,48,128,1,float16,float16,0,0.017125333348910015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,48,4,128,1,float16,fp8,0,0.014853333433469137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,2,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,2,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,48,128,1,float16,fp8,0,0.014783999572197596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,4,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,8,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,8,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,48,48,128,1,float16,float16,0,0.021301334102948506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,48,48,128,1,float16,fp8,0,0.014778666198253632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,48,4,128,1,float16,float16,0,0.022474666436513264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,48,8,128,1,float16,fp8,0,0.02534399926662445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,48,4,128,1,float16,fp8,0,0.019173332800467808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,48,128,1,float16,float16,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,48,4,128,1,float16,float16,0,0.011226666470368704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,48,128,1,float16,float16,0,0.08184533317883809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,48,128,1,float16,fp8,0,0.0641546646753947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,4,128,1,float16,float16,0,0.014991999914248785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,4,128,1,float16,fp8,0,0.014853333433469137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,8,128,1,float16,float16,0,0.018895999838908512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,48,8,128,1,float16,fp8,0,0.01704000060757001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,4,128,1,float16,float16,0,0.012527999778588613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,4,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,8,128,1,float16,float16,0,0.013647999614477158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,48,128,1,float16,float16,0,0.04574400186538696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,8,128,1,float16,fp8,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,4,128,1,float16,float16,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,48,128,1,float16,float16,0,0.02510933329661687
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,4,128,1,float16,fp8,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,8,128,1,float16,fp8,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,48,8,128,1,float16,float16,0,0.029461334149042766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,4,128,1,float16,float16,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,4,128,1,float16,fp8,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,48,128,1,float16,fp8,0,0.01730666682124138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,8,128,1,float16,fp8,0,0.009045333291093508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,48,128,1,float16,fp8,0,0.014890667051076889
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,48,128,1,float16,float16,0,0.013093333691358566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,4,128,1,float16,float16,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,4,128,1,float16,fp8,0,0.008890666688481966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,48,48,128,1,float16,fp8,0,0.03555200000603994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,8,128,1,float16,fp8,0,0.008752000207702318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,48,128,1,float16,fp8,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,48,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,4,128,1,float16,float16,0,0.00903466654320558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,8,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,4,128,1,float16,fp8,0,0.00902399979531765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,8,128,1,float16,float16,0,0.009077333534757296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,48,8,128,1,float16,fp8,0,0.008736000085870424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,48,128,1,float16,fp8,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,48,128,1,float16,float16,0,0.01062400018175443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,48,128,1,float16,float16,0,0.017055999487638474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,4,128,1,float16,float16,0,0.008858666444818178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,4,128,1,float16,fp8,0,0.008799999952316284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,8,128,1,float16,float16,0,0.009050666665037474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,48,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,4,128,1,float16,float16,0,0.008933333059151968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,48,128,1,float16,float16,0,0.010687999427318573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,8,128,1,float16,float16,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,48,8,128,1,float16,float16,0,0.00879466657837232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,8,128,1,float16,fp8,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,48,48,128,1,float16,fp8,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,48,8,128,1,float16,float16,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,48,8,128,1,float16,fp8,0,0.00903466654320558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,48,4,128,1,float16,fp8,0,0.008863999818762144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,1,128,1,float16,fp8,0,14.303877512613932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,2,128,1,float16,fp8,0,14.423269907633463
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,1,128,1,float16,float16,0,19.277114868164062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,2,128,1,float16,float16,0,19.235754648844402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,4,128,1,float16,fp8,0,14.867056528727213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,8,128,1,float16,fp8,0,14.777445475260416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,4,128,1,float16,float16,0,19.167781829833984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,40,8,128,1,float16,float16,0,19.603018442789715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,40,128,1,float16,float16,0,10.133893330891928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,40,128,1,float16,fp8,0,8.082143783569336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,1,128,1,float16,fp8,0,7.254671732584636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,1,128,1,float16,float16,0,9.423887888590494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,2,128,1,float16,fp8,0,7.205493291219075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,2,128,1,float16,float16,0,9.696970621744791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,4,128,1,float16,fp8,0,7.405802408854167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,40,128,1,float16,float16,0,4.9323733647664385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,4,128,1,float16,float16,0,9.666816075642904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,40,128,1,float16,fp8,0,4.539109230041504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,8,128,1,float16,fp8,0,7.282480239868164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,1,128,1,float16,fp8,0,4.65228811899821
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,40,8,128,1,float16,float16,0,9.91156260172526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,2,128,1,float16,float16,0,4.3809919357299805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,2,128,1,float16,fp8,0,3.8124586741129556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,4,128,1,float16,float16,0,4.893338521321614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,4,128,1,float16,fp8,0,3.754122734069824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,8,128,1,float16,fp8,0,3.7781651814778647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,8,128,1,float16,float16,0,4.663722674051921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,1,128,1,float16,float16,0,2.2364586194356284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,2,128,1,float16,float16,0,2.2526399294535318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,40,128,1,float16,float16,0,2.8939412434895835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,40,128,1,float16,fp8,0,2.4313599268595376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,2,128,1,float16,fp8,0,1.874303976694743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,40,1,128,1,float16,float16,0,4.425386746724446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,4,128,1,float16,float16,0,2.6693334579467773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,4,128,1,float16,fp8,0,2.040719985961914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,8,128,1,float16,float16,0,2.3006772994995117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,8,128,1,float16,fp8,0,2.02018674214681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,40,1,128,1,float16,fp8,0,2.3832319577534995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,1,128,1,float16,fp8,0,8.209168116251627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,1,128,1,float16,float16,0,10.995493570963541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,2,128,1,float16,fp8,0,8.368464152018229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,2,128,1,float16,float16,0,11.263381958007812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,4,128,1,float16,fp8,0,8.588127772013346
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,4,128,1,float16,float16,0,11.355962117513021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,8,128,1,float16,float16,0,11.39410146077474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,1,128,1,float16,float16,0,5.235013326009114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,40,128,1,float16,float16,0,5.810618718465169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,1,128,1,float16,fp8,0,4.464570681254069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,40,8,128,1,float16,fp8,0,8.48527971903483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,2,128,1,float16,float16,0,5.245482762654622
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,2,128,1,float16,fp8,0,4.3684641520182295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,40,128,1,float16,fp8,0,4.966335932413737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,4,128,1,float16,float16,0,5.49293835957845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,1,128,1,float16,fp8,0,2.0799360275268555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,1,128,1,float16,float16,0,2.5319040616353354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,8,128,1,float16,fp8,0,4.317861239115397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,4,128,1,float16,fp8,0,4.582437197367351
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,40,128,1,float16,float16,0,2.8106934229532876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,40,128,1,float16,fp8,0,2.8309707641601562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,40,8,128,1,float16,float16,0,5.369359970092773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,2,128,1,float16,float16,0,2.869013468424479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,2,128,1,float16,fp8,0,2.1109066009521484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,4,128,1,float16,fp8,0,2.1998987197875977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,4,128,1,float16,float16,0,2.6264266967773438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,8,128,1,float16,float16,0,2.6241226196289062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,40,128,1,float16,float16,0,1.4613866806030273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,1,128,1,float16,float16,0,1.3028213183085124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,40,8,128,1,float16,fp8,0,2.302501360575358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,1,128,1,float16,fp8,0,1.0910666783650715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,40,128,1,float16,fp8,0,1.480533281962077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,2,128,1,float16,float16,0,1.3048426310221355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,2,128,1,float16,fp8,0,1.2409599622090657
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,4,128,1,float16,float16,0,1.4551733334859211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,4,128,1,float16,fp8,0,1.2038239638010662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,8,128,1,float16,float16,0,1.3469972610473633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,40,8,128,1,float16,fp8,0,1.161199967066447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,1,128,1,float16,fp8,0,5.767701466878255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,2,128,1,float16,fp8,0,5.8680375417073565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,1,128,1,float16,float16,0,7.609888076782227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,2,128,1,float16,float16,0,7.921765645345052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,4,128,1,float16,fp8,0,6.074560165405273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,8,128,1,float16,fp8,0,6.075877507527669
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,4,128,1,float16,float16,0,7.87286376953125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,40,8,128,1,float16,float16,0,8.092704137166342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,40,128,1,float16,fp8,0,3.5880959828694663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,1,128,1,float16,float16,0,4.026181221008301
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,2,128,1,float16,float16,0,3.579098701477051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,1,128,1,float16,fp8,0,3.8289333979288735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,40,128,1,float16,float16,0,4.0491892496744795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,4,128,1,float16,float16,0,3.923098564147949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,4,128,1,float16,fp8,0,3.647120157877604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,1,128,1,float16,float16,0,1.800879955291748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,1,128,1,float16,fp8,0,1.527359962463379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,40,128,1,float16,fp8,0,1.9110933939615886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,8,128,1,float16,float16,0,3.683157285054525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,40,128,1,float16,float16,0,2.0668746630350747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,8,128,1,float16,fp8,0,3.316213289896647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,2,128,1,float16,float16,0,1.8861066500345867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,2,128,1,float16,fp8,0,1.5299839973449707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,4,128,1,float16,float16,0,1.8605119387308757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,4,128,1,float16,fp8,0,1.7502187093098958
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,8,128,1,float16,fp8,0,1.6115946769714355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,40,8,128,1,float16,float16,0,1.8856533368428547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,40,2,128,1,float16,fp8,0,2.935978571573893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,1,128,1,float16,float16,0,0.9332853158315023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,40,128,1,float16,fp8,0,0.9711413383483887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,2,128,1,float16,fp8,0,0.9094399611155192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,2,128,1,float16,float16,0,0.9407893021901449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,40,128,1,float16,float16,0,1.0625226497650146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,4,128,1,float16,float16,0,0.9586239655812582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,4,128,1,float16,fp8,0,0.8612693150838217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,8,128,1,float16,float16,0,0.9592533111572266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,1,128,1,float16,fp8,0,0.8290879726409912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,1,128,1,float16,fp8,0,7.6675465901692705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,40,8,128,1,float16,fp8,0,0.8905813694000244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,2,128,1,float16,fp8,0,7.852191925048828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,1,128,1,float16,float16,0,10.334490458170572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,2,128,1,float16,float16,0,10.128565470377604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,4,128,1,float16,float16,0,10.546719868977865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,4,128,1,float16,fp8,0,8.04213841756185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,1,128,1,float16,float16,0,5.189765294392903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,1,128,1,float16,fp8,0,4.419429461161296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,40,128,1,float16,fp8,0,4.920079867045085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,40,128,1,float16,float16,0,5.516074498494466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,8,128,1,float16,fp8,0,8.125333150227865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,2,128,1,float16,fp8,0,3.8786773681640625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,40,8,128,1,float16,float16,0,10.27572250366211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,4,128,1,float16,fp8,0,4.042624155680339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,4,128,1,float16,float16,0,5.117439905802409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,40,128,1,float16,float16,0,2.729541460673014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,40,128,1,float16,fp8,0,2.476309299468994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,8,128,1,float16,float16,0,4.849781354268392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,1,128,1,float16,float16,0,2.6541333198547363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,8,128,1,float16,fp8,0,4.450245221455892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,1,128,1,float16,fp8,0,1.9263413747151692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,2,128,1,float16,float16,0,2.3483732541402182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,40,2,128,1,float16,float16,0,4.924736022949219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,2,128,1,float16,fp8,0,1.9920585950215657
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,4,128,1,float16,fp8,0,2.2573493321736655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,8,128,1,float16,float16,0,2.4512532552083335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,40,128,1,float16,float16,0,1.3779573440551758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,1,128,1,float16,fp8,0,1.0120000044504802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,2,128,1,float16,float16,0,1.200389305750529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,1,128,1,float16,float16,0,1.1752160390218098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,2,128,1,float16,fp8,0,1.1627573172251384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,40,128,1,float16,fp8,0,1.3745546340942383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,4,128,1,float16,float16,0,1.2349493503570557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,4,128,1,float16,fp8,0,1.077077309290568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,8,128,1,float16,fp8,0,1.0764319896697998
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,40,8,128,1,float16,float16,0,1.249626636505127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,1,128,1,float16,fp8,0,0.5403253237406412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,1,128,1,float16,float16,0,0.618501345316569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,2,128,1,float16,float16,0,0.6220853328704834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,40,128,1,float16,fp8,0,0.6874293486277262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,40,128,1,float16,float16,0,0.7250400384267172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,2,128,1,float16,fp8,0,0.5728266636530558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,4,128,1,float16,float16,0,0.6466346581776937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,4,128,1,float16,fp8,0,0.5738293329874674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,8,128,1,float16,fp8,0,0.5851466655731201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,40,8,128,1,float16,float16,0,0.6461546818415324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,4,128,1,float16,float16,0,2.4129227002461753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,1,128,1,float16,fp8,0,4.467551867167155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,2,128,1,float16,fp8,0,4.540906588236491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,1,128,1,float16,float16,0,5.338437398274739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,2,128,1,float16,float16,0,5.743024190266927
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,4,128,1,float16,fp8,0,4.820709228515625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,4,128,1,float16,float16,0,6.097402572631836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,8,128,1,float16,float16,0,5.759749094645183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,1,128,1,float16,float16,0,2.9101972579956055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,1,128,1,float16,fp8,0,2.2451093991597495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,40,8,128,1,float16,fp8,0,2.0997974077860513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,40,8,128,1,float16,fp8,0,4.884682655334473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,40,128,1,float16,float16,0,3.2758239110310874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,40,128,1,float16,fp8,0,3.2000694274902344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,2,128,1,float16,fp8,0,2.8713385264078775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,4,128,1,float16,fp8,0,2.43996795018514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,4,128,1,float16,float16,0,2.885594685872396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,8,128,1,float16,float16,0,2.8769493103027344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,8,128,1,float16,fp8,0,2.4715147018432617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,1,128,1,float16,float16,0,1.3553120295206706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,40,128,1,float16,fp8,0,1.5631945927937825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,2,128,1,float16,float16,0,1.4506880442301433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,2,128,1,float16,fp8,0,1.1796639760335286
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,4,128,1,float16,float16,0,1.4474186897277832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,4,128,1,float16,fp8,0,1.264469305674235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,8,128,1,float16,float16,0,1.4515946706136067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,8,128,1,float16,fp8,0,1.3314666748046875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,1,128,1,float16,float16,0,0.6958080132802328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,1,128,1,float16,fp8,0,0.6100053389867147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,40,128,1,float16,fp8,0,0.8229653040568033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,40,128,1,float16,float16,0,0.8620320161183676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,2,128,1,float16,fp8,0,0.6308106581370035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,4,128,1,float16,fp8,0,0.6805813312530518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,8,128,1,float16,float16,0,0.7450026671091715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,8,128,1,float16,fp8,0,0.6687093575795492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,40,128,1,float16,float16,0,1.6646080017089844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,40,128,1,float16,float16,0,0.44882134596506756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,40,128,1,float16,fp8,0,0.4403253396352132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,1,128,1,float16,fp8,0,0.33269333839416504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,2,128,1,float16,float16,0,0.3816479841868083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,2,128,1,float16,fp8,0,0.34169598420461017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,4,128,1,float16,float16,0,0.3908640146255493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,2,128,1,float16,float16,0,0.7209226290384928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,40,2,128,1,float16,float16,0,2.8278026580810547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,4,128,1,float16,fp8,0,0.3556053241093953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,8,128,1,float16,float16,0,0.39690665404001874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,8,128,1,float16,fp8,0,0.3647093375523885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,40,4,128,1,float16,float16,0,0.7544053395589193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,40,1,128,1,float16,fp8,0,1.1522986888885498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,1,128,1,float16,fp8,0,4.28059196472168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,40,1,128,1,float16,float16,0,0.380293329556783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,1,128,1,float16,float16,0,5.27184518178304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,2,128,1,float16,float16,0,5.368213017781575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,4,128,1,float16,float16,0,5.49673589070638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,4,128,1,float16,fp8,0,4.80574385325114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,1,128,1,float16,float16,0,2.6923999786376953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,1,128,1,float16,fp8,0,2.381920019785563
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,8,128,1,float16,float16,0,5.804618835449219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,2,128,1,float16,float16,0,2.616917292277018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,40,128,1,float16,float16,0,3.3308159510294595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,8,128,1,float16,fp8,0,4.876794815063477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,40,128,1,float16,fp8,0,3.343690554300944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,2,128,1,float16,fp8,0,2.377621332804362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,40,2,128,1,float16,fp8,0,4.41707197825114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,4,128,1,float16,float16,0,2.7461118698120117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,8,128,1,float16,float16,0,2.7984212239583335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,4,128,1,float16,fp8,0,2.6731465657552085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,1,128,1,float16,float16,0,1.2874879837036133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,1,128,1,float16,fp8,0,1.1032586892445881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,40,8,128,1,float16,fp8,0,2.8017600377400718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,40,128,1,float16,float16,0,1.6857867240905762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,40,128,1,float16,fp8,0,1.6582400004069011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,2,128,1,float16,float16,0,1.3070346514383953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,2,128,1,float16,fp8,0,1.3413066864013672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,4,128,1,float16,float16,0,1.3823413848876953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,1,128,1,float16,float16,0,0.6559679905573527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,8,128,1,float16,float16,0,1.4186560312906902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,8,128,1,float16,fp8,0,1.2721973260243733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,1,128,1,float16,fp8,0,0.6215840180714926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,2,128,1,float16,fp8,0,0.6128960053126017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,40,128,1,float16,float16,0,0.8569280306498209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,40,128,1,float16,fp8,0,0.8474453290303549
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,4,128,1,float16,float16,0,0.7166133721669515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,4,128,1,float16,fp8,0,0.6527200142542521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,8,128,1,float16,float16,0,0.7253333727518717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,1,128,1,float16,float16,0,0.3444746732711792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,1,128,1,float16,fp8,0,0.31653332710266113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,8,128,1,float16,fp8,0,0.6726293563842773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,2,128,1,float16,float16,0,0.35419201850891113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,2,128,1,float16,fp8,0,0.3227679928143819
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,4,128,1,float16,float16,0,0.37457601229349774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,40,128,1,float16,float16,0,0.4503519932428996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,8,128,1,float16,fp8,0,0.348474661509196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,8,128,1,float16,float16,0,0.3785173495610555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,1,128,1,float16,float16,0,0.19282132387161255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,1,128,1,float16,fp8,0,0.18149866660435995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,40,128,1,float16,fp8,0,0.2500159939130147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,2,128,1,float16,float16,0,0.1963520050048828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,40,128,1,float16,float16,0,0.2605066696802775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,4,128,1,float16,float16,0,0.20347734292348227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,4,128,1,float16,fp8,0,0.19103999932607016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,8,128,1,float16,float16,0,0.21054933468500772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,40,4,128,1,float16,fp8,0,1.313109318415324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,40,2,128,1,float16,float16,0,0.6726880073547363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,1,128,1,float16,float16,0,3.05782413482666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,1,128,1,float16,fp8,0,2.6102773348490396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,2,128,1,float16,float16,0,3.1232426961263022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,40,128,1,float16,fp8,0,0.4485066731770833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,2,128,1,float16,fp8,0,2.714127858479818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,2,128,1,float16,fp8,0,0.18368534247080484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,4,128,1,float16,fp8,0,2.998085339864095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,4,128,1,float16,float16,0,3.3981812795003257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,8,128,1,float16,fp8,0,3.0672105153401694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,40,8,128,1,float16,float16,0,3.440042813618978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,40,128,1,float16,float16,0,2.1132747332255044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,1,128,1,float16,float16,0,1.5191466013590496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,1,128,1,float16,fp8,0,1.3944427172342937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,40,128,1,float16,fp8,0,2.141258716583252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,2,128,1,float16,float16,0,1.5739199320475261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,2,128,1,float16,fp8,0,1.5241492589314778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,4,128,1,float16,float16,0,1.6781867345174153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,40,4,128,1,float16,fp8,0,0.3481920162836711
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,4,128,1,float16,fp8,0,1.5184586842854817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,8,128,1,float16,fp8,0,1.5487786928812664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,40,8,128,1,float16,float16,0,1.7246346473693848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,1,128,1,float16,fp8,0,0.6835359732309977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,1,128,1,float16,float16,0,0.7703039646148682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,2,128,1,float16,float16,0,0.7920587062835693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,40,128,1,float16,float16,0,1.080901304880778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,2,128,1,float16,fp8,0,0.7094293435414633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,40,128,1,float16,fp8,0,1.0983306566874187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,4,128,1,float16,float16,0,0.8685440222422282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,40,8,128,1,float16,fp8,0,0.1928000052769979
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,4,128,1,float16,fp8,0,0.7830560207366943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,8,128,1,float16,float16,0,0.8573599656422933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,40,8,128,1,float16,fp8,0,0.8066346645355225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,2,128,1,float16,float16,0,0.40706666310628253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,1,128,1,float16,fp8,0,0.36610134442647296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,4,128,1,float16,float16,0,0.44336533546447754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,40,128,1,float16,fp8,0,0.5645866791407267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,40,128,1,float16,float16,0,0.54966934521993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,2,128,1,float16,fp8,0,0.38123734792073566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,4,128,1,float16,fp8,0,0.40991465250651044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,8,128,1,float16,float16,0,0.4440853198369344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,1,128,1,float16,float16,0,0.21967466672261557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,1,128,1,float16,fp8,0,0.20059200127919516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,2,128,1,float16,float16,0,0.22307199239730835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,2,128,1,float16,fp8,0,0.20519467194875082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,40,128,1,float16,fp8,0,0.30342400074005127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,4,128,1,float16,float16,0,0.23794132471084595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,4,128,1,float16,fp8,0,0.2251466711362203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,40,128,1,float16,float16,0,0.3089653253555298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,8,128,1,float16,float16,0,0.24369066953659058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,40,8,128,1,float16,fp8,0,0.22794665892918906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,1,128,1,float16,fp8,0,0.11785067121187846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,1,128,1,float16,float16,0,0.12734400232632956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,40,128,1,float16,float16,0,0.17891200383504233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,2,128,1,float16,fp8,0,0.12171733379364014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,4,128,1,float16,float16,0,0.13491732875506082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,40,128,1,float16,fp8,0,0.17266666889190674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,4,128,1,float16,fp8,0,0.12837333480517069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,8,128,1,float16,fp8,0,0.13014400005340576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,8,128,1,float16,float16,0,0.13689066966374716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,1,128,1,float16,float16,0,0.40426135063171387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,40,8,128,1,float16,fp8,0,0.4174613157908122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,1,128,1,float16,float16,0,3.0157972971598306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,1,128,1,float16,fp8,0,2.6557706197102866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,2,128,1,float16,fp8,0,2.7962719599405923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,2,128,1,float16,float16,0,3.138607978820801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,4,128,1,float16,float16,0,3.520751953125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,40,2,128,1,float16,float16,0,0.1290720005830129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,4,128,1,float16,fp8,0,3.1723254521687827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,8,128,1,float16,float16,0,3.5183626810709634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,1,128,1,float16,float16,0,1.5893120765686035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,1,128,1,float16,fp8,0,1.433285395304362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,40,128,1,float16,float16,0,2.3071999549865723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,2,128,1,float16,float16,0,1.5823733011881511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,40,8,128,1,float16,fp8,0,3.3067359924316406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,2,128,1,float16,fp8,0,1.4096852938334148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,40,128,1,float16,fp8,0,2.4211732546488443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,4,128,1,float16,float16,0,1.7601280212402344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,4,128,1,float16,fp8,0,1.6683893203735352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,8,128,1,float16,fp8,0,1.6346666018168132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,1,128,1,float16,fp8,0,0.6890827020009359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,1,128,1,float16,float16,0,0.7634452978769938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,40,8,128,1,float16,float16,0,1.7729120254516602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,2,128,1,float16,float16,0,0.8025866349538168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,40,128,1,float16,fp8,0,1.2273866335550945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,40,128,1,float16,float16,0,1.1674346923828125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,2,128,1,float16,fp8,0,0.7291733423868815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,4,128,1,float16,float16,0,0.8744906584421793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,4,128,1,float16,fp8,0,0.8149706522623698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,8,128,1,float16,fp8,0,0.8366399606068929
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,40,8,128,1,float16,float16,0,0.8817546367645264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,1,128,1,float16,float16,0,0.39476267496744794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,1,128,1,float16,fp8,0,0.3641599814097087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,2,128,1,float16,float16,0,0.4026453495025635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,2,128,1,float16,fp8,0,0.3784266710281372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,40,128,1,float16,fp8,0,0.6299786567687988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,4,128,1,float16,float16,0,0.45079998175303143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,4,128,1,float16,fp8,0,0.427562673886617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,8,128,1,float16,fp8,0,0.43590935071309406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,40,128,1,float16,float16,0,0.5931466817855835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,1,128,1,float16,fp8,0,0.19726399580637613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,1,128,1,float16,float16,0,0.21499200661977133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,4,128,1,float16,float16,0,0.24099200963974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,2,128,1,float16,fp8,0,0.2069279948870341
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,4,128,1,float16,fp8,0,0.23009065786997476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,40,128,1,float16,float16,0,0.30799466371536255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,8,128,1,float16,fp8,0,0.23469332853953043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,1,128,1,float16,float16,0,0.12124266227086385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,1,128,1,float16,fp8,0,0.1132480005423228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,2,128,1,float16,float16,0,0.12346133589744568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,2,128,1,float16,fp8,0,0.11622933546702068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,40,128,1,float16,fp8,0,0.18131200472513834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,4,128,1,float16,float16,0,0.13109866778055826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,4,128,1,float16,fp8,0,0.12422933181126912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,8,128,1,float16,float16,0,0.1362666686375936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,8,128,1,float16,fp8,0,0.12594667077064514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,1,128,1,float16,float16,0,0.07316799958546956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,1,128,1,float16,fp8,0,0.06972266733646393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,40,128,1,float16,fp8,0,0.10190932949384053
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,2,128,1,float16,float16,0,0.07542933523654938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,40,8,128,1,float16,float16,0,0.45321067174275714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,4,128,1,float16,float16,0,0.07858133316040039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,4,128,1,float16,fp8,0,0.0746506651242574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,8,128,1,float16,float16,0,0.07821333408355713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,2,128,1,float16,float16,0,0.22022400299708048
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,8,128,1,float16,fp8,0,0.07580266892910004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,40,128,1,float16,fp8,0,0.33161600430806476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,40,8,128,1,float16,float16,0,0.24361066023508707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,1,128,1,float16,float16,0,1.8680480321248372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,40,40,128,1,float16,float16,0,0.17186667521794638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,1,128,1,float16,fp8,0,1.6918187141418457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,2,128,1,float16,float16,0,1.9481066068013508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,2,128,1,float16,fp8,0,1.796079953511556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,40,128,1,float16,float16,0,0.10918399691581726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,4,128,1,float16,float16,0,2.2096266746520996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,4,128,1,float16,fp8,0,2.0761280059814453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,40,2,128,1,float16,fp8,0,0.07149333258469899
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,8,128,1,float16,float16,0,2.260394732157389
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,1,128,1,float16,float16,0,0.9387946923573812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,40,8,128,1,float16,fp8,0,2.1318400700887046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,1,128,1,float16,fp8,0,0.8571146329243978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,4,128,1,float16,float16,0,1.1137440204620361
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,2,128,1,float16,float16,0,0.9933173656463623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,40,128,1,float16,float16,0,1.546170711517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,4,128,1,float16,fp8,0,1.0566453138987224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,8,128,1,float16,float16,0,1.1243253548940022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,8,128,1,float16,fp8,0,1.0752000013987224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,1,128,1,float16,fp8,0,0.4449546734491984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,2,128,1,float16,float16,0,0.5016533136367798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,40,128,1,float16,float16,0,0.7775519688924154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,2,128,1,float16,fp8,0,0.46880535284678143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,40,128,1,float16,fp8,0,0.8449546496073405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,4,128,1,float16,fp8,0,0.5431146621704102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,8,128,1,float16,float16,0,0.5699946482976278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,8,128,1,float16,fp8,0,0.5526933272679647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,1,128,1,float16,float16,0,0.2548000017801921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,1,128,1,float16,fp8,0,0.2376693288485209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,40,128,1,float16,fp8,0,0.43782933553059894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,2,128,1,float16,float16,0,0.26578134298324585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,2,128,1,float16,fp8,0,0.2501973311106364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,4,128,1,float16,float16,0,0.30084800720214844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,40,128,1,float16,fp8,0,1.6688799858093262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,4,128,1,float16,fp8,0,0.28733332951863605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,8,128,1,float16,fp8,0,0.292192002137502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,1,128,1,float16,float16,0,0.4832106828689575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,1,128,1,float16,float16,0,0.144405335187912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,40,128,1,float16,float16,0,0.21490667263666788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,1,128,1,float16,fp8,0,0.13152533769607544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,2,128,1,float16,float16,0,0.14958399534225464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,40,4,128,1,float16,float16,0,0.563157320022583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,2,128,1,float16,fp8,0,0.1399786671002706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,8,128,1,float16,float16,0,0.16359466314315796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,8,128,1,float16,fp8,0,0.16105600198109946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,40,128,1,float16,float16,0,0.13384532928466797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,40,128,1,float16,float16,0,0.40065598487854004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,40,128,1,float16,fp8,0,0.13033066193262735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,1,128,1,float16,float16,0,0.08270399769147237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,40,2,128,1,float16,fp8,0,0.9115626811981201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,1,128,1,float16,fp8,0,0.07707199951012929
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,2,128,1,float16,fp8,0,0.08046933511892955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,40,8,128,1,float16,float16,0,0.30073599020640057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,4,128,1,float16,float16,0,0.09091732899347942
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,4,128,1,float16,fp8,0,0.0867680013179779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,8,128,1,float16,float16,0,0.09187199672063191
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,8,128,1,float16,fp8,0,0.08721599976221721
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,1,128,1,float16,float16,0,0.053914666175842285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,2,128,1,float16,float16,0,0.05416533350944519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,1,128,1,float16,fp8,0,0.050026665131251015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,40,128,1,float16,fp8,0,0.23282132546106973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,40,128,1,float16,float16,0,0.07549333572387695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,40,128,1,float16,fp8,0,0.07494933406511943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,4,128,1,float16,fp8,0,0.05606399973233541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,4,128,1,float16,float16,0,0.16358400384585062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,8,128,1,float16,float16,0,0.05778133372465769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,8,128,1,float16,fp8,0,0.05723733206590017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,1,128,1,float16,float16,0,1.9722347259521484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,1,128,1,float16,fp8,0,1.8294720649719238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,40,2,128,1,float16,float16,0,0.08498666683832805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,2,128,1,float16,float16,0,2.0995200475056968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,2,128,1,float16,fp8,0,1.9731252988179524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,2,128,1,float16,fp8,0,0.052527998884518944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,40,4,128,1,float16,float16,0,0.05787733197212219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,40,4,128,1,float16,fp8,0,0.15713066856066385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,8,128,1,float16,float16,0,2.476293404897054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,4,128,1,float16,fp8,0,2.343557357788086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,4,128,1,float16,float16,0,2.4355732599894204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,1,128,1,float16,float16,0,1.0000212987263997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,40,128,1,float16,float16,0,1.8055466016133626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,1,128,1,float16,fp8,0,0.9244213104248047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,2,128,1,float16,float16,0,1.0532533327738445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,40,128,1,float16,fp8,0,2.0000534057617188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,8,128,1,float16,float16,0,1.2542613347371419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,4,128,1,float16,fp8,0,1.1853439807891846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,4,128,1,float16,float16,0,1.2260053157806396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,2,128,1,float16,fp8,0,0.9956213633219401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,40,8,128,1,float16,fp8,0,1.216538667678833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,40,128,1,float16,fp8,0,1.0120586554209392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,1,128,1,float16,float16,0,0.510202685991923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,2,128,1,float16,fp8,0,0.5163573424021403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,4,128,1,float16,float16,0,0.6253120104471842
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,40,128,1,float16,float16,0,0.9061333338419596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,2,128,1,float16,float16,0,0.5352586507797241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,8,128,1,float16,float16,0,0.6337226629257202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,4,128,1,float16,fp8,0,0.6064426501592001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,1,128,1,float16,fp8,0,0.2490666707356771
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,8,128,1,float16,fp8,0,0.6231893301010132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,40,8,128,1,float16,fp8,0,2.4278079668680825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,2,128,1,float16,float16,0,0.281546672185262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,40,128,1,float16,fp8,0,0.5179359912872314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,2,128,1,float16,fp8,0,0.2703946630160014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,4,128,1,float16,fp8,0,0.3174613316853841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,4,128,1,float16,float16,0,0.324127991994222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,40,128,1,float16,float16,0,0.46325333913167316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,8,128,1,float16,float16,0,0.3294186592102051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,8,128,1,float16,fp8,0,0.32548266649246216
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,1,128,1,float16,float16,0,0.1527253290017446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,1,128,1,float16,fp8,0,0.14121599992116293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,2,128,1,float16,float16,0,0.1564853290716807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,2,128,1,float16,fp8,0,0.1479520003000895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,40,128,1,float16,fp8,0,0.2712000012397766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,4,128,1,float16,fp8,0,0.1697333256403605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,40,128,1,float16,float16,0,0.24279999732971191
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,4,128,1,float16,float16,0,0.17506666978200278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,1,128,1,float16,float16,0,0.08404800295829773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,8,128,1,float16,float16,0,0.18224533398946127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,40,8,128,1,float16,fp8,0,0.182586669921875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,1,128,1,float16,fp8,0,0.07842666904131572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,40,128,1,float16,fp8,0,0.14748266339302063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,2,128,1,float16,fp8,0,0.08162133395671844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,40,1,128,1,float16,fp8,0,0.4761333465576172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,4,128,1,float16,float16,0,0.09589333335558574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,8,128,1,float16,float16,0,0.10131733616193135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,40,128,1,float16,float16,0,0.1344533363978068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,8,128,1,float16,fp8,0,0.09310400485992432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,40,1,128,1,float16,float16,0,0.2671733299891154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,1,128,1,float16,fp8,0,0.047600001096725464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,2,128,1,float16,float16,0,0.05202666421731313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,1,128,1,float16,float16,0,0.05125333368778229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,2,128,1,float16,fp8,0,0.049285332361857094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,40,128,1,float16,fp8,0,0.08054933448632558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,4,128,1,float16,fp8,0,0.05400000015894572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,4,128,1,float16,float16,0,0.05646933118502299
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,40,128,1,float16,float16,0,0.07654933134714763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,8,128,1,float16,float16,0,0.05588266750176748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,40,8,128,1,float16,fp8,0,0.05470933516820272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,1,128,1,float16,float16,0,0.03682133307059606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,1,128,1,float16,fp8,0,0.03510933369398117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,2,128,1,float16,float16,0,0.03762666632731756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,4,128,1,float16,float16,0,0.03961066653331121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,40,128,1,float16,fp8,0,0.04969066878159841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,8,128,1,float16,float16,0,0.03946666667858759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,40,128,1,float16,float16,0,0.04345066845417023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,8,128,1,float16,fp8,0,0.038693333665529885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,2,128,1,float16,float16,0,0.08687466382980347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,40,4,128,1,float16,fp8,0,0.09127466877301534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,1,128,1,float16,float16,0,1.511722723642985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,1,128,1,float16,fp8,0,1.4114185969034831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,2,128,1,float16,float16,0,1.6243947347005208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,2,128,1,float16,fp8,0,1.5529227256774902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,2,128,1,float16,fp8,0,0.036159999668598175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,40,4,128,1,float16,fp8,0,0.03766933331886927
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,4,128,1,float16,float16,0,1.981104056040446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,4,128,1,float16,fp8,0,1.9261813163757324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,1,128,1,float16,float16,0,0.7670026620229086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,1,128,1,float16,fp8,0,0.7185866832733154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,40,128,1,float16,float16,0,1.535530726114909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,40,128,1,float16,fp8,0,1.780959924062093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,2,128,1,float16,float16,0,0.8258826732635498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,8,128,1,float16,float16,0,2.0100159645080566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,2,128,1,float16,fp8,0,0.784501314163208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,4,128,1,float16,float16,0,1.002245346705119
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,40,8,128,1,float16,fp8,0,1.997536023457845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,4,128,1,float16,fp8,0,0.9756266276041666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,1,128,1,float16,float16,0,0.3933493296305339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,1,128,1,float16,fp8,0,0.3699093262354533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,8,128,1,float16,float16,0,1.0079840024312336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,40,8,128,1,float16,fp8,0,1.0049493312835693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,2,128,1,float16,fp8,0,0.404149333635966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,4,128,1,float16,float16,0,0.5064586798350016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,4,128,1,float16,fp8,0,0.49821333090464276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,1,128,1,float16,float16,0,0.21030400196711221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,8,128,1,float16,float16,0,0.5170666774113973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,1,128,1,float16,fp8,0,0.19662932554880777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,40,128,1,float16,float16,0,0.39931201934814453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,8,128,1,float16,fp8,0,0.5152320067087809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,40,128,1,float16,fp8,0,0.45945600668589276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,2,128,1,float16,fp8,0,0.21327465772628784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,2,128,1,float16,float16,0,0.4238133430480957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,4,128,1,float16,float16,0,0.2677333354949951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,2,128,1,float16,float16,0,0.22805867592493692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,4,128,1,float16,fp8,0,0.26368532578150433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,40,128,1,float16,fp8,0,0.89956267674764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,8,128,1,float16,fp8,0,0.26958932479222614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,40,8,128,1,float16,float16,0,0.26958932479222614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,1,128,1,float16,fp8,0,0.10923733313878377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,1,128,1,float16,float16,0,0.11878933509190877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,40,40,128,1,float16,float16,0,0.7778826554616293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,2,128,1,float16,float16,0,0.12473066647847493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,4,128,1,float16,fp8,0,0.13994133472442627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,40,128,1,float16,fp8,0,0.24013332525889078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,40,128,1,float16,float16,0,0.21076265970865884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,2,128,1,float16,fp8,0,0.1181173324584961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,4,128,1,float16,float16,0,0.14570666352907816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,1,128,1,float16,float16,0,0.06618666648864746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,2,128,1,float16,float16,0,0.0685280015071233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,8,128,1,float16,float16,0,0.14813866217931113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,2,128,1,float16,fp8,0,0.06460799773534139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,4,128,1,float16,float16,0,0.07762133578459422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,40,128,1,float16,float16,0,0.11686933040618896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,40,128,1,float16,fp8,0,0.13157866398493448
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,4,128,1,float16,fp8,0,0.07442666590213776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,1,128,1,float16,float16,0,0.041477332512537636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,8,128,1,float16,float16,0,0.08326933284600575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,8,128,1,float16,fp8,0,0.07589333256085713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,1,128,1,float16,fp8,0,0.03783999880154928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,2,128,1,float16,float16,0,0.042810668547948204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,2,128,1,float16,fp8,0,0.04005333284536997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,40,128,1,float16,fp8,0,0.0713866651058197
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,4,128,1,float16,float16,0,0.0454773356517156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,8,128,1,float16,float16,0,0.045882667104403176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,8,128,1,float16,fp8,0,0.04582933088143667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,1,128,1,float16,float16,0,0.029557332396507263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,1,128,1,float16,fp8,0,0.028757333755493164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,40,128,1,float16,float16,0,0.03702933341264725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,40,128,1,float16,fp8,0,0.043578664461771645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,2,128,1,float16,float16,0,0.030832000076770782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,4,128,1,float16,float16,0,0.03221333275238673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,2,128,1,float16,fp8,0,0.029253333806991577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,4,128,1,float16,fp8,0,0.03125333289305369
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,8,128,1,float16,float16,0,0.032138665517171226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,40,8,128,1,float16,fp8,0,0.03230933348337809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,40,1,128,1,float16,fp8,0,0.06181866427262624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,40,8,128,1,float16,fp8,0,0.1460640033086141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,1,128,1,float16,float16,0,0.02102400114138921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,2,128,1,float16,fp8,0,0.02107200026512146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,40,128,1,float16,float16,0,0.02537599951028824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,40,128,1,float16,fp8,0,0.029125332832336426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,2,128,1,float16,float16,0,0.022245332598686218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,4,128,1,float16,float16,0,0.0230880007147789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,8,128,1,float16,float16,0,0.023050665855407715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,4,128,1,float16,fp8,0,0.02319466571013133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,8,128,1,float16,fp8,0,0.02252800017595291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,4,128,1,float16,fp8,0,0.044154668847719826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,1,128,1,float16,float16,0,0.6268479824066162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,1,128,1,float16,fp8,0,0.6011626720428467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,2,128,1,float16,float16,0,0.6847200393676758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,2,128,1,float16,fp8,0,0.6678666273752848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,4,128,1,float16,float16,0,0.8552320003509521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,4,128,1,float16,fp8,0,0.852234681447347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,40,1,128,1,float16,fp8,0,0.02022933339079221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,8,128,1,float16,float16,0,0.8719573020935059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,1,128,1,float16,float16,0,0.3254133264223735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,40,8,128,1,float16,fp8,0,0.88591996828715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,1,128,1,float16,fp8,0,0.30796800057093304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,40,40,128,1,float16,float16,0,0.07415466507275899
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,40,128,1,float16,fp8,0,0.8369386990865072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,2,128,1,float16,float16,0,0.35444267590840656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,4,128,1,float16,float16,0,0.4351840019226074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,8,128,1,float16,float16,0,0.4448373317718506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,8,128,1,float16,fp8,0,0.4512853225072225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,40,128,1,float16,float16,0,0.3641120195388794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,40,128,1,float16,fp8,0,0.4283039967219035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,1,128,1,float16,float16,0,0.17443199952443442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,2,128,1,float16,fp8,0,0.3470240036646525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,4,128,1,float16,fp8,0,0.4341119925181071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,8,128,1,float16,fp8,0,0.23651733001073202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,4,128,1,float16,fp8,0,0.23113600413004556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,2,128,1,float16,fp8,0,0.18659732739130655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,1,128,1,float16,fp8,0,0.1648906668027242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,8,128,1,float16,float16,0,0.2358506719271342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,2,128,1,float16,float16,0,0.18750399351119995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,1,128,1,float16,float16,0,0.09620799620946248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,1,128,1,float16,fp8,0,0.09014933307965596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,40,40,128,1,float16,float16,0,0.7050027052561442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,2,128,1,float16,float16,0,0.10286399722099304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,4,128,1,float16,float16,0,0.12441066900889079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,4,128,1,float16,fp8,0,0.1236853301525116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,40,128,1,float16,fp8,0,0.222053329149882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,40,128,1,float16,float16,0,0.18959999084472656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,8,128,1,float16,float16,0,0.12665067116419473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,8,128,1,float16,fp8,0,0.12603200475374857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,1,128,1,float16,float16,0,0.0553706685702006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,1,128,1,float16,fp8,0,0.052000001072883606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,4,128,1,float16,fp8,0,0.0644160012404124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,2,128,1,float16,float16,0,0.05846933523813883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,4,128,1,float16,float16,0,0.06822933256626129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,8,128,1,float16,fp8,0,0.06621333460013072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,40,4,128,1,float16,float16,0,0.23081066211064658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,40,128,1,float16,float16,0,0.06213866670926412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,40,128,1,float16,fp8,0,0.0660159985224406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,1,128,1,float16,float16,0,0.035589332381884255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,2,128,1,float16,fp8,0,0.03631466627120972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,2,128,1,float16,float16,0,0.037578667203585304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,4,128,1,float16,fp8,0,0.039962666730086006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,8,128,1,float16,float16,0,0.0409706657131513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,8,128,1,float16,fp8,0,0.04131733377774557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,40,128,1,float16,fp8,0,0.12239467104276021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,8,128,1,float16,float16,0,0.07008000214894612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,40,128,1,float16,float16,0,0.03323200096686681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,1,128,1,float16,fp8,0,0.025045332809289295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,40,128,1,float16,fp8,0,0.039706667264302574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,2,128,1,float16,float16,0,0.0271519993742307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,1,128,1,float16,fp8,0,0.035071998834609985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,4,128,1,float16,float16,0,0.028368001182874043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,40,2,128,1,float16,fp8,0,0.09905067086219788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,8,128,1,float16,float16,0,0.029077333708604176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,40,4,128,1,float16,float16,0,0.04021333406368891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,8,128,1,float16,fp8,0,0.029306667546431225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,40,128,1,float16,float16,0,0.022389332453409832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,40,128,1,float16,fp8,0,0.027056001126766205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,2,128,1,float16,fp8,0,0.055770665407180786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,1,128,1,float16,float16,0,0.017269333203633625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,1,128,1,float16,fp8,0,0.01721599946419398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,2,128,1,float16,float16,0,0.017077332983414333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,2,128,1,float16,fp8,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,4,128,1,float16,float16,0,0.01918399954835574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,4,128,1,float16,fp8,0,0.019152000546455383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,8,128,1,float16,float16,0,0.019066666563351948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,1,128,1,float16,float16,0,0.025418666501839954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,1,128,1,float16,float16,0,0.016901332885026932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,2,128,1,float16,fp8,0,0.026975999275843304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,1,128,1,float16,fp8,0,0.017152000218629837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,40,128,1,float16,float16,0,0.01897066707412402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,2,128,1,float16,fp8,0,0.01716800034046173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,40,4,128,1,float16,fp8,0,0.028560000161329906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,4,128,1,float16,float16,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,4,128,1,float16,fp8,0,0.01687466725707054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,8,128,1,float16,float16,0,0.01695466662446658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,8,128,1,float16,fp8,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,1,128,1,float16,float16,0,0.38235199451446533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,1,128,1,float16,fp8,0,0.36214399337768555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,2,128,1,float16,float16,0,0.41289599736531574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,40,40,128,1,float16,float16,0,0.10677867134412129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,40,8,128,1,float16,fp8,0,0.019733333339293797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,2,128,1,float16,fp8,0,0.39316801230112713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,4,128,1,float16,float16,0,0.4927786588668823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,4,128,1,float16,fp8,0,0.48873066902160645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,8,128,1,float16,float16,0,0.5020906527837118
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,2,128,1,float16,float16,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,40,8,128,1,float16,fp8,0,0.5044000148773193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,40,40,128,1,float16,fp8,0,0.020975999534130096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,1,128,1,float16,float16,0,0.2015413244565328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,2,128,1,float16,fp8,0,0.20919466018676758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,1,128,1,float16,fp8,0,0.18885332345962524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,40,128,1,float16,float16,0,0.38788799444834393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,4,128,1,float16,fp8,0,0.2545066674550374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,40,128,1,float16,fp8,0,0.4516693353652954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,2,128,1,float16,float16,0,0.21846399704615274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,4,128,1,float16,float16,0,0.2587466637293498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,1,128,1,float16,float16,0,0.10893332958221436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,40,128,1,float16,fp8,0,0.2355253299077352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,8,128,1,float16,fp8,0,0.26122132937113446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,1,128,1,float16,fp8,0,0.10385599732398987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,2,128,1,float16,fp8,0,0.11449066797892253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,40,128,1,float16,float16,0,0.2022506594657898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,4,128,1,float16,float16,0,0.13606933752695718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,4,128,1,float16,fp8,0,0.1393226683139801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,8,128,1,float16,float16,0,0.1413226624329885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,8,128,1,float16,fp8,0,0.13794133067131042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,40,128,1,float16,fp8,0,0.1250879963239034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,1,128,1,float16,fp8,0,0.05593599875768026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,2,128,1,float16,float16,0,0.06408533453941345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,2,128,1,float16,fp8,0,0.058917333682378135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,40,128,1,float16,float16,0,0.10924800237019856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,4,128,1,float16,fp8,0,0.06771199901898702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,4,128,1,float16,float16,0,0.0769760012626648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,8,128,1,float16,float16,0,0.07597866654396057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,8,128,1,float16,fp8,0,0.07046933472156525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,1,128,1,float16,float16,0,0.03535466641187668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,1,128,1,float16,fp8,0,0.03457599878311157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,2,128,1,float16,float16,0,0.03728533287843069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,40,128,1,float16,float16,0,0.06794133285681407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,2,128,1,float16,fp8,0,0.03654933224121729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,4,128,1,float16,float16,0,0.039333333571751915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,40,128,1,float16,fp8,0,0.06666133304437001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,4,128,1,float16,fp8,0,0.03974399964014689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,8,128,1,float16,float16,0,0.04027733455101649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,40,8,128,1,float16,fp8,0,0.0413973331451416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,1,128,1,float16,float16,0,0.023621333142121632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,1,128,1,float16,fp8,0,0.02295999974012375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,2,128,1,float16,fp8,0,0.02473066747188568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,40,128,1,float16,fp8,0,0.037776000797748566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,4,128,1,float16,float16,0,0.02589866767326991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,40,8,128,1,float16,float16,0,0.26285332441329956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,4,128,1,float16,fp8,0,0.02657066782315572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,40,128,1,float16,float16,0,0.031258667508761086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,8,128,1,float16,float16,0,0.026842666169007618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,1,128,1,float16,float16,0,0.017450666675964992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,1,128,1,float16,fp8,0,0.017850667238235474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,2,128,1,float16,float16,0,0.017781333376963932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,40,1,128,1,float16,float16,0,0.059194669127464294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,2,128,1,float16,fp8,0,0.017914666483799618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,4,128,1,float16,fp8,0,0.019120000302791595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,4,128,1,float16,float16,0,0.019120000302791595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,40,128,1,float16,float16,0,0.022490667800108593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,8,128,1,float16,fp8,0,0.02033599962790807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,8,128,1,float16,float16,0,0.019573333362738293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,1,128,1,float16,float16,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,1,128,1,float16,fp8,0,0.013354666531085968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,2,128,1,float16,float16,0,0.013765333841244379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,40,128,1,float16,fp8,0,0.017103999853134155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,2,128,1,float16,fp8,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,40,128,1,float16,float16,0,0.015333333363135656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,4,128,1,float16,float16,0,0.013781332721312841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,4,128,1,float16,fp8,0,0.013557333499193192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,8,128,1,float16,float16,0,0.013157332936922709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,2,128,1,float16,float16,0,0.025098666548728943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,1,128,1,float16,float16,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,40,2,128,1,float16,float16,0,0.11781866351763408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,1,128,1,float16,fp8,0,0.01313599944114685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,40,8,128,1,float16,fp8,0,0.027242665489514668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,2,128,1,float16,float16,0,0.012975999464591345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,2,128,1,float16,fp8,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,4,128,1,float16,float16,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,4,128,1,float16,fp8,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,8,128,1,float16,float16,0,0.012768000364303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,8,128,1,float16,fp8,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,40,40,128,1,float16,fp8,0,0.02588266630967458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,1,128,1,float16,float16,0,0.299621323744456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,1,128,1,float16,fp8,0,0.2738613287607829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,40,8,128,1,float16,fp8,0,0.014720000326633453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,2,128,1,float16,float16,0,0.31440534194310504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,2,128,1,float16,fp8,0,0.2919680078824361
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,4,128,1,float16,float16,0,0.35813331604003906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,40,128,1,float16,fp8,0,0.015909332782030106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,4,128,1,float16,fp8,0,0.34113065401713055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,8,128,1,float16,fp8,0,0.34678399562835693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,40,8,128,1,float16,float16,0,0.3582880099614461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,40,128,1,float16,float16,0,0.2664159933725993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,1,128,1,float16,fp8,0,0.14732266465822855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,40,128,1,float16,fp8,0,0.27685866753260296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,2,128,1,float16,float16,0,0.16967999935150146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,4,128,1,float16,float16,0,0.18772266308466592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,8,128,1,float16,float16,0,0.19160000483194986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,8,128,1,float16,fp8,0,0.18159466981887817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,1,128,1,float16,float16,0,0.08794132868448894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,40,128,1,float16,float16,0,0.14448533455530801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,40,128,1,float16,fp8,0,0.14692800243695578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,2,128,1,float16,float16,0,0.09083732962608337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,40,40,128,1,float16,float16,0,0.014842666685581207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,1,128,1,float16,fp8,0,0.07832000156243642
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,4,128,1,float16,fp8,0,0.0906986693541209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,4,128,1,float16,float16,0,0.10173333684603374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,8,128,1,float16,fp8,0,0.09493866562843323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,8,128,1,float16,float16,0,0.10143466790517171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,1,128,1,float16,float16,0,0.16156799594561258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,1,128,1,float16,float16,0,0.04785066843032837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,40,128,1,float16,float16,0,0.0802400012811025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,1,128,1,float16,fp8,0,0.04384533564249674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,2,128,1,float16,float16,0,0.05000533163547516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,2,128,1,float16,fp8,0,0.0459199994802475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,40,128,1,float16,fp8,0,0.07692266503969829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,4,128,1,float16,fp8,0,0.04965866605440775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,4,128,1,float16,float16,0,0.05264533559481303
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,8,128,1,float16,float16,0,0.05351999898751577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,40,8,128,1,float16,fp8,0,0.051354666550954185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,1,128,1,float16,float16,0,0.029071999092896778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,1,128,1,float16,fp8,0,0.02720533311367035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,2,128,1,float16,float16,0,0.031146667897701263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,4,128,1,float16,fp8,0,0.1774133245150248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,2,128,1,float16,fp8,0,0.02937600016593933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,4,128,1,float16,float16,0,0.03152533372243246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,40,128,1,float16,float16,0,0.03756266583998998
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,8,128,1,float16,fp8,0,0.031119999786218006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,8,128,1,float16,float16,0,0.03197866678237915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,1,128,1,float16,float16,0,0.019893333315849304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,1,128,1,float16,fp8,0,0.01904533306757609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,40,128,1,float16,fp8,0,0.02852799991766612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,40,128,1,float16,float16,0,0.025077333052953083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,2,128,1,float16,fp8,0,0.01918399954835574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,4,128,1,float16,float16,0,0.020997333029905956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,40,2,128,1,float16,fp8,0,0.08267733454704285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,8,128,1,float16,float16,0,0.02124800036350886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,4,128,1,float16,fp8,0,0.021226666867733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,1,128,1,float16,float16,0,0.015290666371583939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,1,128,1,float16,fp8,0,0.01540800059835116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,2,128,1,float16,float16,0,0.014848000059525171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,40,128,1,float16,float16,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,40,128,1,float16,fp8,0,0.019637333850065868
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,40,2,128,1,float16,fp8,0,0.15507733821868896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,2,128,1,float16,fp8,0,0.015066667149464289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,4,128,1,float16,float16,0,0.016000000139077503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,4,128,1,float16,fp8,0,0.016234666109085083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,8,128,1,float16,fp8,0,0.015925332903862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,40,8,128,1,float16,float16,0,0.016986666868130367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,1,128,1,float16,float16,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,1,128,1,float16,fp8,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,40,128,1,float16,fp8,0,0.04170133173465729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,40,4,128,1,float16,fp8,0,0.031231999397277832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,40,128,1,float16,float16,0,0.013141332815090815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,40,128,1,float16,fp8,0,0.014794666320085526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,4,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,8,128,1,float16,float16,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,4,128,1,float16,float16,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,8,128,1,float16,fp8,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,2,128,1,float16,float16,0,0.020810666183630627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,1,128,1,float16,float16,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,1,128,1,float16,fp8,0,0.011136000355084738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,2,128,1,float16,float16,0,0.011445333560307821
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,2,128,1,float16,fp8,0,0.012069333344697952
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,40,128,1,float16,fp8,0,0.014767999450365702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,4,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,4,128,1,float16,fp8,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,8,128,1,float16,float16,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,8,128,1,float16,fp8,0,0.011695999652147293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,40,8,128,1,float16,fp8,0,0.021029333273569744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,1,128,1,float16,float16,0,0.27964266141255695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,1,128,1,float16,fp8,0,0.2383520007133484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,2,128,1,float16,float16,0,0.28251200914382935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,2,128,1,float16,fp8,0,0.2492426633834839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,2,128,1,float16,float16,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,4,128,1,float16,fp8,0,0.26843200127283734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,8,128,1,float16,float16,0,0.30714666843414307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,8,128,1,float16,fp8,0,0.2738933364550273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,40,40,128,1,float16,float16,0,0.013007999708255133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,1,128,1,float16,float16,0,0.14710932970046997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,40,128,1,float16,float16,0,0.1962133248647054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,40,128,1,float16,fp8,0,0.19182932376861572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,1,128,1,float16,fp8,0,0.12470400333404541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,2,128,1,float16,fp8,0,0.12703466415405273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,8,128,1,float16,float16,0,0.16275733709335327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,8,128,1,float16,fp8,0,0.14195199807484946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,4,128,1,float16,float16,0,0.16364266475041708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,4,128,1,float16,fp8,0,0.13700266679128012
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,40,2,128,1,float16,fp8,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,1,128,1,float16,float16,0,0.07923200229803722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,1,128,1,float16,fp8,0,0.06667733192443848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,2,128,1,float16,float16,0,0.08076266447703044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,40,4,128,1,float16,float16,0,0.3063039978345235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,40,128,1,float16,float16,0,0.10920533537864685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,2,128,1,float16,fp8,0,0.0683786670366923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,4,128,1,float16,float16,0,0.08268799881140391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,4,128,1,float16,fp8,0,0.07251733541488647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,1,128,1,float16,float16,0,0.04451199869314829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,8,128,1,float16,float16,0,0.08403199911117554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,1,128,1,float16,fp8,0,0.03773866593837738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,8,128,1,float16,fp8,0,0.07458666463692983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,40,128,1,float16,fp8,0,0.05176533261934916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,2,128,1,float16,float16,0,0.04584533472855886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,4,128,1,float16,float16,0,0.0476746658484141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,2,128,1,float16,fp8,0,0.03998400022586187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,4,128,1,float16,fp8,0,0.04196799794832865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,40,128,1,float16,float16,0,0.05453866720199585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,8,128,1,float16,fp8,0,0.04218133290608724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,1,128,1,float16,float16,0,0.027269333600997925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,1,128,1,float16,fp8,0,0.024933333198229473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,2,128,1,float16,float16,0,0.02739199995994568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,40,2,128,1,float16,float16,0,0.1523146629333496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,2,128,1,float16,fp8,0,0.025061334172884624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,4,128,1,float16,float16,0,0.028394666810830433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,40,128,1,float16,fp8,0,0.03356799980004629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,8,128,1,float16,float16,0,0.029109333952267964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,8,128,1,float16,fp8,0,0.027327999472618103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,1,128,1,float16,fp8,0,0.017887999614079792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,1,128,1,float16,float16,0,0.01916266605257988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,2,128,1,float16,float16,0,0.018858666221300762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,40,128,1,float16,fp8,0,0.02232533444960912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,2,128,1,float16,fp8,0,0.01878400022784869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,4,128,1,float16,float16,0,0.01904533306757609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,4,128,1,float16,fp8,0,0.018906666586796444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,40,40,128,1,float16,fp8,0,0.10250133275985718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,8,128,1,float16,float16,0,0.01918399954835574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,8,128,1,float16,fp8,0,0.019120000302791595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,1,128,1,float16,float16,0,0.015125333021084467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,1,128,1,float16,fp8,0,0.01488000030318896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,2,128,1,float16,float16,0,0.01482133318980535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,2,128,1,float16,fp8,0,0.015061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,40,128,1,float16,float16,0,0.016117333124081295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,4,128,1,float16,fp8,0,0.015098666151364645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,4,128,1,float16,float16,0,0.014864000181357065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,40,128,1,float16,fp8,0,0.01710933322707812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,40,8,128,1,float16,float16,0,0.046800002455711365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,8,128,1,float16,float16,0,0.015114666273196539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,1,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,1,128,1,float16,fp8,0,0.011370666325092316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,40,128,1,float16,fp8,0,0.014720000326633453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,40,128,1,float16,float16,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,4,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,4,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,8,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,8,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,40,40,128,1,float16,float16,0,0.020842666427294414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,1,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,1,128,1,float16,fp8,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,40,128,1,float16,fp8,0,0.01481066644191742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,40,128,1,float16,float16,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,2,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,4,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,8,128,1,float16,float16,0,0.011098666737476984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,8,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,40,8,128,1,float16,fp8,0,0.01509333277742068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,2,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,4,128,1,float16,fp8,0,0.027024000883102417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,40,2,128,1,float16,fp8,0,0.012960000584522883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,1,128,1,float16,float16,0,0.2727946639060974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,1,128,1,float16,fp8,0,0.21686933437983194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,2,128,1,float16,float16,0,0.2717919945716858
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,2,128,1,float16,fp8,0,0.2193386753400167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,4,128,1,float16,float16,0,0.28032533327738446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,4,128,1,float16,fp8,0,0.22981866200764975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,4,128,1,float16,float16,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,8,128,1,float16,float16,0,0.2863093415896098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,40,8,128,1,float16,fp8,0,0.2361066738764445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,1,128,1,float16,float16,0,0.1423520048459371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,40,128,1,float16,float16,0,0.1702079971631368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,1,128,1,float16,fp8,0,0.114847997824351
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,40,128,1,float16,fp8,0,0.14993066589037576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,2,128,1,float16,float16,0,0.14564266800880432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,2,128,1,float16,fp8,0,0.11549333731333415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,4,128,1,float16,float16,0,0.14588800072669983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,40,40,128,1,float16,float16,0,0.03162133445342382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,4,128,1,float16,fp8,0,0.11940266688664754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,8,128,1,float16,fp8,0,0.11994666854540507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,40,8,128,1,float16,float16,0,0.14993066589037576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,40,128,1,float16,fp8,0,0.07833600044250488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,40,128,1,float16,float16,0,0.09387733538945515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,1,128,1,float16,fp8,0,0.06203199923038483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,2,128,1,float16,float16,0,0.07861333092053731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,2,128,1,float16,fp8,0,0.06429333488146464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,4,128,1,float16,float16,0,0.08051733175913493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,4,128,1,float16,fp8,0,0.06635199983914693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,8,128,1,float16,fp8,0,0.06691200037797292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,1,128,1,float16,float16,0,0.04378133515516917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,1,128,1,float16,fp8,0,0.03745600084463755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,40,2,128,1,float16,fp8,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,2,128,1,float16,float16,0,0.04351999859015147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,40,128,1,float16,fp8,0,0.04572799801826477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,40,128,1,float16,float16,0,0.04789333542188009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,2,128,1,float16,fp8,0,0.037434667348861694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,4,128,1,float16,fp8,0,0.039333333571751915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,4,128,1,float16,float16,0,0.043578664461771645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,8,128,1,float16,float16,0,0.04535999894142151
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,40,8,128,1,float16,fp8,0,0.03969600051641464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,1,128,1,float16,float16,0,0.026949333647886913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,40,128,1,float16,fp8,0,0.02771199991305669
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,40,128,1,float16,float16,0,0.029178666571776073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,4,128,1,float16,float16,0,0.02720000098148982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,4,128,1,float16,fp8,0,0.025114665428797405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,1,128,1,float16,float16,0,0.07656533519426982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,8,128,1,float16,float16,0,0.027429332335789997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,8,128,1,float16,fp8,0,0.023973333338896435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,1,128,1,float16,float16,0,0.019018666197856266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,1,128,1,float16,fp8,0,0.01706133286158244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,2,128,1,float16,float16,0,0.018874666343132656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,40,128,1,float16,float16,0,0.019952000429232914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,40,8,128,1,float16,float16,0,0.07829866806666057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,4,128,1,float16,float16,0,0.018858666221300762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,4,128,1,float16,fp8,0,0.01759999990463257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,8,128,1,float16,float16,0,0.018965333700180054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,8,128,1,float16,fp8,0,0.01711999997496605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,1,128,1,float16,float16,0,0.015008000036080679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,1,128,1,float16,fp8,0,0.014720000326633453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,2,128,1,float16,float16,0,0.014789332946141561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,1,128,1,float16,fp8,0,0.02476266771554947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,40,128,1,float16,fp8,0,0.016976000120242436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,2,128,1,float16,fp8,0,0.014698666830857595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,4,128,1,float16,float16,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,2,128,1,float16,fp8,0,0.02462933212518692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,40,2,128,1,float16,float16,0,0.02737066646416982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,8,128,1,float16,float16,0,0.014773332824309668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,4,128,1,float16,fp8,0,0.014688000082969666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,8,128,1,float16,fp8,0,0.014965333044528961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,1,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,1,128,1,float16,fp8,0,0.010816000401973724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,2,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,2,128,1,float16,fp8,0,0.01146666705608368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,4,128,1,float16,float16,0,0.012149333953857422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,4,128,1,float16,fp8,0,0.01118933285276095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,8,128,1,float16,float16,0,0.011546666423479715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,40,128,1,float16,fp8,0,0.015135999768972397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,1,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,40,128,1,float16,fp8,0,0.021061333517233532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,40,128,1,float16,fp8,0,0.014149333039919535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,40,2,128,1,float16,fp8,0,0.017055999487638474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,2,128,1,float16,float16,0,0.010703999549150467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,2,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,4,128,1,float16,float16,0,0.010762666662534079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,4,128,1,float16,fp8,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,8,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,8,128,1,float16,fp8,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,40,40,128,1,float16,float16,0,0.01568000018596649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,40,4,128,1,float16,float16,0,0.021594665944576263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,40,4,128,1,float16,fp8,0,0.01904533306757609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,40,128,1,float16,float16,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,40,8,128,1,float16,float16,0,0.028399998943010967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,40,128,1,float16,float16,0,0.01268799975514412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,40,1,128,1,float16,fp8,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,40,8,128,1,float16,fp8,0,0.025429333249727886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,40,128,1,float16,float16,0,0.07007466753323872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,4,128,1,float16,float16,0,0.015114666273196539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,4,128,1,float16,fp8,0,0.014725333700577417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,8,128,1,float16,float16,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,8,128,1,float16,fp8,0,0.017237332959969837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,40,128,1,float16,fp8,0,0.03146666785081228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,4,128,1,float16,float16,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,4,128,1,float16,fp8,0,0.011168000598748526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,8,128,1,float16,float16,0,0.013210666676362356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,8,128,1,float16,fp8,0,0.013093333691358566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,40,128,1,float16,fp8,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,40,128,1,float16,float16,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,4,128,1,float16,float16,0,0.009423999736706415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,4,128,1,float16,fp8,0,0.010527999450763067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,8,128,1,float16,float16,0,0.010853332777818045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,40,8,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,40,40,128,1,float16,fp8,0,0.05389333268006643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,40,128,1,float16,fp8,0,0.01685333376129468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,40,128,1,float16,float16,0,0.015589332828919092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,4,128,1,float16,fp8,0,0.009941333283980688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,8,128,1,float16,float16,0,0.00879466657837232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,40,128,1,float16,float16,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,40,128,1,float16,fp8,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,4,128,1,float16,float16,0,0.008757333581646284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,40,40,128,1,float16,float16,0,0.03743999948104223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,4,128,1,float16,fp8,0,0.00901333304742972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,8,128,1,float16,float16,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,40,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,40,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,4,128,1,float16,float16,0,0.008736000085870424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,4,128,1,float16,fp8,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,8,128,1,float16,float16,0,0.009114666531483332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,40,8,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,40,128,1,float16,float16,0,0.010634666929642359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,40,128,1,float16,fp8,0,0.01360000049074491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,4,128,1,float16,float16,0,0.00972800018886725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,40,8,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,8,128,1,float16,float16,0,0.008896000062425932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,8,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,40,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,4,128,1,float16,float16,0,0.00901333304742972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,40,128,1,float16,float16,0,0.010650667051474253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,4,128,1,float16,fp8,0,0.009066666786869368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,8,128,1,float16,float16,0,0.008789333204428354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,40,8,128,1,float16,fp8,0,0.008853333070874214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,40,8,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,40,8,128,1,float16,fp8,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,4,128,1,float16,float16,0,0.008736000085870424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,40,4,128,1,float16,fp8,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,1,128,1,float16,fp8,0,11.268255869547525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,2,128,1,float16,fp8,0,11.376283009847006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,1,128,1,float16,float16,0,15.543290456136068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,2,128,1,float16,float16,0,15.428485870361328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,4,128,1,float16,fp8,0,11.855589548746744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,4,128,1,float16,float16,0,15.68951416015625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,8,128,1,float16,fp8,0,11.797594706217447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,32,128,1,float16,fp8,0,6.357109069824219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,32,8,128,1,float16,float16,0,15.470954895019531
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,32,128,1,float16,float16,0,7.68069330851237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,1,128,1,float16,fp8,0,5.779397328694661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,1,128,1,float16,float16,0,7.515392303466797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,2,128,1,float16,float16,0,7.370458602905273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,2,128,1,float16,fp8,0,5.857231775919597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,4,128,1,float16,float16,0,7.528879801432292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,1,128,1,float16,fp8,0,2.8895413080851235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,4,128,1,float16,fp8,0,5.814661026000977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,1,128,1,float16,float16,0,3.7915948232014975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,32,128,1,float16,float16,0,3.8039306004842124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,32,128,1,float16,fp8,0,3.5945491790771484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,8,128,1,float16,fp8,0,6.05728022257487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,32,8,128,1,float16,float16,0,7.777018864949544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,2,128,1,float16,fp8,0,2.909541447957357
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,2,128,1,float16,float16,0,3.5804265340169272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,4,128,1,float16,float16,0,3.608938535054525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,4,128,1,float16,fp8,0,3.2706826527913413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,8,128,1,float16,fp8,0,2.9980106353759766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,32,8,128,1,float16,float16,0,3.631466547648112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,1,128,1,float16,float16,0,1.7997867266337078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,32,128,1,float16,fp8,0,1.9739200274149578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,32,128,1,float16,float16,0,1.9484853744506836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,1,128,1,float16,fp8,0,1.8961386680603027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,2,128,1,float16,float16,0,1.8973493576049805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,2,128,1,float16,fp8,0,1.5350826581319172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,4,128,1,float16,float16,0,1.842591921488444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,8,128,1,float16,fp8,0,1.6644426981608074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,8,128,1,float16,float16,0,1.932154655456543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,32,4,128,1,float16,fp8,0,1.965237299601237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,1,128,1,float16,fp8,0,6.58192507425944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,2,128,1,float16,fp8,0,6.677082697550456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,1,128,1,float16,float16,0,8.591429392496744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,2,128,1,float16,float16,0,9.051125208536783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,4,128,1,float16,fp8,0,6.914501190185547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,4,128,1,float16,float16,0,9.084421157836914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,8,128,1,float16,float16,0,8.984138488769531
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,1,128,1,float16,fp8,0,3.2757867177327475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,1,128,1,float16,float16,0,4.402672131856282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,2,128,1,float16,float16,0,3.9801066716512046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,32,128,1,float16,float16,0,4.424410820007324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,32,128,1,float16,fp8,0,4.151621182759603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,32,8,128,1,float16,fp8,0,6.8077545166015625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,2,128,1,float16,fp8,0,3.8951199849446616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,4,128,1,float16,float16,0,4.241098721822103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,4,128,1,float16,fp8,0,3.4170827865600586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,1,128,1,float16,fp8,0,1.6969013214111328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,1,128,1,float16,float16,0,2.025968074798584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,32,128,1,float16,float16,0,2.247653325398763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,2,128,1,float16,float16,0,2.256213347117106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,8,128,1,float16,fp8,0,3.7582613627115884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,2,128,1,float16,fp8,0,1.7228533426920574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,32,8,128,1,float16,float16,0,4.719269434611003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,4,128,1,float16,float16,0,2.0742079416910806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,4,128,1,float16,fp8,0,1.7980480194091797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,1,128,1,float16,float16,0,1.2122613588968914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,32,128,1,float16,fp8,0,1.0411732991536458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,8,128,1,float16,float16,0,2.10645325978597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,2,128,1,float16,float16,0,1.0717439651489258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,8,128,1,float16,fp8,0,2.21562131245931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,2,128,1,float16,fp8,0,1.0063306490580242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,4,128,1,float16,float16,0,1.0970293680826824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,4,128,1,float16,fp8,0,1.0418079694112141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,8,128,1,float16,fp8,0,0.9770239988962809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,8,128,1,float16,float16,0,1.1036907037099202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,32,32,128,1,float16,fp8,0,1.9923787117004395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,1,128,1,float16,fp8,0,4.478111902872722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,2,128,1,float16,fp8,0,4.6192318598429365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,1,128,1,float16,float16,0,5.988954544067383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,1,128,1,float16,fp8,0,0.8861760298411051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,2,128,1,float16,float16,0,6.024752298990886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,32,32,128,1,float16,float16,0,1.157706658045451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,4,128,1,float16,float16,0,6.289818445841472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,4,128,1,float16,fp8,0,4.808245340983073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,1,128,1,float16,fp8,0,2.280890623728434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,32,128,1,float16,fp8,0,2.809610684712728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,1,128,1,float16,float16,0,3.2060426076253257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,2,128,1,float16,float16,0,3.1463359196980796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,2,128,1,float16,fp8,0,2.3325279553731284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,32,128,1,float16,float16,0,3.3709920247395835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,8,128,1,float16,fp8,0,4.8895572026570635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,4,128,1,float16,fp8,0,2.4469226201375327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,4,128,1,float16,float16,0,2.9188801447550454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,1,128,1,float16,float16,0,1.4384106000264485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,32,128,1,float16,fp8,0,1.596826712290446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,8,128,1,float16,fp8,0,2.64082670211792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,1,128,1,float16,fp8,0,1.1863359610239665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,2,128,1,float16,fp8,0,1.2462399800618489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,2,128,1,float16,float16,0,1.440970738728841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,4,128,1,float16,float16,0,1.502149264017741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,4,128,1,float16,fp8,0,1.304469347000122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,8,128,1,float16,float16,0,1.4896267255147297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,8,128,1,float16,fp8,0,1.3050453662872314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,1,128,1,float16,float16,0,0.75327467918396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,32,128,1,float16,float16,0,0.8500853379567465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,1,128,1,float16,fp8,0,0.6377653280893961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,32,128,1,float16,fp8,0,0.7855947017669678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,2,128,1,float16,fp8,0,0.6537760098775228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,4,128,1,float16,fp8,0,0.6768693129221598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,32,8,128,1,float16,float16,0,6.353909174601237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,4,128,1,float16,float16,0,0.7821973164876302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,8,128,1,float16,float16,0,0.7879146734873453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,8,128,1,float16,fp8,0,0.6897866725921631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,32,32,128,1,float16,float16,0,1.624117374420166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,32,8,128,1,float16,float16,0,3.479130744934082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,1,128,1,float16,fp8,0,5.975269317626953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,32,2,128,1,float16,float16,0,0.7609066963195801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,2,128,1,float16,fp8,0,6.143338521321614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,1,128,1,float16,float16,0,7.919120152791341
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,2,128,1,float16,float16,0,8.115946451822916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,4,128,1,float16,float16,0,8.264944076538086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,4,128,1,float16,fp8,0,6.376784006754558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,8,128,1,float16,float16,0,8.217440287272135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,1,128,1,float16,float16,0,3.583317438761393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,1,128,1,float16,fp8,0,3.4245548248291016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,32,128,1,float16,fp8,0,3.841493288675944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,32,128,1,float16,float16,0,4.310421307881673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,32,8,128,1,float16,fp8,0,6.531840006510417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,2,128,1,float16,float16,0,4.093082745869954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,4,128,1,float16,float16,0,3.90885861714681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,4,128,1,float16,fp8,0,3.4941492080688477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,1,128,1,float16,float16,0,2.093503952026367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,32,128,1,float16,float16,0,2.165034612019857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,8,128,1,float16,fp8,0,3.2648798624674478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,32,128,1,float16,fp8,0,1.9578666687011719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,8,128,1,float16,float16,0,3.8902454376220703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,1,128,1,float16,fp8,0,1.9487306276957195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,2,128,1,float16,float16,0,1.846725304921468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,2,128,1,float16,fp8,0,1.7645227114359539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,4,128,1,float16,fp8,0,1.6478187243143718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,4,128,1,float16,float16,0,1.9562187194824219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,8,128,1,float16,float16,0,2.050053278605143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,1,128,1,float16,float16,0,0.950767993927002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,32,128,1,float16,float16,0,1.0901920000712078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,32,8,128,1,float16,fp8,0,1.876138687133789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,1,128,1,float16,fp8,0,0.9365973472595215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,32,2,128,1,float16,fp8,0,3.0377651850382485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,32,128,1,float16,fp8,0,1.067850669225057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,2,128,1,float16,float16,0,0.9649813175201416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,2,128,1,float16,fp8,0,0.8129440148671468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,4,128,1,float16,fp8,0,0.8725600242614746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,4,128,1,float16,float16,0,0.9932800134023031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,8,128,1,float16,fp8,0,0.8723093668619791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,1,128,1,float16,fp8,0,0.4370400110880534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,2,128,1,float16,float16,0,0.5055040121078491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,2,128,1,float16,fp8,0,0.4460000197092692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,32,128,1,float16,float16,0,0.5813599824905396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,32,128,1,float16,fp8,0,0.5512533187866211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,4,128,1,float16,float16,0,0.5263093312581381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,4,128,1,float16,fp8,0,0.46883734067281085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,8,128,1,float16,float16,0,0.5307946602503458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,8,128,1,float16,fp8,0,0.4774880011876424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,1,128,1,float16,fp8,0,3.4470240275065103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,1,128,1,float16,float16,0,4.384250640869141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,2,128,1,float16,float16,0,4.232613245646159
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,32,1,128,1,float16,float16,0,0.5004533529281616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,4,128,1,float16,fp8,0,3.835733413696289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,8,128,1,float16,float16,0,4.623648007710774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,32,128,1,float16,float16,0,2.583402633666992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,8,128,1,float16,fp8,0,4.162954648335774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,32,128,1,float16,fp8,0,2.379866600036621
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,1,128,1,float16,float16,0,2.12334934870402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,1,128,1,float16,fp8,0,2.0992159843444824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,2,128,1,float16,float16,0,2.141306718190511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,2,128,1,float16,fp8,0,1.8156533241271973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,2,128,1,float16,fp8,0,3.5566720962524414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,4,128,1,float16,fp8,0,2.238896052042643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,4,128,1,float16,float16,0,2.3030667304992676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,32,8,128,1,float16,float16,0,1.0230133533477783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,1,128,1,float16,float16,0,1.0822186470031738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,1,128,1,float16,fp8,0,0.9032853444417318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,8,128,1,float16,float16,0,2.288661321004232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,32,8,128,1,float16,fp8,0,2.0391732851664224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,2,128,1,float16,float16,0,1.1032426357269287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,32,128,1,float16,float16,0,1.30294402440389
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,2,128,1,float16,fp8,0,0.9627520243326823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,4,128,1,float16,float16,0,1.1970400015513103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,4,128,1,float16,fp8,0,1.0850133101145427
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,8,128,1,float16,float16,0,1.1588266690572102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,1,128,1,float16,float16,0,0.5634986559549967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,8,128,1,float16,fp8,0,1.0209440390268962
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,1,128,1,float16,fp8,0,0.5157546599706014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,32,128,1,float16,float16,0,0.6696426868438721
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,32,128,1,float16,fp8,0,0.645359992980957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,2,128,1,float16,float16,0,0.5671840111414591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,4,128,1,float16,float16,0,0.6027626593907675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,4,128,1,float16,fp8,0,0.5474986632664999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,8,128,1,float16,float16,0,0.6030400196711222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,8,128,1,float16,fp8,0,0.5412266651789347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,32,128,1,float16,float16,0,0.3641493320465088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,1,128,1,float16,fp8,0,0.27266667286554974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,2,128,1,float16,float16,0,0.309168001015981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,32,128,1,float16,fp8,0,0.3510400056838989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,32,4,128,1,float16,float16,0,4.654752095540364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,4,128,1,float16,fp8,0,0.2899199922879537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,2,128,1,float16,fp8,0,0.28064000606536865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,8,128,1,float16,float16,0,0.32812267541885376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,8,128,1,float16,fp8,0,0.29867200056711835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,32,32,128,1,float16,fp8,0,1.2182880242665608
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,32,2,128,1,float16,fp8,0,0.5034186840057373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,1,128,1,float16,fp8,0,3.2758506139119468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,2,128,1,float16,fp8,0,3.4268213907877603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,1,128,1,float16,float16,0,3.996901194254557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,2,128,1,float16,float16,0,4.187904040018718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,1,128,1,float16,float16,0,0.30113067229588825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,4,128,1,float16,float16,0,4.4264373779296875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,32,4,128,1,float16,float16,0,0.32502933343251544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,1,128,1,float16,float16,0,1.9857652982076008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,4,128,1,float16,fp8,0,3.799152056376139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,1,128,1,float16,fp8,0,1.840010643005371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,8,128,1,float16,fp8,0,3.883925437927246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,32,128,1,float16,fp8,0,2.634607950846354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,2,128,1,float16,float16,0,2.0479466120402017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,32,128,1,float16,float16,0,2.6248693466186523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,2,128,1,float16,fp8,0,1.8042240142822266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,4,128,1,float16,float16,0,2.237232049306234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,1,128,1,float16,float16,0,1.0063626766204834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,4,128,1,float16,fp8,0,1.9673867225646973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,8,128,1,float16,float16,0,2.2508586247762046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,32,128,1,float16,fp8,0,1.2861493428548176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,32,8,128,1,float16,fp8,0,1.9802667299906414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,32,128,1,float16,float16,0,1.3225759665171306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,1,128,1,float16,fp8,0,1.0206613540649414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,2,128,1,float16,float16,0,1.038815975189209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,2,128,1,float16,fp8,0,0.9012906551361084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,4,128,1,float16,float16,0,1.1229759852091472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,4,128,1,float16,fp8,0,0.9970560073852539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,8,128,1,float16,float16,0,1.1362026532491047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,32,8,128,1,float16,fp8,0,1.0084746678670247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,1,128,1,float16,float16,0,0.5249706506729126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,32,128,1,float16,fp8,0,0.6607253154118856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,1,128,1,float16,fp8,0,0.4535786708196004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,2,128,1,float16,float16,0,0.5362079938252767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,4,128,1,float16,float16,0,0.5870293378829956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,32,128,1,float16,float16,0,0.6795199712117513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,32,8,128,1,float16,float16,0,4.4247894287109375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,4,128,1,float16,fp8,0,0.5204106569290161
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,8,128,1,float16,fp8,0,0.5266079902648926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,8,128,1,float16,float16,0,0.5799253384272257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,1,128,1,float16,float16,0,0.27806933720906574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,32,128,1,float16,fp8,0,0.35865068435668945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,2,128,1,float16,float16,0,0.2839733362197876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,1,128,1,float16,fp8,0,0.24991466601689658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,2,128,1,float16,fp8,0,0.2587253252665202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,4,128,1,float16,float16,0,0.30351465940475464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,4,128,1,float16,fp8,0,0.28568534056345624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,8,128,1,float16,float16,0,0.3100159962972005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,1,128,1,float16,float16,0,0.1564533313115438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,32,128,1,float16,fp8,0,0.19910933574040732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,32,128,1,float16,float16,0,0.20698134104410806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,2,128,1,float16,float16,0,0.15979199608167013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,2,128,1,float16,fp8,0,0.15003732840220133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,4,128,1,float16,fp8,0,0.1584053337574005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,8,128,1,float16,float16,0,0.1748853325843811
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,8,128,1,float16,fp8,0,0.15987733006477356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,32,2,128,1,float16,fp8,0,0.4724160035451253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,1,128,1,float16,fp8,0,1.9787200291951497
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,2,128,1,float16,float16,0,2.4807039896647134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,1,128,1,float16,float16,0,2.358565330505371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,32,128,1,float16,float16,0,0.35835198561350506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,2,128,1,float16,fp8,0,2.0880212783813477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,32,8,128,1,float16,fp8,0,0.28830933570861816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,1,128,1,float16,fp8,0,0.1469439963499705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,4,128,1,float16,fp8,0,2.5476479530334473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,4,128,1,float16,float16,0,2.6805973052978516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,32,4,128,1,float16,float16,0,0.16715733210245767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,8,128,1,float16,fp8,0,2.433488051096598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,32,8,128,1,float16,float16,0,2.7335840861002603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,1,128,1,float16,float16,0,1.19596266746521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,32,128,1,float16,float16,0,1.6453973452250164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,1,128,1,float16,fp8,0,1.0238666534423828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,2,128,1,float16,float16,0,1.2232960065205891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,32,128,1,float16,fp8,0,1.6358613967895508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,2,128,1,float16,fp8,0,1.0656800270080566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,4,128,1,float16,float16,0,1.3699092864990234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,4,128,1,float16,fp8,0,1.2402666409810383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,1,128,1,float16,float16,0,0.6078453461329142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,8,128,1,float16,float16,0,1.3812586466471355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,2,128,1,float16,float16,0,0.6272213459014893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,32,128,1,float16,fp8,0,0.8820853233337402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,2,128,1,float16,fp8,0,0.5706026554107666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,4,128,1,float16,fp8,0,0.6248746713002523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,4,128,1,float16,float16,0,0.6874667008717855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,8,128,1,float16,float16,0,0.6992053190867106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,8,128,1,float16,fp8,0,0.6413439909617106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,1,128,1,float16,float16,0,0.3161333401997884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,32,128,1,float16,float16,0,0.42841601371765137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,2,128,1,float16,float16,0,0.3305013378461202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,32,128,1,float16,fp8,0,0.4382293224334717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,2,128,1,float16,fp8,0,0.2990826765696208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,4,128,1,float16,float16,0,0.3595679998397827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,4,128,1,float16,fp8,0,0.33347733815511066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,8,128,1,float16,float16,0,0.36105600992838544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,8,128,1,float16,fp8,0,0.3394826650619507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,1,128,1,float16,float16,0,0.17431465784708658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,1,128,1,float16,fp8,0,0.15914133191108704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,2,128,1,float16,fp8,0,0.16360533237457275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,32,128,1,float16,fp8,0,0.23995733261108398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,1,128,1,float16,fp8,0,0.5256693363189697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,4,128,1,float16,float16,0,0.19798400004704794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,4,128,1,float16,fp8,0,0.17750932772954306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,32,8,128,1,float16,fp8,0,1.2311466534932454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,8,128,1,float16,fp8,0,0.18837332725524902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,8,128,1,float16,float16,0,0.19979733228683472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,32,32,128,1,float16,float16,0,0.8415573438008627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,1,128,1,float16,float16,0,0.10492266217867534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,32,128,1,float16,float16,0,0.13926933209101358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,1,128,1,float16,fp8,0,0.09703466296195984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,2,128,1,float16,float16,0,0.10683199763298035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,2,128,1,float16,fp8,0,0.09947733084360759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,4,128,1,float16,fp8,0,0.10742933551470439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,4,128,1,float16,float16,0,0.11346667011578877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,8,128,1,float16,float16,0,0.11143466830253601
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,8,128,1,float16,fp8,0,0.10738666852315266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,1,128,1,float16,fp8,0,1.970069408416748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,2,128,1,float16,float16,0,0.18027200301488241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,32,32,128,1,float16,float16,0,0.23135999838511148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,2,128,1,float16,float16,0,2.474229335784912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,2,128,1,float16,fp8,0,2.1204320589701333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,32,32,128,1,float16,fp8,0,0.13673067092895508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,4,128,1,float16,fp8,0,2.503386656443278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,4,128,1,float16,float16,0,2.773930549621582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,8,128,1,float16,float16,0,2.8145014444986978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,32,128,1,float16,float16,0,1.7850613594055176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,8,128,1,float16,fp8,0,2.5924533208211265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,1,128,1,float16,float16,0,1.1751253604888916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,1,128,1,float16,fp8,0,1.0547040303548176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,2,128,1,float16,float16,0,1.2248799800872803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,32,128,1,float16,fp8,0,1.834368069966634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,32,1,128,1,float16,float16,0,2.3423892656962075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,2,128,1,float16,fp8,0,1.075605312983195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,4,128,1,float16,float16,0,1.3993760744730632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,4,128,1,float16,fp8,0,1.2783093452453613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,8,128,1,float16,fp8,0,1.2978560129801433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,32,8,128,1,float16,float16,0,1.4301600456237793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,1,128,1,float16,fp8,0,0.5188746849695841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,1,128,1,float16,float16,0,0.5947253306706747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,2,128,1,float16,fp8,0,0.5654933452606201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,32,128,1,float16,fp8,0,0.9284640153249105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,2,128,1,float16,float16,0,0.6191360155741373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,4,128,1,float16,fp8,0,0.6489119927088419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,32,1,128,1,float16,fp8,0,0.2846720019976298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,1,128,1,float16,float16,0,0.3083999951680501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,8,128,1,float16,float16,0,0.7183626492818197
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,32,128,1,float16,fp8,0,0.4812693198521932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,32,128,1,float16,float16,0,0.45446932315826416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,4,128,1,float16,float16,0,0.362277348836263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,2,128,1,float16,fp8,0,0.2942933241526286
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,4,128,1,float16,fp8,0,0.34362133344014484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,8,128,1,float16,float16,0,0.36742401123046875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,8,128,1,float16,fp8,0,0.3492106596628825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,32,128,1,float16,float16,0,0.24810133377710977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,32,128,1,float16,float16,0,0.8951679865519205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,1,128,1,float16,float16,0,0.1727679967880249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,32,128,1,float16,fp8,0,0.2561653256416321
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,1,128,1,float16,fp8,0,0.15260266264279684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,2,128,1,float16,float16,0,0.17688000202178955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,8,128,1,float16,fp8,0,0.6624853213628134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,2,128,1,float16,fp8,0,0.1599573294321696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,4,128,1,float16,fp8,0,0.18565332889556885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,8,128,1,float16,fp8,0,0.18894400199254355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,1,128,1,float16,fp8,0,0.27753599484761554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,32,2,128,1,float16,float16,0,0.31968533992767334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,8,128,1,float16,float16,0,0.200762669245402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,1,128,1,float16,float16,0,0.09634666641553243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,1,128,1,float16,fp8,0,0.0900426705678304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,32,128,1,float16,float16,0,0.13674133022626242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,2,128,1,float16,fp8,0,0.0936853289604187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,2,128,1,float16,float16,0,0.09956266482671101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,4,128,1,float16,float16,0,0.10527466734250386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,4,128,1,float16,fp8,0,0.10246400038401286
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,8,128,1,float16,float16,0,0.11244266231854756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,8,128,1,float16,fp8,0,0.10403199990590413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,1,128,1,float16,float16,0,0.06362666686375935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,1,128,1,float16,fp8,0,0.058320000767707825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,2,128,1,float16,float16,0,0.06406400104363759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,2,128,1,float16,fp8,0,0.06003733476003011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,32,128,1,float16,float16,0,0.08053866525491078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,4,128,1,float16,float16,0,0.06731200218200684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,4,128,1,float16,fp8,0,0.06435200075308482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,8,128,1,float16,float16,0,0.06841599941253662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,8,128,1,float16,fp8,0,0.06427200138568878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,32,4,128,1,float16,float16,0,0.7009600003560384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,1,128,1,float16,fp8,0,1.2373279730478923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,32,4,128,1,float16,float16,0,0.19632534186045328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,2,128,1,float16,float16,0,1.5396960576375325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,2,128,1,float16,fp8,0,1.3484266599019368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,4,128,1,float16,float16,0,1.7716479301452637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,32,32,128,1,float16,fp8,0,0.14392000436782837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,4,128,1,float16,fp8,0,1.6278400421142578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,8,128,1,float16,float16,0,1.8063732782999675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,1,128,1,float16,fp8,0,0.634496013323466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,1,128,1,float16,float16,0,0.7291786670684814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,32,32,128,1,float16,fp8,0,0.08124800026416779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,32,128,1,float16,fp8,0,1.253445307413737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,32,128,1,float16,float16,0,1.17958402633667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,8,128,1,float16,fp8,0,1.6917014122009277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,2,128,1,float16,float16,0,0.7650667031606039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,2,128,1,float16,fp8,0,0.6873173713684082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,4,128,1,float16,fp8,0,0.8309600353240967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,4,128,1,float16,float16,0,0.8949173291524252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,8,128,1,float16,fp8,0,0.8559733231862386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,32,1,128,1,float16,float16,0,1.4558347066243489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,1,128,1,float16,float16,0,0.3718986511230469
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,1,128,1,float16,fp8,0,0.33222933610280353
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,32,128,1,float16,float16,0,0.5976106723149618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,4,128,1,float16,fp8,0,0.4297226667404175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,2,128,1,float16,fp8,0,0.3594293196996053
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,4,128,1,float16,float16,0,0.45632000764211017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,32,128,1,float16,fp8,0,0.6384853521982828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,2,128,1,float16,float16,0,0.396234671274821
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,1,128,1,float16,float16,0,0.19628800948460898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,8,128,1,float16,float16,0,0.4633866548538208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,1,128,1,float16,fp8,0,0.1814346710840861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,2,128,1,float16,float16,0,0.2099413275718689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,2,128,1,float16,fp8,0,0.1965493361155192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,4,128,1,float16,float16,0,0.23798400163650513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,32,128,1,float16,float16,0,0.31140265862147015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,32,128,1,float16,fp8,0,0.3338186740875244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,4,128,1,float16,fp8,0,0.23059199253718057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,8,128,1,float16,float16,0,0.24360533555348715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,1,128,1,float16,fp8,0,0.10187733173370361
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,1,128,1,float16,float16,0,0.11227200428644817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,32,8,128,1,float16,fp8,0,0.2382240096728007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,2,128,1,float16,float16,0,0.1167093316713969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,32,128,1,float16,fp8,0,0.18075732390085855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,2,128,1,float16,fp8,0,0.10734933614730835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,4,128,1,float16,float16,0,0.13346133629480997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,32,128,1,float16,float16,0,0.17270932594935098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,4,128,1,float16,fp8,0,0.11975466211636861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,8,128,1,float16,float16,0,0.13521599769592285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,1,128,1,float16,float16,0,0.06664533416430156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,1,128,1,float16,fp8,0,0.06186666587988535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,2,128,1,float16,float16,0,0.06825600067774455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,32,8,128,1,float16,float16,0,0.9056853453318278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,32,128,1,float16,float16,0,0.10658666491508484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,2,128,1,float16,fp8,0,0.06465066472689311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,4,128,1,float16,float16,0,0.07317333420117696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,4,128,1,float16,fp8,0,0.07124799986680348
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,8,128,1,float16,fp8,0,0.072543998559316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,1,128,1,float16,float16,0,0.04384533564249674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,1,128,1,float16,fp8,0,0.04111466556787491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,32,128,1,float16,fp8,0,0.05894933144251505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,2,128,1,float16,float16,0,0.04562133550643921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,32,8,128,1,float16,fp8,0,0.44526398181915283
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,2,128,1,float16,fp8,0,0.04218666752179464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,4,128,1,float16,float16,0,0.04860266546408335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,4,128,1,float16,fp8,0,0.04740799963474274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,8,128,1,float16,float16,0,0.04790399968624115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,8,128,1,float16,fp8,0,0.04755199948946635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,32,8,128,1,float16,fp8,0,0.12990400195121765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,32,128,1,float16,fp8,0,0.10145066181818645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,1,128,1,float16,float16,0,1.527242660522461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,2,128,1,float16,float16,0,1.6396959622701008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,32,8,128,1,float16,float16,0,0.07602666815121968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,32,32,128,1,float16,float16,0,0.05426666637261709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,2,128,1,float16,fp8,0,1.4560267130533855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,8,128,1,float16,float16,0,2.0295467376708984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,4,128,1,float16,float16,0,1.9892160097757976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,1,128,1,float16,float16,0,0.7637973626454672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,32,128,1,float16,float16,0,1.3670239448547363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,4,128,1,float16,fp8,0,1.837813377380371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,8,128,1,float16,fp8,0,1.9096479415893555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,32,128,1,float16,fp8,0,1.4982080459594727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,1,128,1,float16,fp8,0,0.6747519969940186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,2,128,1,float16,fp8,0,0.7378239631652832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,32,1,128,1,float16,fp8,0,1.3146666685740154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,2,128,1,float16,float16,0,0.8271679878234863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,4,128,1,float16,float16,0,0.9882079760233561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,4,128,1,float16,fp8,0,0.9329066276550293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,8,128,1,float16,float16,0,1.0159413019816081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,32,8,128,1,float16,fp8,0,0.9667572975158691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,1,128,1,float16,fp8,0,0.3457706769307454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,32,128,1,float16,fp8,0,0.7569653193155924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,2,128,1,float16,float16,0,0.41844268639882404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,4,128,1,float16,float16,0,0.5044906536738077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,4,128,1,float16,fp8,0,0.4805599848429362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,8,128,1,float16,float16,0,0.5113919973373413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,8,128,1,float16,fp8,0,0.4929493268330892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,1,128,1,float16,float16,0,0.20778133471806845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,1,128,1,float16,fp8,0,0.18781334161758423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,32,128,1,float16,float16,0,0.3508479992548625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,2,128,1,float16,fp8,0,0.20588799317677817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,2,128,1,float16,float16,0,0.21979200839996338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,32,128,1,float16,fp8,0,0.389792005221049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,4,128,1,float16,float16,0,0.2637653350830078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,4,128,1,float16,fp8,0,0.25465067227681476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,8,128,1,float16,fp8,0,0.26243199904759723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,32,8,128,1,float16,float16,0,0.2701920072237651
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,1,128,1,float16,float16,0,0.39070932070414227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,1,128,1,float16,fp8,0,0.10294933120409648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,2,128,1,float16,float16,0,0.12358933687210083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,2,128,1,float16,fp8,0,0.11182933052380879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,32,128,1,float16,float16,0,0.19702933231989542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,32,128,1,float16,fp8,0,0.2059733271598816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,4,128,1,float16,float16,0,0.14149866501490274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,4,128,1,float16,fp8,0,0.13581333557764688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,32,128,1,float16,float16,0,0.6879573663075765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,8,128,1,float16,float16,0,0.14749333262443542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,1,128,1,float16,float16,0,0.06609066824118297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,32,2,128,1,float16,fp8,0,0.38336535294850665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,1,128,1,float16,fp8,0,0.06063466767470042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,32,128,1,float16,float16,0,0.10439466436704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,32,128,1,float16,fp8,0,0.113237331310908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,4,128,1,float16,float16,0,0.07609599828720093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,4,128,1,float16,fp8,0,0.07252266506354015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,8,128,1,float16,float16,0,0.07919999957084656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,8,128,1,float16,fp8,0,0.07624533275763194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,1,128,1,float16,float16,0,0.041589332123597465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,1,128,1,float16,fp8,0,0.03979733337958654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,32,128,1,float16,float16,0,0.06007466713587443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,2,128,1,float16,float16,0,0.04351999859015147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,2,128,1,float16,fp8,0,0.04142933338880539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,4,128,1,float16,float16,0,0.04654933512210846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,4,128,1,float16,fp8,0,0.0455626646677653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,8,128,1,float16,float16,0,0.047423998514811196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,8,128,1,float16,fp8,0,0.04586133360862732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,1,128,1,float16,float16,0,0.02703999976317088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,1,128,1,float16,float16,0,0.11617066462834676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,32,128,1,float16,float16,0,0.031530665854612984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,32,128,1,float16,fp8,0,0.03728533287843069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,2,128,1,float16,float16,0,0.027466667195161183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,2,128,1,float16,fp8,0,0.027104000250498455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,4,128,1,float16,float16,0,0.02917333443959554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,4,128,1,float16,fp8,0,0.02926933268706004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,8,128,1,float16,float16,0,0.029311999678611755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,8,128,1,float16,fp8,0,0.02935466667016347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,2,128,1,float16,float16,0,0.06914666791756947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,32,32,128,1,float16,fp8,0,0.0629066675901413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,1,128,1,float16,float16,0,1.1393280029296875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,1,128,1,float16,fp8,0,0.9878079891204834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,2,128,1,float16,float16,0,1.251589298248291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,32,1,128,1,float16,fp8,0,0.025098666548728943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,32,8,128,1,float16,fp8,0,0.14064533511797586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,2,128,1,float16,fp8,0,1.1281493504842122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,4,128,1,float16,float16,0,1.596010684967041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,4,128,1,float16,fp8,0,1.5061333974202473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,8,128,1,float16,float16,0,1.627295970916748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,1,128,1,float16,float16,0,0.5781279802322388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,32,8,128,1,float16,fp8,0,1.5869654019673665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,32,128,1,float16,fp8,0,1.32150403658549
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,2,128,1,float16,float16,0,0.6405706803003947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,4,128,1,float16,float16,0,0.812613328297933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,32,2,128,1,float16,fp8,0,0.06426666676998138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,2,128,1,float16,fp8,0,0.5764426787694296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,8,128,1,float16,float16,0,0.822533369064331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,8,128,1,float16,fp8,0,0.7958347002665201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,4,128,1,float16,fp8,0,0.7613333066304525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,1,128,1,float16,float16,0,0.30558399359385174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,32,128,1,float16,float16,0,1.175376017888387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,32,128,1,float16,float16,0,0.5914719899495443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,32,128,1,float16,fp8,0,0.6705919901529948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,32,1,128,1,float16,fp8,0,0.5047733386357626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,4,128,1,float16,fp8,0,0.39587199687957764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,2,128,1,float16,fp8,0,0.3025866746902466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,8,128,1,float16,float16,0,0.4200693368911743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,1,128,1,float16,fp8,0,0.2638133366902669
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,2,128,1,float16,float16,0,0.3349173466364543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,8,128,1,float16,fp8,0,0.4099839925765991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,1,128,1,float16,float16,0,0.16193600495656332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,32,128,1,float16,fp8,0,0.34513068199157715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,2,128,1,float16,float16,0,0.17597333590189615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,1,128,1,float16,fp8,0,0.1423306663831075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,2,128,1,float16,fp8,0,0.16051200032234192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,4,128,1,float16,fp8,0,0.21029333273569742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,8,128,1,float16,float16,0,0.22480533520380655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,4,128,1,float16,float16,0,0.22122132778167725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,8,128,1,float16,fp8,0,0.21516267458597818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,32,128,1,float16,float16,0,0.16131200393040976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,32,128,1,float16,fp8,0,0.18144534031550089
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,1,128,1,float16,float16,0,0.09206933776537578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,2,128,1,float16,float16,0,0.09907199939092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,2,128,1,float16,fp8,0,0.08788266777992249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,1,128,1,float16,fp8,0,0.07858666777610779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,4,128,1,float16,float16,0,0.11724266409873962
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,4,128,1,float16,fp8,0,0.11249599854151408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,1,128,1,float16,float16,0,0.05236800014972687
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,32,4,128,1,float16,float16,0,0.4148000081380208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,8,128,1,float16,fp8,0,0.11778133114178975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,1,128,1,float16,fp8,0,0.04623466730117798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,2,128,1,float16,float16,0,0.05416533350944519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,32,8,128,1,float16,float16,0,0.12270399928092957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,32,128,1,float16,float16,0,0.0902239978313446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,2,128,1,float16,fp8,0,0.0499839981396993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,8,128,1,float16,float16,0,0.06640533109505971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,8,128,1,float16,fp8,0,0.06084266801675161
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,1,128,1,float16,float16,0,0.033546666304270424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,1,128,1,float16,fp8,0,0.030106666187445324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,32,32,128,1,float16,float16,0,0.3037760059038798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,2,128,1,float16,float16,0,0.033285332222779594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,32,128,1,float16,float16,0,0.05576533575852712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,32,128,1,float16,fp8,0,0.05398400127887726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,2,128,1,float16,fp8,0,0.03133866687615713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,4,128,1,float16,fp8,0,0.03569599986076355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,8,128,1,float16,float16,0,0.0374293327331543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,1,128,1,float16,float16,0,0.021061333517233532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,32,128,1,float16,float16,0,0.025231999655564625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,1,128,1,float16,fp8,0,0.01912533367673556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,2,128,1,float16,float16,0,0.021290667355060577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,2,128,1,float16,fp8,0,0.020975999534130096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,4,128,1,float16,fp8,0,0.023311999936898548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,8,128,1,float16,float16,0,0.02298133323589961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,32,128,1,float16,fp8,0,0.09983999530474345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,8,128,1,float16,fp8,0,0.02329600105683009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,4,128,1,float16,fp8,0,0.059605335195859276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,32,4,128,1,float16,float16,0,0.06286400059858958
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,1,128,1,float16,float16,0,0.019013332823912304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,1,128,1,float16,fp8,0,0.018992000569899876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,32,128,1,float16,fp8,0,0.025205334027608235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,2,128,1,float16,fp8,0,0.018842666099468868
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,32,128,1,float16,float16,0,0.021205333371957142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,4,128,1,float16,float16,0,0.018944000204404194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,4,128,1,float16,fp8,0,0.02012266715367635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,4,128,1,float16,float16,0,0.03761066744724909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,8,128,1,float16,float16,0,0.019440000255902607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,32,8,128,1,float16,fp8,0,0.037258667250474296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,32,128,1,float16,fp8,0,0.03148799886306127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,1,128,1,float16,float16,0,0.44193601608276367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,1,128,1,float16,fp8,0,0.41015998522440594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,32,4,128,1,float16,float16,0,0.022991999983787537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,2,128,1,float16,float16,0,0.4999573230743408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,2,128,1,float16,fp8,0,0.4774719874064128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,4,128,1,float16,float16,0,0.678879976272583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,2,128,1,float16,float16,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,32,128,1,float16,float16,0,0.5250453154246012
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,32,128,1,float16,fp8,0,0.6177226702372233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,8,128,1,float16,float16,0,0.6875519752502441
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,4,128,1,float16,fp8,0,0.6649493376413981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,32,8,128,1,float16,fp8,0,0.020970667401949566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,1,128,1,float16,fp8,0,0.21676800648371378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,2,128,1,float16,float16,0,0.26659733057022095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,4,128,1,float16,float16,0,0.34454933802286786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,32,8,128,1,float16,fp8,0,0.7039573192596436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,4,128,1,float16,fp8,0,0.34140801429748535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,2,128,1,float16,fp8,0,0.25205334027608234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,8,128,1,float16,float16,0,0.355951984723409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,1,128,1,float16,fp8,0,0.11506666739781697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,1,128,1,float16,float16,0,0.12846400340398154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,32,128,1,float16,float16,0,0.2712426582972209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,2,128,1,float16,fp8,0,0.1314826707045237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,2,128,1,float16,float16,0,0.13909866412480673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,1,128,1,float16,float16,0,0.2311840057373047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,32,128,1,float16,fp8,0,0.31699200471242267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,4,128,1,float16,fp8,0,0.1873813271522522
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,8,128,1,float16,float16,0,0.18820800383885702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,8,128,1,float16,fp8,0,0.18702934185663858
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,1,128,1,float16,float16,0,0.07025066514809926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,1,128,1,float16,fp8,0,0.06253866851329803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,32,128,1,float16,fp8,0,0.1648319959640503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,32,128,1,float16,float16,0,0.14290666580200195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,4,128,1,float16,fp8,0,0.09487467010815938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,4,128,1,float16,float16,0,0.09869333108266194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,2,128,1,float16,float16,0,0.07795733213424683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,2,128,1,float16,fp8,0,0.0716480016708374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,8,128,1,float16,float16,0,0.09903466701507568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,1,128,1,float16,float16,0,0.04083733260631561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,1,128,1,float16,fp8,0,0.03752533346414566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,32,8,128,1,float16,fp8,0,0.10844266414642334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,2,128,1,float16,float16,0,0.043882668018341064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,2,128,1,float16,fp8,0,0.04178666571776072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,4,128,1,float16,float16,0,0.0505920002857844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,32,128,1,float16,fp8,0,0.09286933143933614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,32,128,1,float16,float16,0,0.08073066671689351
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,8,128,1,float16,float16,0,0.05584000051021576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,8,128,1,float16,fp8,0,0.052058666944503784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,32,8,128,1,float16,fp8,0,0.35729066530863446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,1,128,1,float16,float16,0,0.025061334172884624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,2,128,1,float16,float16,0,0.02714666724205017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,1,128,1,float16,fp8,0,0.023546665906906128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,4,128,1,float16,fp8,0,0.029493334392706554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,32,128,1,float16,fp8,0,0.04779199759165446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,4,128,1,float16,float16,0,0.029504001140594482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,32,128,1,float16,float16,0,0.04269866645336151
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,2,128,1,float16,fp8,0,0.025360000630219776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,8,128,1,float16,float16,0,0.02940266579389572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,32,8,128,1,float16,fp8,0,0.03143999973932902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,1,128,1,float16,fp8,0,0.016202667107184727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,2,128,1,float16,float16,0,0.01714133347074191
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,2,128,1,float16,fp8,0,0.017829333742459614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,4,128,1,float16,float16,0,0.019061333189407986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,32,4,128,1,float16,float16,0,0.18395733833312988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,4,128,1,float16,fp8,0,0.019258666783571243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,32,128,1,float16,fp8,0,0.027509334186712902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,32,128,1,float16,float16,0,0.022442666192849476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,8,128,1,float16,fp8,0,0.020992000897725422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,8,128,1,float16,float16,0,0.019280000279347103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,2,128,1,float16,float16,0,0.014853333433469137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,1,128,1,float16,fp8,0,0.015184000134468079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,4,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,8,128,1,float16,float16,0,0.016965333372354507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,32,128,1,float16,float16,0,0.019285333653291065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,4,128,1,float16,fp8,0,0.017045332739750545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,8,128,1,float16,fp8,0,0.017258666455745697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,1,128,1,float16,float16,0,0.014837333311637243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,32,4,128,1,float16,fp8,0,0.0516533354918162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,1,128,1,float16,fp8,0,0.014762666076421738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,32,128,1,float16,float16,0,0.01711999997496605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,2,128,1,float16,fp8,0,0.014746667196353277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,32,128,1,float16,fp8,0,0.017530667285124462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,4,128,1,float16,fp8,0,0.014682666709025701
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,4,128,1,float16,float16,0,0.015216000378131866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,8,128,1,float16,float16,0,0.015402667224407196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,8,128,1,float16,fp8,0,0.014901333798964819
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,32,1,128,1,float16,float16,0,0.01674666628241539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,1,128,1,float16,float16,0,0.014906667172908783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,2,128,1,float16,fp8,0,0.01525866612792015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,2,128,1,float16,float16,0,0.31117866436640423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,2,128,1,float16,fp8,0,0.29098133246103924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,32,2,128,1,float16,float16,0,0.015034666905800501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,1,128,1,float16,float16,0,0.2828906575838725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,1,128,1,float16,fp8,0,0.25969066222508747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,4,128,1,float16,float16,0,0.38863468170166016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,4,128,1,float16,fp8,0,0.3828959862391154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,8,128,1,float16,float16,0,0.39973334471384686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,32,8,128,1,float16,fp8,0,0.40034135182698566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,1,128,1,float16,fp8,0,0.1363146702448527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,1,128,1,float16,float16,0,0.15081600348154703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,32,32,128,1,float16,fp8,0,0.021381333470344543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,32,128,1,float16,fp8,0,0.3378186623255412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,32,128,1,float16,float16,0,0.29054399331410724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,4,128,1,float16,fp8,0,0.19944000244140625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,8,128,1,float16,float16,0,0.20893865823745728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,8,128,1,float16,fp8,0,0.2093013326327006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,2,128,1,float16,fp8,0,0.15541866421699524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,4,128,1,float16,float16,0,0.2048799991607666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,1,128,1,float16,float16,0,0.08268799881140391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,32,128,1,float16,fp8,0,0.1753386656443278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,32,128,1,float16,float16,0,0.15236266454060873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,1,128,1,float16,fp8,0,0.07418666779994965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,4,128,1,float16,float16,0,0.10959466298421223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,2,128,1,float16,float16,0,0.0905013382434845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,2,128,1,float16,fp8,0,0.08302400012811025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,32,2,128,1,float16,float16,0,0.16350932916005453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,8,128,1,float16,fp8,0,0.11117866635322571
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,1,128,1,float16,float16,0,0.045738667249679565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,8,128,1,float16,float16,0,0.11212266484896342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,32,4,128,1,float16,fp8,0,0.10570133725802104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,32,128,1,float16,fp8,0,0.0950933297475179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,32,128,1,float16,float16,0,0.09553600351015727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,2,128,1,float16,float16,0,0.04783466458320618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,1,128,1,float16,fp8,0,0.04146666576464971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,4,128,1,float16,fp8,0,0.05393599967161814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,4,128,1,float16,float16,0,0.05806933343410492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,8,128,1,float16,float16,0,0.06226666768391927
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,1,128,1,float16,float16,0,0.027984000742435455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,1,128,1,float16,fp8,0,0.026421333352724712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,32,128,1,float16,float16,0,0.0516533354918162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,32,128,1,float16,fp8,0,0.0495413343111674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,2,128,1,float16,fp8,0,0.027242665489514668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,4,128,1,float16,float16,0,0.032645332316557564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,8,128,1,float16,fp8,0,0.03340800106525421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,8,128,1,float16,float16,0,0.033258666594823204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,4,128,1,float16,fp8,0,0.03316800047953924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,1,128,1,float16,float16,0,0.017877332866191864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,32,128,1,float16,fp8,0,0.02917333443959554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,32,128,1,float16,float16,0,0.02294933299223582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,2,128,1,float16,float16,0,0.019029332945744198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,2,128,1,float16,fp8,0,0.019167999426523846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,4,128,1,float16,float16,0,0.020842666427294414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,4,128,1,float16,fp8,0,0.021029333273569744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,8,128,1,float16,float16,0,0.02094399929046631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,8,128,1,float16,fp8,0,0.02128000060717265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,32,128,1,float16,float16,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,1,128,1,float16,float16,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,2,128,1,float16,float16,0,0.012858666479587555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,2,128,1,float16,fp8,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,4,128,1,float16,float16,0,0.014805333067973455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,8,128,1,float16,fp8,0,0.0554613322019577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,4,128,1,float16,fp8,0,0.015077333897352219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,8,128,1,float16,float16,0,0.01479999969402949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,32,2,128,1,float16,float16,0,0.029498666524887085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,1,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,32,128,1,float16,float16,0,0.013221333424250284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,1,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,2,128,1,float16,fp8,0,0.012639999389648438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,32,1,128,1,float16,fp8,0,0.017173333714405697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,4,128,1,float16,float16,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,4,128,1,float16,fp8,0,0.013002666334311167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,8,128,1,float16,float16,0,0.012682666381200155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,8,128,1,float16,fp8,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,32,2,128,1,float16,fp8,0,0.04486933350563049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,32,128,1,float16,fp8,0,0.01350933313369751
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,32,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,1,128,1,float16,float16,0,0.011999999483426413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,1,128,1,float16,fp8,0,0.013125333935022354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,2,128,1,float16,float16,0,0.012671999633312225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,2,128,1,float16,fp8,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,4,128,1,float16,float16,0,0.012730666746695837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,4,128,1,float16,fp8,0,0.012650666137536367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,8,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,8,128,1,float16,fp8,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,8,128,1,float16,fp8,0,0.014826666563749313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,32,128,1,float16,fp8,0,0.015103999525308609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,1,128,1,float16,float16,0,0.2406933307647705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,1,128,1,float16,fp8,0,0.20892266432444254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,2,128,1,float16,float16,0,0.2539359927177429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,4,128,1,float16,float16,0,0.29557865858078003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,32,32,128,1,float16,fp8,0,0.019088000059127808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,4,128,1,float16,fp8,0,0.27684799830118817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,8,128,1,float16,float16,0,0.2996053298314412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,8,128,1,float16,fp8,0,0.28204800685246784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,1,128,1,float16,float16,0,0.13130133350690207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,32,128,1,float16,float16,0,0.2079360087712606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,32,128,1,float16,fp8,0,0.21237333615620932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,1,128,1,float16,fp8,0,0.11384532848993938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,2,128,1,float16,float16,0,0.13869333267211914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,2,128,1,float16,fp8,0,0.12036800384521484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,32,2,128,1,float16,float16,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,8,128,1,float16,float16,0,0.15933866302172342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,8,128,1,float16,fp8,0,0.14829333623250326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,4,128,1,float16,float16,0,0.1569813291231791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,32,4,128,1,float16,fp8,0,0.14401599764823914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,32,128,1,float16,float16,0,0.11609066526095073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,1,128,1,float16,float16,0,0.07025599976380666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,1,128,1,float16,fp8,0,0.06043200194835663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,32,128,1,float16,fp8,0,0.11276266972223918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,2,128,1,float16,float16,0,0.07251733541488647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,2,128,1,float16,fp8,0,0.0641546646753947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,4,128,1,float16,fp8,0,0.0731573353211085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,4,128,1,float16,float16,0,0.08277866741021474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,8,128,1,float16,float16,0,0.08619733651479085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,32,8,128,1,float16,fp8,0,0.07470400134722392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,32,1,128,1,float16,fp8,0,0.012703999876976013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,1,128,1,float16,float16,0,0.039674667020638786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,32,128,1,float16,fp8,0,0.057999998331069946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,32,128,1,float16,float16,0,0.0576853354771932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,1,128,1,float16,fp8,0,0.03543466577927271
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,2,128,1,float16,fp8,0,0.037402667105197906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,4,128,1,float16,float16,0,0.04381866753101349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,4,128,1,float16,fp8,0,0.04130133241415024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,8,128,1,float16,float16,0,0.04418133199214935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,8,128,1,float16,fp8,0,0.04170133173465729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,1,128,1,float16,float16,0,0.025066666305065155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,32,128,1,float16,float16,0,0.029477333029111225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,2,128,1,float16,float16,0,0.025381334125995636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,1,128,1,float16,fp8,0,0.023034666975339253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,4,128,1,float16,float16,0,0.027119999130566914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,8,128,1,float16,float16,0,0.027034667630990345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,32,2,128,1,float16,fp8,0,0.22639467318852743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,8,128,1,float16,fp8,0,0.02720533311367035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,1,128,1,float16,float16,0,0.01693333312869072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,1,128,1,float16,fp8,0,0.01718933383623759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,32,128,1,float16,fp8,0,0.022800001005331676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,2,128,1,float16,fp8,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,2,128,1,float16,float16,0,0.016842667013406754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,4,128,1,float16,float16,0,0.017184000462293625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,4,128,1,float16,fp8,0,0.017231999586025875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,8,128,1,float16,float16,0,0.018986667195955913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,8,128,1,float16,fp8,0,0.01766933376590411
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,32,2,128,1,float16,float16,0,0.041434665520985924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,1,128,1,float16,float16,0,0.012703999876976013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,32,128,1,float16,float16,0,0.012874666601419449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,1,128,1,float16,fp8,0,0.012901333471139273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,2,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,32,128,1,float16,fp8,0,0.014826666563749313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,4,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,8,128,1,float16,float16,0,0.012645332763592402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,8,128,1,float16,fp8,0,0.013056000073750814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,32,128,1,float16,fp8,0,0.03535466641187668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,4,128,1,float16,fp8,0,0.026176000634829204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,1,128,1,float16,float16,0,0.012053333222866058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,32,128,1,float16,fp8,0,0.01302933320403099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,1,128,1,float16,fp8,0,0.012448000411192576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,2,128,1,float16,float16,0,0.011994666109482447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,32,128,1,float16,float16,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,2,128,1,float16,fp8,0,0.01267733300725619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,4,128,1,float16,float16,0,0.01179733375708262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,4,128,1,float16,fp8,0,0.012666666259368261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,8,128,1,float16,float16,0,0.01163200040658315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,32,8,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,1,128,1,float16,float16,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,1,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,2,128,1,float16,fp8,0,0.011152000476916632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,32,128,1,float16,float16,0,0.0129120002190272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,4,128,1,float16,float16,0,0.011141333729028702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,4,128,1,float16,fp8,0,0.010805333654085795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,8,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,8,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,4,128,1,float16,float16,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,32,2,128,1,float16,fp8,0,0.025311999022960663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,1,128,1,float16,float16,0,0.2256960074106852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,1,128,1,float16,fp8,0,0.18926399946212769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,2,128,1,float16,float16,0,0.23481067021687826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,2,128,1,float16,fp8,0,0.19662400086720785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,32,32,128,1,float16,float16,0,0.019258666783571243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,4,128,1,float16,float16,0,0.2515146732330322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,2,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,4,128,1,float16,fp8,0,0.21960532665252686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,32,32,128,1,float16,fp8,0,0.012944000462690989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,8,128,1,float16,float16,0,0.2569813330968221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,1,128,1,float16,float16,0,0.11884267131487529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,32,8,128,1,float16,fp8,0,0.22483734289805093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,1,128,1,float16,fp8,0,0.09874133268992107
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,2,128,1,float16,float16,0,0.12585066755612692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,32,128,1,float16,float16,0,0.1614720026652018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,32,2,128,1,float16,fp8,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,2,128,1,float16,fp8,0,0.10242133339246114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,4,128,1,float16,float16,0,0.13250666856765747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,4,128,1,float16,fp8,0,0.11291733384132385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,8,128,1,float16,fp8,0,0.11318399508794148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,1,128,1,float16,float16,0,0.06607466439406078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,1,128,1,float16,fp8,0,0.05410666763782501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,32,128,1,float16,fp8,0,0.07893866797288258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,2,128,1,float16,float16,0,0.06818133095900218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,4,128,1,float16,float16,0,0.07009600102901459
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,2,128,1,float16,fp8,0,0.05611733098824819
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,4,128,1,float16,fp8,0,0.06018133461475372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,8,128,1,float16,float16,0,0.07036266724268596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,8,128,1,float16,fp8,0,0.06144533554712931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,1,128,1,float16,float16,0,0.03763733307520548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,1,128,1,float16,fp8,0,0.03160533308982849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,2,128,1,float16,float16,0,0.039066667358080544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,2,128,1,float16,fp8,0,0.03350933392842611
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,32,128,1,float16,float16,0,0.04380266865094503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,4,128,1,float16,float16,0,0.03937600056330363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,4,128,1,float16,fp8,0,0.035487999518712364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,8,128,1,float16,float16,0,0.039621333281199135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,8,128,1,float16,fp8,0,0.03540800015131632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,1,128,1,float16,float16,0,0.023152001202106476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,32,128,1,float16,fp8,0,0.15039466818173727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,32,128,1,float16,float16,0,0.027119999130566914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,2,128,1,float16,float16,0,0.023050665855407715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,4,128,1,float16,fp8,0,0.023183998962243397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,2,128,1,float16,fp8,0,0.021061333517233532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,8,128,1,float16,float16,0,0.0249493345618248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,32,8,128,1,float16,float16,0,0.13611732920010886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,8,128,1,float16,fp8,0,0.02330133318901062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,32,128,1,float16,fp8,0,0.018042666216691334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,32,128,1,float16,float16,0,0.017130666722853977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,1,128,1,float16,float16,0,0.01578666642308235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,32,32,128,1,float16,fp8,0,0.04358399907747904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,1,128,1,float16,fp8,0,0.014778666198253632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,4,128,1,float16,float16,0,0.015722667177518208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,2,128,1,float16,fp8,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,4,128,1,float16,fp8,0,0.015157333264748255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,8,128,1,float16,float16,0,0.016927999754746754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,8,128,1,float16,fp8,0,0.015072000523408255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,32,128,1,float16,fp8,0,0.028079998989899952
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,1,128,1,float16,fp8,0,0.02130666623512904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,1,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,32,4,128,1,float16,float16,0,0.025205334027608235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,2,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,32,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,32,128,1,float16,fp8,0,0.014122666170199713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,2,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,4,128,1,float16,float16,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,8,128,1,float16,float16,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,4,128,1,float16,fp8,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,8,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,32,32,128,1,float16,float16,0,0.08468266328175862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,1,128,1,float16,fp8,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,2,128,1,float16,fp8,0,0.011173332730929056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,32,128,1,float16,float16,0,0.012949333836634954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,2,128,1,float16,float16,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,32,128,1,float16,fp8,0,0.012997332960367203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,4,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,8,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,4,128,1,float16,fp8,0,0.011152000476916632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,1,128,1,float16,float16,0,0.010832000523805618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,32,2,128,1,float16,float16,0,0.016399999459584553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,1,128,1,float16,fp8,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,2,128,1,float16,float16,0,0.011141333729028702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,32,128,1,float16,float16,0,0.011567999919255575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,2,128,1,float16,fp8,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,4,128,1,float16,float16,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,8,128,1,float16,float16,0,0.010805333654085795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,8,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,32,1,128,1,float16,float16,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,1,128,1,float16,float16,0,0.21986132860183716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,1,128,1,float16,fp8,0,0.17482666174570718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,2,128,1,float16,float16,0,0.2216213345527649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,1,128,1,float16,float16,0,0.01110400011142095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,2,128,1,float16,fp8,0,0.1781866749127706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,4,128,1,float16,float16,0,0.23567465941111246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,4,128,1,float16,fp8,0,0.1899306575457255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,32,8,128,1,float16,fp8,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,8,128,1,float16,float16,0,0.23370667298634848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,32,128,1,float16,fp8,0,0.012719999998807907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,32,8,128,1,float16,fp8,0,0.1949066718419393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,1,128,1,float16,float16,0,0.11730133493741353
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,32,4,128,1,float16,fp8,0,0.011407999942700068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,1,128,1,float16,fp8,0,0.09392000238100688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,32,128,1,float16,fp8,0,0.12060266733169556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,2,128,1,float16,float16,0,0.1164959967136383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,2,128,1,float16,fp8,0,0.09571199615796407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,4,128,1,float16,fp8,0,0.09993066390355428
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,8,128,1,float16,fp8,0,0.0993173321088155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,8,128,1,float16,float16,0,0.12059199810028076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,1,128,1,float16,float16,0,0.06329066554705302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,1,128,1,float16,fp8,0,0.05183466772238413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,32,128,1,float16,fp8,0,0.0637066662311554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,2,128,1,float16,float16,0,0.06474133332570393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,32,128,1,float16,float16,0,0.07358400026957194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,4,128,1,float16,fp8,0,0.05585066477457682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,4,128,1,float16,float16,0,0.06442666550477345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,8,128,1,float16,fp8,0,0.05565866827964783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,8,128,1,float16,float16,0,0.06569066643714905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,32,128,1,float16,fp8,0,0.037445334096749626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,1,128,1,float16,fp8,0,0.03126399964094162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,4,128,1,float16,float16,0,0.03737599899371465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,2,128,1,float16,fp8,0,0.031231999397277832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,4,128,1,float16,fp8,0,0.0332640012105306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,32,128,1,float16,float16,0,0.13939199844996134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,8,128,1,float16,float16,0,0.03757333258787791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,8,128,1,float16,fp8,0,0.03356799980004629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,32,4,128,1,float16,float16,0,0.12016533811887105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,1,128,1,float16,float16,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,32,128,1,float16,fp8,0,0.023397333920001984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,1,128,1,float16,fp8,0,0.020997333029905956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,32,128,1,float16,float16,0,0.025008000433444977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,2,128,1,float16,float16,0,0.023237332701683044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,4,128,1,float16,float16,0,0.02311466634273529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,4,128,1,float16,fp8,0,0.021168000996112823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,32,2,128,1,float16,fp8,0,0.05411200225353241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,8,128,1,float16,float16,0,0.023269332945346832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,8,128,1,float16,fp8,0,0.021061333517233532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,1,128,1,float16,float16,0,0.01613866661985715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,1,128,1,float16,fp8,0,0.015061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,1,128,1,float16,float16,0,0.036943999429543815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,32,128,1,float16,float16,0,0.017231999586025875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,2,128,1,float16,fp8,0,0.014912000546852747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,4,128,1,float16,fp8,0,0.015072000523408255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,4,128,1,float16,float16,0,0.016042667130629223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,2,128,1,float16,float16,0,0.03626133253177007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,8,128,1,float16,float16,0,0.016506666938463848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,8,128,1,float16,fp8,0,0.014959999670584997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,1,128,1,float16,float16,0,0.01210133358836174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,32,128,1,float16,fp8,0,0.013525333255529404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,32,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,2,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,4,128,1,float16,float16,0,0.012768000364303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,2,128,1,float16,fp8,0,0.012698666503032049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,4,128,1,float16,fp8,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,8,128,1,float16,float16,0,0.01292266696691513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,8,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,32,2,128,1,float16,fp8,0,0.021338666478792827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,32,128,1,float16,fp8,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,32,128,1,float16,float16,0,0.01293333371480306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,1,128,1,float16,float16,0,0.010746666540702185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,1,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,2,128,1,float16,float16,0,0.011226666470368704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,4,128,1,float16,float16,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,2,128,1,float16,float16,0,0.01543466622630755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,4,128,1,float16,fp8,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,8,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,8,128,1,float16,fp8,0,0.011994666109482447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,32,32,128,1,float16,float16,0,0.03968533376852671
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,1,128,1,float16,float16,0,0.010853332777818045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,1,128,1,float16,fp8,0,0.011567999919255575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,2,128,1,float16,fp8,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,32,128,1,float16,float16,0,0.012719999998807907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,32,128,1,float16,fp8,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,4,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,4,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,32,1,128,1,float16,fp8,0,0.012058666596810022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,8,128,1,float16,fp8,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,2,128,1,float16,float16,0,0.014959999670584997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,2,128,1,float16,fp8,0,0.014848000059525171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,4,128,1,float16,float16,0,0.018906666586796444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,4,128,1,float16,fp8,0,0.01874133323629697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,32,2,128,1,float16,fp8,0,0.0107893335322539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,8,128,1,float16,float16,0,0.028751999139785767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,32,8,128,1,float16,fp8,0,0.02533866713444392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,2,128,1,float16,float16,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,32,128,1,float16,fp8,0,0.04334400097529093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,32,128,1,float16,float16,0,0.057775999108950295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,2,128,1,float16,float16,0,0.0107893335322539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,2,128,1,float16,fp8,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,32,8,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,4,128,1,float16,float16,0,0.012975999464591345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,4,128,1,float16,fp8,0,0.012960000584522883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,8,128,1,float16,float16,0,0.0189280000825723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,32,8,128,1,float16,fp8,0,0.016890666137139004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,2,128,1,float16,float16,0,0.010672000547250112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,32,128,1,float16,float16,0,0.02951466788848241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,32,128,1,float16,fp8,0,0.027109332382678986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,4,128,1,float16,float16,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,8,128,1,float16,float16,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,4,128,1,float16,fp8,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,2,128,1,float16,float16,0,0.008879999940594038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,2,128,1,float16,fp8,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,4,128,1,float16,float16,0,0.00903466654320558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,32,128,1,float16,float16,0,0.018805333723624546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,8,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,8,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,32,128,1,float16,float16,0,0.01451733335852623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,32,128,1,float16,fp8,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,2,128,1,float16,fp8,0,0.00879466657837232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,4,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,4,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,8,128,1,float16,float16,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,8,128,1,float16,fp8,0,0.009088000282645226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,2,128,1,float16,fp8,0,0.010725333044926325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,32,128,1,float16,fp8,0,0.012639999389648438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,32,8,128,1,float16,fp8,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,2,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,2,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,4,128,1,float16,float16,0,0.00892800030608972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,4,128,1,float16,fp8,0,0.008752000207702318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,8,128,1,float16,float16,0,0.009050666665037474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,8,128,1,float16,fp8,0,0.008869333192706108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,4,128,1,float16,fp8,0,0.008879999940594038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,32,32,128,1,float16,fp8,0,0.01693333312869072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,2,128,1,float16,float16,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,32,128,1,float16,fp8,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,4,128,1,float16,float16,0,0.008613333106040955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,4,128,1,float16,fp8,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,8,128,1,float16,float16,0,0.009642666826645533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,8,128,1,float16,fp8,0,0.008826666822036108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,32,2,128,1,float16,float16,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,32,128,1,float16,float16,0,0.00901333304742972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,32,128,1,float16,fp8,0,0.010677333921194077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,2,128,1,float16,float16,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,2,128,1,float16,fp8,0,0.008879999940594038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,4,128,1,float16,float16,0,0.00874133345981439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,4,128,1,float16,fp8,0,0.008832000195980072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,8,128,1,float16,float16,0,0.008938666433095932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,32,8,128,1,float16,fp8,0,0.009072000160813332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,32,32,128,1,float16,float16,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,32,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,32,128,1,float16,fp8,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,2,128,1,float16,fp8,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,4,128,1,float16,float16,0,0.008858666444818178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,4,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,8,128,1,float16,float16,0,0.008714666590094566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,8,128,1,float16,fp8,0,0.008799999952316284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,32,32,128,1,float16,fp8,0,0.018842666099468868
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,2,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,32,2,128,1,float16,float16,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,1,128,1,float16,fp8,0,8.53213882446289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,32,32,128,1,float16,float16,0,0.010842667271693548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,2,128,1,float16,fp8,0,8.696240107218424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,1,128,1,float16,float16,0,11.539119720458984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,4,128,1,float16,fp8,0,8.964031855265299
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,2,128,1,float16,float16,0,11.75051244099935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,4,128,1,float16,float16,0,11.839317321777344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,24,128,1,float16,fp8,0,4.998666763305664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,1,128,1,float16,fp8,0,4.2669172286987305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,24,128,1,float16,float16,0,5.805039723714192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,2,128,1,float16,float16,0,5.511610666910808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,2,128,1,float16,fp8,0,4.615376154581706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,8,128,1,float16,float16,0,11.98416519165039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,24,8,128,1,float16,fp8,0,8.948170979817709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,4,128,1,float16,fp8,0,4.5043894449869795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,1,128,1,float16,float16,0,2.6415680249532065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,4,128,1,float16,float16,0,5.625573476155599
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,24,128,1,float16,float16,0,2.955712000528971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,24,128,1,float16,fp8,0,2.9856694539388022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,8,128,1,float16,float16,0,5.505541483561198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,8,128,1,float16,fp8,0,5.017210642496745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,1,128,1,float16,fp8,0,2.199168046315511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,24,1,128,1,float16,float16,0,5.311786651611328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,2,128,1,float16,float16,0,2.7159093221028647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,2,128,1,float16,fp8,0,2.4214773178100586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,4,128,1,float16,fp8,0,2.3713547388712564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,4,128,1,float16,float16,0,2.7855199178059897
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,24,128,1,float16,fp8,0,1.3431413968404133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,8,128,1,float16,float16,0,2.8960746129353843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,24,128,1,float16,float16,0,1.5123252868652344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,1,128,1,float16,fp8,0,1.2100053628285725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,2,128,1,float16,float16,0,1.3970774014790852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,2,128,1,float16,fp8,0,1.4490240414937336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,4,128,1,float16,fp8,0,1.3957759539286296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,8,128,1,float16,fp8,0,1.2489866415659587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,4,128,1,float16,float16,0,1.4465333620707195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,8,128,1,float16,float16,0,1.4364746411641438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,1,128,1,float16,fp8,0,4.963055928548177
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,1,128,1,float16,float16,0,6.449663798014323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,24,8,128,1,float16,fp8,0,2.358448028564453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,2,128,1,float16,fp8,0,5.0501813888549805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,24,1,128,1,float16,float16,0,1.556597391764323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,4,128,1,float16,fp8,0,5.305983861287435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,4,128,1,float16,float16,0,6.449424107869466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,1,128,1,float16,fp8,0,2.489807923634847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,1,128,1,float16,float16,0,3.0373013814290366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,24,128,1,float16,float16,0,3.4715840021769204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,24,128,1,float16,fp8,0,3.049440066019694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,2,128,1,float16,float16,0,3.512997309366862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,2,128,1,float16,fp8,0,2.5452426274617515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,8,128,1,float16,float16,0,6.388901392618815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,8,128,1,float16,fp8,0,5.557263692220052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,4,128,1,float16,float16,0,3.2611573537190757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,4,128,1,float16,fp8,0,2.7978080113728843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,24,2,128,1,float16,float16,0,6.018063863118489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,8,128,1,float16,float16,0,3.206480026245117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,1,128,1,float16,float16,0,1.5499413808186848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,24,8,128,1,float16,fp8,0,2.9918346405029297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,1,128,1,float16,fp8,0,1.5307200749715169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,24,128,1,float16,float16,0,1.7470614115397136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,24,128,1,float16,fp8,0,1.6798346837361653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,2,128,1,float16,fp8,0,1.3267359733581543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,2,128,1,float16,float16,0,1.5829440752665203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,4,128,1,float16,fp8,0,1.4144959449768066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,4,128,1,float16,float16,0,1.7861706415812175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,1,128,1,float16,float16,0,0.8076426982879639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,8,128,1,float16,float16,0,1.6505813598632812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,24,8,128,1,float16,fp8,0,1.480570634206136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,2,128,1,float16,float16,0,0.8318186601003011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,24,128,1,float16,float16,0,0.9209706783294678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,2,128,1,float16,fp8,0,0.8101226488749186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,24,128,1,float16,fp8,0,0.88045867284139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,4,128,1,float16,float16,0,0.8512799739837646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,4,128,1,float16,fp8,0,0.7552266915639242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,8,128,1,float16,fp8,0,0.7898453076680502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,8,128,1,float16,float16,0,0.8616267045338949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,1,128,1,float16,fp8,0,3.5914827982584634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,1,128,1,float16,float16,0,4.372202555338542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,2,128,1,float16,fp8,0,3.5590187708536782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,2,128,1,float16,float16,0,4.304986635843913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,4,128,1,float16,fp8,0,3.803605397542318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,4,128,1,float16,float16,0,4.635552088419597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,8,128,1,float16,float16,0,4.506496111551921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,24,1,128,1,float16,fp8,0,0.7038026650746664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,1,128,1,float16,float16,0,2.1467307408650718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,24,128,1,float16,float16,0,2.461066722869873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,1,128,1,float16,fp8,0,1.9760106404622395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,24,128,1,float16,fp8,0,2.492138703664144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,2,128,1,float16,float16,0,2.282719930013021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,2,128,1,float16,fp8,0,1.8314666748046875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,24,8,128,1,float16,fp8,0,4.3393707275390625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,4,128,1,float16,float16,0,2.3027040163675943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,4,128,1,float16,fp8,0,1.9659412701924641
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,1,128,1,float16,float16,0,1.1044267018636067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,8,128,1,float16,float16,0,2.333445390065511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,24,128,1,float16,fp8,0,1.1648320357004802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,24,128,1,float16,float16,0,1.3175360361735027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,1,128,1,float16,fp8,0,1.1158400376637776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,2,128,1,float16,fp8,0,0.9858400026957194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,2,128,1,float16,float16,0,1.1163787047068279
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,4,128,1,float16,fp8,0,1.017850637435913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,4,128,1,float16,float16,0,1.254863977432251
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,8,128,1,float16,float16,0,1.1826186974843342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,24,8,128,1,float16,fp8,0,1.1068053245544434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,24,128,1,float16,float16,0,0.6699253718058268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,1,128,1,float16,fp8,0,0.5342666705449423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,2,128,1,float16,fp8,0,0.5358719825744629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,4,128,1,float16,float16,0,0.6193546851476034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,4,128,1,float16,fp8,0,0.5760800043741862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,8,128,1,float16,fp8,0,0.5609226624170939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,8,128,1,float16,float16,0,0.6292853355407715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,24,8,128,1,float16,fp8,0,1.9978186289469402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,1,128,1,float16,fp8,0,4.567215919494629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,1,128,1,float16,float16,0,5.829717636108398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,24,128,1,float16,fp8,0,0.6347999970118204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,2,128,1,float16,fp8,0,4.694421450297038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,1,128,1,float16,float16,0,0.5890080134073893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,2,128,1,float16,float16,0,5.983194351196289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,24,2,128,1,float16,float16,0,0.5990720192591349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,4,128,1,float16,float16,0,6.3437652587890625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,1,128,1,float16,float16,0,2.8558667500813804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,24,128,1,float16,fp8,0,3.051605224609375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,24,128,1,float16,float16,0,3.3351147969563804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,8,128,1,float16,fp8,0,5.134229342142741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,8,128,1,float16,float16,0,6.379621505737305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,2,128,1,float16,float16,0,2.8610881169637046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,2,128,1,float16,fp8,0,2.9461228052775064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,4,128,1,float16,fp8,0,2.8133598963419595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,4,128,1,float16,float16,0,3.0184958775838218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,8,128,1,float16,float16,0,2.9951305389404297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,8,128,1,float16,fp8,0,2.768890698750814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,24,128,1,float16,float16,0,1.7164853413899739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,1,128,1,float16,float16,0,1.4196586608886719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,1,128,1,float16,fp8,0,1.1928053696950276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,24,128,1,float16,fp8,0,1.6044106483459473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,24,1,128,1,float16,fp8,0,2.3099679946899414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,2,128,1,float16,float16,0,1.4582452774047852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,2,128,1,float16,fp8,0,1.2600213686625164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,4,128,1,float16,fp8,0,1.3467733065287273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,24,4,128,1,float16,fp8,0,5.078250567118327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,4,128,1,float16,float16,0,1.536453406016032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,8,128,1,float16,fp8,0,1.3452000617980957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,24,8,128,1,float16,float16,0,1.5497120221455891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,1,128,1,float16,float16,0,0.7325546741485596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,1,128,1,float16,fp8,0,0.6404159863789877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,2,128,1,float16,fp8,0,0.68831467628479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,4,128,1,float16,float16,0,0.791274627049764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,4,128,1,float16,fp8,0,0.7404853502909342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,8,128,1,float16,fp8,0,0.712277332941691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,1,128,1,float16,float16,0,0.39818668365478516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,24,128,1,float16,float16,0,0.471013347307841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,24,128,1,float16,fp8,0,0.44811733563741046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,2,128,1,float16,float16,0,0.4070506493250529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,4,128,1,float16,float16,0,0.42316798369089764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,4,128,1,float16,fp8,0,0.39212266604105633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,8,128,1,float16,fp8,0,0.3962026834487915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,24,128,1,float16,float16,0,0.8661119937896729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,24,128,1,float16,fp8,0,0.8479893207550049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,1,128,1,float16,float16,0,3.244624137878418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,8,128,1,float16,float16,0,0.7952319780985514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,1,128,1,float16,fp8,0,0.3545653422673543
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,1,128,1,float16,fp8,0,2.6992905934651694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,2,128,1,float16,fp8,0,0.3659306764602661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,2,128,1,float16,float16,0,3.282869338989258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,24,8,128,1,float16,float16,0,0.428928017616272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,2,128,1,float16,fp8,0,3.308490753173828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,4,128,1,float16,float16,0,3.55242125193278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,4,128,1,float16,fp8,0,3.0973227818806968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,1,128,1,float16,float16,0,1.6453973452250164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,8,128,1,float16,fp8,0,3.1648267110188804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,24,8,128,1,float16,float16,0,3.579594612121582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,24,2,128,1,float16,float16,0,0.7551999886830648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,24,128,1,float16,float16,0,2.1476906140645347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,24,128,1,float16,fp8,0,1.933616002400716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,2,128,1,float16,float16,0,1.6780799229939778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,2,128,1,float16,fp8,0,1.6490772565205891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,4,128,1,float16,float16,0,1.8045760790507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,4,128,1,float16,fp8,0,1.846944014231364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,1,128,1,float16,float16,0,0.8478720188140869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,8,128,1,float16,fp8,0,1.6482240358988445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,24,128,1,float16,float16,0,1.0384960174560547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,1,128,1,float16,fp8,0,0.7483786741892496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,2,128,1,float16,float16,0,0.8659413655598959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,2,128,1,float16,fp8,0,0.7639573415120443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,4,128,1,float16,float16,0,0.9189706643422445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,4,128,1,float16,fp8,0,0.8523360093434652
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,8,128,1,float16,fp8,0,0.8479839960734049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,8,128,1,float16,float16,0,0.9372373421986898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,24,128,1,float16,float16,0,0.5415519873301188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,1,128,1,float16,float16,0,0.43851733207702637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,1,128,1,float16,fp8,0,0.3957120180130005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,24,128,1,float16,fp8,0,0.5340906778971354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,1,128,1,float16,fp8,0,1.6046667098999023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,2,128,1,float16,float16,0,0.4532906611760457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,2,128,1,float16,fp8,0,0.4106186628341675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,4,128,1,float16,float16,0,0.4838026762008667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,4,128,1,float16,fp8,0,0.4475146532058716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,8,128,1,float16,float16,0,0.4829599857330322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,24,8,128,1,float16,fp8,0,0.4498879909515381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,1,128,1,float16,float16,0,0.24549333254496256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,24,128,1,float16,float16,0,0.2983306646347046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,2,128,1,float16,float16,0,0.24925865729649863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,24,128,1,float16,fp8,0,0.2970133423805237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,2,128,1,float16,fp8,0,0.23224000136057535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,24,8,128,1,float16,float16,0,1.828181266784668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,4,128,1,float16,fp8,0,0.24355200926462808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,4,128,1,float16,float16,0,0.26657599210739136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,8,128,1,float16,fp8,0,0.25194666783014935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,8,128,1,float16,float16,0,0.2720479965209961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,1,128,1,float16,fp8,0,2.6140267054239907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,2,128,1,float16,float16,0,3.147637367248535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,2,128,1,float16,fp8,0,2.7528533935546875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,4,128,1,float16,float16,0,3.5369065602620444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,24,1,128,1,float16,fp8,0,0.22662399212519327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,4,128,1,float16,fp8,0,3.1476640701293945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,24,24,128,1,float16,fp8,0,0.9949920177459717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,24,128,1,float16,float16,0,2.084885279337565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,8,128,1,float16,float16,0,3.565295855204264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,8,128,1,float16,fp8,0,3.3692798614501953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,24,128,1,float16,fp8,0,2.1736159324645996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,1,128,1,float16,fp8,0,1.3322079976399739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,2,128,1,float16,float16,0,1.6070292790730794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,2,128,1,float16,fp8,0,1.4054452578226726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,4,128,1,float16,float16,0,1.7940053939819336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,4,128,1,float16,fp8,0,1.657898743947347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,24,1,128,1,float16,float16,0,3.091754595438639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,8,128,1,float16,float16,0,1.8058239618937175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,8,128,1,float16,fp8,0,1.6377600034077961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,24,128,1,float16,float16,0,1.0571466286977131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,24,128,1,float16,fp8,0,1.0593492984771729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,1,128,1,float16,fp8,0,0.6915040016174316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,2,128,1,float16,fp8,0,0.7286559740702311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,4,128,1,float16,float16,0,0.9022453625996908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,4,128,1,float16,fp8,0,0.8196266492207845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,8,128,1,float16,fp8,0,0.8408426443735758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,24,128,1,float16,float16,0,0.5459680159886678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,1,128,1,float16,float16,0,0.41088000933329266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,24,128,1,float16,fp8,0,0.5509066581726074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,24,1,128,1,float16,float16,0,1.5380479494730632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,2,128,1,float16,float16,0,0.4240266482035319
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,4,128,1,float16,float16,0,0.4678933223088582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,4,128,1,float16,fp8,0,0.43726933002471924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,8,128,1,float16,float16,0,0.47280001640319824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,8,128,1,float16,fp8,0,0.44527467091878253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,1,128,1,float16,float16,0,0.22587732474009195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,24,128,1,float16,fp8,0,0.29818665981292725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,24,128,1,float16,float16,0,0.29290666182835895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,1,128,1,float16,fp8,0,0.20632533232371011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,2,128,1,float16,float16,0,0.23587733507156372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,2,128,1,float16,fp8,0,0.2144213318824768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,4,128,1,float16,float16,0,0.25301865736643475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,4,128,1,float16,fp8,0,0.23890666166941324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,2,128,1,float16,float16,0,0.8192053635915121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,8,128,1,float16,float16,0,0.2576853235562642
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,24,8,128,1,float16,fp8,0,0.24752533435821533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,1,128,1,float16,float16,0,0.13303466637929282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,1,128,1,float16,fp8,0,0.12503467003504434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,8,128,1,float16,float16,0,0.9053653081258138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,24,128,1,float16,float16,0,0.17082132895787558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,2,128,1,float16,float16,0,0.13495999574661255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,2,128,1,float16,fp8,0,0.12939199805259705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,4,128,1,float16,float16,0,0.14310399691263834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,4,128,1,float16,fp8,0,0.13800000150998434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,1,128,1,float16,fp8,0,0.3733973503112793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,8,128,1,float16,fp8,0,0.1386133333047231
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,8,128,1,float16,float16,0,0.14511467019716898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,24,2,128,1,float16,fp8,0,0.3877280155817668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,1,128,1,float16,fp8,0,1.6063626607259114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,1,128,1,float16,float16,0,1.8411626815795898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,2,128,1,float16,fp8,0,1.7184960047403972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,4,128,1,float16,float16,0,2.1758507092793784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,4,128,1,float16,fp8,0,1.9946880340576172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,24,1,128,1,float16,float16,0,0.789194663365682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,1,128,1,float16,float16,0,0.9323146343231201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,24,128,1,float16,float16,0,1.3454400698343914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,8,128,1,float16,float16,0,2.2266666094462075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,1,128,1,float16,fp8,0,0.8218186696370443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,8,128,1,float16,fp8,0,2.149946689605713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,24,128,1,float16,fp8,0,1.4125919342041016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,2,128,1,float16,float16,0,0.9750400384267172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,2,128,1,float16,fp8,0,0.8765973250071207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,4,128,1,float16,float16,0,1.1096959908803303
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,4,128,1,float16,fp8,0,1.0235626697540283
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,8,128,1,float16,float16,0,1.1135573387145996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,24,8,128,1,float16,fp8,0,1.0435199737548828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,1,128,1,float16,float16,0,0.4782079855600993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,24,128,1,float16,float16,0,0.6826773484547933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,24,128,1,float16,fp8,0,0.7105920314788818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,1,128,1,float16,fp8,0,0.43481600284576416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,2,128,1,float16,float16,0,0.5013333161671957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,24,2,128,1,float16,float16,0,1.9163146018981934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,2,128,1,float16,fp8,0,0.46217600504557294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,4,128,1,float16,fp8,0,0.5362133185068766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,8,128,1,float16,float16,0,0.5679200092951456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,8,128,1,float16,fp8,0,0.543999989827474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,1,128,1,float16,float16,0,0.2571573257446289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,24,128,1,float16,float16,0,0.3574826717376709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,24,128,1,float16,fp8,0,0.3718613386154175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,2,128,1,float16,float16,0,0.2662880023320516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,1,128,1,float16,fp8,0,0.23714667558670044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,2,128,1,float16,fp8,0,0.2520266572634379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,4,128,1,float16,float16,0,0.2997066577275594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,4,128,1,float16,fp8,0,0.2871359984079997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,8,128,1,float16,float16,0,0.30185600121816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,24,8,128,1,float16,fp8,0,0.29393066962560016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,1,128,1,float16,float16,0,0.14680533607800803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,1,128,1,float16,fp8,0,0.1353653371334076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,24,24,128,1,float16,fp8,0,0.1689280072848002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,24,128,1,float16,fp8,0,0.2055893341700236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,2,128,1,float16,fp8,0,0.14130666851997375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,24,128,1,float16,float16,0,0.19735999902089438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,4,128,1,float16,fp8,0,0.15450132886568704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,4,128,1,float16,float16,0,0.16406933466593424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,8,128,1,float16,float16,0,0.1710346738497416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,1,128,1,float16,float16,0,0.09108799695968628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,1,128,1,float16,fp8,0,0.08586133519808452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,2,128,1,float16,float16,0,0.09468799829483032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,2,128,1,float16,fp8,0,0.09033600489298503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,24,128,1,float16,float16,0,0.12001599868138631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,4,128,1,float16,float16,0,0.09847999612490337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,4,128,1,float16,fp8,0,0.09545066952705383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,8,128,1,float16,fp8,0,0.09757333000500996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,8,128,1,float16,float16,0,0.10013332962989807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,24,4,128,1,float16,float16,0,0.564079999923706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,1,128,1,float16,float16,0,1.822314739227295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,1,128,1,float16,fp8,0,1.6320160230000813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,2,128,1,float16,float16,0,1.9444479942321777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,2,128,1,float16,fp8,0,1.771615982055664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,2,128,1,float16,float16,0,0.14873600006103516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,4,128,1,float16,float16,0,2.296895980834961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,24,8,128,1,float16,fp8,0,0.159770667552948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,4,128,1,float16,fp8,0,2.1840747197469077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,24,24,128,1,float16,fp8,0,0.11784533659617107
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,1,128,1,float16,float16,0,0.9268853664398193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,1,128,1,float16,fp8,0,0.831279993057251
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,24,128,1,float16,float16,0,1.4683200518290203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,24,128,1,float16,fp8,0,1.5560213724772136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,8,128,1,float16,fp8,0,2.230079968770345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,2,128,1,float16,float16,0,0.9794293244679769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,2,128,1,float16,fp8,0,0.9017386436462402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,4,128,1,float16,float16,0,1.1572106679280598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,1,128,1,float16,float16,0,0.4715786774953206
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,1,128,1,float16,fp8,0,0.4310773213704427
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,8,128,1,float16,fp8,0,1.1329066753387451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,8,128,1,float16,float16,0,1.168287992477417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,24,128,1,float16,float16,0,0.7414080301920573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,24,128,1,float16,fp8,0,0.7920959790547689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,2,128,1,float16,float16,0,0.49885332584381104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,2,128,1,float16,fp8,0,0.46671998500823975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,4,128,1,float16,float16,0,0.5842666625976562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,4,128,1,float16,fp8,0,0.5607466697692871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,8,128,1,float16,float16,0,0.5934346516927084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,24,8,128,1,float16,fp8,0,0.5770719846089681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,1,128,1,float16,float16,0,0.2523733377456665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,1,128,1,float16,fp8,0,0.23220266898473105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,2,128,1,float16,float16,0,0.2625333269437154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,24,128,1,float16,fp8,0,0.41250133514404297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,24,128,1,float16,float16,0,0.3842879931131999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,4,128,1,float16,fp8,0,0.3019520044326782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,4,128,1,float16,float16,0,0.3104693293571472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,24,8,128,1,float16,float16,0,2.3481547037760415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,8,128,1,float16,float16,0,0.31251200040181476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,8,128,1,float16,fp8,0,0.3100213408470154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,24,128,1,float16,float16,0,0.2071946660677592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,1,128,1,float16,fp8,0,0.13052800297737122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,24,128,1,float16,fp8,0,0.22194133202234903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,2,128,1,float16,float16,0,0.14755733807881674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,2,128,1,float16,fp8,0,0.13563733299573263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,4,128,1,float16,float16,0,0.1706506609916687
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,8,128,1,float16,fp8,0,0.16931732495625815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,4,128,1,float16,fp8,0,0.1616266667842865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,8,128,1,float16,float16,0,0.1738026738166809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,1,128,1,float16,float16,0,0.08364799618721008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,1,128,1,float16,fp8,0,0.07956799864768982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,2,128,1,float16,float16,0,0.0867733359336853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,2,128,1,float16,fp8,0,0.08334933718045552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,24,4,128,1,float16,fp8,0,1.0935253302256267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,4,128,1,float16,float16,0,0.09299199779828389
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,8,128,1,float16,float16,0,0.0944160024325053
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,8,128,1,float16,fp8,0,0.09377599755922954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,24,128,1,float16,float16,0,0.06829866766929626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,1,128,1,float16,float16,0,0.05605333546797434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,24,128,1,float16,fp8,0,0.07223466535409291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,1,128,1,float16,fp8,0,0.054048001766204834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,2,128,1,float16,float16,0,0.057775999108950295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,24,2,128,1,float16,fp8,0,0.25195733706156415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,4,128,1,float16,float16,0,0.0609386662642161
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,4,128,1,float16,fp8,0,0.06030400097370148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,24,1,128,1,float16,float16,0,0.14269333084424338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,8,128,1,float16,float16,0,0.061679999033610024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,8,128,1,float16,fp8,0,0.061573331554730736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,1,128,1,float16,float16,0,1.152234633763631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,24,128,1,float16,float16,0,0.11991999546686809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,1,128,1,float16,fp8,0,1.0417760213216145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,24,128,1,float16,fp8,0,0.12497066458066304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,24,4,128,1,float16,fp8,0,0.09092799822489421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,2,128,1,float16,float16,0,1.2312533060709636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,2,128,1,float16,fp8,0,1.1532479921976726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,4,128,1,float16,fp8,0,1.4380480448404949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,4,128,1,float16,float16,0,1.4962612787882488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,8,128,1,float16,fp8,0,1.4905813535054524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,24,8,128,1,float16,float16,0,1.5330719947814941
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,1,128,1,float16,float16,0,0.5854133367538452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,24,128,1,float16,float16,0,0.9901386896769205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,24,2,128,1,float16,fp8,0,0.055829331278800964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,1,128,1,float16,fp8,0,0.538864016532898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,2,128,1,float16,float16,0,0.625162680943807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,24,128,1,float16,fp8,0,1.0759200255076091
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,4,128,1,float16,float16,0,0.759061336517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,4,128,1,float16,fp8,0,0.7296533584594727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,8,128,1,float16,float16,0,0.7641119956970215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,8,128,1,float16,fp8,0,0.7548267046610514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,1,128,1,float16,float16,0,0.30869332949320477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,2,128,1,float16,float16,0,0.3227840065956116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,24,128,1,float16,float16,0,0.5050666729609171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,24,128,1,float16,fp8,0,0.5512586832046509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,4,128,1,float16,float16,0,0.3885226647059123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,4,128,1,float16,fp8,0,0.3836053212483724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,8,128,1,float16,float16,0,0.3967306613922119
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,1,128,1,float16,float16,0,0.16781334082285562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,8,128,1,float16,fp8,0,0.3938026825586955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,24,128,1,float16,fp8,0,0.2897813320159912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,1,128,1,float16,fp8,0,0.15563199917475382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,24,128,1,float16,float16,0,0.26524800062179565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,2,128,1,float16,float16,0,0.18065067132314047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,2,128,1,float16,fp8,0,0.17139732837677002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,4,128,1,float16,fp8,0,0.20569066206614176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,4,128,1,float16,float16,0,0.20988800128300986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,8,128,1,float16,float16,0,0.2132479945818583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,1,128,1,float16,float16,0,0.09388800462086995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,24,8,128,1,float16,fp8,0,0.21290133396784464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,1,128,1,float16,fp8,0,0.08925867080688477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,2,128,1,float16,float16,0,0.09889066219329834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,24,128,1,float16,fp8,0,0.15850667158762613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,2,128,1,float16,fp8,0,0.09512533744176228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,4,128,1,float16,float16,0,0.11348266402880351
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,4,128,1,float16,fp8,0,0.10737066467603047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,8,128,1,float16,fp8,0,0.11370666821797688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,24,2,128,1,float16,fp8,0,0.5886720021565756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,1,128,1,float16,float16,0,0.05986666679382324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,24,128,1,float16,fp8,0,0.08614400029182434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,24,128,1,float16,float16,0,0.08841066559155782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,2,128,1,float16,float16,0,0.06238933404286703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,1,128,1,float16,fp8,0,0.05615466833114624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,2,128,1,float16,fp8,0,0.05872533222039541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,4,128,1,float16,float16,0,0.06591466565926869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,4,128,1,float16,fp8,0,0.06545599798361461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,8,128,1,float16,float16,0,0.06668800115585327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,24,8,128,1,float16,fp8,0,0.06657066444555919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,1,128,1,float16,float16,0,0.03584533433119456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,1,128,1,float16,fp8,0,0.033413333197434746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,2,128,1,float16,float16,0,0.037530665596326195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,24,128,1,float16,float16,0,0.04320533573627472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,24,128,1,float16,fp8,0,0.04844266672929128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,2,128,1,float16,fp8,0,0.035317334036032356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,4,128,1,float16,float16,0,0.03951466580231985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,4,128,1,float16,fp8,0,0.03930133332808813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,8,128,1,float16,fp8,0,0.03963200002908707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,24,8,128,1,float16,float16,0,0.03961600114901861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,24,128,1,float16,float16,0,0.14613866806030273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,1,128,1,float16,float16,0,1.2224480311075847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,1,128,1,float16,fp8,0,1.1321067015329997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,24,8,128,1,float16,float16,0,0.12070932984352112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,2,128,1,float16,float16,0,1.3273759682973225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,2,128,1,float16,fp8,0,1.2775039672851562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,1,128,1,float16,fp8,0,0.282970666885376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,24,2,128,1,float16,fp8,0,0.3127359946568807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,1,128,1,float16,float16,0,0.6213013331095377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,8,128,1,float16,float16,0,1.7238027254740398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,4,128,1,float16,float16,0,1.6903252601623535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,8,128,1,float16,fp8,0,1.7252000172932942
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,24,128,1,float16,float16,0,1.158138672510783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,1,128,1,float16,fp8,0,0.5774986743927002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,24,128,1,float16,fp8,0,1.2966667016347249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,2,128,1,float16,float16,0,0.6782506306966146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,2,128,1,float16,fp8,0,0.6487413247426351
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,4,128,1,float16,float16,0,0.854032039642334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,8,128,1,float16,float16,0,0.8623039722442627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,4,128,1,float16,fp8,0,0.8357760111490885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,1,128,1,float16,float16,0,0.3233333428700765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,24,8,128,1,float16,fp8,0,0.869370698928833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,24,128,1,float16,float16,0,0.5876160065333048
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,1,128,1,float16,fp8,0,0.30235199133555096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,2,128,1,float16,fp8,0,0.34332799911499023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,24,128,1,float16,fp8,0,0.6591840187708536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,2,128,1,float16,float16,0,0.35633599758148193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,4,128,1,float16,float16,0,0.4336640040079753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,4,128,1,float16,fp8,0,0.43297600746154785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,8,128,1,float16,fp8,0,0.4481866757074992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,24,8,128,1,float16,float16,0,0.44385067621866864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,1,128,1,float16,float16,0,0.1751413345336914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,2,128,1,float16,float16,0,0.18938666582107544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,24,128,1,float16,float16,0,0.3055573304494222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,1,128,1,float16,fp8,0,0.1660319964090983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,24,128,1,float16,fp8,0,0.34151466687520343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,2,128,1,float16,fp8,0,0.18320000171661377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,4,128,1,float16,float16,0,0.2305013338724772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,24,4,128,1,float16,fp8,0,1.6522773106892903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,4,128,1,float16,fp8,0,0.23109332720438638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,8,128,1,float16,float16,0,0.2355413238207499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,1,128,1,float16,float16,0,0.09846400221188863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,2,128,1,float16,float16,0,0.1072266697883606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,24,128,1,float16,fp8,0,0.18297600746154785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,4,128,1,float16,float16,0,0.12623467048009238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,1,128,1,float16,fp8,0,0.09161067008972168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,8,128,1,float16,fp8,0,0.13125333189964294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,8,128,1,float16,float16,0,0.1316373348236084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,1,128,1,float16,float16,0,0.05826666454474131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,1,128,1,float16,fp8,0,0.05499733487764994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,24,128,1,float16,float16,0,0.09404800335566203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,2,128,1,float16,fp8,0,0.05916266640027364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,24,128,1,float16,fp8,0,0.10562666257222493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,4,128,1,float16,float16,0,0.06734933455785115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,8,128,1,float16,float16,0,0.07062399884064992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,8,128,1,float16,fp8,0,0.06969599922498067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,24,128,1,float16,fp8,0,0.05656533439954122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,24,128,1,float16,float16,0,0.05156800150871277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,1,128,1,float16,float16,0,0.040720000863075256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,2,128,1,float16,float16,0,0.04197866717974345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,1,128,1,float16,fp8,0,0.038719999293486275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,24,8,128,1,float16,fp8,0,0.24014933904012045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,2,128,1,float16,fp8,0,0.04144533226887385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,4,128,1,float16,float16,0,0.044821331898371376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,4,128,1,float16,fp8,0,0.04534933467706045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,8,128,1,float16,float16,0,0.04571733375390371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,2,128,1,float16,fp8,0,0.09918399651845296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,24,8,128,1,float16,fp8,0,0.04584000011285146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,1,128,1,float16,float16,0,0.027136000494162243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,1,128,1,float16,fp8,0,0.02645866572856903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,4,128,1,float16,fp8,0,0.11818666259447734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,24,128,1,float16,fp8,0,0.03559466699759165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,2,128,1,float16,float16,0,0.02809600035349528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,2,128,1,float16,fp8,0,0.028197333216667175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,4,128,1,float16,float16,0,0.030234667162100475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,4,128,1,float16,fp8,0,0.02932800104220708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,8,128,1,float16,float16,0,0.02992533395687739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,2,128,1,float16,float16,0,0.06229333579540253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,24,4,128,1,float16,fp8,0,0.0682239979505539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,1,128,1,float16,float16,0,0.9316319624582926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,1,128,1,float16,fp8,0,0.8830613295237223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,2,128,1,float16,float16,0,1.0570240020751953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,2,128,1,float16,fp8,0,1.0231040318806965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,24,24,128,1,float16,float16,0,0.16504533092180887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,4,128,1,float16,float16,0,1.4107573827107747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,24,128,1,float16,float16,0,0.032111999889214836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,24,8,128,1,float16,fp8,0,0.031146667897701263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,4,128,1,float16,fp8,0,1.3910719553629558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,24,128,1,float16,fp8,0,1.1645066738128662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,1,128,1,float16,fp8,0,0.4532959858576457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,24,128,1,float16,float16,0,1.0075466632843018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,1,128,1,float16,float16,0,0.4803359905878703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,8,128,1,float16,float16,0,1.4397279421488445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,24,8,128,1,float16,fp8,0,1.4694719314575195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,2,128,1,float16,float16,0,0.5371093352635702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,1,128,1,float16,float16,0,0.25198932488759357
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,4,128,1,float16,fp8,0,0.7057279745737711
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,2,128,1,float16,fp8,0,0.5211893320083618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,4,128,1,float16,float16,0,0.7131840387980143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,1,128,1,float16,fp8,0,0.23732799291610718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,2,128,1,float16,fp8,0,0.2775733272234599
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,8,128,1,float16,float16,0,0.7263680299123129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,24,8,128,1,float16,fp8,0,0.739029328028361
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,4,128,1,float16,float16,0,0.36591466267903644
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,2,128,1,float16,float16,0,0.2898079951604207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,4,128,1,float16,fp8,0,0.36562132835388184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,1,128,1,float16,float16,0,0.13868799805641174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,8,128,1,float16,float16,0,0.37336532274882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,1,128,1,float16,fp8,0,0.12922666470209757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,2,128,1,float16,float16,0,0.15358933806419373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,24,128,1,float16,fp8,0,0.3065653244654338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,24,128,1,float16,float16,0,0.5128746827443441
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,4,128,1,float16,float16,0,0.19323732455571493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,24,128,1,float16,fp8,0,0.5909973382949829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,24,8,128,1,float16,fp8,0,0.3818826675415039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,4,128,1,float16,fp8,0,0.19464532534281412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,2,128,1,float16,fp8,0,0.14723199605941772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,1,128,1,float16,fp8,0,0.07218666871388753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,8,128,1,float16,float16,0,0.1974453330039978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,8,128,1,float16,fp8,0,0.20433066288630167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,24,128,1,float16,fp8,0,0.16293866435686746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,24,128,1,float16,float16,0,0.14478400349617004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,4,128,1,float16,fp8,0,0.10105599959691365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,2,128,1,float16,fp8,0,0.07803733150164287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,1,128,1,float16,float16,0,0.07863999903202057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,24,24,128,1,float16,float16,0,0.26866666475931805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,8,128,1,float16,float16,0,0.1109386682510376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,8,128,1,float16,fp8,0,0.11019200086593628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,1,128,1,float16,float16,0,0.04619733492533366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,1,128,1,float16,fp8,0,0.04376000165939331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,24,128,1,float16,fp8,0,0.09015466769536336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,24,128,1,float16,float16,0,0.08100266754627228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,4,128,1,float16,float16,0,0.05467733244101206
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,4,128,1,float16,fp8,0,0.057162667314211525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,8,128,1,float16,float16,0,0.05743999779224396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,8,128,1,float16,fp8,0,0.05816000203291575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,1,128,1,float16,float16,0,0.03149333347876867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,1,128,1,float16,fp8,0,0.03129599988460541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,24,128,1,float16,float16,0,0.04136000076929728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,2,128,1,float16,float16,0,0.033088001112143196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,2,128,1,float16,fp8,0,0.03253333270549774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,4,128,1,float16,float16,0,0.0356480007370313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,4,128,1,float16,fp8,0,0.03605866680542628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,2,128,1,float16,float16,0,0.08562133709589641
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,8,128,1,float16,float16,0,0.0360000009338061
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,8,128,1,float16,fp8,0,0.03735466549793879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,1,128,1,float16,float16,0,0.02128000060717265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,1,128,1,float16,fp8,0,0.020992000897725422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,24,4,128,1,float16,float16,0,0.10699733098347981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,24,128,1,float16,float16,0,0.027109332382678986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,2,128,1,float16,float16,0,0.023130667706330616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,2,128,1,float16,fp8,0,0.022991999983787537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,4,128,1,float16,float16,0,0.02333866556485494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,4,128,1,float16,fp8,0,0.02437866727511088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,8,128,1,float16,float16,0,0.0230880007147789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,8,128,1,float16,fp8,0,0.025301332275072735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,2,128,1,float16,fp8,0,0.04722133278846741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,24,2,128,1,float16,float16,0,0.04920533299446106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,1,128,1,float16,fp8,0,0.019786667078733444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,24,128,1,float16,fp8,0,0.0252960001428922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,2,128,1,float16,float16,0,0.021040000021457672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,24,128,1,float16,float16,0,0.02313599983851115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,4,128,1,float16,float16,0,0.021301334102948506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,4,128,1,float16,fp8,0,0.021104000508785248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,8,128,1,float16,float16,0,0.02145066608985265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,24,24,128,1,float16,fp8,0,0.04824000100294749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,1,128,1,float16,float16,0,0.3962666591008504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,1,128,1,float16,fp8,0,0.3830133279164632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,2,128,1,float16,float16,0,0.4524639844894409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,2,128,1,float16,fp8,0,0.4495413303375244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,24,24,128,1,float16,fp8,0,0.030847998956839245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,4,128,1,float16,float16,0,0.6217919985453287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,4,128,1,float16,fp8,0,0.6381333271662394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,1,128,1,float16,float16,0,0.01988799994190534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,2,128,1,float16,fp8,0,0.019274666905403137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,24,128,1,float16,float16,0,0.4678453207015991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,8,128,1,float16,float16,0,0.6372213363647461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,24,8,128,1,float16,fp8,0,0.021397332350413006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,1,128,1,float16,fp8,0,0.19831999142964682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,24,8,128,1,float16,fp8,0,0.667306661605835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,1,128,1,float16,float16,0,0.20765332380930582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,2,128,1,float16,float16,0,0.23996800184249878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,2,128,1,float16,fp8,0,0.2386773427327474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,4,128,1,float16,fp8,0,0.3307253321011861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,24,128,1,float16,fp8,0,0.552565336227417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,24,128,1,float16,float16,0,0.24474666515986124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,1,128,1,float16,float16,0,0.1151146690050761
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,24,128,1,float16,fp8,0,0.28570665915807086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,8,128,1,float16,float16,0,0.32755200068155926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,1,128,1,float16,fp8,0,0.11107200384140015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,4,128,1,float16,float16,0,0.3206560015678406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,24,8,128,1,float16,fp8,0,0.3428106705347697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,2,128,1,float16,float16,0,0.12924266854921976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,2,128,1,float16,fp8,0,0.1285653313000997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,4,128,1,float16,float16,0,0.16937599579493204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,4,128,1,float16,fp8,0,0.17546133200327554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,1,128,1,float16,float16,0,0.06633600095907848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,2,128,1,float16,float16,0,0.07442133128643036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,2,128,1,float16,fp8,0,0.06829333305358887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,4,128,1,float16,fp8,0,0.08822932839393616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,4,128,1,float16,float16,0,0.09397332866986592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,24,128,1,float16,float16,0,0.13321600357691446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,24,128,1,float16,fp8,0,0.15306133031845093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,8,128,1,float16,float16,0,0.09718400239944458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,8,128,1,float16,fp8,0,0.10061867038408916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,24,128,1,float16,fp8,0,0.08095466593901317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,24,128,1,float16,float16,0,0.07306666672229767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,1,128,1,float16,fp8,0,0.0373333344856898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,2,128,1,float16,fp8,0,0.039674667020638786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,4,128,1,float16,float16,0,0.04649066428343455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,8,128,1,float16,float16,0,0.04967466493447622
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,8,128,1,float16,fp8,0,0.04972266654173533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,24,128,1,float16,fp8,0,0.04548266530036926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,24,128,1,float16,float16,0,0.037018666664759316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,1,128,1,float16,float16,0,0.02731200059254964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,1,128,1,float16,fp8,0,0.027024000883102417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,2,128,1,float16,float16,0,0.02918400118748347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,24,1,128,1,float16,fp8,0,0.0613919993241628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,8,128,1,float16,float16,0,0.1741866668065389
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,4,128,1,float16,float16,0,0.03151999910672506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,4,128,1,float16,fp8,0,0.03336533407370249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,8,128,1,float16,float16,0,0.03126399964094162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,8,128,1,float16,fp8,0,0.03347733368476232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,1,128,1,float16,float16,0,0.03736000011364619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,24,128,1,float16,float16,0,0.0233599990606308
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,1,128,1,float16,float16,0,0.017301333447297413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,1,128,1,float16,fp8,0,0.016986666868130367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,2,128,1,float16,float16,0,0.04050666590531667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,2,128,1,float16,float16,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,2,128,1,float16,fp8,0,0.019274666905403137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,4,128,1,float16,float16,0,0.020848001043001812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,24,4,128,1,float16,fp8,0,0.049695998430252075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,4,128,1,float16,fp8,0,0.021002667645613354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,8,128,1,float16,float16,0,0.021141332884629566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,8,128,1,float16,fp8,0,0.021701333423455555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,1,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,1,128,1,float16,fp8,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,2,128,1,float16,float16,0,0.016976000120242436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,2,128,1,float16,fp8,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,4,128,1,float16,float16,0,0.017498667041460674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,4,128,1,float16,fp8,0,0.01883200059334437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,8,128,1,float16,float16,0,0.01894933357834816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,8,128,1,float16,fp8,0,0.019248000035683315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,24,2,128,1,float16,fp8,0,0.029103999336560566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,24,8,128,1,float16,fp8,0,0.18217599391937256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,1,128,1,float16,float16,0,0.016965333372354507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,24,128,1,float16,float16,0,0.018986667195955913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,24,128,1,float16,fp8,0,0.018826667219400406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,2,128,1,float16,float16,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,2,128,1,float16,fp8,0,0.01729600007335345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,4,128,1,float16,float16,0,0.016789333273967106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,4,128,1,float16,fp8,0,0.016949333250522614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,8,128,1,float16,float16,0,0.017077332983414333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,8,128,1,float16,fp8,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,24,24,128,1,float16,fp8,0,0.027109332382678986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,24,128,1,float16,fp8,0,0.02293866624434789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,1,128,1,float16,float16,0,0.24290666977564493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,1,128,1,float16,fp8,0,0.2290239930152893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,2,128,1,float16,float16,0,0.27160000801086426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,2,128,1,float16,fp8,0,0.2680213252703349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,4,128,1,float16,float16,0,0.35738666852315265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,24,1,128,1,float16,fp8,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,8,128,1,float16,float16,0,0.3608693281809489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,4,128,1,float16,fp8,0,0.35636266072591144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,24,8,128,1,float16,fp8,0,0.37522133191426593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,24,128,1,float16,float16,0,0.25878934065500897
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,1,128,1,float16,float16,0,0.1306933363278707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,24,128,1,float16,fp8,0,0.2988160053888957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,1,128,1,float16,fp8,0,0.12365333239237468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,2,128,1,float16,float16,0,0.1441920002301534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,8,128,1,float16,float16,0,0.19216533501942953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,24,24,128,1,float16,float16,0,0.01956266661485036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,24,128,1,float16,float16,0,0.1365173359711965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,2,128,1,float16,fp8,0,0.1421226660410563
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,24,128,1,float16,fp8,0,0.15797332922617593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,1,128,1,float16,fp8,0,0.06676800052324931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,2,128,1,float16,fp8,0,0.07635200023651123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,1,128,1,float16,float16,0,0.07402133444945018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,4,128,1,float16,fp8,0,0.0990826686223348
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,4,128,1,float16,float16,0,0.18544000387191772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,2,128,1,float16,float16,0,0.0820906658967336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,8,128,1,float16,fp8,0,0.19597333669662476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,24,4,128,1,float16,fp8,0,0.1892533302307129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,8,128,1,float16,float16,0,0.10482666889826457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,4,128,1,float16,float16,0,0.10078932841618855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,24,8,128,1,float16,fp8,0,0.10524266958236694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,1,128,1,float16,float16,0,0.0415040006240209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,24,128,1,float16,fp8,0,0.08667733271916707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,4,128,1,float16,fp8,0,0.05348266661167145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,24,128,1,float16,float16,0,0.07776533563931783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,4,128,1,float16,float16,0,0.05046399931112925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,2,128,1,float16,fp8,0,0.04382933179537455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,2,128,1,float16,float16,0,0.04418666660785675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,8,128,1,float16,fp8,0,0.054058666030565895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,1,128,1,float16,float16,0,0.02517866591612498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,1,128,1,float16,fp8,0,0.039733332892258964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,24,128,1,float16,fp8,0,0.04324266811211904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,24,128,1,float16,float16,0,0.03860266755024592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,2,128,1,float16,fp8,0,0.027248000105222065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,4,128,1,float16,float16,0,0.02958933264017105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,4,128,1,float16,fp8,0,0.031061333914597828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,8,128,1,float16,float16,0,0.03123733401298523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,1,128,1,float16,float16,0,0.018837332725524902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,24,128,1,float16,float16,0,0.022986667851607006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,1,128,1,float16,fp8,0,0.018751999984184902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,2,128,1,float16,fp8,0,0.02067733307679494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,4,128,1,float16,float16,0,0.02094399929046631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,4,128,1,float16,fp8,0,0.022874665757020313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,8,128,1,float16,float16,0,0.02096533278624217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,8,128,1,float16,fp8,0,0.023189333577950794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,24,128,1,float16,float16,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,24,128,1,float16,fp8,0,0.018853332847356796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,1,128,1,float16,float16,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,1,128,1,float16,fp8,0,0.013807999591032663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,2,128,1,float16,float16,0,0.013418667018413544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,2,128,1,float16,fp8,0,0.014650666465361914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,4,128,1,float16,float16,0,0.015024000157912573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,4,128,1,float16,fp8,0,0.01544533297419548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,8,128,1,float16,float16,0,0.015114666273196539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,24,8,128,1,float16,float16,0,0.05340800185998281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,1,128,1,float16,fp8,0,0.025306666890780132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,24,8,128,1,float16,fp8,0,0.015210667004187902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,1,128,1,float16,float16,0,0.0129120002190272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,8,128,1,float16,fp8,0,0.03142933299144109
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,1,128,1,float16,fp8,0,0.013072000195582708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,24,128,1,float16,float16,0,0.013701333353916803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,2,128,1,float16,float16,0,0.019215999792019527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,24,24,128,1,float16,fp8,0,0.027552001178264618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,4,128,1,float16,float16,0,0.012928000340859095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,4,128,1,float16,fp8,0,0.013130666067202887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,8,128,1,float16,fp8,0,0.01312000056107839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,8,128,1,float16,float16,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,1,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,24,128,1,float16,fp8,0,0.01469333345691363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,24,128,1,float16,float16,0,0.013493333011865616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,1,128,1,float16,fp8,0,0.013125333935022354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,2,128,1,float16,float16,0,0.013023999830087027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,4,128,1,float16,float16,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,2,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,4,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,8,128,1,float16,float16,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,24,8,128,1,float16,fp8,0,0.013157332936922709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,24,2,128,1,float16,float16,0,0.027087998886903126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,1,128,1,float16,float16,0,0.1917333404223124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,1,128,1,float16,fp8,0,0.17499200503031412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,2,128,1,float16,float16,0,0.2087413271268209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,2,128,1,float16,fp8,0,0.19219734271367392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,24,128,1,float16,fp8,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,2,128,1,float16,fp8,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,24,2,128,1,float16,float16,0,0.012970666090647379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,4,128,1,float16,float16,0,0.24994132916132608
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,4,128,1,float16,fp8,0,0.23812800645828247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,1,128,1,float16,float16,0,0.10479467113812764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,8,128,1,float16,float16,0,0.25167999664942425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,1,128,1,float16,fp8,0,0.09220799803733826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,24,128,1,float16,float16,0,0.16965333620707193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,24,8,128,1,float16,fp8,0,0.24712000290552774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,24,128,1,float16,fp8,0,0.18264534076054892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,2,128,1,float16,float16,0,0.11265599727630615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,2,128,1,float16,fp8,0,0.10106133421262105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,4,128,1,float16,float16,0,0.132533331712087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,8,128,1,float16,float16,0,0.13568533460299173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,4,128,1,float16,fp8,0,0.12403200070063274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,24,8,128,1,float16,fp8,0,0.13131733735402426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,1,128,1,float16,fp8,0,0.052485331892967224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,2,128,1,float16,float16,0,0.05959466596444448
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,24,128,1,float16,fp8,0,0.09844799836476643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,4,128,1,float16,float16,0,0.0659093310435613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,2,128,1,float16,fp8,0,0.05646933118502299
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,4,128,1,float16,fp8,0,0.06459733347098033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,8,128,1,float16,float16,0,0.07050666709740956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,8,128,1,float16,fp8,0,0.06599999964237213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,1,128,1,float16,float16,0,0.034272000193595886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,2,128,1,float16,float16,0,0.03561066587766012
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,1,128,1,float16,fp8,0,0.032272001107533775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,4,128,1,float16,float16,0,0.038634667793909706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,24,128,1,float16,float16,0,0.045034666856129967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,4,128,1,float16,fp8,0,0.037461332976818085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,8,128,1,float16,float16,0,0.0394400010506312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,8,128,1,float16,fp8,0,0.03941866755485535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,24,128,1,float16,fp8,0,0.029135999580224354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,1,128,1,float16,fp8,0,0.02083733429511388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,24,128,1,float16,float16,0,0.02718399961789449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,2,128,1,float16,float16,0,0.02293866624434789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,2,128,1,float16,fp8,0,0.02197866638501485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,4,128,1,float16,float16,0,0.02436800052722295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,4,128,1,float16,fp8,0,0.02327466756105423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,8,128,1,float16,float16,0,0.023946667710940044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,8,128,1,float16,fp8,0,0.024879999458789825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,24,128,1,float16,float16,0,0.01913600042462349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,1,128,1,float16,float16,0,0.016890666137139004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,1,128,1,float16,float16,0,0.056736002365748085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,24,128,1,float16,fp8,0,0.021173333128293354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,1,128,1,float16,fp8,0,0.01642666632930438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,2,128,1,float16,float16,0,0.01676799977819125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,2,128,1,float16,fp8,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,4,128,1,float16,float16,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,24,24,128,1,float16,float16,0,0.09080533186594646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,4,128,1,float16,fp8,0,0.018810667097568512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,8,128,1,float16,float16,0,0.017258666455745697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,1,128,1,float16,fp8,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,24,128,1,float16,float16,0,0.013408000270525614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,24,128,1,float16,fp8,0,0.01488000030318896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,24,128,1,float16,fp8,0,0.05049600203831991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,2,128,1,float16,fp8,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,4,128,1,float16,float16,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,4,128,1,float16,fp8,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,8,128,1,float16,fp8,0,0.012768000364303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,24,1,128,1,float16,float16,0,0.021301334102948506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,24,128,1,float16,fp8,0,0.01309866706530253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,24,128,1,float16,float16,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,1,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,1,128,1,float16,fp8,0,0.01179733375708262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,2,128,1,float16,fp8,0,0.012794667234023413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,4,128,1,float16,float16,0,0.011370666325092316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,2,128,1,float16,float16,0,0.011381333072980246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,4,128,1,float16,fp8,0,0.012634667257467905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,8,128,1,float16,float16,0,0.011424000064531961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,24,8,128,1,float16,fp8,0,0.012714666624863943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,24,128,1,float16,float16,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,24,128,1,float16,fp8,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,1,128,1,float16,float16,0,0.011861333002646765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,24,8,128,1,float16,fp8,0,0.01887999971707662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,2,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,2,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,4,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,4,128,1,float16,fp8,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,24,2,128,1,float16,fp8,0,0.03400533397992452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,8,128,1,float16,float16,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,8,128,1,float16,float16,0,0.012655999511480331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,1,128,1,float16,float16,0,0.17587733268737793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,1,128,1,float16,fp8,0,0.1492533286412557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,2,128,1,float16,float16,0,0.18013334274291992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,2,128,1,float16,fp8,0,0.15628266334533691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,1,128,1,float16,fp8,0,0.012602667013804117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,4,128,1,float16,float16,0,0.20082134008407593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,1,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,8,128,1,float16,float16,0,0.2023520072301229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,4,128,1,float16,fp8,0,0.1819360057512919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,24,2,128,1,float16,float16,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,1,128,1,float16,float16,0,0.0942133367061615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,24,8,128,1,float16,fp8,0,0.01209066684047381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,24,128,1,float16,fp8,0,0.12569066882133484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,24,128,1,float16,float16,0,0.127018670241038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,2,128,1,float16,float16,0,0.09711466232935588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,2,128,1,float16,fp8,0,0.08282133440176646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,4,128,1,float16,float16,0,0.1043893297513326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,4,128,1,float16,fp8,0,0.09223467111587524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,8,128,1,float16,float16,0,0.1118986705938975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,24,128,1,float16,float16,0,0.07167466481526692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,8,128,1,float16,fp8,0,0.09505599737167358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,1,128,1,float16,fp8,0,0.04588800172011057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,24,128,1,float16,fp8,0,0.06438399851322174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,2,128,1,float16,float16,0,0.053946668903032936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,2,128,1,float16,fp8,0,0.047925333182017006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,4,128,1,float16,float16,0,0.056927998860677086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,8,128,1,float16,float16,0,0.057487999399503074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,8,128,1,float16,fp8,0,0.05208533505598704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,24,128,1,float16,fp8,0,0.037402667105197906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,1,128,1,float16,float16,0,0.03136000037193298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,24,128,1,float16,float16,0,0.03602666656176249
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,1,128,1,float16,fp8,0,0.027797333896160126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,2,128,1,float16,fp8,0,0.02939733366171519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,2,128,1,float16,float16,0,0.03332799921433131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,4,128,1,float16,fp8,0,0.03126933425664902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,24,8,128,1,float16,fp8,0,0.18708266814549765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,8,128,1,float16,float16,0,0.033589333295822144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,8,128,1,float16,fp8,0,0.0315786674618721
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,24,1,128,1,float16,fp8,0,0.07998933394749959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,1,128,1,float16,float16,0,0.019706666469573975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,1,128,1,float16,fp8,0,0.019018666197856266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,24,128,1,float16,float16,0,0.022991999983787537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,24,128,1,float16,fp8,0,0.02362666775782903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,2,128,1,float16,float16,0,0.020303999384244282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,2,128,1,float16,fp8,0,0.01893866683046023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,4,128,1,float16,float16,0,0.02125866711139679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,4,128,1,float16,fp8,0,0.020954666038354237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,8,128,1,float16,float16,0,0.020949333906173706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,24,8,128,1,float16,fp8,0,0.021125334004561108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,1,128,1,float16,fp8,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,24,128,1,float16,float16,0,0.017055999487638474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,24,128,1,float16,fp8,0,0.017312000195185345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,2,128,1,float16,fp8,0,0.015125333021084467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,4,128,1,float16,float16,0,0.015135999768972397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,4,128,1,float16,fp8,0,0.051813334226608276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,4,128,1,float16,fp8,0,0.015018666783968607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,8,128,1,float16,float16,0,0.015200000256299973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,8,128,1,float16,fp8,0,0.015189333508412043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,1,128,1,float16,float16,0,0.012128000458081564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,2,128,1,float16,float16,0,0.011770666887362799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,1,128,1,float16,fp8,0,0.011178666104873022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,24,128,1,float16,float16,0,0.012661332885424295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,2,128,1,float16,fp8,0,0.012149333953857422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,4,128,1,float16,float16,0,0.012879999975363413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,4,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,24,4,128,1,float16,float16,0,0.0337119996547699
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,8,128,1,float16,fp8,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,1,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,1,128,1,float16,fp8,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,24,128,1,float16,fp8,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,24,128,1,float16,float16,0,0.012026666353146235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,2,128,1,float16,float16,0,0.011215999722480774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,2,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,4,128,1,float16,float16,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,4,128,1,float16,fp8,0,0.011205332974592844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,8,128,1,float16,float16,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,24,8,128,1,float16,fp8,0,0.011285333583752314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,1,128,1,float16,float16,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,24,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,24,1,128,1,float16,float16,0,0.05332266787687937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,24,128,1,float16,float16,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,1,128,1,float16,float16,0,0.015263999501864115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,2,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,2,128,1,float16,fp8,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,4,128,1,float16,float16,0,0.01080000028014183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,4,128,1,float16,fp8,0,0.011381333072980246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,8,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,8,128,1,float16,fp8,0,0.011306667079528173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,24,128,1,float16,fp8,0,0.013072000195582708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,1,128,1,float16,fp8,0,0.13568533460299173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,1,128,1,float16,float16,0,0.1698933243751526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,2,128,1,float16,float16,0,0.1730133295059204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,24,8,128,1,float16,float16,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,2,128,1,float16,fp8,0,0.13799466689427695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,4,128,1,float16,float16,0,0.17861332496007284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,4,128,1,float16,fp8,0,0.14824000000953674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,8,128,1,float16,float16,0,0.18738667170206705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,24,8,128,1,float16,fp8,0,0.15226667126019797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,24,128,1,float16,float16,0,0.1088106632232666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,24,2,128,1,float16,float16,0,0.01516266663869222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,1,128,1,float16,float16,0,0.09030399719874065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,1,128,1,float16,fp8,0,0.07442666590213776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,24,128,1,float16,fp8,0,0.09629333019256592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,2,128,1,float16,float16,0,0.09354666868845622
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,2,128,1,float16,fp8,0,0.07601599891980489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,4,128,1,float16,fp8,0,0.08055999875068665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,4,128,1,float16,float16,0,0.09310400485992432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,8,128,1,float16,float16,0,0.09612799684206645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,24,8,128,1,float16,fp8,0,0.08050666749477386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,24,128,1,float16,fp8,0,0.05176533261934916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,1,128,1,float16,fp8,0,0.041802664597829185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,2,128,1,float16,fp8,0,0.04383466641108195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,2,128,1,float16,float16,0,0.05165866514046987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,24,128,1,float16,float16,0,0.055786664287249245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,4,128,1,float16,float16,0,0.0528053343296051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,4,128,1,float16,fp8,0,0.045642669002215065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,8,128,1,float16,fp8,0,0.04574400186538696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,8,128,1,float16,float16,0,0.05202133456865946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,1,128,1,float16,float16,0,0.02975466599067052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,24,128,1,float16,float16,0,0.03356266766786575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,1,128,1,float16,fp8,0,0.02683199942111969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,2,128,1,float16,float16,0,0.03050133337577184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,2,128,1,float16,fp8,0,0.027210667729377747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,4,128,1,float16,float16,0,0.03136533250411352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,4,128,1,float16,fp8,0,0.029093332588672638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,8,128,1,float16,float16,0,0.03147733211517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,1,128,1,float16,float16,0,0.018981333822011948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,24,128,1,float16,float16,0,0.02086399992307027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,2,128,1,float16,float16,0,0.018901333212852478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,24,1,128,1,float16,fp8,0,0.01109333336353302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,4,128,1,float16,float16,0,0.018885333091020584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,2,128,1,float16,fp8,0,0.018911999960740406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,4,128,1,float16,fp8,0,0.018922666708628338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,8,128,1,float16,float16,0,0.019093333433071773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,8,128,1,float16,fp8,0,0.018922666708628338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,1,128,1,float16,float16,0,0.014831999937693277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,24,128,1,float16,float16,0,0.01613333324591319
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,24,128,1,float16,fp8,0,0.01659199967980385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,1,128,1,float16,fp8,0,0.014906667172908783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,2,128,1,float16,fp8,0,0.014778666198253632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,24,1,128,1,float16,float16,0,0.050240000089009605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,2,128,1,float16,float16,0,0.015040000279744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,4,128,1,float16,float16,0,0.014885333677132925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,4,128,1,float16,fp8,0,0.014848000059525171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,8,128,1,float16,float16,0,0.014794666320085526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,24,8,128,1,float16,fp8,0,0.015013333410024643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,1,128,1,float16,float16,0,0.011829332758982977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,1,128,1,float16,fp8,0,0.01268799975514412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,24,128,1,float16,fp8,0,0.012960000584522883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,24,128,1,float16,fp8,0,0.03164266546567281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,2,128,1,float16,float16,0,0.012202666451533636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,2,128,1,float16,fp8,0,0.011242666592200598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,4,128,1,float16,float16,0,0.011402666568756104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,4,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,8,128,1,float16,float16,0,0.012042666474978128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,8,128,1,float16,fp8,0,0.011765333513418833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,1,128,1,float16,fp8,0,0.010693332801262537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,24,8,128,1,float16,fp8,0,0.02923733244339625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,24,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,24,128,1,float16,float16,0,0.012874666601419449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,2,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,2,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,4,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,4,128,1,float16,fp8,0,0.011098666737476984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,24,128,1,float16,fp8,0,0.020954666038354237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,8,128,1,float16,float16,0,0.01121066634853681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,1,128,1,float16,fp8,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,1,128,1,float16,float16,0,0.010672000547250112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,2,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,24,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,24,128,1,float16,float16,0,0.011749333391586939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,2,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,4,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,4,128,1,float16,float16,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,8,128,1,float16,float16,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,24,8,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,2,128,1,float16,float16,0,0.014943999548753103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,2,128,1,float16,fp8,0,0.014853333433469137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,4,128,1,float16,float16,0,0.01907733331123988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,24,24,128,1,float16,float16,0,0.012768000364303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,4,128,1,float16,fp8,0,0.016794666647911072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,24,128,1,float16,fp8,0,0.035301332672437034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,8,128,1,float16,float16,0,0.029088000456492107
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,24,8,128,1,float16,fp8,0,0.025306666890780132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,24,1,128,1,float16,fp8,0,0.018874666343132656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,2,128,1,float16,fp8,0,0.0122079998254776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,24,8,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,2,128,1,float16,float16,0,0.011760000139474869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,4,128,1,float16,float16,0,0.013045333325862885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,4,128,1,float16,fp8,0,0.012784000486135483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,8,128,1,float16,fp8,0,0.01711999997496605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,8,128,1,float16,float16,0,0.018816000471512478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,24,128,1,float16,fp8,0,0.023013333479563396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,2,128,1,float16,float16,0,0.00966933307548364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,2,128,1,float16,fp8,0,0.0106133334338665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,4,128,1,float16,fp8,0,0.01062400018175443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,4,128,1,float16,float16,0,0.011146667102972666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,8,128,1,float16,float16,0,0.01312000056107839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,24,128,1,float16,float16,0,0.01676799977819125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,24,128,1,float16,fp8,0,0.016885332763195038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,2,128,1,float16,float16,0,0.009050666665037474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,2,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,4,128,1,float16,fp8,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,8,128,1,float16,float16,0,0.011109333485364914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,8,128,1,float16,fp8,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,24,24,128,1,float16,float16,0,0.04604800045490265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,24,128,1,float16,float16,0,0.012863999853531519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,2,128,1,float16,float16,0,0.00867733359336853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,24,128,1,float16,fp8,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,4,128,1,float16,float16,0,0.008890666688481966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,4,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,8,128,1,float16,float16,0,0.009077333534757296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,8,128,1,float16,fp8,0,0.00884799969693025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,24,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,24,128,1,float16,float16,0,0.024085332949956257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,24,128,1,float16,fp8,0,0.010821333775917688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,2,128,1,float16,fp8,0,0.00914666677514712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,4,128,1,float16,fp8,0,0.009082666908701261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,8,128,1,float16,float16,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,24,128,1,float16,float16,0,0.010469333579142889
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,24,128,1,float16,fp8,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,24,4,128,1,float16,float16,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,2,128,1,float16,fp8,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,2,128,1,float16,float16,0,0.008656000097592672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,4,128,1,float16,float16,0,0.00877333308259646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,4,128,1,float16,fp8,0,0.009077333534757296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,8,128,1,float16,fp8,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,24,8,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,24,2,128,1,float16,fp8,0,0.00891733355820179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,24,128,1,float16,float16,0,0.009621333330869675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,24,128,1,float16,fp8,0,0.010709332923094431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,2,128,1,float16,float16,0,0.00884799969693025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,2,128,1,float16,fp8,0,0.008885333314538002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,4,128,1,float16,float16,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,4,128,1,float16,fp8,0,0.00902399979531765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,8,128,1,float16,fp8,0,0.008837333569924036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,24,128,1,float16,float16,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,24,128,1,float16,fp8,0,0.010666667173306147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,2,128,1,float16,float16,0,0.008650666723648706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,2,128,1,float16,fp8,0,0.009253333633144697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,2,128,1,float16,float16,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,4,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,24,8,128,1,float16,fp8,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,8,128,1,float16,float16,0,0.009125333279371262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,8,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,24,8,128,1,float16,float16,0,0.008778666456540426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,1,128,1,float16,fp8,0,5.571840286254883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,2,128,1,float16,fp8,0,5.842981338500977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,4,128,1,float16,float16,0,0.008896000062425932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,24,4,128,1,float16,float16,0,0.00878399983048439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,1,128,1,float16,float16,0,7.5427093505859375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,2,128,1,float16,float16,0,7.650608062744141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,24,8,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,1,128,1,float16,fp8,0,2.852927843729655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,1,128,1,float16,float16,0,3.7353493372599282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,16,128,1,float16,float16,0,3.834320068359375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,16,128,1,float16,fp8,0,3.5399465560913086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,4,128,1,float16,fp8,0,6.095530827840169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,4,128,1,float16,float16,0,7.880837122599284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,8,128,1,float16,fp8,0,6.196207682291667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,16,8,128,1,float16,float16,0,7.91977055867513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,2,128,1,float16,float16,0,3.616922696431478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,2,128,1,float16,fp8,0,3.6454668045043945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,4,128,1,float16,fp8,0,3.151434580485026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,4,128,1,float16,float16,0,3.707680066426595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,16,128,1,float16,float16,0,1.9911786715189617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,8,128,1,float16,float16,0,3.7494026819864907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,1,128,1,float16,float16,0,1.8827199935913086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,16,8,128,1,float16,fp8,0,3.550421396891276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,1,128,1,float16,fp8,0,1.5217386881510417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,16,128,1,float16,fp8,0,2.2799413998921714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,2,128,1,float16,float16,0,1.8370399475097656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,2,128,1,float16,fp8,0,1.6986506779988606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,4,128,1,float16,float16,0,1.9263787269592285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,4,128,1,float16,fp8,0,1.8527466456095378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,8,128,1,float16,float16,0,1.955349286397298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,1,128,1,float16,float16,0,0.9538986682891846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,16,8,128,1,float16,fp8,0,1.6781813303629558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,16,128,1,float16,fp8,0,0.9353333314259847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,16,128,1,float16,float16,0,1.0609172979990642
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,1,128,1,float16,fp8,0,0.9048799673716227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,2,128,1,float16,float16,0,0.9797386328379313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,2,128,1,float16,fp8,0,0.936288038889567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,4,128,1,float16,float16,0,1.0146293640136719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,8,128,1,float16,fp8,0,0.889695962270101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,8,128,1,float16,float16,0,1.024469296137492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,1,128,1,float16,fp8,0,3.2888107299804688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,1,128,1,float16,float16,0,4.077146530151367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,2,128,1,float16,fp8,0,3.3703254063924155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,2,128,1,float16,float16,0,4.081184069315593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,4,128,1,float16,fp8,0,3.6760212580362954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,4,128,1,float16,float16,0,4.4030561447143555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,1,128,1,float16,float16,0,2.0116373697916665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,1,128,1,float16,fp8,0,1.6815253893534343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,16,128,1,float16,float16,0,2.2881174087524414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,16,4,128,1,float16,fp8,0,0.8791680335998535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,2,128,1,float16,float16,0,2.212384064992269
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,8,128,1,float16,fp8,0,3.972655932108561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,16,8,128,1,float16,float16,0,4.369173367818196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,2,128,1,float16,fp8,0,1.9613173802693684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,4,128,1,float16,float16,0,2.2030879656473794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,4,128,1,float16,fp8,0,1.9071146647135417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,8,128,1,float16,float16,0,2.204410711924235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,16,128,1,float16,float16,0,1.195578654607137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,16,128,1,float16,fp8,0,1.217194636662801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,1,128,1,float16,float16,0,1.058773358662923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,8,128,1,float16,fp8,0,2.226970672607422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,1,128,1,float16,fp8,0,0.8910933335622152
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,2,128,1,float16,float16,0,1.0748533407847087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,2,128,1,float16,fp8,0,0.9185439745585123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,4,128,1,float16,fp8,0,1.038805325826009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,4,128,1,float16,float16,0,1.1493546962738037
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,1,128,1,float16,float16,0,0.5940906604131063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,8,128,1,float16,float16,0,1.142416000366211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,16,128,1,float16,float16,0,0.6377706527709961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,16,128,1,float16,fp8,0,0.5883626540501913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,1,128,1,float16,fp8,0,0.5030773480733236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,2,128,1,float16,float16,0,0.5808266798655192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,2,128,1,float16,fp8,0,0.5678826570510864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,4,128,1,float16,float16,0,0.6095573504765829
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,4,128,1,float16,fp8,0,0.5558133522669474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,8,128,1,float16,float16,0,0.6115839878718058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,16,8,128,1,float16,fp8,0,0.5659306844075521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,16,16,128,1,float16,fp8,0,2.341263930002848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,1,128,1,float16,fp8,0,2.344693342844645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,1,128,1,float16,float16,0,2.8484319051106772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,2,128,1,float16,fp8,0,2.416389306386312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,2,128,1,float16,float16,0,2.9158719380696616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,4,128,1,float16,fp8,0,2.6535305976867676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,4,128,1,float16,float16,0,3.0906880696614585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,8,128,1,float16,float16,0,3.149200121561686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,1,128,1,float16,float16,0,1.4554986953735352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,16,128,1,float16,float16,0,1.8488853772481282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,1,128,1,float16,fp8,0,1.2074506282806396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,16,128,1,float16,fp8,0,1.5163733164469402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,16,8,128,1,float16,fp8,0,2.6988585789998374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,2,128,1,float16,float16,0,1.4770347277323406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,16,8,128,1,float16,fp8,0,1.026037295659383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,2,128,1,float16,fp8,0,1.2659839789072673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,4,128,1,float16,float16,0,1.681536038716634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,1,128,1,float16,float16,0,0.756165345509847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,1,128,1,float16,fp8,0,0.7048640251159668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,8,128,1,float16,fp8,0,1.3970774014790852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,4,128,1,float16,fp8,0,1.4945440292358398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,16,8,128,1,float16,float16,0,1.6020213762919109
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,16,128,1,float16,fp8,0,0.8367466926574707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,2,128,1,float16,fp8,0,0.6689173380533854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,2,128,1,float16,float16,0,0.775935967763265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,4,128,1,float16,float16,0,0.8376693725585938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,4,128,1,float16,fp8,0,0.7322719891866049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,8,128,1,float16,fp8,0,0.7397546768188477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,8,128,1,float16,float16,0,0.8445066610972086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,1,128,1,float16,float16,0,0.41120533148447674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,16,128,1,float16,float16,0,0.4774666627248128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,1,128,1,float16,fp8,0,0.37914665540059406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,2,128,1,float16,float16,0,0.41994667053222656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,4,128,1,float16,float16,0,0.4535200198491414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,4,128,1,float16,fp8,0,0.4049546718597412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,8,128,1,float16,float16,0,0.45278934637705487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,8,128,1,float16,fp8,0,0.42718935012817383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,1,128,1,float16,fp8,0,3.0479307174682617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,1,128,1,float16,float16,0,3.8176212310791016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,2,128,1,float16,float16,0,3.7459999720255532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,2,128,1,float16,fp8,0,3.1723254521687827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,16,16,128,1,float16,float16,0,0.8629493713378906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,4,128,1,float16,float16,0,4.1559146245320635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,16,128,1,float16,fp8,0,0.44411198298136395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,16,2,128,1,float16,fp8,0,0.37434132893880206
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,1,128,1,float16,float16,0,1.8779892921447754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,4,128,1,float16,fp8,0,3.553002675374349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,16,128,1,float16,float16,0,2.2083627382914224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,16,128,1,float16,fp8,0,2.0360159873962402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,8,128,1,float16,float16,0,4.215615908304851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,1,128,1,float16,fp8,0,1.5509227116902669
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,2,128,1,float16,float16,0,1.9248959223429363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,4,128,1,float16,float16,0,2.0825600624084473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,4,128,1,float16,fp8,0,1.8151572545369465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,8,128,1,float16,fp8,0,1.8460960388183594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,1,128,1,float16,float16,0,0.9517920017242432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,16,128,1,float16,float16,0,1.1324693361918132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,8,128,1,float16,float16,0,2.137882709503174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,16,128,1,float16,fp8,0,1.1840533415476482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,1,128,1,float16,fp8,0,0.8114720185597738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,2,128,1,float16,float16,0,0.985530694325765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,2,128,1,float16,fp8,0,0.8844799995422363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,4,128,1,float16,float16,0,1.0896639823913574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,4,128,1,float16,fp8,0,0.9758773644765218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,8,128,1,float16,fp8,0,0.9608266353607178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,1,128,1,float16,float16,0,0.5076053142547607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,16,8,128,1,float16,float16,0,1.0754559834798176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,16,128,1,float16,float16,0,0.5989653269449869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,16,128,1,float16,fp8,0,0.5755946636199951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,1,128,1,float16,fp8,0,0.4415466785430908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,2,128,1,float16,float16,0,0.5221759875615438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,2,128,1,float16,fp8,0,0.46350399653116864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,4,128,1,float16,float16,0,0.5689760049184164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,16,8,128,1,float16,fp8,0,3.7712319691975913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,4,128,1,float16,fp8,0,0.5214720169703165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,8,128,1,float16,float16,0,0.5645066499710083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,16,8,128,1,float16,fp8,0,0.522709329922994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,1,128,1,float16,float16,0,0.2860373258590698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,16,128,1,float16,fp8,0,0.32178666194279987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,2,128,1,float16,fp8,0,0.26635199785232544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,4,128,1,float16,fp8,0,0.2840053240458171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,8,128,1,float16,float16,0,0.31781333684921265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,8,128,1,float16,fp8,0,0.29604800542195636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,1,128,1,float16,fp8,0,1.797610600789388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,1,128,1,float16,float16,0,2.1456000010172525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,2,128,1,float16,float16,0,2.2423680623372397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,2,128,1,float16,fp8,0,1.9051146507263184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,16,2,128,1,float16,fp8,0,1.6203840573628743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,1,128,1,float16,fp8,0,0.26057066520055133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,16,128,1,float16,float16,0,0.3285599946975708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,4,128,1,float16,float16,0,2.551701386769613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,4,128,1,float16,fp8,0,2.2583840688069663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,2,128,1,float16,float16,0,0.285808006922404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,16,4,128,1,float16,float16,0,0.31143999099731445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,1,128,1,float16,float16,0,1.127834637959798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,1,128,1,float16,fp8,0,0.9298186302185059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,16,128,1,float16,float16,0,1.3627413113911946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,2,128,1,float16,float16,0,1.1355573336283367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,16,128,1,float16,fp8,0,1.291914701461792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,2,128,1,float16,fp8,0,0.9821973641713461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,8,128,1,float16,fp8,0,2.251178741455078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,16,8,128,1,float16,float16,0,2.5354026158650718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,4,128,1,float16,float16,0,1.2834080060323079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,4,128,1,float16,fp8,0,1.1238346894582112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,1,128,1,float16,float16,0,0.5692373514175415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,8,128,1,float16,float16,0,1.2822453180948894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,16,8,128,1,float16,fp8,0,1.1527093251546223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,16,128,1,float16,fp8,0,0.6769973436991373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,1,128,1,float16,fp8,0,0.5073813199996948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,2,128,1,float16,float16,0,0.5932159821192423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,4,128,1,float16,float16,0,0.6521066824595133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,4,128,1,float16,fp8,0,0.609941323598226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,8,128,1,float16,fp8,0,0.6082880099614462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,8,128,1,float16,float16,0,0.6553440093994141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,16,128,1,float16,float16,0,0.37741867701212567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,1,128,1,float16,float16,0,0.3068319956461589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,16,128,1,float16,fp8,0,0.3763519922892253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,1,128,1,float16,fp8,0,0.28228267033894855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,2,128,1,float16,float16,0,0.3205973307291667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,4,128,1,float16,float16,0,0.35209067662556964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,4,128,1,float16,fp8,0,0.32706665992736816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,8,128,1,float16,float16,0,0.3590986728668213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,1,128,1,float16,float16,0,0.17805866400400797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,16,128,1,float16,float16,0,0.2143626610438029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,16,128,1,float16,fp8,0,0.2130133310953776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,1,128,1,float16,fp8,0,0.16888533035914102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,2,128,1,float16,float16,0,0.1806186636288961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,2,128,1,float16,fp8,0,0.17223999897638956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,4,128,1,float16,fp8,0,0.18643200397491455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,8,128,1,float16,float16,0,0.2021226684252421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,8,128,1,float16,fp8,0,0.1875306765238444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,1,128,1,float16,float16,0,2.033957322438558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,16,128,1,float16,float16,0,0.6983520189921061
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,1,128,1,float16,fp8,0,1.7232319513956706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,2,128,1,float16,float16,0,2.1318346659342446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,2,128,1,float16,fp8,0,0.2877333362897237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,2,128,1,float16,fp8,0,2.063408056894938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,16,8,128,1,float16,fp8,0,0.33710400263468426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,4,128,1,float16,float16,0,2.494042714436849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,4,128,1,float16,fp8,0,2.2502613067626953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,8,128,1,float16,float16,0,2.512997309366862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,16,4,128,1,float16,float16,0,0.1919040083885193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,1,128,1,float16,float16,0,1.0353333155314128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,1,128,1,float16,fp8,0,0.882256031036377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,16,128,1,float16,float16,0,1.383962631225586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,16,128,1,float16,fp8,0,1.3738560676574707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,2,128,1,float16,float16,0,1.0992693106333415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,2,128,1,float16,fp8,0,0.9576319853464762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,4,128,1,float16,float16,0,1.2632906436920166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,4,128,1,float16,fp8,0,1.1487253506978352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,8,128,1,float16,float16,0,1.273909330368042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,16,2,128,1,float16,fp8,0,0.5238240162531534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,16,128,1,float16,float16,0,0.7069493134816488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,16,8,128,1,float16,fp8,0,1.2106080055236816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,1,128,1,float16,float16,0,0.5297173261642456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,2,128,1,float16,float16,0,0.5609440008799235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,1,128,1,float16,fp8,0,0.46618131796518963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,16,128,1,float16,fp8,0,0.7115893363952637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,2,128,1,float16,fp8,0,0.5044533411661783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,4,128,1,float16,float16,0,0.6429813305536906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,4,128,1,float16,fp8,0,0.5972053209940592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,8,128,1,float16,fp8,0,0.6154666741689047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,16,8,128,1,float16,float16,0,0.6520160039265951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,1,128,1,float16,float16,0,0.2837120095888774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,16,128,1,float16,fp8,0,0.37673068046569824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,1,128,1,float16,fp8,0,0.2546880046526591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,16,128,1,float16,float16,0,0.36929066975911456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,2,128,1,float16,float16,0,0.29636265834172565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,2,128,1,float16,fp8,0,0.2772480050722758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,4,128,1,float16,float16,0,0.3410773277282715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,4,128,1,float16,fp8,0,0.32256533702214557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,8,128,1,float16,float16,0,0.3420106569925944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,16,8,128,1,float16,fp8,0,0.33397332827250165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,1,128,1,float16,float16,0,0.16105600198109946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,1,128,1,float16,fp8,0,0.1497066617012024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,16,128,1,float16,float16,0,0.20618667205174765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,2,128,1,float16,float16,0,0.16565866271654764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,16,128,1,float16,fp8,0,0.21175465981165567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,4,128,1,float16,fp8,0,0.1748746633529663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,4,128,1,float16,float16,0,0.18888000647226968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,16,8,128,1,float16,fp8,0,2.3306825955708823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,8,128,1,float16,float16,0,0.19326400756835938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,8,128,1,float16,fp8,0,0.18252799908320108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,1,128,1,float16,float16,0,0.10493866602579753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,16,128,1,float16,float16,0,0.12504000465075174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,16,128,1,float16,fp8,0,0.12172266840934753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,2,128,1,float16,fp8,0,0.10115733742713928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,4,128,1,float16,float16,0,0.11269866426785786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,4,128,1,float16,fp8,0,0.10922666390736897
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,8,128,1,float16,float16,0,0.11403733491897583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,1,128,1,float16,float16,0,1.2171146869659424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,1,128,1,float16,fp8,0,1.0553279717763264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,2,128,1,float16,float16,0,1.3211413224538167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,2,128,1,float16,fp8,0,1.1620480219523113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,4,128,1,float16,fp8,0,1.441882610321045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,4,128,1,float16,float16,0,1.5587679545084636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,8,128,1,float16,float16,0,1.6044212977091472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,16,2,128,1,float16,fp8,0,0.15711999932924905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,16,8,128,1,float16,fp8,0,1.506384054819743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,1,128,1,float16,float16,0,0.6224480072657267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,1,128,1,float16,fp8,0,0.5477386713027954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,16,128,1,float16,float16,0,0.884709358215332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,2,128,1,float16,float16,0,0.10705600182215373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,2,128,1,float16,float16,0,0.6661973396937052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,2,128,1,float16,fp8,0,0.6036159992218018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,8,128,1,float16,fp8,0,0.11153067151705424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,4,128,1,float16,fp8,0,0.7416533629099528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,8,128,1,float16,float16,0,0.8002933661142985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,8,128,1,float16,fp8,0,0.770032008488973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,1,128,1,float16,float16,0,0.3243093291918437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,16,128,1,float16,fp8,0,0.4726506471633911
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,2,128,1,float16,float16,0,0.3471786578496297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,1,128,1,float16,fp8,0,0.294320007165273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,2,128,1,float16,fp8,0,0.321669340133667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,4,128,1,float16,float16,0,0.40807998180389404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,4,128,1,float16,fp8,0,0.39166398843129474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,8,128,1,float16,float16,0,0.41463998953501385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,8,128,1,float16,fp8,0,0.405839999516805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,1,128,1,float16,float16,0,0.17964265743891397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,16,128,1,float16,float16,0,0.2437386711438497
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,16,128,1,float16,fp8,0,0.25600000222524005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,1,128,1,float16,fp8,0,0.1646346648534139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,2,128,1,float16,float16,0,0.1922773321469625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,2,128,1,float16,fp8,0,0.17372800906499228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,4,128,1,float16,float16,0,0.22360533475875854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,4,128,1,float16,fp8,0,0.21196266015370688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,8,128,1,float16,float16,0,0.2259946664174398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,16,8,128,1,float16,fp8,0,0.22197866439819336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,16,1,128,1,float16,fp8,0,0.09801066915194194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,16,128,1,float16,float16,0,0.14054399728775024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,16,128,1,float16,fp8,0,0.14442666371663412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,1,128,1,float16,fp8,0,0.09982400139172871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,2,128,1,float16,float16,0,0.10999466975529988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,4,128,1,float16,float16,0,0.11956800023714702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,4,128,1,float16,fp8,0,0.11824533343315125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,4,128,1,float16,float16,0,0.7937493324279785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,8,128,1,float16,float16,0,0.12981333335240683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,16,128,1,float16,float16,0,0.08111999928951263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,1,128,1,float16,float16,0,0.07042133311430614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,16,16,128,1,float16,float16,0,0.45500266551971436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,16,128,1,float16,fp8,0,0.08451199531555176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,1,128,1,float16,fp8,0,0.06637866795063019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,4,128,1,float16,fp8,0,0.07622399926185608
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,8,128,1,float16,float16,0,0.07830933233102162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,8,128,1,float16,fp8,0,0.07649066547552745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,16,16,128,1,float16,fp8,0,0.907632033030192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,1,128,1,float16,float16,0,0.10685867071151733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,1,128,1,float16,float16,0,1.2251946926116943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,2,128,1,float16,fp8,0,0.10622933506965637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,16,8,128,1,float16,fp8,0,0.12121599912643433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,1,128,1,float16,fp8,0,1.0609386761983235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,2,128,1,float16,float16,0,1.3436907132466633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,2,128,1,float16,float16,0,0.07216000060240428
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,2,128,1,float16,fp8,0,1.2104106744130452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,4,128,1,float16,float16,0,1.6780586242675781
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,4,128,1,float16,fp8,0,1.582906723022461
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,8,128,1,float16,fp8,0,1.6649120648701985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,16,128,1,float16,float16,0,0.9694879849751791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,16,8,128,1,float16,float16,0,1.7167305946350098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,1,128,1,float16,float16,0,0.6114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,16,128,1,float16,fp8,0,1.0313493410746257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,1,128,1,float16,fp8,0,0.5473866860071818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,2,128,1,float16,float16,0,0.6708426475524902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,4,128,1,float16,fp8,0,0.8090559641520182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,4,128,1,float16,float16,0,0.8410027027130127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,1,128,1,float16,float16,0,0.3158400058746338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,8,128,1,float16,float16,0,0.856170654296875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,16,128,1,float16,float16,0,0.48839465777079266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,16,128,1,float16,fp8,0,0.5273919900258383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,1,128,1,float16,fp8,0,0.28757333755493164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,8,128,1,float16,fp8,0,0.840399980545044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,2,128,1,float16,float16,0,0.34436798095703125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,2,128,1,float16,fp8,0,0.3311946590741475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,4,128,1,float16,float16,0,0.42987199624379474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,4,128,1,float16,fp8,0,0.4245866537094116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,8,128,1,float16,float16,0,0.4431573152542114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,1,128,1,float16,float16,0,0.17281599839528403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,16,8,128,1,float16,fp8,0,0.4375946521759033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,1,128,1,float16,fp8,0,0.1618773341178894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,16,128,1,float16,fp8,0,0.2791573405265808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,16,128,1,float16,float16,0,0.25884799162546795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,2,128,1,float16,float16,0,0.18926932414372763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,2,128,1,float16,fp8,0,0.1781866749127706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,4,128,1,float16,float16,0,0.2300800085067749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,8,128,1,float16,float16,0,0.2351626753807068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,8,128,1,float16,fp8,0,0.23598933219909668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,1,128,1,float16,float16,0,0.09918399651845296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,1,128,1,float16,fp8,0,0.09302399555842082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,16,128,1,float16,float16,0,0.14421866337458292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,2,128,1,float16,float16,0,0.10403199990590413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,2,128,1,float16,fp8,0,0.09912000099817912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,16,2,128,1,float16,fp8,0,0.6184426546096802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,4,128,1,float16,float16,0,0.1249066690603892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,4,128,1,float16,fp8,0,0.1167093316713969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,8,128,1,float16,float16,0,0.13292266925175986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,8,128,1,float16,fp8,0,0.12829333543777466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,16,128,1,float16,fp8,0,0.0844533344109853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,16,128,1,float16,float16,0,0.0846720039844513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,1,128,1,float16,float16,0,0.06364266574382782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,1,128,1,float16,fp8,0,0.06002133091290792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,2,128,1,float16,float16,0,0.06645866731802623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,2,128,1,float16,fp8,0,0.06361599763234456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,4,128,1,float16,float16,0,0.07698666552702586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,4,128,1,float16,float16,0,0.072543998559316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,8,128,1,float16,float16,0,0.07467733323574066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,8,128,1,float16,fp8,0,0.07339199880758922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,1,128,1,float16,float16,0,0.04116799930731455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,1,128,1,float16,fp8,0,0.0391146664818128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,16,128,1,float16,fp8,0,0.05030933519204458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,2,128,1,float16,float16,0,0.04177600145339966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,2,128,1,float16,fp8,0,0.03963200002908707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,4,128,1,float16,float16,0,0.0452159990866979
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,4,128,1,float16,fp8,0,0.045642669002215065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,8,128,1,float16,float16,0,0.04444266855716705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,8,128,1,float16,fp8,0,0.045509333411852516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,16,4,128,1,float16,fp8,0,0.22728532552719116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,1,128,1,float16,float16,0,0.7541013558705648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,1,128,1,float16,fp8,0,0.6784799893697103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,16,16,128,1,float16,fp8,0,0.15396799643834433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,2,128,1,float16,float16,0,0.8414506912231445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,2,128,1,float16,fp8,0,0.7834773063659668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,4,128,1,float16,float16,0,1.1049173672993977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,4,128,1,float16,fp8,0,1.064186652501424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,8,128,1,float16,float16,0,1.1342399915059407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,16,2,128,1,float16,fp8,0,0.06980800131956737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,16,4,128,1,float16,fp8,0,0.0721973329782486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,1,128,1,float16,float16,0,0.386624018351237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,1,128,1,float16,fp8,0,0.352944016456604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,16,128,1,float16,float16,0,0.6491626501083374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,16,8,128,1,float16,fp8,0,1.1222240130106609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,16,16,128,1,float16,float16,0,0.04649066428343455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,2,128,1,float16,float16,0,0.42976534366607666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,2,128,1,float16,fp8,0,0.40750400225321454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,4,128,1,float16,float16,0,0.5603413184483846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,4,128,1,float16,fp8,0,0.5480373303095499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,8,128,1,float16,float16,0,0.5741546551386515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,1,128,1,float16,float16,0,0.2077066699663798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,8,128,1,float16,fp8,0,0.574399987856547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,16,128,1,float16,float16,0,0.33373332023620605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,1,128,1,float16,fp8,0,0.1933120091756185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,2,128,1,float16,float16,0,0.23093867301940918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,2,128,1,float16,fp8,0,0.22161600987116495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,4,128,1,float16,float16,0,0.29338665803273517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,8,128,1,float16,float16,0,0.29713600873947144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,8,128,1,float16,fp8,0,0.3006880084673564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,1,128,1,float16,float16,0,0.11458133657773335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,1,128,1,float16,fp8,0,0.10621333122253418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,16,128,1,float16,float16,0,0.17998933792114258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,2,128,1,float16,float16,0,0.12875200311342874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,16,128,1,float16,fp8,0,0.1976319948832194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,2,128,1,float16,fp8,0,0.11864533027013142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,4,128,1,float16,float16,0,0.15929599603017172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,4,128,1,float16,fp8,0,0.15761599938074747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,8,128,1,float16,float16,0,0.16179200013478598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,16,8,128,1,float16,fp8,0,0.1643893321355184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,16,128,1,float16,fp8,0,0.10857599973678589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,1,128,1,float16,float16,0,0.06853866577148438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,1,128,1,float16,fp8,0,0.06464000046253204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,2,128,1,float16,fp8,0,0.07004799942175548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,2,128,1,float16,float16,0,0.07273600002129872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,4,128,1,float16,float16,0,0.0827946662902832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,16,16,128,1,float16,fp8,0,0.7139519850413004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,4,128,1,float16,fp8,0,0.08266133566697438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,8,128,1,float16,fp8,0,0.08543999989827473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,1,128,1,float16,float16,0,0.04576000074545542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,16,128,1,float16,fp8,0,0.06066133578618368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,16,128,1,float16,float16,0,0.056090667843818665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,1,128,1,float16,fp8,0,0.042837331692377724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,2,128,1,float16,float16,0,0.04750399788220724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,2,128,1,float16,fp8,0,0.04585599899291992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,4,128,1,float16,float16,0,0.051776001850763954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,4,128,1,float16,fp8,0,0.051829333106676735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,16,128,1,float16,fp8,0,0.36929599444071454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,8,128,1,float16,float16,0,0.05215999980767568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,16,4,128,1,float16,fp8,0,0.2889866630236308
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,16,8,128,1,float16,fp8,0,0.05389333268006643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,1,128,1,float16,float16,0,0.03188266605138779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,1,128,1,float16,fp8,0,0.031290667752424874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,16,128,1,float16,fp8,0,0.04110399881998698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,16,128,1,float16,float16,0,0.03735466549793879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,2,128,1,float16,fp8,0,0.03230933348337809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,4,128,1,float16,float16,0,0.035599999129772186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,4,128,1,float16,fp8,0,0.036533333361148834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,8,128,1,float16,float16,0,0.03696533292531967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,8,128,1,float16,fp8,0,0.037445334096749626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,16,128,1,float16,float16,0,0.10215999682744344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,1,128,1,float16,fp8,0,0.7281440099080404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,1,128,1,float16,float16,0,0.8114026387532552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,2,128,1,float16,float16,0,0.928938627243042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,2,128,1,float16,fp8,0,0.8685812950134277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,4,128,1,float16,float16,0,1.267690658569336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,4,128,1,float16,fp8,0,1.2482293446858723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,16,2,128,1,float16,float16,0,0.03364266703526179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,8,128,1,float16,float16,0,1.3055573304494221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,16,128,1,float16,float16,0,0.7601439952850342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,16,128,1,float16,fp8,0,0.8539306322733561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,1,128,1,float16,float16,0,0.41670934359232586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,2,128,1,float16,float16,0,0.46804265181223553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,2,128,1,float16,fp8,0,0.4474879900614421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,1,128,1,float16,fp8,0,0.37841065724690753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,4,128,1,float16,float16,0,0.6453760067621866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,4,128,1,float16,fp8,0,0.6316693226496378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,8,128,1,float16,float16,0,0.6577706734339396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,16,8,128,1,float16,float16,0,0.0922826627890269
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,16,128,1,float16,fp8,0,0.4403680165608724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,16,128,1,float16,float16,0,0.3868693510691325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,1,128,1,float16,float16,0,0.21818133195241293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,2,128,1,float16,float16,0,0.2488159934679667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,2,128,1,float16,fp8,0,0.23620800177256265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,1,128,1,float16,fp8,0,0.20424532890319824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,4,128,1,float16,float16,0,0.32870399951934814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,4,128,1,float16,fp8,0,0.33371198177337646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,8,128,1,float16,fp8,0,0.3462453285853068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,16,8,128,1,float16,fp8,0,1.3134880065917969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,16,128,1,float16,fp8,0,0.22986666361490884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,2,128,1,float16,float16,0,0.13524799545605978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,1,128,1,float16,float16,0,0.12225066622098286
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,2,128,1,float16,fp8,0,0.13156267007191977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,4,128,1,float16,float16,0,0.1781866749127706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,4,128,1,float16,fp8,0,0.17722666263580322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,8,128,1,float16,float16,0,0.1816320021947225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,8,128,1,float16,fp8,0,0.18595733245213827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,1,128,1,float16,float16,0,0.0680266668399175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,16,128,1,float16,fp8,0,0.12387733658154805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,1,128,1,float16,fp8,0,0.06418133278687795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,16,128,1,float16,float16,0,0.11196266611417134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,2,128,1,float16,float16,0,0.07446933289368947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,2,128,1,float16,fp8,0,0.0701386680205663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,4,128,1,float16,fp8,0,0.08662933111190796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,4,128,1,float16,float16,0,0.09467732906341553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,8,128,1,float16,float16,0,0.10174399614334106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,16,8,128,1,float16,fp8,0,0.6668480237325033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,1,128,1,float16,float16,0,0.04367466767628988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,1,128,1,float16,fp8,0,0.041637333730856575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,2,128,1,float16,float16,0,0.04577066500981649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,16,128,1,float16,fp8,0,0.06599466502666473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,2,128,1,float16,fp8,0,0.044863998889923096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,4,128,1,float16,float16,0,0.05192000170548757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,4,128,1,float16,fp8,0,0.05398400127887726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,16,8,128,1,float16,float16,0,0.3407573302586873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,8,128,1,float16,float16,0,0.05301866432030996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,8,128,1,float16,fp8,0,0.05584000051021576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,1,128,1,float16,fp8,0,0.0273333340883255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,1,128,1,float16,float16,0,0.02757866680622101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,16,128,1,float16,float16,0,0.034117333590984344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,2,128,1,float16,float16,0,0.02918400118748347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,16,128,1,float16,float16,0,0.203167994817098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,16,1,128,1,float16,fp8,0,0.11231467127799988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,2,128,1,float16,fp8,0,0.02916266769170761
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,4,128,1,float16,float16,0,0.0328053335348765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,4,128,1,float16,fp8,0,0.03332266708215078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,8,128,1,float16,fp8,0,0.03364799916744232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,8,128,1,float16,float16,0,0.03312533348798752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,16,128,1,float16,float16,0,0.02922133356332779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,1,128,1,float16,float16,0,0.025120000044504803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,1,128,1,float16,fp8,0,0.02497066557407379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,16,128,1,float16,fp8,0,0.030080000559488933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,2,128,1,float16,fp8,0,0.025749333202838898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,2,128,1,float16,float16,0,0.02696000039577484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,4,128,1,float16,float16,0,0.027461332579453785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,4,128,1,float16,fp8,0,0.027530667682488758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,8,128,1,float16,fp8,0,0.029365333418051403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,16,8,128,1,float16,float16,0,0.02759466568628947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,16,8,128,1,float16,fp8,0,0.09726933638254802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,16,16,128,1,float16,float16,0,0.06246933341026306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,1,128,1,float16,float16,0,0.6234826644261678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,1,128,1,float16,fp8,0,0.5625439882278442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,2,128,1,float16,float16,0,0.7392906347910563
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,2,128,1,float16,fp8,0,0.7027520338694254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,16,16,128,1,float16,fp8,0,0.03741333385308584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,4,128,1,float16,float16,0,1.085861365000407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,1,128,1,float16,float16,0,0.31983999411265057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,4,128,1,float16,fp8,0,1.077898661295573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,1,128,1,float16,fp8,0,0.2939680020014445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,2,128,1,float16,fp8,0,0.36640532811482746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,16,128,1,float16,float16,0,0.6633866628011068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,16,128,1,float16,fp8,0,0.7657226721445719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,2,128,1,float16,float16,0,0.3874239921569824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,4,128,1,float16,float16,0,0.5519733428955078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,8,128,1,float16,float16,0,1.1136106650034587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,4,128,1,float16,fp8,0,0.5474613507588705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,16,8,128,1,float16,fp8,0,1.1561546325683594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,16,128,1,float16,float16,0,0.34003734588623047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,1,128,1,float16,float16,0,0.1716053287188212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,8,128,1,float16,float16,0,0.5642613172531128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,16,8,128,1,float16,fp8,0,0.5825546582539877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,2,128,1,float16,float16,0,0.20519999663035074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,4,128,1,float16,float16,0,0.28414400418599445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,4,128,1,float16,fp8,0,0.28378132979075116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,16,128,1,float16,float16,0,0.1785973310470581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,16,128,1,float16,fp8,0,0.20485333601633707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,1,128,1,float16,float16,0,0.09621866544087727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,8,128,1,float16,float16,0,0.2946079969406128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,1,128,1,float16,fp8,0,0.086709330479304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,8,128,1,float16,fp8,0,0.3006239930788676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,2,128,1,float16,fp8,0,0.10737066467603047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,1,128,1,float16,fp8,0,0.15731733043988547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,4,128,1,float16,fp8,0,0.15052800377209982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,4,128,1,float16,float16,0,0.15201066931088766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,2,128,1,float16,fp8,0,0.19106133778889975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,8,128,1,float16,float16,0,0.15728533267974854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,8,128,1,float16,fp8,0,0.1604426701863607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,16,128,1,float16,float16,0,0.09842666983604431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,1,128,1,float16,fp8,0,0.04981866478919983
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,1,128,1,float16,float16,0,0.053786665201187134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,2,128,1,float16,fp8,0,0.05608533322811127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,2,128,1,float16,float16,0,0.059418668349583946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,16,16,128,1,float16,fp8,0,0.39155201117197674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,4,128,1,float16,float16,0,0.08018133540948232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,4,128,1,float16,fp8,0,0.07421333094437917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,8,128,1,float16,float16,0,0.08748799562454224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,1,128,1,float16,float16,0,0.033615998923778534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,1,128,1,float16,fp8,0,0.03141333411137263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,16,128,1,float16,fp8,0,0.05606933434804281
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,2,128,1,float16,float16,0,0.03606933355331421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,2,128,1,float16,fp8,0,0.03521066655715307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,4,128,1,float16,float16,0,0.04174399872620901
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,4,128,1,float16,fp8,0,0.04460800190766653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,8,128,1,float16,float16,0,0.043706665436426796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,8,128,1,float16,fp8,0,0.045824001232783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,16,128,1,float16,fp8,0,0.03325333446264267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,1,128,1,float16,fp8,0,0.020997333029905956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,2,128,1,float16,float16,0,0.02293866624434789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,2,128,1,float16,fp8,0,0.023210667073726654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,16,2,128,1,float16,float16,0,0.11114133397738139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,4,128,1,float16,float16,0,0.02526933451493581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,4,128,1,float16,fp8,0,0.02741333345572154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,8,128,1,float16,fp8,0,0.02743999908367793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,16,128,1,float16,float16,0,0.02182399978240331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,16,128,1,float16,fp8,0,0.023951999843120575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,1,128,1,float16,fp8,0,0.017477333545684814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,16,128,1,float16,fp8,0,0.10995200276374817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,2,128,1,float16,float16,0,0.019258666783571243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,2,128,1,float16,fp8,0,0.02103466788927714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,4,128,1,float16,float16,0,0.02107200026512146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,4,128,1,float16,fp8,0,0.023034666975339253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,16,16,128,1,float16,float16,0,0.0543146679798762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,8,128,1,float16,float16,0,0.02107200026512146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,8,128,1,float16,fp8,0,0.02294933299223582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,1,128,1,float16,float16,0,0.018810667097568512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,16,128,1,float16,float16,0,0.01905599981546402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,16,128,1,float16,fp8,0,0.020954666038354237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,1,128,1,float16,fp8,0,0.01738133281469345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,2,128,1,float16,fp8,0,0.01791999985774358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,2,128,1,float16,float16,0,0.01800000046690305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,4,128,1,float16,float16,0,0.018922666708628338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,16,128,1,float16,float16,0,0.027322667340437572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,4,128,1,float16,fp8,0,0.019285333653291065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,8,128,1,float16,float16,0,0.019130667050679524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,16,8,128,1,float16,fp8,0,0.019461333751678467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,8,128,1,float16,float16,0,0.02682666728893916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,16,1,128,1,float16,float16,0,0.01899733394384384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,1,128,1,float16,float16,0,0.25681066513061523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,16,8,128,1,float16,fp8,0,0.0836906631787618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,1,128,1,float16,fp8,0,0.24150933821996054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,2,128,1,float16,float16,0,0.3176906704902649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,2,128,1,float16,fp8,0,0.30990399916966754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,16,1,128,1,float16,float16,0,0.021002667645613354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,4,128,1,float16,float16,0,0.48154131571451825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,4,128,1,float16,fp8,0,0.49450135231018066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,16,128,1,float16,fp8,0,0.36451733112335205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,16,128,1,float16,float16,0,0.30557866891225177
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,1,128,1,float16,fp8,0,0.12760532895723978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,1,128,1,float16,float16,0,0.13700800140698752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,8,128,1,float16,float16,0,0.494485338528951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,16,8,128,1,float16,fp8,0,0.5288053353627523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,2,128,1,float16,float16,0,0.16455466548601785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,2,128,1,float16,fp8,0,0.16674667596817017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,8,128,1,float16,float16,0,0.2589120070139567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,4,128,1,float16,float16,0,0.2504853407541911
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,4,128,1,float16,fp8,0,0.25572800636291504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,1,128,1,float16,fp8,0,0.07004799942175548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,4,128,1,float16,float16,0,0.1327786644299825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,2,128,1,float16,fp8,0,0.0885653297106425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,2,128,1,float16,float16,0,0.09283199906349182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,4,128,1,float16,fp8,0,0.1362613340218862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,16,128,1,float16,float16,0,0.1593226691087087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,16,128,1,float16,fp8,0,0.18714133898417154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,8,128,1,float16,float16,0,0.13827733198801676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,16,128,1,float16,fp8,0,0.10188800096511841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,8,128,1,float16,fp8,0,0.1439520021279653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,16,128,1,float16,float16,0,0.08849066495895386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,1,128,1,float16,float16,0,0.04178133110205332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,2,128,1,float16,fp8,0,0.04800533254941305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,4,128,1,float16,fp8,0,0.06542400022347768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,16,8,128,1,float16,fp8,0,0.2709333300590515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,4,128,1,float16,float16,0,0.07048533360163371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,2,128,1,float16,float16,0,0.050517335534095764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,8,128,1,float16,float16,0,0.07709866762161255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,8,128,1,float16,fp8,0,0.0768746683994929
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,1,128,1,float16,float16,0,0.02664000044266383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,16,128,1,float16,float16,0,0.04753066599369049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,1,128,1,float16,fp8,0,0.025583999852339428
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,2,128,1,float16,fp8,0,0.02941333254178365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,2,128,1,float16,float16,0,0.029365333418051403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,4,128,1,float16,fp8,0,0.03941333293914795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,16,1,128,1,float16,float16,0,0.07553066809972127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,8,128,1,float16,float16,0,0.03770666569471359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,16,128,1,float16,fp8,0,0.04976533353328705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,16,128,1,float16,fp8,0,0.029338667790095013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,1,128,1,float16,float16,0,0.01710933322707812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,16,128,1,float16,float16,0,0.023381332556406658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,1,128,1,float16,fp8,0,0.018789333601792652
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,2,128,1,float16,float16,0,0.01924266666173935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,2,128,1,float16,fp8,0,0.01953599974513054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,4,128,1,float16,fp8,0,0.024906667570273083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,8,128,1,float16,fp8,0,0.025008000433444977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,8,128,1,float16,float16,0,0.02319466571013133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,16,128,1,float16,float16,0,0.018901333212852478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,1,128,1,float16,float16,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,1,128,1,float16,fp8,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,2,128,1,float16,float16,0,0.01711999997496605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,2,128,1,float16,fp8,0,0.0169813334941864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,4,128,1,float16,float16,0,0.017136000096797943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,4,128,1,float16,fp8,0,0.018810667097568512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,8,128,1,float16,float16,0,0.018954666952292126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,16,1,128,1,float16,fp8,0,0.04145599901676178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,16,128,1,float16,float16,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,16,128,1,float16,fp8,0,0.01735466718673706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,1,128,1,float16,float16,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,1,128,1,float16,fp8,0,0.015040000279744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,2,128,1,float16,fp8,0,0.015029333531856537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,4,128,1,float16,float16,0,0.01691199963291486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,4,128,1,float16,fp8,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,4,128,1,float16,float16,0,0.03563733398914337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,8,128,1,float16,float16,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,16,8,128,1,float16,fp8,0,0.0397119993964831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,16,128,1,float16,float16,0,0.015109332899252573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,1,128,1,float16,float16,0,0.015386667102575302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,16,128,1,float16,fp8,0,0.01522133375207583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,16,4,128,1,float16,float16,0,0.02130666623512904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,1,128,1,float16,fp8,0,0.01481066644191742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,2,128,1,float16,float16,0,0.014741333822409311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,2,128,1,float16,fp8,0,0.014864000181357065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,4,128,1,float16,float16,0,0.015184000134468079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,4,128,1,float16,fp8,0,0.014677333335081736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,8,128,1,float16,float16,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,16,8,128,1,float16,fp8,0,0.015018666783968607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,16,128,1,float16,fp8,0,0.021242665747801464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,16,8,128,1,float16,fp8,0,0.019098666807015736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,1,128,1,float16,float16,0,0.1604159971078237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,1,128,1,float16,fp8,0,0.15018133322397867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,2,128,1,float16,float16,0,0.19431465864181519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,2,128,1,float16,float16,0,0.01509333277742068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,4,128,1,float16,float16,0,0.2718506654103597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,16,8,128,1,float16,fp8,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,4,128,1,float16,fp8,0,0.27666133642196655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,16,128,1,float16,float16,0,0.16910932461420694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,8,128,1,float16,float16,0,0.28171734015146893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,1,128,1,float16,float16,0,0.08935999870300293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,8,128,1,float16,fp8,0,0.2945653398831685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,16,128,1,float16,fp8,0,0.19771732886632284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,4,128,1,float16,float16,0,0.14689600467681885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,2,128,1,float16,float16,0,0.10273067156473796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,1,128,1,float16,fp8,0,0.08269866804281871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,8,128,1,float16,fp8,0,0.15523200233777365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,4,128,1,float16,fp8,0,0.14778133233388266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,16,2,128,1,float16,fp8,0,0.18468799193700156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,1,128,1,float16,float16,0,0.049125333627065025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,16,128,1,float16,float16,0,0.0909493366877238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,2,128,1,float16,float16,0,0.055029332637786865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,1,128,1,float16,fp8,0,0.044826666514078774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,4,128,1,float16,float16,0,0.07675733168919881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,2,128,1,float16,fp8,0,0.05173333485921224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,4,128,1,float16,fp8,0,0.06851199766000111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,8,128,1,float16,float16,0,0.08100800216197968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,8,128,1,float16,fp8,0,0.0812853326400121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,16,128,1,float16,float16,0,0.05097599824269613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,16,128,1,float16,fp8,0,0.05298133194446564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,1,128,1,float16,float16,0,0.029493334392706554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,1,128,1,float16,fp8,0,0.028362666567166645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,2,128,1,float16,float16,0,0.03222399950027466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,4,128,1,float16,fp8,0,0.0409706657131513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,4,128,1,float16,float16,0,0.039408000806967415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,2,128,1,float16,fp8,0,0.03233599911133448
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,8,128,1,float16,float16,0,0.03972266614437103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,16,8,128,1,float16,fp8,0,0.041562666495641075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,1,128,1,float16,fp8,0,0.019002666076024372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,16,128,1,float16,fp8,0,0.029738667110602062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,16,128,1,float16,float16,0,0.02513599892457326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,2,128,1,float16,float16,0,0.020938667158285778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,2,128,1,float16,fp8,0,0.021029333273569744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,4,128,1,float16,float16,0,0.023029332359631855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,4,128,1,float16,fp8,0,0.025050667424996693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,8,128,1,float16,float16,0,0.15061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,8,128,1,float16,float16,0,0.023168000082174938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,16,2,128,1,float16,fp8,0,0.09897067149480183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,16,128,1,float16,fp8,0,0.01912533367673556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,16,128,1,float16,float16,0,0.016757333030303318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,1,128,1,float16,fp8,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,2,128,1,float16,fp8,0,0.01479999969402949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,2,128,1,float16,float16,0,0.014912000546852747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,4,128,1,float16,float16,0,0.015402667224407196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,4,128,1,float16,fp8,0,0.017152000218629837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,16,16,128,1,float16,fp8,0,0.1051093339920044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,8,128,1,float16,float16,0,0.016885332763195038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,8,128,1,float16,fp8,0,0.016986666868130367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,1,128,1,float16,fp8,0,0.012847999731699625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,16,128,1,float16,float16,0,0.014741333822409311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,16,128,1,float16,fp8,0,0.01522133375207583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,1,128,1,float16,float16,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,2,128,1,float16,fp8,0,0.012847999731699625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,4,128,1,float16,float16,0,0.01404800017674764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,4,128,1,float16,fp8,0,0.015061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,8,128,1,float16,fp8,0,0.014853333433469137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,8,128,1,float16,float16,0,0.013072000195582708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,1,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,16,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,16,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,1,128,1,float16,fp8,0,0.012005332857370377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,2,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,4,128,1,float16,float16,0,0.012442667037248611
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,2,128,1,float16,fp8,0,0.012986666212479273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,8,128,1,float16,float16,0,0.012896000097195307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,1,128,1,float16,float16,0,0.019274666905403137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,8,128,1,float16,fp8,0,0.012714666624863943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,1,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,1,128,1,float16,fp8,0,0.011296000331640244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,16,128,1,float16,fp8,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,16,128,1,float16,float16,0,0.012719999998807907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,2,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,2,128,1,float16,fp8,0,0.011333333949247995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,4,128,1,float16,float16,0,0.011183999478816986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,4,128,1,float16,fp8,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,16,8,128,1,float16,fp8,0,0.026399999856948853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,8,128,1,float16,float16,0,0.011834666132926941
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,16,1,128,1,float16,float16,0,0.013157332936922709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,1,128,1,float16,float16,0,0.13620266318321228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,1,128,1,float16,fp8,0,0.11983999609947205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,2,128,1,float16,float16,0,0.14776000380516052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,16,2,128,1,float16,float16,0,0.012858666479587555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,2,128,1,float16,fp8,0,0.1365013321240743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,4,128,1,float16,float16,0,0.18982400496800741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,4,128,1,float16,fp8,0,0.18437866369883218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,16,128,1,float16,fp8,0,0.12351999680201213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,16,4,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,16,128,1,float16,float16,0,0.11572266618410747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,8,128,1,float16,float16,0,0.1943626602490743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,1,128,1,float16,float16,0,0.07214933137098949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,1,128,1,float16,fp8,0,0.06406400104363759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,2,128,1,float16,float16,0,0.07859733204046886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,2,128,1,float16,fp8,0,0.07026666899522145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,16,8,128,1,float16,fp8,0,0.012682666381200155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,4,128,1,float16,float16,0,0.09692266583442688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,4,128,1,float16,fp8,0,0.09249066313107808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,8,128,1,float16,float16,0,0.10602133472760518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,1,128,1,float16,float16,0,0.0415040006240209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,16,8,128,1,float16,fp8,0,0.10006399949391682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,16,128,1,float16,fp8,0,0.062133332093556724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,2,128,1,float16,fp8,0,0.040789333482583366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,4,128,1,float16,float16,0,0.05014933149019877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,4,128,1,float16,fp8,0,0.04977599779764811
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,8,128,1,float16,float16,0,0.05162133276462555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,8,128,1,float16,fp8,0,0.051354666550954185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,16,128,1,float16,float16,0,0.031445334355036415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,1,128,1,float16,float16,0,0.025413334369659424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,16,128,1,float16,fp8,0,0.035631999373435974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,2,128,1,float16,float16,0,0.02735466758410136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,1,128,1,float16,fp8,0,0.025285333395004272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,2,128,1,float16,fp8,0,0.027034667630990345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,16,8,128,1,float16,fp8,0,0.19075733423233032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,4,128,1,float16,float16,0,0.029696000119050343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,4,128,1,float16,fp8,0,0.031386665999889374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,8,128,1,float16,float16,0,0.03121600051720937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,16,8,128,1,float16,fp8,0,0.03145600110292435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,16,128,1,float16,float16,0,0.019285333653291065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,16,128,1,float16,fp8,0,0.023178666830062866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,1,128,1,float16,fp8,0,0.01648533344268799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,2,128,1,float16,float16,0,0.01770666614174843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,2,128,1,float16,fp8,0,0.018794666975736618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,4,128,1,float16,float16,0,0.019130667050679524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,16,128,1,float16,float16,0,0.06211199859778086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,1,128,1,float16,fp8,0,0.03755733370780945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,4,128,1,float16,fp8,0,0.01929066702723503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,16,2,128,1,float16,float16,0,0.043749332427978516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,8,128,1,float16,fp8,0,0.02088533341884613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,1,128,1,float16,float16,0,0.012944000462690989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,8,128,1,float16,float16,0,0.01940800001223882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,2,128,1,float16,fp8,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,2,128,1,float16,float16,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,1,128,1,float16,fp8,0,0.012671999633312225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,4,128,1,float16,float16,0,0.014997333288192749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,16,128,1,float16,fp8,0,0.015429332852363586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,16,128,1,float16,float16,0,0.014725333700577417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,4,128,1,float16,fp8,0,0.014848000059525171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,8,128,1,float16,float16,0,0.014762666076421738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,1,128,1,float16,float16,0,0.011328000575304031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,16,8,128,1,float16,fp8,0,0.014752000570297241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,1,128,1,float16,fp8,0,0.012661332885424295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,2,128,1,float16,float16,0,0.011109333485364914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,16,128,1,float16,fp8,0,0.012613333761692047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,16,128,1,float16,float16,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,4,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,4,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,8,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,8,128,1,float16,fp8,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,1,128,1,float16,float16,0,0.011173332730929056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,1,128,1,float16,fp8,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,16,128,1,float16,fp8,0,0.011893333246310553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,2,128,1,float16,fp8,0,0.011498666057984034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,4,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,4,128,1,float16,fp8,0,0.011264000087976456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,8,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,8,128,1,float16,fp8,0,0.011152000476916632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,1,128,1,float16,float16,0,0.011146667102972666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,16,128,1,float16,float16,0,0.010890666395425797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,16,128,1,float16,fp8,0,0.011733333269755045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,2,128,1,float16,float16,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,4,128,1,float16,float16,0,0.011125333607196808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,2,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,4,128,1,float16,fp8,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,16,1,128,1,float16,float16,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,8,128,1,float16,float16,0,0.011359999577204386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,1,128,1,float16,fp8,0,0.10079999764760335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,1,128,1,float16,float16,0,0.12320533394813538
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,2,128,1,float16,float16,0,0.1327839990456899
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,16,2,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,2,128,1,float16,fp8,0,0.10717333356539409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,2,128,1,float16,float16,0,0.01184533288081487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,16,16,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,4,128,1,float16,fp8,0,0.12917866309483847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,1,128,1,float16,fp8,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,8,128,1,float16,float16,0,0.15365866820017496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,1,128,1,float16,float16,0,0.06743999818960826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,16,128,1,float16,fp8,0,0.08299200236797333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,16,128,1,float16,float16,0,0.088837335507075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,8,128,1,float16,fp8,0,0.13995200395584106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,16,8,128,1,float16,fp8,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,1,128,1,float16,fp8,0,0.05612266560395559
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,2,128,1,float16,float16,0,0.07010133564472198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,2,128,1,float16,fp8,0,0.05975466469923655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,4,128,1,float16,float16,0,0.07646933197975159
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,4,128,1,float16,fp8,0,0.0703359991312027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,8,128,1,float16,fp8,0,0.07046933472156525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,16,128,1,float16,fp8,0,0.045552000403404236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,16,8,128,1,float16,float16,0,0.08029333253701527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,1,128,1,float16,fp8,0,0.033413333197434746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,2,128,1,float16,float16,0,0.03961066653331121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,2,128,1,float16,fp8,0,0.035605333745479584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,4,128,1,float16,float16,0,0.04369066655635834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,4,128,1,float16,fp8,0,0.03947199881076813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,8,128,1,float16,float16,0,0.043568000197410583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,8,128,1,float16,fp8,0,0.0400693342089653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,1,128,1,float16,float16,0,0.02333866556485494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,16,128,1,float16,fp8,0,0.02738133321205775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,16,128,1,float16,float16,0,0.02716800073782603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,2,128,1,float16,float16,0,0.02513066679239273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,4,128,1,float16,float16,0,0.025978667040665943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,2,128,1,float16,fp8,0,0.023103999594847362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,16,4,128,1,float16,float16,0,0.15396799643834433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,4,128,1,float16,fp8,0,0.02499199906984965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,8,128,1,float16,float16,0,0.025519999365011852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,8,128,1,float16,fp8,0,0.02518400053183238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,16,128,1,float16,fp8,0,0.018277333428462345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,1,128,1,float16,float16,0,0.014965333044528961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,16,128,1,float16,float16,0,0.017103999853134155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,1,128,1,float16,fp8,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,2,128,1,float16,fp8,0,0.014874666929244995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,4,128,1,float16,float16,0,0.017114666601022083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,4,128,1,float16,fp8,0,0.017077332983414333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,8,128,1,float16,float16,0,0.01722666621208191
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,8,128,1,float16,fp8,0,0.017071999609470367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,16,128,1,float16,float16,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,1,128,1,float16,float16,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,16,128,1,float16,float16,0,0.04555733501911163
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,1,128,1,float16,fp8,0,0.013002666334311167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,2,128,1,float16,float16,0,0.013002666334311167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,2,128,1,float16,fp8,0,0.012730666746695837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,4,128,1,float16,float16,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,4,128,1,float16,fp8,0,0.012986666212479273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,8,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,8,128,1,float16,fp8,0,0.012650666137536367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,1,128,1,float16,float16,0,0.01119999960064888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,1,128,1,float16,fp8,0,0.011760000139474869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,16,128,1,float16,float16,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,16,128,1,float16,fp8,0,0.011882666498422623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,16,1,128,1,float16,fp8,0,0.021525333325068157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,2,128,1,float16,float16,0,0.010816000401973724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,2,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,4,128,1,float16,float16,0,0.010816000401973724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,4,128,1,float16,fp8,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,8,128,1,float16,float16,0,0.011141333729028702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,16,8,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,1,128,1,float16,float16,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,1,128,1,float16,fp8,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,16,128,1,float16,fp8,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,2,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,16,128,1,float16,float16,0,0.010762666662534079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,2,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,4,128,1,float16,float16,0,0.010768000036478043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,8,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,8,128,1,float16,fp8,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,16,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,16,128,1,float16,float16,0,0.010821333775917688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,16,2,128,1,float16,float16,0,0.01600533351302147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,2,128,1,float16,float16,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,2,128,1,float16,fp8,0,0.010842667271693548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,4,128,1,float16,float16,0,0.01109333336353302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,4,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,8,128,1,float16,float16,0,0.010768000036478043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,8,128,1,float16,fp8,0,0.01110400011142095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,16,16,128,1,float16,fp8,0,0.012784000486135483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,16,1,128,1,float16,float16,0,0.038959999879201256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,1,128,1,float16,float16,0,0.1163093348344167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,2,128,1,float16,float16,0,0.11852266391118367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,1,128,1,float16,fp8,0,0.09529067079226176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,2,128,1,float16,fp8,0,0.09897599617640178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,4,128,1,float16,float16,0,0.12639466921488443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,4,128,1,float16,fp8,0,0.10706667105356853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,16,128,1,float16,float16,0,0.07635733485221863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,16,4,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,8,128,1,float16,fp8,0,0.11219200491905212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,16,8,128,1,float16,float16,0,0.13390933473904928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,16,128,1,float16,fp8,0,0.06472533444563548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,1,128,1,float16,float16,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,1,128,1,float16,float16,0,0.06419200201829274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,1,128,1,float16,fp8,0,0.05403199791908264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,2,128,1,float16,float16,0,0.06622399886449178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,2,128,1,float16,fp8,0,0.05615466833114624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,4,128,1,float16,float16,0,0.06865600248177846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,4,128,1,float16,fp8,0,0.059893334905306496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,8,128,1,float16,float16,0,0.07005333403746287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,16,8,128,1,float16,fp8,0,0.059989333152770996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,16,128,1,float16,float16,0,0.04112533231576284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,1,128,1,float16,fp8,0,0.031301334500312805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,2,128,1,float16,float16,0,0.03773866593837738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,16,128,1,float16,fp8,0,0.03760000069936117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,2,128,1,float16,fp8,0,0.0336053321758906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,4,128,1,float16,float16,0,0.0393653338154157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,4,128,1,float16,fp8,0,0.035530666510264076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,8,128,1,float16,fp8,0,0.035642666121323906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,1,128,1,float16,float16,0,0.023183998962243397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,16,128,1,float16,fp8,0,0.024688000480333965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,1,128,1,float16,fp8,0,0.021141332884629566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,2,128,1,float16,fp8,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,16,128,1,float16,float16,0,0.025231999655564625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,4,128,1,float16,float16,0,0.02521066615978877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,4,128,1,float16,fp8,0,0.02316266546646754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,8,128,1,float16,float16,0,0.02498133232196172
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,16,128,1,float16,float16,0,0.01700266698996226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,1,128,1,float16,float16,0,0.016117333124081295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,2,128,1,float16,float16,0,0.016666666915019352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,2,128,1,float16,fp8,0,0.014991999914248785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,16,1,128,1,float16,fp8,0,0.010661333799362183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,4,128,1,float16,float16,0,0.017024000485738117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,4,128,1,float16,fp8,0,0.01516266663869222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,8,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,8,128,1,float16,fp8,0,0.014970666418472925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,1,128,1,float16,float16,0,0.03729599962631861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,16,128,1,float16,float16,0,0.012613333761692047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,16,128,1,float16,fp8,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,16,8,128,1,float16,float16,0,0.0397119993964831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,1,128,1,float16,float16,0,0.012560000022252401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,1,128,1,float16,fp8,0,0.012602667013804117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,2,128,1,float16,fp8,0,0.012682666381200155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,2,128,1,float16,float16,0,0.012810666114091873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,2,128,1,float16,float16,0,0.02330133318901062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,4,128,1,float16,float16,0,0.012703999876976013
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,4,128,1,float16,fp8,0,0.012719999998807907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,8,128,1,float16,fp8,0,0.01267733300725619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,16,8,128,1,float16,float16,0,0.012874666601419449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,16,8,128,1,float16,fp8,0,0.02345066765944163
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,1,128,1,float16,float16,0,0.011503999431928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,16,128,1,float16,fp8,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,16,128,1,float16,float16,0,0.011770666887362799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,2,128,1,float16,float16,0,0.011424000064531961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,2,128,1,float16,fp8,0,0.011658667276302973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,4,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,1,128,1,float16,fp8,0,0.015125333021084467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,4,128,1,float16,fp8,0,0.011114666859308878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,8,128,1,float16,float16,0,0.011285333583752314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,8,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,1,128,1,float16,float16,0,0.010746666540702185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,16,128,1,float16,fp8,0,0.01080000028014183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,16,128,1,float16,float16,0,0.010687999427318573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,2,128,1,float16,float16,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,1,128,1,float16,fp8,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,4,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,4,128,1,float16,fp8,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,8,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,8,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,16,128,1,float16,fp8,0,0.010842667271693548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,1,128,1,float16,float16,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,1,128,1,float16,fp8,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,2,128,1,float16,float16,0,0.010618666807810465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,2,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,4,128,1,float16,float16,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,4,128,1,float16,fp8,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,8,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,8,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,1,128,1,float16,float16,0,0.012389333297808966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,2,128,1,float16,float16,0,0.013199999928474426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,1,128,1,float16,fp8,0,0.012042666474978128
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,16,16,128,1,float16,fp8,0,0.016410666207472484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,16,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,2,128,1,float16,fp8,0,0.013130666067202887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,4,128,1,float16,float16,0,0.018933333456516266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,4,128,1,float16,fp8,0,0.016906666258970898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,16,2,128,1,float16,fp8,0,0.01080000028014183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,16,128,1,float16,float16,0,0.02812800059715907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,8,128,1,float16,fp8,0,0.025055999557177227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,16,8,128,1,float16,float16,0,0.028853334486484528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,1,128,1,float16,float16,0,0.01055466632048289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,16,128,1,float16,fp8,0,0.025386666258176167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,16,16,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,1,128,1,float16,fp8,0,0.010666667173306147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,2,128,1,float16,float16,0,0.010821333775917688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,2,128,1,float16,fp8,0,0.010698666175206503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,4,128,1,float16,float16,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,4,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,1,128,1,float16,float16,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,8,128,1,float16,float16,0,0.01889066646496455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,16,8,128,1,float16,fp8,0,0.017210666090250015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,1,128,1,float16,fp8,0,0.009952000031868616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,16,128,1,float16,fp8,0,0.016917333006858826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,2,128,1,float16,float16,0,0.00979200005531311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,2,128,1,float16,fp8,0,0.009226666763424873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,4,128,1,float16,float16,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,8,128,1,float16,float16,0,0.013130666067202887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,1,128,1,float16,float16,0,0.009066666786869368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,1,128,1,float16,fp8,0,0.009050666665037474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,16,128,1,float16,fp8,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,16,128,1,float16,float16,0,0.013157332936922709
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,2,128,1,float16,float16,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,2,128,1,float16,fp8,0,0.008805333326260248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,4,128,1,float16,float16,0,0.00902399979531765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,4,128,1,float16,fp8,0,0.010234666367371878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,8,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,16,8,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,1,128,1,float16,float16,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,16,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,1,128,1,float16,fp8,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,16,128,1,float16,fp8,0,0.010672000547250112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,2,128,1,float16,fp8,0,0.00919999989370505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,4,128,1,float16,float16,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,4,128,1,float16,fp8,0,0.008997333546479544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,8,128,1,float16,fp8,0,0.009957333405812582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,8,128,1,float16,float16,0,0.009765333185593287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,1,128,1,float16,float16,0,0.008858666444818178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,1,128,1,float16,fp8,0,0.008762666955590248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,16,128,1,float16,float16,0,0.009109333157539368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,2,128,1,float16,float16,0,0.008853333070874214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,2,128,1,float16,fp8,0,0.009066666786869368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,4,128,1,float16,float16,0,0.00892800030608972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,4,128,1,float16,fp8,0,0.008933333059151968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,8,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,16,128,1,float16,float16,0,0.00915733352303505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,16,128,1,float16,float16,0,0.018986667195955913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,16,128,1,float16,fp8,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,4,128,1,float16,fp8,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,1,128,1,float16,fp8,0,0.008863999818762144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,2,128,1,float16,float16,0,0.008639999975760778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,16,8,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,2,128,1,float16,fp8,0,0.008992000172535578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,4,128,1,float16,float16,0,0.008629333227872849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,4,128,1,float16,fp8,0,0.00878399983048439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,8,128,1,float16,float16,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,8,128,1,float16,fp8,0,0.009045333291093508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,1,128,1,float16,float16,0,0.008629333227872849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,16,128,1,float16,float16,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,16,128,1,float16,fp8,0,0.008901333436369896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,16,2,128,1,float16,float16,0,0.008832000195980072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,2,128,1,float16,fp8,0,0.00902399979531765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,2,128,1,float16,float16,0,0.008672000219424566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,4,128,1,float16,float16,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,4,128,1,float16,fp8,0,0.008949333180983862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,8,128,1,float16,float16,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,16,128,1,float16,fp8,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,16,128,1,float16,float16,0,0.009056000038981438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,16,128,1,float16,fp8,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,1,128,1,float16,fp8,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,2,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,2,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,16,8,128,1,float16,fp8,0,0.008816000074148178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,4,128,1,float16,float16,0,0.00891733355820179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,4,128,1,float16,fp8,0,0.00855466661353906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,8,128,1,float16,float16,0,0.008842666943868002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,8,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,16,1,128,1,float16,float16,0,0.008821333448092142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,1,128,1,float16,fp8,0,0.009029333169261614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,1,128,1,float16,fp8,0,4.307344118754069
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,2,128,1,float16,fp8,0,4.440954526265462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,16,8,128,1,float16,fp8,0,0.008997333546479544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,1,128,1,float16,float16,0,5.470773061116536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,2,128,1,float16,float16,0,5.467098871866862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,16,1,128,1,float16,float16,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,4,128,1,float16,float16,0,5.846922556559245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,1,128,1,float16,float16,0,2.697258631388346
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,1,128,1,float16,fp8,0,2.7021331787109375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,12,128,1,float16,fp8,0,2.6476640701293945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,2,128,1,float16,fp8,0,2.293509324391683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,2,128,1,float16,float16,0,2.9678773880004883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,4,128,1,float16,float16,0,2.9274400075276694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,12,128,1,float16,fp8,0,1.387781302134196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,1,128,1,float16,float16,0,1.3902826309204102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,4,128,1,float16,fp8,0,2.6958185831705728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,1,128,1,float16,fp8,0,1.405413309733073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,2,128,1,float16,fp8,0,1.211722691853841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,4,128,1,float16,float16,0,1.5058132807413738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,4,128,1,float16,fp8,0,1.3581867218017578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,12,128,1,float16,float16,0,0.8373920122782389
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,12,128,1,float16,fp8,0,0.757962703704834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,1,128,1,float16,float16,0,0.7489706675211588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,1,128,1,float16,fp8,0,0.6741066773732504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,2,128,1,float16,float16,0,0.768064022064209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,2,128,1,float16,fp8,0,0.6657600005467733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,4,128,1,float16,fp8,0,0.7309280236562093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,12,4,128,1,float16,float16,0,0.8158880074818929
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,12,12,128,1,float16,float16,0,3.0287040074666343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,12,4,128,1,float16,fp8,0,4.8345387776692705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,1,128,1,float16,float16,0,3.0372800827026367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,1,128,1,float16,fp8,0,2.5350613594055176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,12,128,1,float16,float16,0,1.558362642923991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,2,128,1,float16,fp8,0,2.7319199244181314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,2,128,1,float16,float16,0,3.158624013264974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,4,128,1,float16,float16,0,3.404325485229492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,12,2,128,1,float16,float16,0,1.42959992090861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,12,4,128,1,float16,fp8,0,2.9352426528930664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,1,128,1,float16,float16,0,1.5596213340759277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,12,128,1,float16,fp8,0,1.6409707069396973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,1,128,1,float16,fp8,0,1.319050629933675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,12,128,1,float16,float16,0,1.8315307299296062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,2,128,1,float16,float16,0,1.670698642730713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,2,128,1,float16,fp8,0,1.472314675649007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,4,128,1,float16,fp8,0,1.518437385559082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,12,4,128,1,float16,float16,0,1.7190826733907063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,1,128,1,float16,fp8,0,0.7814880212148031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,12,128,1,float16,fp8,0,0.8705066839853922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,12,128,1,float16,float16,0,0.9395093123118082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,2,128,1,float16,float16,0,0.8797439734141032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,2,128,1,float16,fp8,0,0.7623733679453532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,1,128,1,float16,fp8,0,0.46323200066884357
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,4,128,1,float16,fp8,0,0.8198239803314209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,1,128,1,float16,float16,0,0.44868799050649005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,12,128,1,float16,fp8,0,0.48666131496429443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,12,128,1,float16,float16,0,0.5301706790924072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,2,128,1,float16,float16,0,0.4689013163248698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,2,128,1,float16,fp8,0,0.4410239855448405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,4,128,1,float16,float16,0,0.49861868222554523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,12,4,128,1,float16,fp8,0,0.44280000527699787
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,1,128,1,float16,float16,0,2.1711360613505044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,1,128,1,float16,fp8,0,1.9057812690734863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,2,128,1,float16,fp8,0,1.9719947179158528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,2,128,1,float16,float16,0,2.25436798731486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,4,128,1,float16,fp8,0,2.1502505938212075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,12,4,128,1,float16,float16,0,2.4345812797546387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,1,128,1,float16,fp8,0,0.9600000381469727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,1,128,1,float16,float16,0,1.1262239615122478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,12,128,1,float16,fp8,0,1.2253226439158122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,12,128,1,float16,float16,0,1.3480106989542644
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,2,128,1,float16,float16,0,1.1499359607696533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,1,128,1,float16,float16,0,0.823743979136149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,2,128,1,float16,fp8,0,1.0850880146026611
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,4,128,1,float16,float16,0,1.2677653630574544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,12,4,128,1,float16,fp8,0,1.1225173473358154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,12,128,1,float16,float16,0,0.7036746342976888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,1,128,1,float16,float16,0,0.5941226482391357
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,1,128,1,float16,fp8,0,0.5406293471654257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,12,4,128,1,float16,float16,0,0.9103786945343018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,2,128,1,float16,fp8,0,0.5465759833653768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,4,128,1,float16,float16,0,0.6699519952138265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,4,128,1,float16,fp8,0,0.608245333035787
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,1,128,1,float16,float16,0,0.3299786647160848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,12,128,1,float16,float16,0,0.39508267243703205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,1,128,1,float16,fp8,0,0.3128533363342285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,2,128,1,float16,float16,0,0.33793067932128906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,2,128,1,float16,fp8,0,0.3126346667607625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,4,128,1,float16,float16,0,0.36955734093983966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,4,128,1,float16,fp8,0,0.34224534034729004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,1,128,1,float16,float16,0,2.8301121393839517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,1,128,1,float16,fp8,0,2.3723732630411782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,2,128,1,float16,fp8,0,2.518069267272949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,2,128,1,float16,float16,0,2.9500373204549155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,12,128,1,float16,fp8,0,0.6766239802042643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,1,128,1,float16,float16,0,1.4540212949117024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,12,2,128,1,float16,float16,0,0.6203200022379557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,12,128,1,float16,float16,0,1.7589759826660156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,12,128,1,float16,fp8,0,1.65283203125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,12,12,128,1,float16,fp8,0,0.37301866213480633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,4,128,1,float16,fp8,0,3.1161012649536133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,12,4,128,1,float16,float16,0,3.2487147649129233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,2,128,1,float16,fp8,0,1.2992160320281982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,2,128,1,float16,float16,0,1.5198240280151367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,4,128,1,float16,float16,0,1.6578933397928874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,4,128,1,float16,fp8,0,1.48688538869222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,12,128,1,float16,float16,0,0.9104320208231608
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,1,128,1,float16,float16,0,0.7408320109049479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,12,128,1,float16,fp8,0,0.8797067006429037
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,1,128,1,float16,fp8,0,0.6469173431396484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,2,128,1,float16,float16,0,0.7902719974517822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,2,128,1,float16,fp8,0,0.6905386447906494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,4,128,1,float16,float16,0,0.858949343363444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,1,128,1,float16,float16,0,0.4033706585566203
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,1,128,1,float16,fp8,0,0.37145066261291504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,12,4,128,1,float16,fp8,0,0.782416025797526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,12,128,1,float16,float16,0,0.5126453240712484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,12,128,1,float16,fp8,0,0.47440532843271893
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,2,128,1,float16,float16,0,0.4175306558609009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,2,128,1,float16,fp8,0,0.3781813383102417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,4,128,1,float16,float16,0,0.45931732654571533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,1,128,1,float16,float16,0,0.2339893380800883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,1,128,1,float16,fp8,0,0.21712533632914224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,12,128,1,float16,float16,0,0.2744693358739217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,2,128,1,float16,float16,0,0.2396799921989441
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,2,128,1,float16,fp8,0,0.22396800915400186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,4,128,1,float16,float16,0,0.2566773295402527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,4,128,1,float16,fp8,0,0.2406346599260966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,1,128,1,float16,fp8,0,1.431349277496338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,1,128,1,float16,float16,0,1.6801546414693196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,2,128,1,float16,float16,0,1.7679840723673503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,12,1,128,1,float16,fp8,0,1.2253599961598713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,4,128,1,float16,float16,0,2.005093256632487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,1,128,1,float16,float16,0,0.8511093457539877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,12,128,1,float16,float16,0,1.1019946734110515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,4,128,1,float16,fp8,0,1.8262826601664226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,1,128,1,float16,fp8,0,0.7448853651682535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,12,128,1,float16,fp8,0,1.1175626913706462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,2,128,1,float16,float16,0,0.898751974105835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,2,128,1,float16,fp8,0,0.9013866583506266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,4,128,1,float16,float16,0,1.0244373480478923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,12,4,128,1,float16,fp8,0,0.9778133233388265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,12,128,1,float16,float16,0,0.5755946636199951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,1,128,1,float16,float16,0,0.4552533229192098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,12,4,128,1,float16,fp8,0,0.435205340385437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,2,128,1,float16,float16,0,0.47224001089731854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,12,128,1,float16,fp8,0,0.5690026680628458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,12,12,128,1,float16,fp8,0,0.2807520031929016
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,2,128,1,float16,fp8,0,0.43587732315063477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,1,128,1,float16,float16,0,0.2523253361384074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,12,128,1,float16,fp8,0,0.3148426612218221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,2,128,1,float16,float16,0,0.26293333371480304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,2,128,1,float16,fp8,0,0.24362132946650186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,4,128,1,float16,float16,0,0.2967039942741394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,12,2,128,1,float16,fp8,0,1.5373493830362956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,12,128,1,float16,float16,0,0.18847467501958212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,12,128,1,float16,fp8,0,0.18134933710098267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,1,128,1,float16,float16,0,0.1575146714846293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,1,128,1,float16,fp8,0,0.1486186683177948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,2,128,1,float16,float16,0,0.16170133153597513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,2,128,1,float16,fp8,0,0.1536960005760193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,4,128,1,float16,float16,0,0.17101333538691202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,12,4,128,1,float16,fp8,0,0.16676799456278482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,1,128,1,float16,fp8,0,0.4068373441696167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,1,128,1,float16,float16,0,1.5791254043579102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,4,128,1,float16,fp8,0,0.5055626630783081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,1,128,1,float16,fp8,0,1.391152064005534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,12,128,1,float16,float16,0,0.31409066915512085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,2,128,1,float16,float16,0,1.709770679473877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,4,128,1,float16,fp8,0,0.27746667464574176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,2,128,1,float16,fp8,0,1.5290506680806477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,4,128,1,float16,float16,0,2.056842645009359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,12,4,128,1,float16,fp8,0,1.9152053197224934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,1,128,1,float16,float16,0,0.8060426712036133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,12,128,1,float16,float16,0,1.135632038116455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,1,128,1,float16,fp8,0,0.7178293069203695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,12,128,1,float16,fp8,0,1.1453226407368977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,2,128,1,float16,float16,0,0.8686453501383463
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,2,128,1,float16,fp8,0,0.7913546562194824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,4,128,1,float16,float16,0,1.0367413361867268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,12,4,128,1,float16,fp8,0,0.9856746991475424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,12,128,1,float16,float16,0,0.5806133349736532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,1,128,1,float16,float16,0,0.41945600509643555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,12,128,1,float16,fp8,0,0.5955946842829386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,1,128,1,float16,fp8,0,0.3850613435109456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,2,128,1,float16,float16,0,0.45349331696828205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,2,128,1,float16,fp8,0,0.42316265900929767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,4,128,1,float16,float16,0,0.5324480136235555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,12,1,128,1,float16,fp8,0,0.2325119972229004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,12,128,1,float16,float16,0,0.31040533383687335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,1,128,1,float16,float16,0,0.23000532388687134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,1,128,1,float16,fp8,0,0.2135146657625834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,12,128,1,float16,fp8,0,0.3203199903170268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,2,128,1,float16,float16,0,0.24518932898839316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,2,128,1,float16,fp8,0,0.23017066717147827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,4,128,1,float16,float16,0,0.29044799009958905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,12,128,1,float16,float16,0,0.1779680053393046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,1,128,1,float16,fp8,0,0.12854933738708496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,2,128,1,float16,float16,0,0.14267733693122864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,2,128,1,float16,fp8,0,0.13391466935475668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,4,128,1,float16,float16,0,0.16053866346677145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,12,128,1,float16,float16,0,0.10320533315340678
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,12,128,1,float16,fp8,0,0.10470400253931682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,1,128,1,float16,float16,0,0.09040533502896626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,1,128,1,float16,fp8,0,0.08434666196505229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,2,128,1,float16,float16,0,0.0925546685854594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,2,128,1,float16,fp8,0,0.08847467104593913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,4,128,1,float16,float16,0,0.0983679989973704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,12,4,128,1,float16,fp8,0,0.5139146645863851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,12,4,128,1,float16,fp8,0,0.09706133604049683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,1,128,1,float16,float16,0,0.9715092976888021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,12,4,128,1,float16,fp8,0,0.28306132555007935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,2,128,1,float16,float16,0,1.0597600142161052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,1,128,1,float16,fp8,0,0.8642880121866862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,1,128,1,float16,float16,0,0.13616533080736795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,12,128,1,float16,fp8,0,0.179967999458313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,2,128,1,float16,fp8,0,0.9777226448059082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,1,128,1,float16,float16,0,0.49861868222554523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,12,4,128,1,float16,fp8,0,0.15154666701952615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,12,128,1,float16,fp8,0,0.7702933152516683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,4,128,1,float16,fp8,0,1.2591893672943115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,12,128,1,float16,float16,0,0.7377706368764242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,1,128,1,float16,fp8,0,0.45451732476552326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,2,128,1,float16,float16,0,0.5414026578267416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,2,128,1,float16,fp8,0,0.5091520150502523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,4,128,1,float16,fp8,0,0.6511040131251017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,12,4,128,1,float16,float16,0,0.6698986689249674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,12,128,1,float16,float16,0,0.38486401240030926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,1,128,1,float16,float16,0,0.2656586567560832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,12,128,1,float16,fp8,0,0.4050453503926595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,2,128,1,float16,float16,0,0.28773866097132367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,4,128,1,float16,float16,0,0.34939734141031903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,4,128,1,float16,fp8,0,0.3450933297475179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,12,128,1,float16,float16,0,0.21106133858362833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,1,128,1,float16,float16,0,0.14855999747912088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,12,4,128,1,float16,float16,0,0.5436480045318604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,12,128,1,float16,fp8,0,0.21998933951059976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,2,128,1,float16,float16,0,0.1625333329041799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,2,128,1,float16,fp8,0,0.15016000469525656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,4,128,1,float16,float16,0,0.19293866554896036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,1,128,1,float16,float16,0,0.0943999985853831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,12,128,1,float16,fp8,0,0.12099732955296834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,12,128,1,float16,float16,0,0.1252959966659546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,1,128,1,float16,fp8,0,0.08967999617258708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,2,128,1,float16,float16,0,0.09773332873980205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,2,128,1,float16,fp8,0,0.09542399644851685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,4,128,1,float16,float16,0,0.1076639990011851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,12,4,128,1,float16,fp8,0,0.10777067144711812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,12,128,1,float16,fp8,0,0.06833600004514058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,1,128,1,float16,float16,0,0.05745600163936615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,12,4,128,1,float16,float16,0,1.3071040312449138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,1,128,1,float16,fp8,0,0.053717335065205894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,2,128,1,float16,float16,0,0.05927466849486033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,2,128,1,float16,fp8,0,0.057162667314211525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,4,128,1,float16,float16,0,0.06331733365853627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,1,128,1,float16,fp8,0,0.24662399291992188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,12,2,128,1,float16,fp8,0,0.27480532725652057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,1,128,1,float16,float16,0,0.9668906529744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,1,128,1,float16,fp8,0,0.8935786883036295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,1,128,1,float16,fp8,0,0.14070399602254233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,2,128,1,float16,float16,0,1.092293341954549
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,12,4,128,1,float16,fp8,0,0.18269866704940796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,2,128,1,float16,fp8,0,1.033354679743449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,4,128,1,float16,float16,0,1.4336800575256348
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,12,4,128,1,float16,fp8,0,1.4138026237487793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,1,128,1,float16,float16,0,0.4980800151824951
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,12,128,1,float16,float16,0,0.8247413635253906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,1,128,1,float16,fp8,0,0.4668960173924764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,2,128,1,float16,float16,0,0.5516320069630941
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,12,128,1,float16,float16,0,0.06585599978764851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,2,128,1,float16,fp8,0,0.5362026691436768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,12,128,1,float16,fp8,0,0.8819999694824219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,4,128,1,float16,float16,0,0.7271146774291992
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,12,4,128,1,float16,fp8,0,0.06420266628265381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,12,128,1,float16,fp8,0,0.45740799109141034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,1,128,1,float16,fp8,0,0.24849067131678262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,2,128,1,float16,float16,0,0.29224000374476117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,1,128,1,float16,float16,0,0.2640106678009033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,12,4,128,1,float16,fp8,0,0.7195573647816976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,12,128,1,float16,float16,0,0.4233440160751343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,2,128,1,float16,fp8,0,0.28147733211517334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,4,128,1,float16,float16,0,0.37887998421986896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,1,128,1,float16,float16,0,0.14645333091417947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,12,4,128,1,float16,fp8,0,0.37671999136606854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,1,128,1,float16,fp8,0,0.13596799969673157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,12,128,1,float16,fp8,0,0.24447999397913614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,2,128,1,float16,float16,0,0.164682666460673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,12,128,1,float16,float16,0,0.2273706595102946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,4,128,1,float16,float16,0,0.20516266425450644
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,2,128,1,float16,fp8,0,0.1521813372770945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,12,4,128,1,float16,fp8,0,0.20541866620381674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,1,128,1,float16,float16,0,0.08661333719889323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,12,128,1,float16,fp8,0,0.13421866297721863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,2,128,1,float16,fp8,0,0.08959466218948364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,12,128,1,float16,float16,0,0.12865599989891052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,4,128,1,float16,fp8,0,0.10764799515406291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,4,128,1,float16,float16,0,0.1083519955476125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,1,128,1,float16,float16,0,0.0589279979467392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,1,128,1,float16,fp8,0,0.055888002117474876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,12,128,1,float16,fp8,0,0.07553599774837494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,12,128,1,float16,float16,0,0.07092800239721934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,2,128,1,float16,fp8,0,0.05996266504128774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,4,128,1,float16,fp8,0,0.06844800213972728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,12,128,1,float16,float16,0,0.047781333327293396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,1,128,1,float16,float16,0,0.04161600023508072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,12,128,1,float16,fp8,0,0.0496373325586319
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,1,128,1,float16,fp8,0,0.03961066653331121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,2,128,1,float16,float16,0,0.04214400053024292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,2,128,1,float16,fp8,0,0.041589332123597465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,4,128,1,float16,float16,0,0.04573333263397217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,12,4,128,1,float16,fp8,0,0.045791998505592346
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,1,128,1,float16,float16,0,0.6182346741358439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,1,128,1,float16,fp8,0,0.5791413386662801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,2,128,1,float16,float16,0,0.7112639745076498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,2,128,1,float16,fp8,0,0.6852959791819254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,4,128,1,float16,float16,0,0.9620800018310547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,12,4,128,1,float16,fp8,0,0.9674560228983561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,2,128,1,float16,float16,0,0.09195733070373535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,12,128,1,float16,float16,0,0.563093344370524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,1,128,1,float16,float16,0,0.3267040054003398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,1,128,1,float16,fp8,0,0.30665600299835205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,2,128,1,float16,float16,0,0.06002666552861532
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,2,128,1,float16,float16,0,0.3670399983723958
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,12,128,1,float16,fp8,0,0.6181973218917847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,12,4,128,1,float16,float16,0,0.06731200218200684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,2,128,1,float16,fp8,0,0.3624426523844401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,4,128,1,float16,fp8,0,0.4983946482340495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,12,4,128,1,float16,float16,0,0.4982293446858724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,12,128,1,float16,float16,0,0.2941439946492513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,1,128,1,float16,float16,0,0.17893334229787192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,2,128,1,float16,float16,0,0.2021226684252421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,12,128,1,float16,fp8,0,0.322437326113383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,2,128,1,float16,fp8,0,0.1953279972076416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,4,128,1,float16,float16,0,0.26445867617925006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,4,128,1,float16,fp8,0,0.2662293314933777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,1,128,1,float16,float16,0,0.09971200426419576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,12,128,1,float16,float16,0,0.16051733493804932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,1,128,1,float16,fp8,0,0.09467732906341553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,2,128,1,float16,float16,0,0.1120693286259969
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,12,128,1,float16,fp8,0,0.17363200585047403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,2,128,1,float16,fp8,0,0.10357866684595744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,4,128,1,float16,float16,0,0.14442132910092673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,12,4,128,1,float16,fp8,0,0.1378720005353292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,1,128,1,float16,float16,0,0.06171200176080068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,1,128,1,float16,fp8,0,0.0594400018453598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,2,128,1,float16,float16,0,0.06607999900976817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,12,128,1,float16,float16,0,0.09294933080673218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,12,128,1,float16,fp8,0,0.09231999516487122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,2,128,1,float16,fp8,0,0.06439466774463654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,4,128,1,float16,fp8,0,0.07682133217652638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,12,4,128,1,float16,float16,0,0.07576533158620198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,1,128,1,float16,float16,0,0.03741333385308584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,1,128,1,float16,fp8,0,0.03555200000603994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,12,128,1,float16,fp8,0,0.051557332277297974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,2,128,1,float16,float16,0,0.03965333352486292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,2,128,1,float16,fp8,0,0.03961066653331121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,12,128,1,float16,float16,0,0.04582933088143667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,4,128,1,float16,float16,0,0.04424533247947693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,1,128,1,float16,float16,0,0.03346133232116699
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,1,128,1,float16,fp8,0,0.03323200096686681
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,12,1,128,1,float16,fp8,0,0.08341333270072937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,2,128,1,float16,fp8,0,0.03515200068553289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,4,128,1,float16,float16,0,0.03846933444341024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,4,128,1,float16,fp8,0,0.03760000069936117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,1,128,1,float16,float16,0,0.6687680085500082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,12,1,128,1,float16,fp8,0,0.16927999258041382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,1,128,1,float16,fp8,0,0.638159990310669
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,2,128,1,float16,float16,0,0.7847786744435629
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,2,128,1,float16,fp8,0,0.785535971323649
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,4,128,1,float16,float16,0,1.1431199709574382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,12,4,128,1,float16,fp8,0,1.15064533551534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,12,4,128,1,float16,fp8,0,0.0458186666170756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,12,128,1,float16,float16,0,0.03763733307520548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,2,128,1,float16,float16,0,0.03542399903138479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,12,128,1,float16,float16,0,0.668170690536499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,1,128,1,float16,fp8,0,0.33389333883921307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,2,128,1,float16,fp8,0,0.40249598026275635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,2,128,1,float16,float16,0,0.4065440098444621
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,1,128,1,float16,float16,0,0.34858667850494385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,4,128,1,float16,float16,0,0.5855413277943929
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,4,128,1,float16,fp8,0,0.5902666648228964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,12,12,128,1,float16,fp8,0,0.751530647277832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,1,128,1,float16,float16,0,0.1881386637687683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,12,128,1,float16,fp8,0,0.3869440158208211
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,12,128,1,float16,float16,0,0.34565333525339764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,2,128,1,float16,fp8,0,0.215013325214386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,2,128,1,float16,float16,0,0.21990400552749634
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,4,128,1,float16,float16,0,0.3014986713727315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,12,128,1,float16,float16,0,0.18431466817855835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,4,128,1,float16,fp8,0,0.3069653312365214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,2,128,1,float16,float16,0,0.1200213332970937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,12,128,1,float16,fp8,0,0.20497065782546997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,1,128,1,float16,fp8,0,0.0981280008951823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,2,128,1,float16,fp8,0,0.11549866199493408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,4,128,1,float16,float16,0,0.16461867094039917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,12,12,128,1,float16,fp8,0,0.040181333820025124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,4,128,1,float16,fp8,0,0.16325333714485168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,1,128,1,float16,float16,0,0.06113600234190623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,12,128,1,float16,float16,0,0.10380799571673076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,1,128,1,float16,fp8,0,0.060191998879114784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,2,128,1,float16,float16,0,0.06821866830190022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,2,128,1,float16,fp8,0,0.06630933284759521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,4,128,1,float16,float16,0,0.08272533118724823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,1,128,1,float16,float16,0,0.04144000013669332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,4,128,1,float16,fp8,0,0.08431999882062276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,12,128,1,float16,float16,0,0.053616002202034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,1,128,1,float16,fp8,0,0.04005866746107737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,2,128,1,float16,float16,0,0.044490665197372437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,2,128,1,float16,fp8,0,0.044079999128977455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,4,128,1,float16,float16,0,0.05087999999523163
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,4,128,1,float16,fp8,0,0.05310933291912079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,12,128,1,float16,float16,0,0.033957332372665405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,1,128,1,float16,float16,0,0.029264000554879505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,2,128,1,float16,float16,0,0.02959466725587845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,2,128,1,float16,fp8,0,0.030245333909988403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,4,128,1,float16,float16,0,0.033226666351159416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,4,128,1,float16,fp8,0,0.03387200087308884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,12,128,1,float16,float16,0,0.029418667157491047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,12,1,128,1,float16,fp8,0,0.17860267559687296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,1,128,1,float16,float16,0,0.02735466758410136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,1,128,1,float16,fp8,0,0.025349333882331848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,2,128,1,float16,float16,0,0.02734400083621343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,12,1,128,1,float16,float16,0,0.10751466949780782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,2,128,1,float16,fp8,0,0.027376001079877216
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,4,128,1,float16,float16,0,0.02906133234500885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,4,128,1,float16,fp8,0,0.029370665550231934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,12,12,128,1,float16,fp8,0,0.11162666479746501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,12,12,128,1,float16,fp8,0,0.06004266440868378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,1,128,1,float16,float16,0,0.5348693529764811
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,1,128,1,float16,fp8,0,0.5133066574732462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,12,128,1,float16,fp8,0,0.03794133414824804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,2,128,1,float16,float16,0,0.652896006902059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,2,128,1,float16,fp8,0,0.6535733143488566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,12,12,128,1,float16,fp8,0,0.03140799949566523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,1,128,1,float16,float16,0,0.27803200483322144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,12,128,1,float16,fp8,0,0.682581345240275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,12,128,1,float16,float16,0,0.5941760142644247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,4,128,1,float16,float16,0,0.9999252955118815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,1,128,1,float16,fp8,0,0.26969067255655926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,12,4,128,1,float16,fp8,0,1.0236480236053467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,2,128,1,float16,fp8,0,0.33860798676808673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,2,128,1,float16,float16,0,0.33658134937286377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,12,128,1,float16,fp8,0,0.3506773312886556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,4,128,1,float16,float16,0,0.5079520146052042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,1,128,1,float16,fp8,0,0.144405335187912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,2,128,1,float16,float16,0,0.1803200046221415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,2,128,1,float16,fp8,0,0.1793173352877299
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,12,1,128,1,float16,fp8,0,0.02882666637500127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,1,128,1,float16,float16,0,0.1518400013446808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,4,128,1,float16,fp8,0,0.27240532636642456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,12,128,1,float16,float16,0,0.3055199980735779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,1,128,1,float16,float16,0,0.0846666693687439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,1,128,1,float16,fp8,0,0.07958933214346568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,12,128,1,float16,fp8,0,0.18464533487955728
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,2,128,1,float16,float16,0,0.10077866911888123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,2,128,1,float16,fp8,0,0.09461333354314168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,4,128,1,float16,float16,0,0.14243732889493307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,4,128,1,float16,fp8,0,0.14823466539382935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,1,128,1,float16,float16,0,0.04863466819127401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,12,128,1,float16,float16,0,0.09026666482289632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,1,128,1,float16,fp8,0,0.04765866696834564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,2,128,1,float16,float16,0,0.05489600201447805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,2,128,1,float16,fp8,0,0.05373333394527435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,4,128,1,float16,float16,0,0.07002666592597961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,4,128,1,float16,fp8,0,0.07257600128650665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,1,128,1,float16,float16,0,0.033557333052158356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,1,128,1,float16,fp8,0,0.033439998825391136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,12,128,1,float16,float16,0,0.04584000011285146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,2,128,1,float16,float16,0,0.035674666364987694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,2,128,1,float16,fp8,0,0.03621866554021835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,12,4,128,1,float16,fp8,0,0.524895985921224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,4,128,1,float16,fp8,0,0.04574933151404063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,12,128,1,float16,fp8,0,0.03156266609827677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,12,128,1,float16,float16,0,0.027776000400384266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,1,128,1,float16,float16,0,0.02293866624434789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,1,128,1,float16,fp8,0,0.02317333221435547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,2,128,1,float16,float16,0,0.023226665953795116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,12,12,128,1,float16,float16,0,0.16270400087038675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,2,128,1,float16,fp8,0,0.025013332565625507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,4,128,1,float16,float16,0,0.027072000006834667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,12,4,128,1,float16,fp8,0,0.029264000554879505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,12,128,1,float16,fp8,0,0.0252960001428922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,1,128,1,float16,float16,0,0.020879998803138733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,1,128,1,float16,fp8,0,0.02091199904680252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,2,128,1,float16,float16,0,0.02126399924357732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,12,12,128,1,float16,fp8,0,0.09929600358009338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,2,128,1,float16,fp8,0,0.022687998910744984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,4,128,1,float16,float16,0,0.02325333406527837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,4,128,1,float16,fp8,0,0.023423999547958374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,12,128,1,float16,float16,0,0.021253332495689392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,1,128,1,float16,float16,0,0.019248000035683315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,1,128,1,float16,fp8,0,0.019082666685183842
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,2,128,1,float16,float16,0,0.02060266708334287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,12,128,1,float16,fp8,0,0.052671998739242554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,2,128,1,float16,fp8,0,0.019018666197856266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,4,128,1,float16,fp8,0,0.021957332889238994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,4,128,1,float16,float16,0,0.02117866774400075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,12,4,128,1,float16,float16,0,0.0421066681543986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,12,4,128,1,float16,float16,0,0.26629867156346637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,12,12,128,1,float16,float16,0,0.023520000278949738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,1,128,1,float16,float16,0,0.23244265715281168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,2,128,1,float16,float16,0,0.2954026659329732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,1,128,1,float16,fp8,0,0.22826133171717325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,12,12,128,1,float16,fp8,0,0.022277332842350006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,2,128,1,float16,fp8,0,0.2975040078163147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,4,128,1,float16,float16,0,0.4602400064468384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,12,128,1,float16,fp8,0,0.33049599329630536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,12,4,128,1,float16,fp8,0,0.48473600546518963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,1,128,1,float16,float16,0,0.12730133533477783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,2,128,1,float16,float16,0,0.16054399808247885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,1,128,1,float16,fp8,0,0.12504000465075174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,12,128,1,float16,float16,0,0.28390934069951373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,2,128,1,float16,fp8,0,0.16292267044385275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,4,128,1,float16,float16,0,0.23945599794387817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,12,128,1,float16,fp8,0,0.17468265692392984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,1,128,1,float16,fp8,0,0.06826133529345195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,2,128,1,float16,fp8,0,0.08238400022188823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,12,128,1,float16,float16,0,0.1523413360118866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,12,128,1,float16,float16,0,0.08266666531562805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,4,128,1,float16,float16,0,0.13180800278981528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,12,128,1,float16,fp8,0,0.09273067116737366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,4,128,1,float16,fp8,0,0.13660800457000732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,12,4,128,1,float16,fp8,0,0.2556533416112264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,2,128,1,float16,float16,0,0.04586133360862732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,1,128,1,float16,fp8,0,0.041509332756201424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,2,128,1,float16,float16,0,0.08899199962615967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,2,128,1,float16,fp8,0,0.04738133152325948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,4,128,1,float16,float16,0,0.06428266565004985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,1,128,1,float16,float16,0,0.040607998768488564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,12,4,128,1,float16,fp8,0,0.06634133557478587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,1,128,1,float16,fp8,0,0.029205332199732464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,1,128,1,float16,float16,0,0.029285334050655365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,12,128,1,float16,fp8,0,0.04869333406289419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,12,128,1,float16,float16,0,0.04186666508515676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,2,128,1,float16,float16,0,0.03156266609827677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,2,128,1,float16,fp8,0,0.03332799921433131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,4,128,1,float16,fp8,0,0.04178133110205332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,1,128,1,float16,fp8,0,0.019280000279347103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,2,128,1,float16,float16,0,0.02123733361562093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,12,4,128,1,float16,float16,0,0.039066667358080544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,12,128,1,float16,float16,0,0.025077333052953083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,12,128,1,float16,fp8,0,0.02961066613594691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,1,128,1,float16,float16,0,0.01930133377512296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,4,128,1,float16,float16,0,0.02329600105683009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,1,128,1,float16,float16,0,0.017024000485738117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,12,128,1,float16,fp8,0,0.022266666094462078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,12,128,1,float16,float16,0,0.020986666282018025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,2,128,1,float16,fp8,0,0.01918399954835574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,12,1,128,1,float16,float16,0,0.07355199754238129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,4,128,1,float16,float16,0,0.019178666174411774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,4,128,1,float16,fp8,0,0.020954666038354237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,12,128,1,float16,float16,0,0.019018666197856266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,1,128,1,float16,float16,0,0.016789333273967106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,12,128,1,float16,fp8,0,0.018895999838908512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,2,128,1,float16,float16,0,0.01722666621208191
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,2,128,1,float16,fp8,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,4,128,1,float16,float16,0,0.018672000616788864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,4,128,1,float16,fp8,0,0.019093333433071773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,12,128,1,float16,float16,0,0.01716800034046173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,12,128,1,float16,fp8,0,0.01720000058412552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,2,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,1,128,1,float16,fp8,0,0.017184000462293625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,2,128,1,float16,fp8,0,0.017029333859682083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,4,128,1,float16,float16,0,0.01674666628241539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,4,128,1,float16,fp8,0,0.01687466725707054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,2,128,1,float16,fp8,0,0.021242665747801464
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,1,128,1,float16,float16,0,0.142767995595932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,12,4,128,1,float16,fp8,0,0.02555199960867564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,1,128,1,float16,fp8,0,0.01687466725707054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,1,128,1,float16,fp8,0,0.13792533675829569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,12,2,128,1,float16,float16,0,0.01877333347996076
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,12,1,128,1,float16,fp8,0,0.01724799970785777
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,2,128,1,float16,float16,0,0.17115734020868936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,12,1,128,1,float16,float16,0,0.016751999656359356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,4,128,1,float16,fp8,0,0.2643093268076579
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,2,128,1,float16,fp8,0,0.175546665986379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,12,4,128,1,float16,float16,0,0.254421333471934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,12,128,1,float16,fp8,0,0.17882132530212402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,1,128,1,float16,float16,0,0.08062399923801422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,1,128,1,float16,fp8,0,0.07300800085067749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,2,128,1,float16,float16,0,0.09390399853388469
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,2,128,1,float16,fp8,0,0.09319466352462769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,12,128,1,float16,float16,0,0.08611733714739482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,4,128,1,float16,fp8,0,0.14359999696413675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,4,128,1,float16,float16,0,0.13726400335629782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,12,128,1,float16,fp8,0,0.09468266367912292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,1,128,1,float16,float16,0,0.043882668018341064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,1,128,1,float16,fp8,0,0.043765331308046974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,2,128,1,float16,float16,0,0.05003199974695841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,2,128,1,float16,fp8,0,0.050986667474110924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,4,128,1,float16,fp8,0,0.06801066795984904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,12,4,128,1,float16,float16,0,0.06799466907978058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,1,128,1,float16,float16,0,0.027162666122118633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,12,128,1,float16,float16,0,0.039690665900707245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,12,128,1,float16,fp8,0,0.045663997530937195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,1,128,1,float16,fp8,0,0.027258666853109997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,4,128,1,float16,float16,0,0.036874666810035706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,2,128,1,float16,float16,0,0.030378667016824085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,4,128,1,float16,fp8,0,0.03991466760635376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,1,128,1,float16,fp8,0,0.021007999777793884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,1,128,1,float16,float16,0,0.020682666450738907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,2,128,1,float16,float16,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,12,128,1,float16,fp8,0,0.029365333418051403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,2,128,1,float16,fp8,0,0.021903999149799347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,4,128,1,float16,float16,0,0.024613333245118458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,4,128,1,float16,fp8,0,0.027029333015282948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,1,128,1,float16,float16,0,0.014783999572197596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,12,12,128,1,float16,float16,0,0.15620799859364828
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,12,128,1,float16,float16,0,0.017071999609470367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,1,128,1,float16,fp8,0,0.013253333667914072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,12,128,1,float16,fp8,0,0.01912533367673556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,2,128,1,float16,float16,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,2,128,1,float16,fp8,0,0.015173333386580149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,4,128,1,float16,fp8,0,0.017685333887736004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,12,128,1,float16,fp8,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,12,128,1,float16,float16,0,0.014906667172908783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,2,128,1,float16,float16,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,12,4,128,1,float16,float16,0,0.01692266638080279
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,2,128,1,float16,fp8,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,4,128,1,float16,float16,0,0.014943999548753103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,4,128,1,float16,fp8,0,0.01509333277742068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,1,128,1,float16,float16,0,0.01292266696691513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,12,128,1,float16,float16,0,0.013125333935022354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,2,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,2,128,1,float16,fp8,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,4,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,4,128,1,float16,fp8,0,0.013088000317414602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,12,128,1,float16,float16,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,12,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,1,128,1,float16,float16,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,1,128,1,float16,fp8,0,0.01309866706530253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,2,128,1,float16,float16,0,0.013088000317414602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,2,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,4,128,1,float16,float16,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,12,2,128,1,float16,fp8,0,0.031221332649389904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,12,4,128,1,float16,fp8,0,0.013007999708255133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,12,12,128,1,float16,float16,0,0.025248001019159954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,1,128,1,float16,float16,0,0.012917333592971167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,1,128,1,float16,float16,0,0.11094933748245239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,1,128,1,float16,fp8,0,0.10044800241788228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,12,1,128,1,float16,fp8,0,0.012981332838535309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,2,128,1,float16,float16,0,0.12367467085520427
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,2,128,1,float16,fp8,0,0.11866666873296101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,12,128,1,float16,fp8,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,1,128,1,float16,float16,0,0.05973866581916809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,12,128,1,float16,float16,0,0.10038399696350098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,12,128,1,float16,fp8,0,0.10703466335932414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,1,128,1,float16,fp8,0,0.05592533449331919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,4,128,1,float16,fp8,0,0.1679733395576477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,12,4,128,1,float16,float16,0,0.1683466633160909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,2,128,1,float16,float16,0,0.06426666676998138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,2,128,1,float16,fp8,0,0.062080000837643944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,4,128,1,float16,float16,0,0.08571733037630717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,12,128,1,float16,float16,0,0.04966933528582255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,1,128,1,float16,fp8,0,0.03417599946260452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,2,128,1,float16,float16,0,0.03835200021664301
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,12,4,128,1,float16,fp8,0,0.08045866588751475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,1,128,1,float16,float16,0,0.03550933301448822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,2,128,1,float16,fp8,0,0.037978666524092354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,4,128,1,float16,float16,0,0.04567466676235199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,1,128,1,float16,float16,0,0.023024000227451324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,4,128,1,float16,fp8,0,0.047237331668535866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,12,128,1,float16,fp8,0,0.031290667752424874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,1,128,1,float16,fp8,0,0.022976001103719074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,2,128,1,float16,fp8,0,0.02347733328739802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,2,128,1,float16,float16,0,0.025301332275072735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,4,128,1,float16,float16,0,0.027456000447273254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,4,128,1,float16,fp8,0,0.028618666032950085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,1,128,1,float16,float16,0,0.01720000058412552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,12,128,1,float16,float16,0,0.01922133316596349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,12,128,1,float16,fp8,0,0.021322667598724365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,2,128,1,float16,float16,0,0.018538666268189747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,2,128,1,float16,fp8,0,0.018885333091020584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,4,128,1,float16,float16,0,0.019093333433071773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,4,128,1,float16,fp8,0,0.021066665649414062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,12,128,1,float16,float16,0,0.01414399966597557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,12,128,1,float16,fp8,0,0.015040000279744467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,1,128,1,float16,fp8,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,2,128,1,float16,float16,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,2,128,1,float16,fp8,0,0.012858666479587555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,4,128,1,float16,float16,0,0.01312000056107839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,4,128,1,float16,fp8,0,0.014746667196353277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,12,128,1,float16,float16,0,0.012378666549921036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,12,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,12,12,128,1,float16,fp8,0,0.05421333511670431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,1,128,1,float16,float16,0,0.011136000355084738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,1,128,1,float16,fp8,0,0.012421333541472753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,2,128,1,float16,float16,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,4,128,1,float16,float16,0,0.01258133351802826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,4,128,1,float16,fp8,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,12,12,128,1,float16,float16,0,0.028592000404993694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,1,128,1,float16,float16,0,0.011509332805871964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,12,128,1,float16,float16,0,0.011658667276302973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,2,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,1,128,1,float16,fp8,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,2,128,1,float16,fp8,0,0.011855999628702799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,4,128,1,float16,float16,0,0.012138667205969492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,4,128,1,float16,fp8,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,12,1,128,1,float16,fp8,0,0.0170666662355264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,1,128,1,float16,float16,0,0.0124746672809124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,12,128,1,float16,float16,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,12,128,1,float16,fp8,0,0.012629333883523941
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,1,128,1,float16,fp8,0,0.011792000383138657
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,2,128,1,float16,float16,0,0.011653333902359009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,2,128,1,float16,fp8,0,0.01109333336353302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,4,128,1,float16,fp8,0,0.012458667159080505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,12,4,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,12,1,128,1,float16,fp8,0,0.013093333691358566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,1,128,1,float16,float16,0,0.0943946639696757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,1,128,1,float16,fp8,0,0.08261333405971527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,2,128,1,float16,fp8,0,0.089519997437795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,2,128,1,float16,float16,0,0.10206933816274007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,4,128,1,float16,float16,0,0.12612266341845194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,1,128,1,float16,float16,0,0.05397333204746246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,12,2,128,1,float16,fp8,0,0.01293333371480306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,12,128,1,float16,float16,0,0.07239466905593872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,12,4,128,1,float16,fp8,0,0.10666666428248088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,1,128,1,float16,fp8,0,0.047584002216657005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,12,12,128,1,float16,fp8,0,0.01379199946920077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,2,128,1,float16,float16,0,0.0561653325955073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,2,128,1,float16,fp8,0,0.05133333305517832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,4,128,1,float16,float16,0,0.062261333068211876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,12,128,1,float16,float16,0,0.03764266769091288
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,4,128,1,float16,fp8,0,0.060266668597857155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,12,128,1,float16,fp8,0,0.039221333960692085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,1,128,1,float16,float16,0,0.03306666761636734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,1,128,1,float16,fp8,0,0.02940800040960312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,2,128,1,float16,float16,0,0.03316266586383184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,2,128,1,float16,fp8,0,0.031290667752424874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,4,128,1,float16,float16,0,0.03598399957021078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,12,4,128,1,float16,fp8,0,0.035349334279696144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,12,128,1,float16,fp8,0,0.02330133318901062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,1,128,1,float16,float16,0,0.02093333254257838
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,12,1,128,1,float16,float16,0,0.013002666334311167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,1,128,1,float16,fp8,0,0.01921066641807556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,2,128,1,float16,float16,0,0.021157334248224895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,2,128,1,float16,fp8,0,0.021168000996112823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,4,128,1,float16,float16,0,0.023071999351183575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,4,128,1,float16,fp8,0,0.022965334355831146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,1,128,1,float16,fp8,0,0.014709333578745524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,12,128,1,float16,fp8,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,12,128,1,float16,float16,0,0.017103999853134155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,2,128,1,float16,float16,0,0.015573333948850632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,2,128,1,float16,fp8,0,0.01540800059835116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,4,128,1,float16,fp8,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,1,128,1,float16,float16,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,1,128,1,float16,fp8,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,12,128,1,float16,float16,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,2,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,2,128,1,float16,fp8,0,0.012698666503032049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,4,128,1,float16,float16,0,0.011328000575304031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,4,128,1,float16,fp8,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,12,128,1,float16,fp8,0,0.011237333218256632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,1,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,12,12,128,1,float16,fp8,0,0.06794133285681407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,2,128,1,float16,float16,0,0.011146667102972666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,2,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,4,128,1,float16,float16,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,12,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,12,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,1,128,1,float16,float16,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,1,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,2,128,1,float16,float16,0,0.010698666175206503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,2,128,1,float16,fp8,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,12,12,128,1,float16,float16,0,0.023077333966890972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,4,128,1,float16,fp8,0,0.01116266722480456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,1,128,1,float16,float16,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,12,128,1,float16,float16,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,12,4,128,1,float16,float16,0,0.01704000060757001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,12,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,1,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,1,128,1,float16,fp8,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,2,128,1,float16,float16,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,12,12,128,1,float16,fp8,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,2,128,1,float16,fp8,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,4,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,12,4,128,1,float16,fp8,0,0.010778666784365972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,12,128,1,float16,float16,0,0.011146667102972666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,1,128,1,float16,float16,0,0.09308800101280212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,1,128,1,float16,fp8,0,0.07554133236408234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,2,128,1,float16,float16,0,0.09387200077374776
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,4,128,1,float16,float16,0,0.10037333766619365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,12,128,1,float16,float16,0,0.05775466561317444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,12,128,1,float16,fp8,0,0.05193066596984863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,4,128,1,float16,fp8,0,0.08855467041333516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,12,4,128,1,float16,float16,0,0.010725333044926325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,1,128,1,float16,float16,0,0.051914667089780174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,2,128,1,float16,float16,0,0.05193600058555603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,1,128,1,float16,fp8,0,0.043925335009892784
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,2,128,1,float16,fp8,0,0.04565866788228353
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,4,128,1,float16,float16,0,0.05606399973233541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,12,4,128,1,float16,fp8,0,0.04987733562787374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,12,128,1,float16,fp8,0,0.0315733328461647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,12,128,1,float16,float16,0,0.03313600023587545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,1,128,1,float16,fp8,0,0.027306665976842243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,2,128,1,float16,float16,0,0.031173333525657654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,2,128,1,float16,fp8,0,0.02867199977238973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,4,128,1,float16,float16,0,0.03321066747109095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,4,128,1,float16,fp8,0,0.02942399928967158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,1,128,1,float16,float16,0,0.019146667172511418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,12,128,1,float16,float16,0,0.021013334393501282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,12,128,1,float16,fp8,0,0.020821332931518555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,1,128,1,float16,fp8,0,0.018960000326236088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,2,128,1,float16,float16,0,0.019258666783571243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,2,128,1,float16,fp8,0,0.018965333700180054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,4,128,1,float16,float16,0,0.021082667013009388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,12,4,128,1,float16,fp8,0,0.019258666783571243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,12,4,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,12,128,1,float16,fp8,0,0.015018666783968607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,12,128,1,float16,float16,0,0.014949332922697067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,1,128,1,float16,fp8,0,0.014783999572197596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,2,128,1,float16,float16,0,0.015253332753976187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,2,128,1,float16,fp8,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,4,128,1,float16,float16,0,0.014773332824309668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,4,128,1,float16,fp8,0,0.014730667074521383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,1,128,1,float16,fp8,0,0.011226666470368704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,12,128,1,float16,float16,0,0.011711999773979187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,2,128,1,float16,float16,0,0.01110400011142095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,2,128,1,float16,fp8,0,0.011296000331640244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,4,128,1,float16,float16,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,4,128,1,float16,fp8,0,0.011711999773979187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,1,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,12,128,1,float16,fp8,0,0.011221333096424738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,12,128,1,float16,float16,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,1,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,2,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,2,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,4,128,1,float16,float16,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,12,4,128,1,float16,fp8,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,12,128,1,float16,float16,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,1,128,1,float16,float16,0,0.010693332801262537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,1,128,1,float16,fp8,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,12,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,2,128,1,float16,float16,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,2,128,1,float16,fp8,0,0.010794666906197866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,4,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,12,4,128,1,float16,fp8,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,12,1,128,1,float16,float16,0,0.031328000128269196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,1,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,12,128,1,float16,float16,0,0.010714666297038397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,2,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,4,128,1,float16,float16,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,4,128,1,float16,fp8,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,12,2,128,1,float16,fp8,0,0.07872533301512401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,12,1,128,1,float16,float16,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,1,128,1,float16,float16,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,2,128,1,float16,float16,0,0.01413333291808764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,1,128,1,float16,fp8,0,0.013007999708255133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,12,12,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,2,128,1,float16,fp8,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,4,128,1,float16,float16,0,0.01893866683046023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,12,4,128,1,float16,fp8,0,0.01706133286158244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,12,128,1,float16,float16,0,0.02312533309062322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,12,128,1,float16,fp8,0,0.02149333308140437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,1,128,1,float16,float16,0,0.010703999549150467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,1,128,1,float16,fp8,0,0.010682666053374609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,2,128,1,float16,float16,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,2,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,4,128,1,float16,float16,0,0.01369599997997284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,12,4,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,12,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,1,128,1,float16,float16,0,0.009957333405812582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,12,128,1,float16,fp8,0,0.015098666151364645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,12,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,12,2,128,1,float16,float16,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,2,128,1,float16,fp8,0,0.00884799969693025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,4,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,4,128,1,float16,fp8,0,0.01081066702802976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,12,128,1,float16,fp8,0,0.011594666788975397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,12,128,1,float16,float16,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,1,128,1,float16,fp8,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,2,128,1,float16,float16,0,0.008863999818762144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,2,128,1,float16,fp8,0,0.008837333569924036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,4,128,1,float16,float16,0,0.008805333326260248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,4,128,1,float16,fp8,0,0.010106666634480158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,12,128,1,float16,float16,0,0.010656000425418219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,1,128,1,float16,float16,0,0.008938666433095932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,12,128,1,float16,fp8,0,0.011765333513418833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,1,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,2,128,1,float16,float16,0,0.008869333192706108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,2,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,4,128,1,float16,float16,0,0.00914666677514712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,12,4,128,1,float16,fp8,0,0.009029333169261614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,1,128,1,float16,float16,0,0.008656000097592672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,12,128,1,float16,fp8,0,0.009205333267649015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,12,128,1,float16,float16,0,0.00961599995692571
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,2,128,1,float16,float16,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,2,128,1,float16,fp8,0,0.008922666932145754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,4,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,4,128,1,float16,fp8,0,0.00879466657837232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,12,128,1,float16,fp8,0,0.008826666822036108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,12,128,1,float16,float16,0,0.009061333412925402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,2,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,1,128,1,float16,fp8,0,0.009685333197315535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,2,128,1,float16,fp8,0,0.009056000038981438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,4,128,1,float16,float16,0,0.008885333314538002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,4,128,1,float16,fp8,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,1,128,1,float16,fp8,0,0.009365333244204521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,12,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,1,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,2,128,1,float16,float16,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,1,128,1,float16,fp8,0,0.009253333633144697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,12,1,128,1,float16,float16,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,2,128,1,float16,fp8,0,0.009029333169261614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,4,128,1,float16,fp8,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,4,128,1,float16,float16,0,0.009061333412925402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,1,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,1,128,1,float16,fp8,0,0.009066666786869368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,2,128,1,float16,float16,0,0.008634666601816813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,2,128,1,float16,fp8,0,0.008816000074148178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,4,128,1,float16,float16,0,0.008634666601816813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,4,128,1,float16,fp8,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,12,1,128,1,float16,fp8,0,0.00891733355820179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,12,1,128,1,float16,float16,0,0.009306666751702627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,1,128,1,float16,fp8,0,2.9130080540974936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,1,128,1,float16,float16,0,3.549056053161621
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,12,2,128,1,float16,float16,0,0.009914666414260864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,12,12,128,1,float16,fp8,0,0.008709333216150602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,2,128,1,float16,float16,0,3.643850644429525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,2,128,1,float16,fp8,0,3.0568745930989585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,12,128,1,float16,float16,0,0.008586666857202848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,8,128,1,float16,float16,0,2.0837173461914062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,1,128,1,float16,float16,0,1.8325600624084473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,8,128,1,float16,fp8,0,1.8230560620625813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,1,128,1,float16,fp8,0,1.5228320757548015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,4,128,1,float16,float16,0,4.0273332595825195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,8,4,128,1,float16,fp8,0,3.798741340637207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,2,128,1,float16,float16,0,1.8724320729573567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,2,128,1,float16,fp8,0,1.6078453063964844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,4,128,1,float16,fp8,0,1.785568078358968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,4,128,1,float16,float16,0,2.0726186434427896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,8,128,1,float16,float16,0,1.0952746868133545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,1,128,1,float16,float16,0,0.9697226683298746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,8,128,1,float16,float16,0,2.059546629587809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,8,8,128,1,float16,fp8,0,1.9616319338480632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,1,128,1,float16,fp8,0,0.8237386544545492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,2,128,1,float16,float16,0,1.0544479688008626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,2,128,1,float16,fp8,0,0.8826026916503906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,4,128,1,float16,float16,0,1.0819520155588787
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,4,128,1,float16,fp8,0,1.009973367055257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,8,128,1,float16,fp8,0,0.9781226317087809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,8,128,1,float16,float16,0,0.6330986817677816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,8,128,1,float16,float16,0,1.0915093421936035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,8,128,1,float16,fp8,0,0.555621345837911
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,1,128,1,float16,fp8,0,0.4874346653620402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,1,128,1,float16,float16,0,0.5324853261311849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,2,128,1,float16,float16,0,0.5495253403981527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,2,128,1,float16,fp8,0,0.5351626475652059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,4,128,1,float16,float16,0,0.5974986553192139
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,4,128,1,float16,fp8,0,0.5404266516367594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,12,12,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,8,128,1,float16,fp8,0,0.5541066726048788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,8,8,128,1,float16,float16,0,0.6011146704355875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,1,128,1,float16,fp8,0,1.7592426935831706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,1,128,1,float16,float16,0,2.0841919581095376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,2,128,1,float16,fp8,0,1.8350240389506023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,2,128,1,float16,float16,0,2.1564319928487143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,8,128,1,float16,float16,0,1.2486186822255452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,4,128,1,float16,float16,0,2.4122559229532876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,8,4,128,1,float16,fp8,0,2.118000030517578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,1,128,1,float16,float16,0,1.0824159781138103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,1,128,1,float16,fp8,0,0.9117653369903564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,8,128,1,float16,fp8,0,1.1625226338704426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,2,128,1,float16,float16,0,1.1827306747436523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,8,8,128,1,float16,fp8,0,1.0423680146535237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,2,128,1,float16,fp8,0,0.9656213124593099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,4,128,1,float16,float16,0,1.2481013139088948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,4,128,1,float16,fp8,0,1.1108853022257488
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,8,128,1,float16,float16,0,0.6688533624013265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,1,128,1,float16,float16,0,0.5721013148625692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,8,128,1,float16,fp8,0,1.137712001800537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,8,8,128,1,float16,float16,0,1.3223573366800945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,8,128,1,float16,fp8,0,0.6196159919102987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,1,128,1,float16,fp8,0,0.506928006807963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,2,128,1,float16,float16,0,0.6016159852345785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,2,128,1,float16,fp8,0,0.5347679853439331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,4,128,1,float16,float16,0,0.6605066855748495
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,8,128,1,float16,fp8,0,0.6208213170369467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,8,128,1,float16,float16,0,0.6775626341501871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,8,128,1,float16,fp8,0,0.35873599847157794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,1,128,1,float16,fp8,0,0.3035413424173991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,2,128,1,float16,float16,0,0.3327839970588684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,2,128,1,float16,fp8,0,0.31779734293619794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,4,128,1,float16,float16,0,0.36988266309102374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,4,128,1,float16,fp8,0,0.34427201747894287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,8,128,1,float16,float16,0,0.3786826531092326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,8,128,1,float16,fp8,0,0.3641546567281087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,1,128,1,float16,float16,0,1.4862772623697917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,2,128,1,float16,fp8,0,1.3338507016499836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,2,128,1,float16,float16,0,1.5373066266377766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,4,128,1,float16,float16,0,1.767936070760091
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,8,128,1,float16,float16,0,0.928447961807251
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,4,128,1,float16,fp8,0,1.5739572842915852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,1,128,1,float16,float16,0,0.7826613585154215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,1,128,1,float16,fp8,0,0.6644800106684366
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,8,4,128,1,float16,fp8,0,0.6065813302993774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,8,128,1,float16,float16,0,0.38165334860483807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,2,128,1,float16,float16,0,0.806933323542277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,8,1,128,1,float16,float16,0,0.3251199920972188
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,2,128,1,float16,fp8,0,0.7329440116882324
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,4,128,1,float16,fp8,0,0.8311893145243326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,4,128,1,float16,float16,0,0.9159359931945801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,8,128,1,float16,float16,0,0.5000693400700887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,8,128,1,float16,float16,0,0.9272426764170328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,8,128,1,float16,fp8,0,0.8520479996999105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,1,128,1,float16,float16,0,0.4185440142949422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,1,128,1,float16,fp8,0,0.37193600336710614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,2,128,1,float16,float16,0,0.4375946521759033
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,2,128,1,float16,fp8,0,0.39318398634592694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,4,128,1,float16,fp8,0,0.45905601978302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,4,128,1,float16,float16,0,0.4919733206431071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,8,1,128,1,float16,fp8,0,1.2441013654073079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,8,128,1,float16,float16,0,0.5000053246816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,8,128,1,float16,fp8,0,0.4729599952697754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,1,128,1,float16,fp8,0,0.23538132508595785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,8,128,1,float16,fp8,0,0.27774399518966675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,8,128,1,float16,float16,0,0.295909325281779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,2,128,1,float16,float16,0,0.25811733802159625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,2,128,1,float16,fp8,0,0.2434933384259542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,4,128,1,float16,float16,0,0.2836479942003886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,8,128,1,float16,float16,0,0.29603199164072674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,8,8,128,1,float16,fp8,0,0.8526293436686198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,1,128,1,float16,fp8,0,1.6046560605367024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,1,128,1,float16,float16,0,1.9015199343363445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,2,128,1,float16,float16,0,2.0332694053649902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,2,128,1,float16,fp8,0,1.7567839622497559
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,8,8,128,1,float16,fp8,0,0.4867946704228719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,8,128,1,float16,float16,0,1.2554346720377605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,4,128,1,float16,fp8,0,2.2426932652791343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,8,4,128,1,float16,float16,0,2.355994701385498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,8,128,1,float16,fp8,0,1.1458826859792073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,1,128,1,float16,float16,0,0.9893973668416342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,1,128,1,float16,float16,0,0.25356266895929974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,1,128,1,float16,fp8,0,0.8507093588511149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,2,128,1,float16,fp8,0,0.9152746995290121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,2,128,1,float16,float16,0,1.0374933083852131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,4,128,1,float16,float16,0,1.2048052946726482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,4,128,1,float16,fp8,0,0.26359466711680096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,8,8,128,1,float16,fp8,0,0.2751306692759196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,1,128,1,float16,float16,0,0.515994668006897
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,8,128,1,float16,fp8,0,0.6205013195673624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,8,128,1,float16,fp8,0,1.1590452988942463
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,2,128,1,float16,float16,0,0.5522826512654623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,2,128,1,float16,fp8,0,0.4962879816691081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,4,128,1,float16,float16,0,0.6323306560516357
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,4,128,1,float16,fp8,0,0.5943040053049723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,8,128,1,float16,float16,0,0.6432373523712158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,8,128,1,float16,fp8,0,0.6053119897842407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,8,128,1,float16,fp8,0,0.3440213203430176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,8,128,1,float16,float16,0,0.35417600472768146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,1,128,1,float16,float16,0,0.28703467051188153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,1,128,1,float16,fp8,0,0.2659306724866231
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,2,128,1,float16,float16,0,0.3034293254216512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,2,128,1,float16,fp8,0,0.2763413389523824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,4,128,1,float16,float16,0,0.3500479857126872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,4,128,1,float16,fp8,0,0.32869333028793335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,8,128,1,float16,fp8,0,0.3431253433227539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,8,8,128,1,float16,float16,0,0.35572266578674316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,8,128,1,float16,fp8,0,0.20109333594640097
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,1,128,1,float16,float16,0,0.1819360057512919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,1,128,1,float16,fp8,0,0.17259732882181802
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,2,128,1,float16,float16,0,0.1881813406944275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,2,128,1,float16,fp8,0,0.1771999994913737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,4,128,1,float16,float16,0,0.20068800449371338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,4,128,1,float16,fp8,0,0.19601066907246908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,4,128,1,float16,fp8,0,1.108181317647298
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,8,128,1,float16,float16,0,0.2161653240521749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,8,128,1,float16,fp8,0,0.2010879913965861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,1,128,1,float16,fp8,0,0.4593386650085449
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,8,8,128,1,float16,float16,0,1.2297066847483318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,1,128,1,float16,float16,0,1.1413599650065105
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,1,128,1,float16,fp8,0,0.9725493590037028
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,2,128,1,float16,float16,0,1.216000000635783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,2,128,1,float16,fp8,0,1.0851199626922607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,4,128,1,float16,float16,0,1.477776050567627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,1,128,1,float16,float16,0,0.5857173204421997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,8,4,128,1,float16,fp8,0,1.3693920771280925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,1,128,1,float16,fp8,0,0.5163733164469401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,2,128,1,float16,float16,0,0.6303466558456421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,2,128,1,float16,fp8,0,0.5702879826227824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,8,8,128,1,float16,float16,0,0.21300800641377768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,4,128,1,float16,float16,0,0.759557326634725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,4,128,1,float16,fp8,0,0.711850643157959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,8,128,1,float16,float16,0,0.7758506933848063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,8,128,1,float16,fp8,0,0.39907201131184894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,8,128,1,float16,float16,0,0.41017599900563556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,8,128,1,float16,fp8,0,0.7384906609853109
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,1,128,1,float16,float16,0,0.31753067175547284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,1,128,1,float16,fp8,0,0.2888960043589274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,2,128,1,float16,float16,0,0.34166932106018066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,2,128,1,float16,fp8,0,0.3161440094312032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,4,128,1,float16,fp8,0,0.38838398456573486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,8,128,1,float16,float16,0,0.41361065705617267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,8,128,1,float16,fp8,0,0.40117335319519043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,8,128,1,float16,float16,0,0.23088000218073526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,8,128,1,float16,fp8,0,0.22842667500178018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,1,128,1,float16,float16,0,0.18309332927068075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,1,128,1,float16,fp8,0,0.17182934284210205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,2,128,1,float16,fp8,0,0.18138132492701212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,4,128,1,float16,float16,0,0.22499199708302817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,4,128,1,float16,fp8,0,0.20616000890731812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,8,128,1,float16,float16,0,0.7684373060862223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,8,8,128,1,float16,fp8,0,0.7397813002268473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,8,128,1,float16,float16,0,0.23283199469248453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,8,128,1,float16,fp8,0,0.22721600532531738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,8,128,1,float16,float16,0,0.13732266426086426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,8,128,1,float16,fp8,0,0.13410133123397827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,1,128,1,float16,float16,0,0.12090133627255757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,1,128,1,float16,fp8,0,0.11337066690127055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,2,128,1,float16,float16,0,0.12482133507728577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,2,128,1,float16,fp8,0,0.11939199765523274
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,4,128,1,float16,fp8,0,0.1323466698328654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,8,128,1,float16,fp8,0,0.1344106694062551
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,8,128,1,float16,float16,0,0.13737066586812338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,1,128,1,float16,fp8,0,0.939903974533081
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,1,128,1,float16,float16,0,1.0852693716684978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,2,128,1,float16,float16,0,1.203114668528239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,2,128,1,float16,fp8,0,1.0883839925130208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,8,8,128,1,float16,float16,0,0.6525706847508749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,8,4,128,1,float16,float16,0,0.40569067001342773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,8,128,1,float16,float16,0,0.807253360748291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,8,128,1,float16,fp8,0,0.7965333461761475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,1,128,1,float16,float16,0,0.5499146779378256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,8,2,128,1,float16,float16,0,0.18822934230168661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,1,128,1,float16,fp8,0,0.49720533688863117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,4,128,1,float16,float16,0,1.5383626619974773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,2,128,1,float16,fp8,0,0.5697120030721029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,4,128,1,float16,float16,0,0.781877358754476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,4,128,1,float16,fp8,0,0.7523893515268961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,8,128,1,float16,fp8,0,0.7969386577606201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,8,128,1,float16,float16,0,0.8080373605092367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,1,128,1,float16,float16,0,0.2956533432006836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,8,4,128,1,float16,float16,0,0.13429866234461466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,8,128,1,float16,fp8,0,0.42122666041056317
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,1,128,1,float16,fp8,0,0.27034133672714233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,2,128,1,float16,float16,0,0.3237066666285197
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,2,128,1,float16,fp8,0,0.3097440004348755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,4,128,1,float16,float16,0,0.40837868054707843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,4,128,1,float16,fp8,0,0.40413331985473633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,8,128,1,float16,fp8,0,0.41973332564036053
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,8,128,1,float16,float16,0,0.415887991587321
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,1,128,1,float16,float16,0,0.16692266861597696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,8,128,1,float16,fp8,0,0.23174933592478433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,1,128,1,float16,fp8,0,0.15523200233777365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,2,128,1,float16,float16,0,0.1823306679725647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,4,128,1,float16,float16,0,0.22831465800603232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,4,128,1,float16,fp8,0,0.21732266743977866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,8,128,1,float16,fp8,0,0.23138666152954102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,8,128,1,float16,float16,0,0.23170133431752524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,8,128,1,float16,float16,0,0.13859732945760092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,1,128,1,float16,float16,0,0.10583466291427612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,8,128,1,float16,fp8,0,0.12970133622487387
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,1,128,1,float16,fp8,0,0.10081066687901814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,2,128,1,float16,float16,0,0.11127466956774394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,2,128,1,float16,fp8,0,0.10734400153160095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,8,4,128,1,float16,fp8,0,1.4637120564778645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,4,128,1,float16,float16,0,0.12602133552233377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,4,128,1,float16,fp8,0,0.12563199798266092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,8,2,128,1,float16,float16,0,0.6171626647313436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,8,128,1,float16,float16,0,0.14033066232999167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,8,8,128,1,float16,fp8,0,0.1300106644630432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,1,128,1,float16,float16,0,0.06623999774456024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,8,128,1,float16,fp8,0,0.07821866869926453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,1,128,1,float16,fp8,0,0.06425599753856659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,2,128,1,float16,float16,0,0.07006933291753133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,2,128,1,float16,fp8,0,0.0680213322242101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,4,128,1,float16,float16,0,0.07551466425259908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,8,8,128,1,float16,float16,0,0.41808001200358075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,4,128,1,float16,fp8,0,0.07629866898059845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,8,128,1,float16,fp8,0,0.07809066772460938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,8,128,1,float16,float16,0,0.0766186664501826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,1,128,1,float16,float16,0,0.660378654797872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,1,128,1,float16,fp8,0,0.5920746723810831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,2,128,1,float16,float16,0,0.7489013671875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,2,128,1,float16,fp8,0,0.7015466690063477
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,2,128,1,float16,fp8,0,0.16781866550445557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,8,128,1,float16,fp8,0,0.5365493297576904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,8,128,1,float16,float16,0,0.5297333399454752
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,4,128,1,float16,float16,0,1.001482645670573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,1,128,1,float16,float16,0,0.3421493371327718
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,8,4,128,1,float16,fp8,0,0.9773653348286947
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,1,128,1,float16,fp8,0,0.31572800874710083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,2,128,1,float16,float16,0,0.38796265920003253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,2,128,1,float16,fp8,0,0.37132267157236737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,4,128,1,float16,fp8,0,0.5077759822209676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,4,128,1,float16,float16,0,0.5134933392206827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,8,128,1,float16,float16,0,0.28059200445810956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,8,128,1,float16,fp8,0,0.5371359984079996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,1,128,1,float16,float16,0,0.19262933731079102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,8,8,128,1,float16,float16,0,0.5306933323542277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,8,128,1,float16,fp8,0,0.28811200459798175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,8,8,128,1,float16,float16,0,0.07830933233102162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,1,128,1,float16,fp8,0,0.1749173402786255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,2,128,1,float16,float16,0,0.21258666117986044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,2,128,1,float16,fp8,0,0.20249066750208536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,4,128,1,float16,float16,0,0.2747360070546468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,4,128,1,float16,fp8,0,0.27666133642196655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,8,128,1,float16,float16,0,0.16040533781051636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,8,128,1,float16,float16,0,0.28091732660929364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,8,128,1,float16,fp8,0,0.16182933251063028
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,1,128,1,float16,fp8,0,0.10517332951227824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,2,128,1,float16,float16,0,0.11853866775830586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,2,128,1,float16,fp8,0,0.11473600069681804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,4,128,1,float16,float16,0,0.1527359982331594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,4,128,1,float16,fp8,0,0.1397599975268046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,8,8,128,1,float16,float16,0,0.2332586646080017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,8,128,1,float16,fp8,0,0.09012800455093384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,8,128,1,float16,float16,0,0.08844799796740214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,8,128,1,float16,fp8,0,0.16099199652671814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,2,128,1,float16,float16,0,0.07635200023651123
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,1,128,1,float16,fp8,0,0.07018666466077168
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,2,128,1,float16,fp8,0,0.07478400071461995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,4,128,1,float16,float16,0,0.08682133754094441
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,4,128,1,float16,fp8,0,0.08797333637873332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,8,128,1,float16,float16,0,0.08886399865150452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,8,128,1,float16,fp8,0,0.09057066837946574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,8,128,1,float16,float16,0,0.06001066664854685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,8,128,1,float16,fp8,0,0.06188266475995382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,1,128,1,float16,float16,0,0.053183997670809426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,2,128,1,float16,float16,0,0.054469332098960876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,2,128,1,float16,fp8,0,0.0537013312180837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,4,128,1,float16,float16,0,0.060165335734685264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,4,128,1,float16,fp8,0,0.06006399790445963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,8,128,1,float16,float16,0,0.060175999999046326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,8,128,1,float16,fp8,0,0.06225599845250448
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,1,128,1,float16,float16,0,0.661733349164327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,1,128,1,float16,fp8,0,0.6036159992218018
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,1,128,1,float16,float16,0,0.11063466469446818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,2,128,1,float16,float16,0,0.7837226390838623
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,2,128,1,float16,fp8,0,0.7488906383514404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,8,1,128,1,float16,float16,0,0.07241599758466084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,8,8,128,1,float16,float16,0,0.16055466731389365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,4,128,1,float16,float16,0,1.1273492972056072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,8,4,128,1,float16,fp8,0,1.1235199769337971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,8,128,1,float16,float16,0,0.5951626698176066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,1,128,1,float16,float16,0,0.3449653387069702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,1,128,1,float16,fp8,0,0.32129599650700885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,8,128,1,float16,fp8,0,0.6079573233922323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,4,128,1,float16,float16,0,0.5707253217697144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,2,128,1,float16,fp8,0,0.38729600111643475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,4,128,1,float16,fp8,0,0.5779253244400024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,8,128,1,float16,float16,0,0.305893341700236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,8,128,1,float16,fp8,0,0.6085813442866007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,8,128,1,float16,fp8,0,0.32123200098673504
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,8,128,1,float16,float16,0,0.5909706751505533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,1,128,1,float16,fp8,0,0.1755519906679789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,4,128,1,float16,float16,0,0.29873067140579224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,2,128,1,float16,float16,0,0.2183039983113607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,2,128,1,float16,fp8,0,0.21304533878962198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,4,128,1,float16,fp8,0,0.3054293394088745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,8,8,128,1,float16,fp8,0,0.2864426573117574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,8,128,1,float16,float16,0,0.3102399905522664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,8,128,1,float16,float16,0,0.17137066523234049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,8,128,1,float16,fp8,0,0.31989866495132446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,8,128,1,float16,fp8,0,0.17614932854970297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,1,128,1,float16,float16,0,0.10426132877667744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,1,128,1,float16,fp8,0,0.09981333216031392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,2,128,1,float16,float16,0,0.1218293309211731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,2,128,1,float16,fp8,0,0.1109920044740041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,4,128,1,float16,float16,0,0.16563199957211813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,4,128,1,float16,fp8,0,0.16148799657821655
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,8,128,1,float16,float16,0,0.16933866341908774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,8,8,128,1,float16,fp8,0,0.17714667320251465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,1,128,1,float16,float16,0,0.06632533172766368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,8,128,1,float16,fp8,0,0.0925546685854594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,8,128,1,float16,float16,0,0.09867733716964722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,1,128,1,float16,fp8,0,0.06390400230884552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,2,128,1,float16,fp8,0,0.07035199801127116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,4,128,1,float16,float16,0,0.08493333061536153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,4,128,1,float16,fp8,0,0.08859200278917949
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,8,2,128,1,float16,float16,0,0.4051733414332072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,8,128,1,float16,fp8,0,0.09213333328564961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,8,1,128,1,float16,fp8,0,0.05021866659323374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,8,128,1,float16,float16,0,0.09729066491127014
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,1,128,1,float16,float16,0,0.041663999358812966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,8,128,1,float16,float16,0,0.05194133520126343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,1,128,1,float16,fp8,0,0.04171200096607208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,2,128,1,float16,float16,0,0.04382933179537455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,8,128,1,float16,fp8,0,0.055685331424077354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,2,128,1,float16,fp8,0,0.043866669138272606
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,4,128,1,float16,float16,0,0.05092266698678335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,4,128,1,float16,fp8,0,0.05387733379999796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,8,1,128,1,float16,float16,0,0.18593066930770874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,8,128,1,float16,float16,0,0.0517439991235733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,1,128,1,float16,float16,0,0.03915733347336451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,8,128,1,float16,float16,0,0.0435146689414978
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,8,128,1,float16,fp8,0,0.04442666471004486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,1,128,1,float16,fp8,0,0.037578667203585304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,2,128,1,float16,float16,0,0.03925866633653641
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,2,128,1,float16,fp8,0,0.0395413339138031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,4,128,1,float16,float16,0,0.04215999941031138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,4,128,1,float16,fp8,0,0.043696001172065735
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,8,128,1,float16,float16,0,0.043509334325790405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,8,8,128,1,float16,fp8,0,0.043706665436426796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,1,128,1,float16,float16,0,0.42548267046610516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,1,128,1,float16,fp8,0,0.39509332180023193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,2,128,1,float16,float16,0,0.5128533442815145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,4,128,1,float16,float16,0,0.7737279733022054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,2,128,1,float16,fp8,0,0.506442666053772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,8,128,1,float16,float16,0,0.4097706476847331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,1,128,1,float16,float16,0,0.22518932819366455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,8,2,128,1,float16,float16,0,0.07312533259391785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,1,128,1,float16,fp8,0,0.21568000316619873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,8,128,1,float16,fp8,0,0.43113064765930176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,2,128,1,float16,fp8,0,0.26685865720113117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,2,128,1,float16,float16,0,0.27297067642211914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,4,128,1,float16,fp8,0,0.4069226582845052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,8,128,1,float16,fp8,0,0.4321226676305135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,8,128,1,float16,float16,0,0.40997334321339923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,1,128,1,float16,float16,0,0.127402663230896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,8,8,128,1,float16,fp8,0,0.05555200080076853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,8,128,1,float16,fp8,0,0.2294506629308065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,8,128,1,float16,float16,0,0.2156426707903544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,1,128,1,float16,fp8,0,0.1181066632270813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,2,128,1,float16,float16,0,0.14985600113868713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,2,128,1,float16,fp8,0,0.1435413360595703
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,4,128,1,float16,fp8,0,0.21525333325068155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,4,128,1,float16,float16,0,0.21437867482503256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,8,128,1,float16,float16,0,0.21712533632914224
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,8,8,128,1,float16,fp8,0,0.23210134108861288
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,8,128,1,float16,fp8,0,0.1246613363424937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,1,128,1,float16,float16,0,0.07267199953397115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,2,128,1,float16,float16,0,0.0795253316561381
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,4,128,1,float16,float16,0,0.11377066373825073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,4,128,1,float16,fp8,0,0.10702932874361674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,8,128,1,float16,fp8,0,0.12677866220474243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,8,128,1,float16,float16,0,0.06433600187301636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,1,128,1,float16,float16,0,0.04790399968624115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,8,128,1,float16,fp8,0,0.06650133430957794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,8,4,128,1,float16,fp8,0,0.7794506549835205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,2,128,1,float16,float16,0,0.05205333232879639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,2,128,1,float16,fp8,0,0.05162666738033295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,4,128,1,float16,float16,0,0.061978667974472046
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,4,128,1,float16,fp8,0,0.06409066418806712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,8,4,128,1,float16,float16,0,0.39678935209910077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,8,128,1,float16,float16,0,0.06495466828346252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,8,128,1,float16,fp8,0,0.0662773350874583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,8,128,1,float16,float16,0,0.04171733558177948
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,1,128,1,float16,float16,0,0.03346666693687439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,1,128,1,float16,fp8,0,0.03333866596221924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,2,128,1,float16,float16,0,0.03559466699759165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,2,128,1,float16,fp8,0,0.03572266548871994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,4,128,1,float16,float16,0,0.04140799989302953
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,4,128,1,float16,fp8,0,0.04211199780305227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,8,128,1,float16,float16,0,0.12196266651153564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,8,128,1,float16,float16,0,0.04174399872620901
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,8,128,1,float16,fp8,0,0.043824002146720886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,2,128,1,float16,fp8,0,0.07997333506743114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,8,128,1,float16,fp8,0,0.03587199995915095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,8,128,1,float16,float16,0,0.03559466699759165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,1,128,1,float16,fp8,0,0.029472000896930695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,2,128,1,float16,float16,0,0.0322026660044988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,2,128,1,float16,fp8,0,0.031471999982992806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,8,128,1,float16,float16,0,0.12146666646003723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,4,128,1,float16,float16,0,0.035189333061377205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,4,128,1,float16,fp8,0,0.03542399903138479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,8,128,1,float16,fp8,0,0.03566399961709976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,8,128,1,float16,float16,0,0.035631999373435974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,8,1,128,1,float16,fp8,0,0.04562133550643921
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,1,128,1,float16,float16,0,0.4618346691131592
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,1,128,1,float16,fp8,0,0.43294934431711835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,2,128,1,float16,float16,0,0.5831466515858968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,2,128,1,float16,fp8,0,0.5781546831130981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,8,8,128,1,float16,fp8,0,0.043765331308046974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,4,128,1,float16,float16,0,0.928218682607015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,8,128,1,float16,float16,0,0.4906239906946818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,1,128,1,float16,float16,0,0.24588799476623535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,8,4,128,1,float16,fp8,0,0.9476959705352783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,8,1,128,1,float16,fp8,0,0.06913599868615468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,8,128,1,float16,fp8,0,0.5205920139948527
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,2,128,1,float16,float16,0,0.3026346762975057
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,2,128,1,float16,fp8,0,0.2983413338661194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,1,128,1,float16,fp8,0,0.22899200518925986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,8,128,1,float16,float16,0,0.4894560178120931
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,4,128,1,float16,float16,0,0.47806398073832196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,4,128,1,float16,fp8,0,0.48817598819732666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,8,128,1,float16,float16,0,0.2544800043106079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,8,128,1,float16,fp8,0,0.27105067173639935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,8,8,128,1,float16,fp8,0,0.5256906747817993
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,1,128,1,float16,fp8,0,0.12607466181119284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,2,128,1,float16,fp8,0,0.16285866498947144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,2,128,1,float16,float16,0,0.16730666160583496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,4,128,1,float16,float16,0,0.24642133712768555
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,4,128,1,float16,fp8,0,0.2533386747042338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,8,128,1,float16,float16,0,0.2563520073890686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,8,128,1,float16,fp8,0,0.2693546613057454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,8,128,1,float16,float16,0,0.1376479963461558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,1,128,1,float16,float16,0,0.07354666789372762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,1,128,1,float16,fp8,0,0.07019199927647908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,8,128,1,float16,fp8,0,0.14777599771817526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,2,128,1,float16,float16,0,0.08508267005284627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,2,128,1,float16,fp8,0,0.08102400104204814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,4,128,1,float16,float16,0,0.13344533244768778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,4,128,1,float16,fp8,0,0.13217066725095114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,8,128,1,float16,float16,0,0.13898133238156637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,8,8,128,1,float16,fp8,0,0.14776000380516052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,1,128,1,float16,float16,0,0.046240001916885376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,8,128,1,float16,float16,0,0.07420800129572551
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,1,128,1,float16,fp8,0,0.044938668608665466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,2,128,1,float16,float16,0,0.05203733344872793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,8,128,1,float16,fp8,0,0.07435733576615651
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,2,128,1,float16,fp8,0,0.051669334371884666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,4,128,1,float16,float16,0,0.06438933312892914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,4,128,1,float16,fp8,0,0.06939733525117238
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,8,128,1,float16,fp8,0,0.07448000212510426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,8,128,1,float16,float16,0,0.039093332986036934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,8,8,128,1,float16,float16,0,0.07712000111738841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,1,128,1,float16,float16,0,0.029450667401154835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,2,128,1,float16,float16,0,0.03212266663710276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,8,128,1,float16,fp8,0,0.04345066845417023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,2,128,1,float16,fp8,0,0.031856000423431396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,4,128,1,float16,float16,0,0.03751466671625773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,4,128,1,float16,fp8,0,0.04171200096607208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,8,1,128,1,float16,float16,0,0.13384532928466797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,8,128,1,float16,fp8,0,0.043434664607048035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,1,128,1,float16,float16,0,0.02682666728893916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,8,128,1,float16,fp8,0,0.03325333446264267
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,1,128,1,float16,fp8,0,0.025439999997615814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,2,128,1,float16,float16,0,0.027045334378878277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,2,128,1,float16,fp8,0,0.02716800073782603
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,4,128,1,float16,float16,0,0.03149333347876867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,4,128,1,float16,fp8,0,0.031258667508761086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,8,128,1,float16,fp8,0,0.033386667569478355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,8,128,1,float16,float16,0,0.031173333525657654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,8,128,1,float16,float16,0,0.02741333345572154
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,1,128,1,float16,fp8,0,0.02319466571013133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,2,128,1,float16,float16,0,0.025402667621771496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,4,128,1,float16,float16,0,0.02701866626739502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,4,128,1,float16,fp8,0,0.02731200059254964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,8,128,1,float16,float16,0,0.027162666122118633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,8,1,128,1,float16,float16,0,0.03151999910672506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,8,128,1,float16,fp8,0,0.027744000156720478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,1,128,1,float16,fp8,0,0.029093332588672638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,8,8,128,1,float16,float16,0,0.03965333352486292
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,8,8,128,1,float16,float16,0,0.031221332649389904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,1,128,1,float16,float16,0,0.3700053294499715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,1,128,1,float16,fp8,0,0.3532266616821289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,8,128,1,float16,fp8,0,0.027445333699385326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,2,128,1,float16,fp8,0,0.025231999655564625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,2,128,1,float16,float16,0,0.4888586600621541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,8,128,1,float16,float16,0,0.4371680021286011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,1,128,1,float16,float16,0,0.19812800486882529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,2,128,1,float16,fp8,0,0.49012800057729083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,4,128,1,float16,fp8,0,0.858949343363444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,2,128,1,float16,float16,0,0.2625653346379598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,8,128,1,float16,fp8,0,0.47037867705027264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,2,128,1,float16,fp8,0,0.25764266649882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,8,4,128,1,float16,float16,0,0.8389973640441895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,8,1,128,1,float16,float16,0,0.024288001159826916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,8,128,1,float16,float16,0,0.23186665773391724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,8,128,1,float16,fp8,0,0.47019731998443604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,8,128,1,float16,float16,0,0.43718401590983075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,4,128,1,float16,float16,0,0.4257440169652303
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,4,128,1,float16,fp8,0,0.43786664803822833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,1,128,1,float16,fp8,0,0.10244266192118327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,4,128,1,float16,float16,0,0.22564266125361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,8,1,128,1,float16,fp8,0,0.18713066975275675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,4,128,1,float16,fp8,0,0.2320906718571981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,8,128,1,float16,float16,0,0.22848532597223917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,2,128,1,float16,float16,0,0.13757866621017456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,2,128,1,float16,fp8,0,0.13796266913414001
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,8,128,1,float16,fp8,0,0.24861866235733032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,8,128,1,float16,fp8,0,0.13212266564369202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,2,128,1,float16,float16,0,0.07521066566308339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,1,128,1,float16,fp8,0,0.055904000997543335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,1,128,1,float16,float16,0,0.058703998724619545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,4,128,1,float16,float16,0,0.1218826671441396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,1,128,1,float16,float16,0,0.10971732934315999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,4,128,1,float16,fp8,0,0.11784000198046367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,8,128,1,float16,float16,0,0.12351466218630473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,8,128,1,float16,fp8,0,0.13150399923324585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,8,128,1,float16,fp8,0,0.06506666541099548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,1,128,1,float16,float16,0,0.036271999279658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,1,128,1,float16,fp8,0,0.03545066714286804
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,2,128,1,float16,float16,0,0.04251199960708618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,2,128,1,float16,fp8,0,0.041877334316571556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,4,128,1,float16,float16,0,0.05468800167242686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,4,128,1,float16,fp8,0,0.060165335734685264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,8,128,1,float16,float16,0,0.0682773341735204
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,8,128,1,float16,fp8,0,0.06445333361625671
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,8,128,1,float16,float16,0,0.033471999069054924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,8,128,1,float16,fp8,0,0.03734400123357773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,1,128,1,float16,float16,0,0.022895999252796173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,1,128,1,float16,fp8,0,0.023056000471115112
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,2,128,1,float16,float16,0,0.025402667621771496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,2,128,1,float16,fp8,0,0.027114666998386383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,4,128,1,float16,float16,0,0.03145600110292435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,4,128,1,float16,fp8,0,0.0352960005402565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,8,128,1,float16,float16,0,0.03276800115903219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,8,8,128,1,float16,fp8,0,0.03736000011364619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,8,128,1,float16,float16,0,0.12296000123023987
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,8,128,1,float16,fp8,0,0.027258666853109997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,1,128,1,float16,float16,0,0.019167999426523846
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,1,128,1,float16,fp8,0,0.020799999435742695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,2,128,1,float16,float16,0,0.02107200026512146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,4,128,1,float16,fp8,0,0.02696000039577484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,8,2,128,1,float16,fp8,0,0.0682826687892278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,8,128,1,float16,float16,0,0.025077333052953083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,8,128,1,float16,float16,0,0.021984001000722248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,8,128,1,float16,fp8,0,0.02313599983851115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,8,8,128,1,float16,float16,0,0.06833066542943318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,1,128,1,float16,float16,0,0.018058666338523228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,1,128,1,float16,fp8,0,0.017290666699409485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,2,128,1,float16,float16,0,0.018986667195955913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,2,128,1,float16,fp8,0,0.019178666174411774
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,4,128,1,float16,float16,0,0.02096533278624217
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,4,128,1,float16,fp8,0,0.02109866589307785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,8,128,1,float16,float16,0,0.021253332495689392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,8,128,1,float16,float16,0,0.019071999937295914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,8,8,128,1,float16,fp8,0,0.24442134300867716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,1,128,1,float16,float16,0,0.01720000058412552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,8,128,1,float16,fp8,0,0.020154666155576706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,2,128,1,float16,float16,0,0.018144000321626663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,2,128,1,float16,fp8,0,0.017386666188637417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,1,128,1,float16,fp8,0,0.01798933371901512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,4,128,1,float16,float16,0,0.01916266605257988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,4,128,1,float16,fp8,0,0.01903466631968816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,8,128,1,float16,fp8,0,0.019930666933457058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,8,8,128,1,float16,float16,0,0.018837332725524902
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,8,128,1,float16,float16,0,0.025018667181332905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,4,128,1,float16,float16,0,0.023183998962243397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,8,128,1,float16,fp8,0,0.027301333844661713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,1,128,1,float16,float16,0,0.16209066907564798
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,1,128,1,float16,fp8,0,0.15875200430552164
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,2,128,1,float16,float16,0,0.2255893349647522
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,2,128,1,float16,fp8,0,0.22828799486160278
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,8,8,128,1,float16,fp8,0,0.023007998863856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,8,128,1,float16,float16,0,0.20885332425435385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,4,128,1,float16,float16,0,0.3880213499069214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,8,128,1,float16,fp8,0,0.22803199291229248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,8,4,128,1,float16,fp8,0,0.40855467319488525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,1,128,1,float16,float16,0,0.08884800473848979
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,1,128,1,float16,fp8,0,0.08542933066685994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,2,128,1,float16,float16,0,0.11716799934705098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,8,128,1,float16,fp8,0,0.12364266316095988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,2,128,1,float16,fp8,0,0.124399999777476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,4,128,1,float16,fp8,0,0.2157813310623169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,8,2,128,1,float16,fp8,0,0.02125866711139679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,8,128,1,float16,fp8,0,0.23175466060638428
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,1,128,1,float16,float16,0,0.04758933186531067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,2,128,1,float16,fp8,0,0.0609386662642161
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,2,128,1,float16,float16,0,0.06261333326498668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,4,128,1,float16,fp8,0,0.11077866951624553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,4,128,1,float16,float16,0,0.11001599828402202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,8,128,1,float16,float16,0,0.11429333686828613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,8,128,1,float16,fp8,0,0.12422933181126912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,8,128,1,float16,float16,0,0.06154666841030121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,8,128,1,float16,fp8,0,0.058490668733914696
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,1,128,1,float16,float16,0,0.029189333319664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,1,128,1,float16,fp8,0,0.02922666569550832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,2,128,1,float16,fp8,0,0.035562666753927864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,4,128,1,float16,float16,0,0.047594666481018066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,8,128,1,float16,float16,0,0.060720001657803856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,8,128,1,float16,fp8,0,0.05791999896367391
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,8,128,1,float16,float16,0,0.02905600021282832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,8,128,1,float16,float16,0,0.1135093371073405
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,8,128,1,float16,fp8,0,0.033701332906881966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,1,128,1,float16,float16,0,0.01911466692884763
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,4,128,1,float16,float16,0,0.20538665850957236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,2,128,1,float16,fp8,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,1,128,1,float16,fp8,0,0.021066665649414062
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,4,128,1,float16,float16,0,0.02844800055027008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,8,1,128,1,float16,fp8,0,0.047872001926104225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,4,128,1,float16,fp8,0,0.03311466674009959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,8,128,1,float16,float16,0,0.029279999434947968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,8,128,1,float16,fp8,0,0.03349866718053818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,8,128,1,float16,float16,0,0.02128000060717265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,8,128,1,float16,fp8,0,0.023578666150569916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,1,128,1,float16,float16,0,0.017114666601022083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,1,128,1,float16,fp8,0,0.01709866647919019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,2,128,1,float16,float16,0,0.01886933296918869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,2,128,1,float16,fp8,0,0.019141333798567455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,4,128,1,float16,float16,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,2,128,1,float16,float16,0,0.035375999907652535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,4,128,1,float16,fp8,0,0.023226665953795116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,8,4,128,1,float16,fp8,0,0.05395199855168661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,8,128,1,float16,float16,0,0.02107200026512146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,8,128,1,float16,float16,0,0.01754133279124896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,1,128,1,float16,float16,0,0.015119999647140503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,1,128,1,float16,fp8,0,0.01516266663869222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,8,128,1,float16,fp8,0,0.019013332823912304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,2,128,1,float16,float16,0,0.015354666858911514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,2,128,1,float16,fp8,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,4,128,1,float16,float16,0,0.016890666137139004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,4,128,1,float16,fp8,0,0.01882133384545644
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,8,128,1,float16,float16,0,0.017349333812793095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,8,8,128,1,float16,float16,0,0.2127573291460673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,8,8,128,1,float16,fp8,0,0.019002666076024372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,8,128,1,float16,float16,0,0.016842667013406754
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,8,2,128,1,float16,float16,0,0.0220320001244545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,1,128,1,float16,fp8,0,0.014954666296641031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,8,128,1,float16,fp8,0,0.01714666684468587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,2,128,1,float16,float16,0,0.015050667027632395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,2,128,1,float16,fp8,0,0.014912000546852747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,4,128,1,float16,float16,0,0.015184000134468079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,4,128,1,float16,fp8,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,8,128,1,float16,float16,0,0.01684800038735072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,1,128,1,float16,float16,0,0.014783999572197596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,1,128,1,float16,float16,0,0.015141333142916361
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,8,128,1,float16,float16,0,0.01504533365368843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,2,128,1,float16,fp8,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,4,128,1,float16,float16,0,0.015146666516860327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,4,128,1,float16,fp8,0,0.014933332800865173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,8,128,1,float16,float16,0,0.015194666882356008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,8,128,1,float16,fp8,0,0.015146666516860327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,8,8,128,1,float16,fp8,0,0.025061334172884624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,1,128,1,float16,float16,0,0.10088533163070679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,1,128,1,float16,fp8,0,0.09567999839782715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,2,128,1,float16,float16,0,0.12849600116411844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,2,128,1,float16,fp8,0,0.13351466258366904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,8,8,128,1,float16,fp8,0,0.017071999609470367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,4,128,1,float16,float16,0,0.21468265851338705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,8,128,1,float16,fp8,0,0.014970666418472925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,8,128,1,float16,float16,0,0.11719999710718791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,8,4,128,1,float16,fp8,0,0.22600533564885458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,8,128,1,float16,fp8,0,0.12566399574279785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,1,128,1,float16,fp8,0,0.051039998730023704
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,1,128,1,float16,float16,0,0.05390933156013489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,2,128,1,float16,float16,0,0.07014399766921997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,4,128,1,float16,float16,0,0.1129813293615977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,4,128,1,float16,fp8,0,0.11477866768836975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,8,128,1,float16,float16,0,0.11744532982508342
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,8,128,1,float16,float16,0,0.0622026671965917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,8,128,1,float16,fp8,0,0.06218666831652323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,1,128,1,float16,float16,0,0.03299200038115183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,8,128,1,float16,fp8,0,0.12764267126719156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,1,128,1,float16,fp8,0,0.03150933235883713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,2,128,1,float16,float16,0,0.037647999823093414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,2,128,1,float16,fp8,0,0.039333333571751915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,4,128,1,float16,float16,0,0.05109333495299021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,4,128,1,float16,fp8,0,0.056890666484832764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,8,128,1,float16,fp8,0,0.06012799839178721
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,8,8,128,1,float16,float16,0,0.06417599817117055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,8,128,1,float16,fp8,0,0.0352906659245491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,8,128,1,float16,float16,0,0.031173333525657654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,1,128,1,float16,float16,0,0.021290667355060577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,1,128,1,float16,fp8,0,0.014890667051076889
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,2,128,1,float16,float16,0,0.023354666928450268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,4,128,1,float16,float16,0,0.02958400050799052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,4,128,1,float16,fp8,0,0.03346133232116699
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,8,128,1,float16,float16,0,0.03088533381621043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,8,128,1,float16,fp8,0,0.035173334181308746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,8,128,1,float16,fp8,0,0.021962667504946392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,1,128,1,float16,float16,0,0.014869333555301031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,2,128,1,float16,float16,0,0.015002666662136713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,2,128,1,float16,fp8,0,0.016906666258970898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,4,128,1,float16,float16,0,0.019253333409627277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,4,128,1,float16,fp8,0,0.021221332252025604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,8,2,128,1,float16,fp8,0,0.06402666866779327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,8,128,1,float16,float16,0,0.01923199991385142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,8,128,1,float16,fp8,0,0.021557333568731945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,8,128,1,float16,float16,0,0.014991999914248785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,1,128,1,float16,float16,0,0.012794667234023413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,1,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,2,128,1,float16,float16,0,0.013327999661366144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,2,128,1,float16,fp8,0,0.014826666563749313
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,4,128,1,float16,float16,0,0.015066667149464289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,4,128,1,float16,fp8,0,0.016879999389251072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,8,128,1,float16,float16,0,0.015109332899252573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,8,128,1,float16,fp8,0,0.01718933383623759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,8,2,128,1,float16,float16,0,0.01470400020480156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,8,128,1,float16,float16,0,0.013477332890033722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,2,128,1,float16,fp8,0,0.025098666548728943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,8,1,128,1,float16,fp8,0,0.02094399929046631
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,8,128,1,float16,fp8,0,0.014874666929244995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,1,128,1,float16,float16,0,0.01201066623131434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,2,128,1,float16,float16,0,0.012847999731699625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,2,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,4,128,1,float16,float16,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,8,128,1,float16,float16,0,0.019152000546455383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,4,128,1,float16,fp8,0,0.014805333067973455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,8,128,1,float16,fp8,0,0.01481066644191742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,8,128,1,float16,float16,0,0.013194666554530462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,8,1,128,1,float16,fp8,0,0.015103999525308609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,8,128,1,float16,float16,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,1,128,1,float16,float16,0,0.011482667177915573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,2,128,1,float16,float16,0,0.012752000242471695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,1,128,1,float16,fp8,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,2,128,1,float16,fp8,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,4,128,1,float16,float16,0,0.01251199965675672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,4,128,1,float16,fp8,0,0.01268799975514412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,8,128,1,float16,float16,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,8,128,1,float16,fp8,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,8,128,1,float16,fp8,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,8,128,1,float16,float16,0,0.01258133351802826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,1,128,1,float16,fp8,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,1,128,1,float16,float16,0,0.011301333705584208
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,2,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,8,8,128,1,float16,fp8,0,0.017152000218629837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,2,128,1,float16,fp8,0,0.012357333054145178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,4,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,4,128,1,float16,fp8,0,0.011744000017642975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,8,128,1,float16,float16,0,0.01240533341964086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,8,8,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,1,128,1,float16,float16,0,0.077824001510938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,1,128,1,float16,fp8,0,0.06983466446399689
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,2,128,1,float16,float16,0,0.09267733494440715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,2,128,1,float16,fp8,0,0.0876586635907491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,4,128,1,float16,float16,0,0.1377226710319519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,8,1,128,1,float16,fp8,0,0.012309333930412928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,8,4,128,1,float16,fp8,0,0.1332266628742218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,8,128,1,float16,fp8,0,0.06950399776299794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,8,128,1,float16,float16,0,0.0735040009021759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,1,128,1,float16,float16,0,0.043653334180514015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,1,128,1,float16,fp8,0,0.04062933226426443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,8,8,128,1,float16,fp8,0,0.012853333105643591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,2,128,1,float16,fp8,0,0.04755199948946635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,2,128,1,float16,float16,0,0.049914668003718056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,4,128,1,float16,float16,0,0.06286933521429698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,4,128,1,float16,fp8,0,0.06593599915504456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,8,128,1,float16,float16,0,0.03728000074625015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,8,128,1,float16,float16,0,0.07610133290290833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,8,8,128,1,float16,fp8,0,0.07034666836261749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,1,128,1,float16,float16,0,0.02714666724205017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,8,128,1,float16,fp8,0,0.04015466570854187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,1,128,1,float16,fp8,0,0.026005332668622334
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,2,128,1,float16,float16,0,0.029525332152843475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,4,128,1,float16,float16,0,0.03565866748491923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,2,128,1,float16,fp8,0,0.02958400050799052
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,4,128,1,float16,fp8,0,0.03963200002908707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,8,128,1,float16,fp8,0,0.03961066653331121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,8,128,1,float16,fp8,0,0.025248001019159954
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,1,128,1,float16,float16,0,0.017909333109855652
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,8,128,1,float16,float16,0,0.02310933421055476
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,8,8,128,1,float16,float16,0,0.03731200098991394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,2,128,1,float16,fp8,0,0.01926400015751521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,4,128,1,float16,float16,0,0.022986667851607006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,4,128,1,float16,fp8,0,0.023237332701683044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,8,128,1,float16,fp8,0,0.025274666647116344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,8,128,1,float16,float16,0,0.023071999351183575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,8,128,1,float16,float16,0,0.01509333277742068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,1,128,1,float16,float16,0,0.013088000317414602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,1,128,1,float16,fp8,0,0.013056000073750814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,2,128,1,float16,float16,0,0.014837333311637243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,4,128,1,float16,float16,0,0.01516266663869222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,2,128,1,float16,fp8,0,0.014906667172908783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,4,128,1,float16,fp8,0,0.016794666647911072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,8,128,1,float16,fp8,0,0.01720000058412552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,8,128,1,float16,float16,0,0.01534933348496755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,1,128,1,float16,float16,0,0.011354666203260422
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,8,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,1,128,1,float16,fp8,0,0.01180800050497055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,2,128,1,float16,fp8,0,0.012650666137536367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,4,128,1,float16,float16,0,0.01312000056107839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,4,128,1,float16,fp8,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,8,128,1,float16,float16,0,0.01309866706530253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,8,128,1,float16,fp8,0,0.014922666052977243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,8,128,1,float16,float16,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,8,128,1,float16,fp8,0,0.012682666381200155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,1,128,1,float16,fp8,0,0.011178666104873022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,2,128,1,float16,float16,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,2,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,4,128,1,float16,float16,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,4,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,8,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,8,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,1,128,1,float16,fp8,0,0.016938666502634685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,8,2,128,1,float16,float16,0,0.019007999449968338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,8,128,1,float16,fp8,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,8,128,1,float16,float16,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,1,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,2,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,1,128,1,float16,fp8,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,2,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,4,128,1,float16,float16,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,8,8,128,1,float16,fp8,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,4,128,1,float16,fp8,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,8,128,1,float16,float16,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,8,8,128,1,float16,fp8,0,0.011941333611806234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,8,128,1,float16,fp8,0,0.01109333336353302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,8,128,1,float16,float16,0,0.011381333072980246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,1,128,1,float16,float16,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,1,128,1,float16,fp8,0,0.011365332951148352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,2,128,1,float16,float16,0,0.010858666151762009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,4,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,2,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,8,128,1,float16,fp8,0,0.014858666807413101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,8,128,1,float16,float16,0,0.010816000401973724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,8,2,128,1,float16,float16,0,0.013013333082199097
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,8,128,1,float16,fp8,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,8,1,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,1,128,1,float16,float16,0,0.06901866694291432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,1,128,1,float16,fp8,0,0.06010133524735769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,2,128,1,float16,float16,0,0.07564799984296162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,2,128,1,float16,fp8,0,0.06653866668542226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,4,128,1,float16,float16,0,0.09478933612505595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,8,128,1,float16,float16,0,0.04987733562787374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,8,128,1,float16,fp8,0,0.04960533479849497
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,1,128,1,float16,float16,0,0.03977599988381068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,1,128,1,float16,fp8,0,0.035487999518712364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,2,128,1,float16,float16,0,0.041802664597829185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,8,4,128,1,float16,fp8,0,0.08441600203514099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,2,128,1,float16,fp8,0,0.03930133332808813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,4,128,1,float16,float16,0,0.04934933284918467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,4,128,1,float16,fp8,0,0.047695999344189964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,8,128,1,float16,float16,0,0.029626667499542236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,8,128,1,float16,fp8,0,0.049029335379600525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,8,8,128,1,float16,float16,0,0.04985066751639048
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,8,128,1,float16,fp8,0,0.030282666285832722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,1,128,1,float16,float16,0,0.02508266766866048
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,1,128,1,float16,fp8,0,0.023376000424226124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,8,4,128,1,float16,fp8,0,0.01166933278242747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,2,128,1,float16,float16,0,0.02719466636578242
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,2,128,1,float16,fp8,0,0.02516799916823705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,4,128,1,float16,float16,0,0.029359998802344005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,4,128,1,float16,fp8,0,0.02939733366171519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,8,128,1,float16,fp8,0,0.030746666093667347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,1,128,1,float16,float16,0,0.016837333639462788
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,8,8,128,1,float16,float16,0,0.02922666569550832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,8,128,1,float16,float16,0,0.018853332847356796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,1,128,1,float16,fp8,0,0.01595199977358182
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,2,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,2,128,1,float16,fp8,0,0.01720000058412552
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,4,128,1,float16,float16,0,0.019199999670187633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,4,128,1,float16,fp8,0,0.018906666586796444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,8,128,1,float16,fp8,0,0.01926933353145917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,8,128,1,float16,fp8,0,0.014783999572197596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,8,128,1,float16,float16,0,0.01894933357834816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,1,128,1,float16,float16,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,8,128,1,float16,float16,0,0.014773332824309668
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,1,128,1,float16,fp8,0,0.012784000486135483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,2,128,1,float16,float16,0,0.012634667257467905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,2,128,1,float16,fp8,0,0.01267733300725619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,4,128,1,float16,float16,0,0.01302933320403099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,4,128,1,float16,fp8,0,0.014709333578745524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,8,128,1,float16,fp8,0,0.014922666052977243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,1,128,1,float16,float16,0,0.011349332829316458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,8,8,128,1,float16,float16,0,0.013477332890033722
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,1,128,1,float16,fp8,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,2,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,8,128,1,float16,float16,0,0.012442667037248611
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,4,128,1,float16,float16,0,0.012736000120639801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,2,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,8,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,8,128,1,float16,float16,0,0.012335999558369318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,8,128,1,float16,float16,0,0.011285333583752314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,1,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,2,128,1,float16,float16,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,2,128,1,float16,fp8,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,4,128,1,float16,fp8,0,0.011391999820868174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,4,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,8,128,1,float16,float16,0,0.011168000598748526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,8,128,1,float16,fp8,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,8,128,1,float16,float16,0,0.010693332801262537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,8,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,1,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,2,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,1,128,1,float16,fp8,0,0.010714666297038397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,2,128,1,float16,fp8,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,4,128,1,float16,float16,0,0.0107893335322539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,4,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,8,128,1,float16,float16,0,0.011109333485364914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,8,8,128,1,float16,fp8,0,0.011247999966144562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,8,8,128,1,float16,fp8,0,0.01956266661485036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,1,128,1,float16,float16,0,0.010703999549150467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,8,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,1,128,1,float16,fp8,0,0.011173332730929056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,2,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,2,128,1,float16,fp8,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,4,128,1,float16,float16,0,0.010714666297038397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,4,128,1,float16,fp8,0,0.011247999966144562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,8,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,8,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,8,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,8,4,128,1,float16,fp8,0,0.011525332927703857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,1,128,1,float16,float16,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,1,128,1,float16,float16,0,0.06427200138568878
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,1,128,1,float16,fp8,0,0.056159997979799904
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,2,128,1,float16,float16,0,0.06809066732724507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,2,128,1,float16,fp8,0,0.060005332032839455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,8,128,1,float16,float16,0,0.043290664752324425
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,4,128,1,float16,float16,0,0.07505066692829132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,8,4,128,1,float16,fp8,0,0.06809600194295247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,1,128,1,float16,float16,0,0.039077334105968475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,8,8,128,1,float16,float16,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,1,128,1,float16,fp8,0,0.03345066557327906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,2,128,1,float16,float16,0,0.03977066775163015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,4,128,1,float16,float16,0,0.04271466533342997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,4,128,1,float16,fp8,0,0.03972800076007843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,8,128,1,float16,float16,0,0.02526933451493581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,8,128,1,float16,fp8,0,0.02508266766866048
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,1,128,1,float16,float16,0,0.02311466634273529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,1,128,1,float16,fp8,0,0.021274665991465252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,2,128,1,float16,float16,0,0.025008000433444977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,8,8,128,1,float16,fp8,0,0.01116266722480456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,2,128,1,float16,fp8,0,0.023061332603295643
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,4,128,1,float16,float16,0,0.02536533276240031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,4,128,1,float16,fp8,0,0.02497600018978119
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,8,128,1,float16,float16,0,0.026922665536403656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,8,8,128,1,float16,fp8,0,0.025221332907676697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,8,128,1,float16,fp8,0,0.01721599946419398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,8,128,1,float16,float16,0,0.01709866647919019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,1,128,1,float16,float16,0,0.015594666202863058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,1,128,1,float16,fp8,0,0.014805333067973455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,2,128,1,float16,float16,0,0.016821333517630894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,2,128,1,float16,fp8,0,0.015024000157912573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,8,128,1,float16,fp8,0,0.03984000037113825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,4,128,1,float16,float16,0,0.016858667135238647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,4,128,1,float16,fp8,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,8,128,1,float16,fp8,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,8,8,128,1,float16,float16,0,0.017184000462293625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,8,128,1,float16,fp8,0,0.012576000144084295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,1,128,1,float16,float16,0,0.012714666624863943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,1,128,1,float16,fp8,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,2,128,1,float16,float16,0,0.012784000486135483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,2,128,1,float16,fp8,0,0.012655999511480331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,2,128,1,float16,fp8,0,0.03554133325815201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,4,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,4,128,1,float16,fp8,0,0.012890666723251343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,8,128,1,float16,fp8,0,0.012746666868527731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,8,128,1,float16,float16,0,0.012746666868527731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,8,128,1,float16,float16,0,0.011695999652147293
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,1,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,8,128,1,float16,fp8,0,0.012901333471139273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,2,128,1,float16,float16,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,2,128,1,float16,fp8,0,0.011765333513418833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,4,128,1,float16,float16,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,8,128,1,float16,fp8,0,0.040362666050593056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,8,128,1,float16,float16,0,0.011749333391586939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,8,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,8,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,1,128,1,float16,float16,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,1,128,1,float16,fp8,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,2,128,1,float16,float16,0,0.010858666151762009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,2,128,1,float16,fp8,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,4,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,4,128,1,float16,fp8,0,0.011317333827416102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,8,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,8,8,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,8,128,1,float16,fp8,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,8,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,1,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,1,128,1,float16,fp8,0,0.010778666784365972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,2,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,2,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,4,128,1,float16,float16,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,4,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,8,128,1,float16,float16,0,0.010768000036478043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,8,8,128,1,float16,fp8,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,8,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,1,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,2,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,2,128,1,float16,fp8,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,4,128,1,float16,float16,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,4,128,1,float16,fp8,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,8,8,128,1,float16,float16,0,0.012784000486135483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,8,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,8,8,128,1,float16,float16,0,0.04348266621430715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,1,128,1,float16,float16,0,0.011312000453472137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,8,128,1,float16,fp8,0,0.011146667102972666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,1,128,1,float16,fp8,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,2,128,1,float16,float16,0,0.014752000570297241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,2,128,1,float16,fp8,0,0.012885333349307379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,4,128,1,float16,float16,0,0.019029332945744198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,8,4,128,1,float16,fp8,0,0.017077332983414333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,8,128,1,float16,float16,0,0.019007999449968338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,8,128,1,float16,fp8,0,0.01717866708834966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,2,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,2,128,1,float16,fp8,0,0.010709332923094431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,4,128,1,float16,float16,0,0.013173333058754602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,4,128,1,float16,fp8,0,0.013045333325862885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,8,128,1,float16,float16,0,0.019039999693632126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,8,4,128,1,float16,fp8,0,0.012709333250919977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,8,128,1,float16,float16,0,0.013957332819700241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,8,128,1,float16,fp8,0,0.01301866645614306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,1,128,1,float16,float16,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,2,128,1,float16,float16,0,0.009535999968647957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,2,128,1,float16,fp8,0,0.010005333150426546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,8,128,1,float16,fp8,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,4,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,1,128,1,float16,float16,0,0.010725333044926325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,1,128,1,float16,fp8,0,0.010650667051474253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,4,128,1,float16,fp8,0,0.01071999967098236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,8,128,1,float16,float16,0,0.0145066666106383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,8,128,1,float16,float16,0,0.01081066702802976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,8,128,1,float16,fp8,0,0.010661333799362183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,1,128,1,float16,float16,0,0.008869333192706108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,1,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,2,128,1,float16,float16,0,0.008938666433095932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,2,128,1,float16,fp8,0,0.009984000275532404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,4,128,1,float16,float16,0,0.008992000172535578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,4,128,1,float16,fp8,0,0.009056000038981438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,8,8,128,1,float16,fp8,0,0.016895999511082966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,8,128,1,float16,fp8,0,0.0106133334338665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,8,8,128,1,float16,float16,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,8,128,1,float16,float16,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,8,128,1,float16,fp8,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,2,128,1,float16,float16,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,2,128,1,float16,fp8,0,0.009039999917149544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,4,128,1,float16,float16,0,0.008634666601816813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,1,128,1,float16,fp8,0,0.009328000247478485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,8,128,1,float16,float16,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,8,128,1,float16,fp8,0,0.008853333070874214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,8,128,1,float16,float16,0,0.009072000160813332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,8,128,1,float16,fp8,0,0.009050666665037474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,1,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,1,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,2,128,1,float16,fp8,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,2,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,4,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,4,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,8,128,1,float16,float16,0,0.008752000207702318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,8,8,128,1,float16,fp8,0,0.008746666833758354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,8,8,128,1,float16,fp8,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,8,128,1,float16,float16,0,0.00860799973209699
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,8,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,8,8,128,1,float16,fp8,0,0.010858666151762009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,1,128,1,float16,float16,0,0.008703999842206636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,1,128,1,float16,fp8,0,0.008650666723648706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,2,128,1,float16,float16,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,2,128,1,float16,fp8,0,0.008789333204428354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,4,128,1,float16,float16,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,8,128,1,float16,float16,0,0.008656000097592672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,8,128,1,float16,float16,0,0.008826666822036108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,1,128,1,float16,float16,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,1,128,1,float16,fp8,0,0.008714666590094566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,2,128,1,float16,float16,0,0.008736000085870424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,2,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,4,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,1,128,1,float16,float16,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,4,128,1,float16,fp8,0,0.009072000160813332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,8,128,1,float16,fp8,0,0.00892800030608972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,8,128,1,float16,float16,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,4,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,8,128,1,float16,float16,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,1,128,1,float16,float16,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,8,128,1,float16,fp8,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,1,128,1,float16,fp8,0,0.008762666955590248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,2,128,1,float16,fp8,0,0.008656000097592672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,2,128,1,float16,float16,0,0.0086666668454806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,4,128,1,float16,float16,0,0.008682666967312494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,4,128,1,float16,fp8,0,0.008826666822036108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,8,128,1,float16,fp8,0,0.00891733355820179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,8,8,128,1,float16,float16,0,0.00877333308259646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,4,1,128,1,float16,fp8,0,1.5824480056762695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,4,1,128,1,float16,float16,0,1.8916640281677246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,4,2,128,1,float16,fp8,0,1.7291253407796223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,4,2,128,1,float16,float16,0,1.9859573046366374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,4,128,1,float16,fp8,0,0.008778666456540426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,8,8,128,1,float16,fp8,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,4,128,1,float16,fp8,0,1.1164320309956868
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,4,128,1,float16,float16,0,1.227552016576131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,8,8,128,1,float16,fp8,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,1,128,1,float16,float16,0,0.9802772998809814
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,1,128,1,float16,fp8,0,0.9104800224304199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,2,128,1,float16,float16,0,1.0537226994832356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,2,128,1,float16,fp8,0,0.9316799640655518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,1,128,1,float16,float16,0,0.5566346645355225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,8,1,128,1,float16,fp8,0,0.00901333304742972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,4,128,1,float16,fp8,0,1.1232906977335613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,4,4,128,1,float16,float16,0,1.2226293087005615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,1,128,1,float16,fp8,0,0.486789345741272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,2,128,1,float16,float16,0,0.5835839907328287
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,2,128,1,float16,fp8,0,0.5501013199488322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,4,128,1,float16,float16,0,0.3935519854227702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,4,128,1,float16,float16,0,0.6663839817047119
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,4,128,1,float16,fp8,0,0.6254826784133911
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,4,128,1,float16,fp8,0,0.3633973201115926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,1,128,1,float16,float16,0,0.33318400382995605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,1,128,1,float16,fp8,0,0.31594133377075195
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,2,128,1,float16,float16,0,0.3438719908396403
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,2,128,1,float16,fp8,0,0.3239786624908447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,4,128,1,float16,float16,0,0.39660799503326416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,4,1,128,1,float16,float16,0,1.1136853694915771
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,4,1,128,1,float16,fp8,0,0.9679520130157471
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,4,2,128,1,float16,float16,0,1.2034080028533936
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,4,2,128,1,float16,fp8,0,1.066650629043579
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,4,128,1,float16,float16,0,0.7704586982727051
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,4,128,1,float16,fp8,0,0.7322826385498047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,1,128,1,float16,fp8,0,0.5294346809387207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,2,128,1,float16,float16,0,0.642959992090861
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,4,128,1,float16,float16,0,0.7704799969991049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,4,128,1,float16,fp8,0,0.7269546985626221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,4,128,1,float16,float16,0,0.42812267939249676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,4,128,1,float16,fp8,0,0.6248693466186523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,4,128,1,float16,fp8,0,0.41099198659261066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,1,128,1,float16,fp8,0,0.3130026658376058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,2,128,1,float16,float16,0,0.36141331990559894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,2,128,1,float16,fp8,0,0.3354506492614746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,4,128,1,float16,float16,0,0.4275039831797282
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,4,128,1,float16,fp8,0,0.4105386734008789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,4,4,128,1,float16,fp8,0,0.3577599922815959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,1,128,1,float16,float16,0,0.21572800477345785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,1,128,1,float16,fp8,0,0.2019946575164795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,2,128,1,float16,float16,0,0.2239039937655131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,4,128,1,float16,float16,0,0.2495786746342977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,1,128,1,float16,float16,0,0.5998080174128214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,4,128,1,float16,fp8,0,0.2379093368848165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,4,2,128,1,float16,fp8,0,0.5907520055770874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,4,1,128,1,float16,float16,0,0.80841596921285
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,4,1,128,1,float16,fp8,0,0.7043146292368571
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,4,2,128,1,float16,float16,0,0.8726933002471924
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,4,4,128,1,float16,float16,0,0.6665440003077189
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,4,2,128,1,float16,fp8,0,0.7916746934254965
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,4,1,128,1,float16,float16,0,0.3354933261871338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,1,128,1,float16,float16,0,0.4353333314259847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,1,128,1,float16,fp8,0,0.389845331509908
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,2,128,1,float16,float16,0,0.47495468457539874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,2,128,1,float16,fp8,0,0.44263466199239093
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,4,128,1,float16,float16,0,0.5802079836527506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,4,128,1,float16,float16,0,0.33588798840840656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,4,128,1,float16,fp8,0,0.5606613159179688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,4,128,1,float16,fp8,0,0.23865600426991782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,4,128,1,float16,fp8,0,0.31214932600657147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,1,128,1,float16,float16,0,0.25710399945576984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,2,128,1,float16,fp8,0,0.2108693321545919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,2,128,1,float16,float16,0,0.26918933788935345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,2,128,1,float16,fp8,0,0.2577173312505086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,4,128,1,float16,float16,0,0.3362666765848796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,4,128,1,float16,float16,0,0.18332266807556152
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,4,128,1,float16,fp8,0,0.32012800375620526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,4,128,1,float16,fp8,0,0.1800266702969869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,1,128,1,float16,fp8,0,0.14918399850527445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,1,128,1,float16,float16,0,0.16006933649381003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,2,128,1,float16,float16,0,0.16841065883636475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,2,128,1,float16,fp8,0,0.1588746706644694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,4,128,1,float16,fp8,0,0.1792373259862264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,4,4,128,1,float16,float16,0,0.18264534076054892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,4,1,128,1,float16,float16,0,1.03438401222229
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,4,1,128,1,float16,fp8,0,0.9006079832712809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,4,2,128,1,float16,fp8,0,1.045093297958374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,4,2,128,1,float16,float16,0,1.1416746775309246
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,4,128,1,float16,float16,0,0.5798559983571371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,4,4,128,1,float16,fp8,0,0.5588159958521525
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,4,128,1,float16,fp8,0,0.7464213371276855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,1,128,1,float16,float16,0,0.5416533152262369
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,1,128,1,float16,fp8,0,0.488047997156779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,2,128,1,float16,float16,0,0.607093334197998
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,2,128,1,float16,fp8,0,0.5630826552708944
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,4,128,1,float16,float16,0,0.4176746606826782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,4,128,1,float16,fp8,0,0.41201599438985187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,4,4,128,1,float16,float16,0,0.24791467189788818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,4,128,1,float16,fp8,0,0.7519733111063639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,1,128,1,float16,float16,0,0.3012320001920064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,1,128,1,float16,fp8,0,0.2758400042851766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,2,128,1,float16,float16,0,0.3354560136795044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,2,128,1,float16,fp8,0,0.3115680019060771
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,4,128,1,float16,float16,0,0.4181600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,4,1,128,1,float16,fp8,0,0.24177600940068564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,4,128,1,float16,float16,0,0.24619199832280478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,1,128,1,float16,float16,0,0.18889067570368448
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,4,128,1,float16,fp8,0,0.23416000604629517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,1,128,1,float16,fp8,0,0.1761066714922587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,2,128,1,float16,float16,0,0.19658132394154867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,2,128,1,float16,fp8,0,0.18954133987426758
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,4,128,1,float16,float16,0,0.24824533859888712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,4,4,128,1,float16,fp8,0,0.23429866631825766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,4,128,1,float16,float16,0,0.13917332887649536
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,1,128,1,float16,float16,0,0.12011200189590454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,1,128,1,float16,fp8,0,0.11395200093587239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,2,128,1,float16,float16,0,0.12501866618792215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,2,128,1,float16,fp8,0,0.12075733145078023
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,4,128,1,float16,float16,0,0.13846932848294577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,4,128,1,float16,fp8,0,0.1383626659711202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,4,1,128,1,float16,float16,0,0.6262293259302775
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,4,1,128,1,float16,fp8,0,0.5610613425572714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,4,2,128,1,float16,float16,0,0.7062986691792806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,4,2,128,1,float16,fp8,0,0.6679840087890625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,4,128,1,float16,float16,0,0.508895993232727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,4,128,1,float16,fp8,0,0.5034346580505371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,1,128,1,float16,fp8,0,0.31278934081395465
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,1,128,1,float16,float16,0,0.3368053436279297
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,4,128,1,float16,float16,0,0.7756853103637695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,2,128,1,float16,float16,0,0.3818613290786743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,2,128,1,float16,fp8,0,0.36572265625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,4,128,1,float16,float16,0,0.508842666943868
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,4,4,128,1,float16,fp8,0,0.503056009610494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,4,128,1,float16,float16,0,0.2822773257891337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,4,4,128,1,float16,float16,0,0.7732533613840739
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,1,128,1,float16,float16,0,0.19178666671117148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,1,128,1,float16,fp8,0,0.18032532930374146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,2,128,1,float16,float16,0,0.21335999170939127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,2,128,1,float16,fp8,0,0.19569599628448486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,4,128,1,float16,float16,0,0.28273600339889526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,4,128,1,float16,float16,0,0.158053328593572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,4,4,128,1,float16,fp8,0,0.41329065958658856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,4,128,1,float16,fp8,0,0.15260266264279684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,1,128,1,float16,float16,0,0.125408003727595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,2,128,1,float16,float16,0,0.1311893363793691
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,2,128,1,float16,fp8,0,0.12713600198427835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,4,128,1,float16,fp8,0,0.15236799915631613
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,4,4,128,1,float16,fp8,0,0.1385546624660492
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,4,128,1,float16,fp8,0,0.10729599992434184
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,1,128,1,float16,float16,0,0.091648002465566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,1,128,1,float16,fp8,0,0.0885813335577647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,2,128,1,float16,float16,0,0.09702933828035991
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,2,128,1,float16,fp8,0,0.09325866897900899
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,4,128,1,float16,float16,0,0.10689600308736165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,4,128,1,float16,fp8,0,0.10729066530863444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,4,1,128,1,float16,float16,0,0.6048426628112793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,4,1,128,1,float16,fp8,0,0.5522933403650919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,4,128,1,float16,fp8,0,0.27928000688552856
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,4,2,128,1,float16,float16,0,0.7258240381876627
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,4,2,128,1,float16,fp8,0,0.6948106288909912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,4,128,1,float16,float16,0,0.5528586705525717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,4,128,1,float16,fp8,0,0.556714653968811
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,4,4,128,1,float16,fp8,0,0.28033600250879925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,1,128,1,float16,fp8,0,0.11929600437482198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,2,128,1,float16,float16,0,0.38447999954223633
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,2,128,1,float16,fp8,0,0.37563733259836835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,4,4,128,1,float16,float16,0,0.15685333808263144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,4,128,1,float16,float16,0,0.5516159931818644
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,1,128,1,float16,float16,0,0.3236959973971049
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,4,4,128,1,float16,float16,0,0.1055466632048289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,4,128,1,float16,fp8,0,0.55731201171875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,4,128,1,float16,float16,0,0.29869333902994794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,4,128,1,float16,fp8,0,0.2997386654218038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,1,128,1,float16,float16,0,0.18227734168370566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,1,128,1,float16,fp8,0,0.166703999042511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,2,128,1,float16,fp8,0,0.20350400606791177
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,2,128,1,float16,float16,0,0.21198934316635132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,4,128,1,float16,float16,0,0.29554667075475055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,4,128,1,float16,float16,0,0.17177067200342813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,1,128,1,float16,float16,0,0.11172266801198323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,4,128,1,float16,fp8,0,0.16544000307718912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,1,128,1,float16,fp8,0,0.10668266812960307
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,2,128,1,float16,float16,0,0.12119999527931213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,2,128,1,float16,fp8,0,0.11968533198038737
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,4,128,1,float16,float16,0,0.1711946725845337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,4,4,128,1,float16,fp8,0,0.154639999071757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,4,128,1,float16,float16,0,0.08897599577903748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,4,128,1,float16,fp8,0,0.09262933333714803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,1,128,1,float16,fp8,0,0.06806399921576183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,2,128,1,float16,float16,0,0.07508799930413564
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,2,128,1,float16,fp8,0,0.07417599856853485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,4,128,1,float16,float16,0,0.08893332878748576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,4,128,1,float16,fp8,0,0.09266133109728496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,4,128,1,float16,float16,0,0.07246399919191997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,4,128,1,float16,fp8,0,0.07443200051784515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,1,128,1,float16,fp8,0,0.062165334820747375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,2,128,1,float16,float16,0,0.06705600023269653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,2,128,1,float16,fp8,0,0.06611200173695882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,4,128,1,float16,float16,0,0.07274133463700612
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,4,128,1,float16,fp8,0,0.07507733503977458
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,4,1,128,1,float16,float16,0,0.38116268316904706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,4,1,128,1,float16,fp8,0,0.35760001341501874
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,4,2,128,1,float16,float16,0,0.46652265389760333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,4,2,128,1,float16,fp8,0,0.4644746780395508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,1,128,1,float16,float16,0,0.20814400911331177
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,1,128,1,float16,fp8,0,0.2021333376566569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,4,128,1,float16,fp8,0,0.39075199762980145
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,2,128,1,float16,float16,0,0.255295991897583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,2,128,1,float16,fp8,0,0.25485867261886597
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,4,1,128,1,float16,fp8,0,0.30021866162618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,4,128,1,float16,float16,0,0.38091735045115155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,4,128,1,float16,float16,0,0.20921599864959717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,4,128,1,float16,fp8,0,0.2134666641553243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,1,128,1,float16,float16,0,0.1172213355700175
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,4,4,128,1,float16,fp8,0,0.2999573349952698
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,1,128,1,float16,fp8,0,0.1127786636352539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,2,128,1,float16,float16,0,0.1414293348789215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,2,128,1,float16,fp8,0,0.1360106666882833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,4,1,128,1,float16,float16,0,0.06934399902820587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,4,128,1,float16,float16,0,0.10586133599281311
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,4,128,1,float16,fp8,0,0.10894399881362915
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,1,128,1,float16,float16,0,0.07589333256085713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,1,128,1,float16,fp8,0,0.0742986649274826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,2,128,1,float16,float16,0,0.084714670976003
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,4,1,128,1,float16,float16,0,0.06383466720581055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,2,128,1,float16,fp8,0,0.08381332953770955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,4,128,1,float16,fp8,0,0.10858666896820068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,4,128,1,float16,float16,0,0.06828799843788147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,4,128,1,float16,fp8,0,0.07241599758466084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,1,128,1,float16,float16,0,0.05429333448410034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,1,128,1,float16,fp8,0,0.05407466491063436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,2,128,1,float16,float16,0,0.059530665477116905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,2,128,1,float16,fp8,0,0.05835733314355215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,4,128,1,float16,float16,0,0.06850133339564006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,4,4,128,1,float16,fp8,0,0.07215466598669688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,4,128,1,float16,float16,0,0.05824000140031179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,1,128,1,float16,float16,0,0.05123733480771383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,4,128,1,float16,float16,0,0.37992000579833984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,1,128,1,float16,fp8,0,0.0495413343111674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,2,128,1,float16,float16,0,0.053317333261171974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,2,128,1,float16,fp8,0,0.0536053329706192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,4,128,1,float16,float16,0,0.05813866853713989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,4,128,1,float16,fp8,0,0.05978133281071981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,4,4,128,1,float16,fp8,0,0.3892853260040283
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,4,1,128,1,float16,float16,0,0.3934719959894816
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,4,128,1,float16,float16,0,0.2095253268877665
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,4,1,128,1,float16,fp8,0,0.37510931491851807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,4,4,128,1,float16,fp8,0,0.21277866760889688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,4,2,128,1,float16,float16,0,0.5125066836675009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,4,128,1,float16,float16,0,0.4411626656850179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,1,128,1,float16,float16,0,0.21583465735117593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,4,4,128,1,float16,float16,0,0.10967999696731567
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,4,2,128,1,float16,fp8,0,0.5154560009638468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,1,128,1,float16,fp8,0,0.2066133419672648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,2,128,1,float16,fp8,0,0.27348800500233966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,2,128,1,float16,float16,0,0.2688693404197693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,4,128,1,float16,float16,0,0.43908798694610596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,4,128,1,float16,float16,0,0.234117329120636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,4,128,1,float16,fp8,0,0.45554133256276447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,1,128,1,float16,fp8,0,0.10938133796056111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,1,128,1,float16,float16,0,0.12005333105723064
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,4,128,1,float16,fp8,0,0.2427306572596232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,2,128,1,float16,float16,0,0.15266133348147073
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,2,128,1,float16,fp8,0,0.14891200264294943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,4,4,128,1,float16,fp8,0,0.059818665186564125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,4,128,1,float16,float16,0,0.23640533288319907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,4,4,128,1,float16,fp8,0,0.24304000536600748
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,4,128,1,float16,float16,0,0.12971733013788858
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,4,128,1,float16,fp8,0,0.11560533444086711
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,1,128,1,float16,float16,0,0.07239999870459239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,1,128,1,float16,fp8,0,0.06968000034491222
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,2,128,1,float16,float16,0,0.0809386670589447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,2,128,1,float16,fp8,0,0.08125333487987518
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,4,128,1,float16,float16,0,0.1285706659158071
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,4,4,128,1,float16,fp8,0,0.11943999926249187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,1,128,1,float16,float16,0,0.043706665436426796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,4,128,1,float16,float16,0,0.06381333371003468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,1,128,1,float16,fp8,0,0.04364799956480662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,2,128,1,float16,float16,0,0.049973333875338234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,2,128,1,float16,fp8,0,0.051738664507865906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,4,128,1,float16,float16,0,0.06423466900984447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,4,128,1,float16,fp8,0,0.06902933120727539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,4,128,1,float16,float16,0,0.049322664737701416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,4,128,1,float16,fp8,0,0.05171733101209005
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,1,128,1,float16,fp8,0,0.03944533318281174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,2,128,1,float16,float16,0,0.042021334171295166
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,2,128,1,float16,fp8,0,0.04354666670163473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,4,128,1,float16,float16,0,0.049253334601720176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,1,128,1,float16,float16,0,0.037674665451049805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,4,128,1,float16,fp8,0,0.043653334180514015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,1,128,1,float16,fp8,0,0.03735466549793879
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,2,128,1,float16,float16,0,0.03925333420435587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,4,128,1,float16,float16,0,0.04166933397452036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,4,128,1,float16,fp8,0,0.043391997615496315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,4,4,128,1,float16,fp8,0,0.45812265078226727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,4,1,128,1,float16,float16,0,0.26707732677459717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,4,1,128,1,float16,fp8,0,0.25464000304539997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,4,2,128,1,float16,float16,0,0.35037867228190106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,4,128,1,float16,float16,0,0.3162720004717509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,4,2,128,1,float16,fp8,0,0.36458667119344074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,1,128,1,float16,float16,0,0.14748266339302063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,4,128,1,float16,fp8,0,0.33215999603271484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,1,128,1,float16,fp8,0,0.1416000028451284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,4,4,128,1,float16,fp8,0,0.06811200082302094
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,2,128,1,float16,float16,0,0.1925493280092875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,2,128,1,float16,fp8,0,0.19361066818237305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,1,128,1,float16,float16,0,0.039520000418027244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,4,128,1,float16,float16,0,0.32128000259399414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,4,4,128,1,float16,fp8,0,0.3309440016746521
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,4,4,128,1,float16,fp8,0,0.051872000098228455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,1,128,1,float16,float16,0,0.08060266574223836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,4,128,1,float16,float16,0,0.04197333256403605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,1,128,1,float16,fp8,0,0.07774400214354198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,4,128,1,float16,fp8,0,0.17726399501164755
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,4,128,1,float16,float16,0,0.17090133825937906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,2,128,1,float16,float16,0,0.10393066207567851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,4,2,128,1,float16,fp8,0,0.039621333281199135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,2,128,1,float16,fp8,0,0.09571733077367146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,1,128,1,float16,float16,0,0.05205333232879639
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,4,128,1,float16,fp8,0,0.17566933234532675
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,4,128,1,float16,float16,0,0.0846666693687439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,4,4,128,1,float16,float16,0,0.17023466030756632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,1,128,1,float16,fp8,0,0.05005866785844167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,2,128,1,float16,float16,0,0.06047999858856201
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,2,128,1,float16,fp8,0,0.06042666733264923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,4,128,1,float16,fp8,0,0.08653866251309712
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,1,128,1,float16,float16,0,0.03559466699759165
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,4,128,1,float16,float16,0,0.08363733688990276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,4,128,1,float16,fp8,0,0.055455997586250305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,1,128,1,float16,fp8,0,0.03583999971548716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,2,128,1,float16,float16,0,0.039605334401130676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,2,128,1,float16,fp8,0,0.04168533285458883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,4,128,1,float16,fp8,0,0.055445333321889244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,4,128,1,float16,float16,0,0.04961599906285604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,4,128,1,float16,fp8,0,0.04173333446184794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,1,128,1,float16,float16,0,0.03344533344109853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,1,128,1,float16,fp8,0,0.03148799886306127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,2,128,1,float16,float16,0,0.035455999275048576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,4,128,1,float16,float16,0,0.039701332648595176
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,4,128,1,float16,fp8,0,0.04167999823888143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,4,128,1,float16,float16,0,0.03566399961709976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,4,128,1,float16,fp8,0,0.03530666728814443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,1,128,1,float16,float16,0,0.031317333380381264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,1,128,1,float16,fp8,0,0.029317334294319153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,2,128,1,float16,float16,0,0.031199999153614044
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,2,128,1,float16,fp8,0,0.03139200061559677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,4,128,1,float16,float16,0,0.03528533379236857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,4,4,128,1,float16,fp8,0,0.0353973334034284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,4,1,128,1,float16,float16,0,0.2929973403612773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,4,1,128,1,float16,fp8,0,0.28596266110738117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,4,4,128,1,float16,fp8,0,0.08582933743794759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,4,2,128,1,float16,float16,0,0.40620799859364826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,4,4,128,1,float16,float16,0,0.05008000135421753
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,4,2,128,1,float16,fp8,0,0.4291573365529378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,4,128,1,float16,fp8,0,0.4060426553090413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,4,128,1,float16,float16,0,0.03942399968703588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,1,128,1,float16,float16,0,0.15811733404795328
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,4,128,1,float16,float16,0,0.38966933886210126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,1,128,1,float16,fp8,0,0.15440000096956888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,2,128,1,float16,float16,0,0.22098666429519653
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,4,128,1,float16,float16,0,0.3945653438568115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,1,128,1,float16,float16,0,0.08707732955614726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,2,128,1,float16,fp8,0,0.22694400946299234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,4,128,1,float16,fp8,0,0.21437333027521768
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,1,128,1,float16,fp8,0,0.08091199894746144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,4,4,128,1,float16,fp8,0,0.4103786547978719
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,2,128,1,float16,float16,0,0.1223306655883789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,2,128,1,float16,fp8,0,0.11809066931406657
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,4,128,1,float16,fp8,0,0.21368000904719034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,4,128,1,float16,float16,0,0.20252267519632974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,4,128,1,float16,float16,0,0.10690133770306905
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,1,128,1,float16,float16,0,0.050442665815353394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,1,128,1,float16,fp8,0,0.0513919989267985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,4,128,1,float16,float16,0,0.11071466406186421
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,2,128,1,float16,float16,0,0.06190933287143707
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,2,128,1,float16,fp8,0,0.06400533517201741
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,4,128,1,float16,fp8,0,0.09833066662152608
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,1,128,1,float16,float16,0,0.031130666534105938
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,4,128,1,float16,fp8,0,0.05709866682688395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,1,128,1,float16,fp8,0,0.03294399877389272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,2,128,1,float16,float16,0,0.03736000011364619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,2,128,1,float16,fp8,0,0.038890667259693146
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,4,128,1,float16,float16,0,0.0498986691236496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,4,128,1,float16,float16,0,0.03723733375469843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,4,128,1,float16,fp8,0,0.03988266736268997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,1,128,1,float16,float16,0,0.027429332335789997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,1,128,1,float16,fp8,0,0.02712533374627431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,2,128,1,float16,float16,0,0.02998399982849757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,2,128,1,float16,fp8,0,0.03156266609827677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,4,128,1,float16,fp8,0,0.03997866561015447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,4,4,128,1,float16,float16,0,0.03738133360942205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,4,128,1,float16,float16,0,0.029546665648619335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,4,2,128,1,float16,fp8,0,0.03516799956560135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,4,128,1,float16,fp8,0,0.03141866624355316
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,1,128,1,float16,fp8,0,0.02534399926662445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,4,4,128,1,float16,float16,0,0.20342934131622314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,2,128,1,float16,float16,0,0.02696000039577484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,2,128,1,float16,fp8,0,0.027386667827765148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,4,128,1,float16,float16,0,0.029274667302767437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,4,128,1,float16,float16,0,0.027066667874654133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,4,128,1,float16,fp8,0,0.027301333844661713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,4,128,1,float16,fp8,0,0.031248000760873158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,1,128,1,float16,float16,0,0.023290666441122692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,1,128,1,float16,fp8,0,0.023210667073726654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,2,128,1,float16,float16,0,0.025050667424996693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,2,128,1,float16,fp8,0,0.02515200028816859
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,4,128,1,float16,float16,0,0.02720533311367035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,4,4,128,1,float16,fp8,0,0.02697066714366277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,4,128,1,float16,float16,0,0.0499893327554067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,4,4,128,1,float16,fp8,0,0.09698133667310078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,4,4,128,1,float16,fp8,0,0.05728533367315928
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,4,1,128,1,float16,float16,0,0.25014932950337726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,4,1,128,1,float16,fp8,0,0.240282674630483
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,4,2,128,1,float16,float16,0,0.3617120186487834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,4,2,128,1,float16,fp8,0,0.3821973403294881
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,4,128,1,float16,fp8,0,0.38632531960805255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,1,128,1,float16,float16,0,0.13342932860056558
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,4,1,128,1,float16,float16,0,0.025311999022960663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,1,128,1,float16,fp8,0,0.13285332918167114
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,2,128,1,float16,float16,0,0.19763733943303427
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,4,128,1,float16,float16,0,0.187717338403066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,2,128,1,float16,fp8,0,0.20361065864562988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,4,128,1,float16,fp8,0,0.19936533768971762
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,1,128,1,float16,float16,0,0.0740586668252945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,1,128,1,float16,fp8,0,0.06630399823188782
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,4,128,1,float16,float16,0,0.36448001861572266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,2,128,1,float16,float16,0,0.10684266686439514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,4,128,1,float16,float16,0,0.09938133756319682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,4,128,1,float16,float16,0,0.18819733460744223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,2,128,1,float16,fp8,0,0.1069493293762207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,4,4,128,1,float16,fp8,0,0.19979733228683472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,1,128,1,float16,float16,0,0.041008000572522484
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,4,128,1,float16,fp8,0,0.0888426701227824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,2,128,1,float16,float16,0,0.051818668842315674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,1,128,1,float16,fp8,0,0.04177066683769226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,2,128,1,float16,fp8,0,0.054005334774653115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,4,128,1,float16,fp8,0,0.08801066875457764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,4,4,128,1,float16,float16,0,0.09698667128880818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,4,128,1,float16,float16,0,0.04364799956480662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,1,128,1,float16,float16,0,0.02515200028816859
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,4,128,1,float16,fp8,0,0.051589335004488625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,1,128,1,float16,fp8,0,0.027077332139015198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,2,128,1,float16,float16,0,0.031136001149813335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,2,128,1,float16,fp8,0,0.03334933271010717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,4,128,1,float16,fp8,0,0.05161599814891815
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,4,4,128,1,float16,float16,0,0.04375466704368591
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,4,128,1,float16,float16,0,0.031119999786218006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,4,128,1,float16,fp8,0,0.035258665680885315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,1,128,1,float16,float16,0,0.021055998901526134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,1,128,1,float16,fp8,0,0.02293333411216736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,2,128,1,float16,float16,0,0.024304000039895374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,2,128,1,float16,fp8,0,0.025205334027608235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,4,128,1,float16,fp8,0,0.03491200009981791
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,4,4,128,1,float16,float16,0,0.030991998811562855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,4,128,1,float16,float16,0,0.3590773344039917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,1,128,1,float16,float16,0,0.019050666441520054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,1,128,1,float16,fp8,0,0.01926933353145917
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,2,128,1,float16,float16,0,0.020997333029905956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,4,128,1,float16,float16,0,0.023269332945346832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,4,128,1,float16,fp8,0,0.025311999022960663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,4,128,1,float16,float16,0,0.02111999938885371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,4,128,1,float16,fp8,0,0.021173333128293354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,2,128,1,float16,fp8,0,0.021253332495689392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,1,128,1,float16,float16,0,0.01725333308180173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,1,128,1,float16,fp8,0,0.01739199956258138
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,2,128,1,float16,float16,0,0.01916266605257988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,2,128,1,float16,fp8,0,0.019285333653291065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,4,4,128,1,float16,fp8,0,0.38736534118652344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,4,128,1,float16,float16,0,0.02091199904680252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,4,4,128,1,float16,fp8,0,0.021055998901526134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,4,128,1,float16,float16,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,4,128,1,float16,fp8,0,0.01899733394384384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,1,128,1,float16,fp8,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,2,128,1,float16,float16,0,0.01716800034046173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,2,128,1,float16,fp8,0,0.017194667210181553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,4,128,1,float16,float16,0,0.018911999960740406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,4,128,1,float16,fp8,0,0.018933333456516266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,4,1,128,1,float16,float16,0,0.017242666333913803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,4,128,1,float16,float16,0,0.17830934127171835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,4,128,1,float16,fp8,0,0.19134400288263956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,4,1,128,1,float16,fp8,0,0.1153600017229716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,4,2,128,1,float16,float16,0,0.1703946590423584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,4,2,128,1,float16,fp8,0,0.18242132663726807
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,4,128,1,float16,float16,0,0.023226665953795116
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,1,128,1,float16,float16,0,0.06374933322270711
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,2,128,1,float16,float16,0,0.09796800216039021
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,2,128,1,float16,fp8,0,0.09759466846783955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,4,128,1,float16,float16,0,0.0918880005677541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,4,128,1,float16,float16,0,0.17787200212478638
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,4,128,1,float16,fp8,0,0.08233066896597545
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,4,128,1,float16,fp8,0,0.19157334168752035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,1,128,1,float16,float16,0,0.033546666304270424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,1,128,1,float16,fp8,0,0.035429333647092186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,2,128,1,float16,float16,0,0.04451733330885569
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,2,128,1,float16,fp8,0,0.047770669062932335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,4,128,1,float16,float16,0,0.04022933294375738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,4,128,1,float16,fp8,0,0.08249600231647491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,4,128,1,float16,fp8,0,0.047824000318845115
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,1,128,1,float16,float16,0,0.021568000316619873
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,1,128,1,float16,fp8,0,0.023232000569502514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,2,128,1,float16,fp8,0,0.029946667452653248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,4,128,1,float16,float16,0,0.04107200105985006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,4,128,1,float16,fp8,0,0.047882666190465294
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,4,128,1,float16,float16,0,0.027189334233601887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,4,128,1,float16,fp8,0,0.03150933235883713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,1,128,1,float16,float16,0,0.017269333203633625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,1,128,1,float16,fp8,0,0.01930133377512296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,2,128,1,float16,float16,0,0.02086399992307027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,4,1,128,1,float16,float16,0,0.11258133252461751
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,4,128,1,float16,float16,0,0.02741866558790207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,4,128,1,float16,fp8,0,0.031317333380381264
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,4,128,1,float16,fp8,0,0.02333866556485494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,4,4,128,1,float16,fp8,0,0.025274666647116344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,1,128,1,float16,float16,0,0.016010666886965435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,2,128,1,float16,float16,0,0.016810666769742966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,2,128,1,float16,fp8,0,0.01886933296918869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,4,1,128,1,float16,fp8,0,0.06032533446947733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,4,128,1,float16,float16,0,0.020848001043001812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,4,128,1,float16,fp8,0,0.02294933299223582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,4,128,1,float16,float16,0,0.01709866647919019
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,1,128,1,float16,float16,0,0.014943999548753103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,1,128,1,float16,fp8,0,0.015066667149464289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,2,128,1,float16,float16,0,0.015365333606799444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,2,128,1,float16,fp8,0,0.017055999487638474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,4,128,1,float16,float16,0,0.01706133286158244
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,4,128,1,float16,fp8,0,0.018992000569899876
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,4,4,128,1,float16,float16,0,0.08875200152397156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,4,128,1,float16,float16,0,0.016970666746298473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,4,128,1,float16,fp8,0,0.017071999609470367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,1,128,1,float16,float16,0,0.01482133318980535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,1,128,1,float16,fp8,0,0.015125333021084467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,4,2,128,1,float16,float16,0,0.02737066646416982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,2,128,1,float16,float16,0,0.015194666882356008
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,2,128,1,float16,fp8,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,4,128,1,float16,float16,0,0.016714667280515034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,4,4,128,1,float16,fp8,0,0.01676799977819125
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,4,128,1,float16,float16,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,4,128,1,float16,fp8,0,0.014991999914248785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,1,128,1,float16,fp8,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,2,128,1,float16,float16,0,0.014837333311637243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,4,2,128,1,float16,fp8,0,0.02309866746266683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,2,128,1,float16,fp8,0,0.015279999623696009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,4,128,1,float16,float16,0,0.014767999450365702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,4,128,1,float16,fp8,0,0.014912000546852747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,4,128,1,float16,float16,0,0.02107733239730199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,4,1,128,1,float16,fp8,0,0.017157333592573803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,4,1,128,1,float16,float16,0,0.06853866577148438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,4,4,128,1,float16,fp8,0,0.018853332847356796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,4,1,128,1,float16,fp8,0,0.06192000210285187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,4,2,128,1,float16,float16,0,0.10129066308339436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,4,128,1,float16,float16,0,0.09525866309801738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,4,2,128,1,float16,fp8,0,0.09982933600743611
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,1,128,1,float16,float16,0,0.037231999138991036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,1,128,1,float16,fp8,0,0.03770666569471359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,2,128,1,float16,fp8,0,0.050293331344922386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,2,128,1,float16,float16,0,0.04791999856630961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,4,128,1,float16,float16,0,0.09599467118581136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,4,128,1,float16,float16,0,0.04167999823888143
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,4,128,1,float16,fp8,0,0.08643733461697896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,4,128,1,float16,fp8,0,0.04925866425037384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,1,128,1,float16,float16,0,0.023200000325838726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,1,128,1,float16,fp8,0,0.025146665672461193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,2,128,1,float16,fp8,0,0.03142933299144109
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,2,128,1,float16,float16,0,0.029109333952267964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,4,1,128,1,float16,float16,0,0.01481066644191742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,4,128,1,float16,float16,0,0.025466665625572205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,4,128,1,float16,fp8,0,0.049322664737701416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,1,128,1,float16,float16,0,0.01655999943614006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,4,128,1,float16,fp8,0,0.03030399978160858
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,2,128,1,float16,float16,0,0.018863999595244724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,2,128,1,float16,fp8,0,0.020917333662509918
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,4,128,1,float16,float16,0,0.025087999800841015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,4,128,1,float16,fp8,0,0.029669334491093952
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,4,128,1,float16,float16,0,0.018976000448067982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,4,128,1,float16,fp8,0,0.02090666691462199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,1,128,1,float16,float16,0,0.013077333569526672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,2,128,1,float16,float16,0,0.015098666151364645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,2,128,1,float16,fp8,0,0.016906666258970898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,4,128,1,float16,float16,0,0.01887999971707662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,4,128,1,float16,fp8,0,0.02120000123977661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,4,128,1,float16,fp8,0,0.01685333376129468
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,1,128,1,float16,float16,0,0.012645332763592402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,1,128,1,float16,fp8,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,2,128,1,float16,float16,0,0.013104000439246496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,2,128,1,float16,fp8,0,0.014901333798964819
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,4,128,1,float16,float16,0,0.014890667051076889
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,4,128,1,float16,fp8,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,4,128,1,float16,float16,0,0.013151999562978745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,4,128,1,float16,fp8,0,0.01320533330241839
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,1,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,1,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,2,128,1,float16,float16,0,0.011770666887362799
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,4,4,128,1,float16,fp8,0,0.08949866890907288
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,2,128,1,float16,fp8,0,0.012655999511480331
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,4,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,4,4,128,1,float16,float16,0,0.012981332838535309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,4,128,1,float16,fp8,0,0.011994666109482447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,4,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,1,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,1,128,1,float16,fp8,0,0.012666666259368261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,2,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,2,128,1,float16,fp8,0,0.011152000476916632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,4,128,1,float16,float16,0,0.011365332951148352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,4,128,1,float16,float16,0,0.011722666521867117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,4,128,1,float16,fp8,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,1,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,1,128,1,float16,fp8,0,0.011498666057984034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,2,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,2,128,1,float16,fp8,0,0.011733333269755045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,4,4,128,1,float16,float16,0,0.041402667760849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,4,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,4,4,128,1,float16,fp8,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,4,1,128,1,float16,fp8,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,4,1,128,1,float16,fp8,0,0.015018666783968607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,4,128,1,float16,float16,0,0.04789866507053375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,4,1,128,1,float16,float16,0,0.0489279975493749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,4,1,128,1,float16,fp8,0,0.04659733176231384
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,4,4,128,1,float16,float16,0,0.014954666296641031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,4,2,128,1,float16,float16,0,0.05795733133951823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,4,2,128,1,float16,fp8,0,0.05880000193913778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,4,128,1,float16,fp8,0,0.05403733253479004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,1,128,1,float16,fp8,0,0.02951466788848241
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,2,128,1,float16,float16,0,0.03542399903138479
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,2,128,1,float16,fp8,0,0.037077332536379494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,4,128,1,float16,float16,0,0.04763199885686239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,4,128,1,float16,fp8,0,0.05416533350944519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,4,128,1,float16,float16,0,0.028512001037597656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,4,128,1,float16,fp8,0,0.0332640012105306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,1,128,1,float16,fp8,0,0.020314666132132213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,2,128,1,float16,float16,0,0.02232533444960912
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,2,128,1,float16,fp8,0,0.023647998770078022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,4,128,1,float16,float16,0,0.029152000943819683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,4,128,1,float16,fp8,0,0.03338133295377096
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,4,4,128,1,float16,fp8,0,0.012869333227475485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,4,128,1,float16,fp8,0,0.021365332106749218
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,1,128,1,float16,float16,0,0.014949332922697067
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,1,128,1,float16,fp8,0,0.014720000326633453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,2,128,1,float16,float16,0,0.015082667271296183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,2,128,1,float16,fp8,0,0.01699200024207433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,4,128,1,float16,float16,0,0.018810667097568512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,4,128,1,float16,fp8,0,0.020848001043001812
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,4,128,1,float16,fp8,0,0.01578666642308235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,4,128,1,float16,float16,0,0.015135999768972397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,1,128,1,float16,float16,0,0.012805332740147909
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,1,128,1,float16,fp8,0,0.012671999633312225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,2,128,1,float16,float16,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,2,128,1,float16,fp8,0,0.014917333920796713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,4,128,1,float16,float16,0,0.014965333044528961
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,4,4,128,1,float16,fp8,0,0.015072000523408255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,4,128,1,float16,float16,0,0.012863999853531519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,4,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,1,128,1,float16,float16,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,1,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,2,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,2,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,4,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,4,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,4,128,1,float16,fp8,0,0.011120000233252844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,1,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,1,128,1,float16,fp8,0,0.010853332777818045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,4,1,128,1,float16,float16,0,0.02937600016593933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,2,128,1,float16,float16,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,2,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,4,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,4,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,4,4,128,1,float16,fp8,0,0.011365332951148352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,4,128,1,float16,float16,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,1,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,1,128,1,float16,fp8,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,2,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,4,1,128,1,float16,float16,0,0.01912533367673556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,2,128,1,float16,fp8,0,0.01116266722480456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,4,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,4,4,128,1,float16,fp8,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,1,128,1,float16,float16,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,4,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,4,128,1,float16,fp8,0,0.011557333171367645
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,1,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,2,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,2,128,1,float16,fp8,0,0.011130666981140772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,4,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,4,4,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,4,4,128,1,float16,float16,0,0.018911999960740406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,4,128,1,float16,float16,0,0.03522133330504099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,4,128,1,float16,fp8,0,0.037589333951473236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,4,1,128,1,float16,fp8,0,0.039408000806967415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,4,1,128,1,float16,float16,0,0.04312000175317129
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,4,2,128,1,float16,float16,0,0.047914668917655945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,4,2,128,1,float16,fp8,0,0.045834665497144066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,1,128,1,float16,fp8,0,0.025429333249727886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,1,128,1,float16,float16,0,0.025493333737055462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,2,128,1,float16,float16,0,0.029130667448043823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,2,128,1,float16,fp8,0,0.02940800040960312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,4,128,1,float16,float16,0,0.02117866774400075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,4,128,1,float16,float16,0,0.035589332381884255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,4,4,128,1,float16,fp8,0,0.03796799977620443
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,4,4,128,1,float16,float16,0,0.012768000364303589
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,1,128,1,float16,float16,0,0.01682666689157486
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,1,128,1,float16,fp8,0,0.017136000096797943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,2,128,1,float16,float16,0,0.019050666441520054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,2,128,1,float16,fp8,0,0.018965333700180054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,4,128,1,float16,float16,0,0.022645334402720135
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,4,128,1,float16,fp8,0,0.02332799881696701
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,4,128,1,float16,float16,0,0.014917333920796713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,1,128,1,float16,float16,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,1,128,1,float16,fp8,0,0.012714666624863943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,2,128,1,float16,float16,0,0.013738666971524557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,2,128,1,float16,fp8,0,0.01492799942692121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,1,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,4,128,1,float16,float16,0,0.015237333873907724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,4,128,1,float16,fp8,0,0.01381333296497663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,1,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,2,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,2,128,1,float16,fp8,0,0.012608000387748083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,4,128,1,float16,float16,0,0.012714666624863943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,4,128,1,float16,fp8,0,0.012842666357755661
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,4,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,4,128,1,float16,fp8,0,0.011610666910807291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,1,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,1,128,1,float16,fp8,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,2,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,2,128,1,float16,fp8,0,0.010842667271693548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,4,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,4,4,128,1,float16,fp8,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,4,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,1,128,1,float16,float16,0,0.01110400011142095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,2,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,2,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,4,128,1,float16,float16,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,4,128,1,float16,fp8,0,0.010960000256697336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,4,128,1,float16,float16,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,4,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,1,128,1,float16,float16,0,0.01080000028014183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,4,4,128,1,float16,fp8,0,0.02325333406527837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,1,128,1,float16,fp8,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,2,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,2,128,1,float16,fp8,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,4,128,1,float16,float16,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,4,4,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,4,128,1,float16,fp8,0,0.01684800038735072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,4,4,128,1,float16,float16,0,0.012906666845083237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,4,128,1,float16,float16,0,0.010757333288590113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,4,128,1,float16,fp8,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,1,128,1,float16,float16,0,0.010666667173306147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,1,128,1,float16,fp8,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,2,128,1,float16,float16,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,2,128,1,float16,fp8,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,4,128,1,float16,float16,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,4,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,4,1,128,1,float16,fp8,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,4,1,128,1,float16,float16,0,0.03945599993069967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,4,1,128,1,float16,fp8,0,0.03565333286921183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,4,2,128,1,float16,float16,0,0.04229333500067393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,4,128,1,float16,float16,0,0.02935466667016347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,4,128,1,float16,fp8,0,0.029504001140594482
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,1,128,1,float16,float16,0,0.025055999557177227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,4,2,128,1,float16,fp8,0,0.0394400010506312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,1,128,1,float16,fp8,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,2,128,1,float16,float16,0,0.025381334125995636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,2,128,1,float16,fp8,0,0.025242666403452556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,4,128,1,float16,fp8,0,0.02917333443959554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,4,128,1,float16,fp8,0,0.019146667172511418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,4,4,128,1,float16,fp8,0,0.016208000481128693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,1,128,1,float16,float16,0,0.017050666113694508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,1,128,1,float16,fp8,0,0.01492799942692121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,2,128,1,float16,fp8,0,0.016895999511082966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,4,128,1,float16,float16,0,0.01884799947341283
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,4,128,1,float16,fp8,0,0.019002666076024372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,4,4,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,1,128,1,float16,float16,0,0.012719999998807907
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,1,128,1,float16,fp8,0,0.013066666821638743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,2,128,1,float16,float16,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,2,128,1,float16,fp8,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,4,128,1,float16,float16,0,0.0141546664138635
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,4,128,1,float16,fp8,0,0.014736000448465347
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,4,128,1,float16,float16,0,0.011402666568756104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,4,128,1,float16,fp8,0,0.011365332951148352
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,1,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,1,128,1,float16,fp8,0,0.011125333607196808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,2,128,1,float16,float16,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,2,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,4,128,1,float16,float16,0,0.011125333607196808
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,4,4,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,4,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,4,128,1,float16,fp8,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,1,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,1,128,1,float16,fp8,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,2,128,1,float16,float16,0,0.010757333288590113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,2,128,1,float16,fp8,0,0.012618667135636011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,4,128,1,float16,float16,0,0.019071999937295914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,4,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,4,4,128,1,float16,float16,0,0.029253333806991577
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,1,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,4,128,1,float16,fp8,0,0.01119999960064888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,1,128,1,float16,fp8,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,2,128,1,float16,float16,0,0.010858666151762009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,2,128,1,float16,fp8,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,4,2,128,1,float16,float16,0,0.01711999997496605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,4,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,4,4,128,1,float16,fp8,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,4,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,4,128,1,float16,fp8,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,4,128,1,float16,float16,0,0.013445333888133367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,1,128,1,float16,float16,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,4,4,128,1,float16,fp8,0,0.014874666929244995
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,2,128,1,float16,float16,0,0.010773333410422007
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,2,128,1,float16,fp8,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,4,128,1,float16,fp8,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,4,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,4,128,1,float16,float16,0,0.010768000036478043
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,1,128,1,float16,fp8,0,0.010757333288590113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,2,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,2,128,1,float16,fp8,0,0.010650667051474253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,4,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,4,128,1,float16,fp8,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,4,1,128,1,float16,float16,0,0.011066666493813196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,4,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,4,2,128,1,float16,float16,0,0.014170666535695394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,4,2,128,1,float16,fp8,0,0.012789333860079447
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,4,128,1,float16,float16,0,0.012975999464591345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,4,128,1,float16,fp8,0,0.013162666310866674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,1,128,1,float16,float16,0,0.009375999992092451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,4,128,1,float16,float16,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,2,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,2,128,1,float16,fp8,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,4,128,1,float16,float16,0,0.013061333447694778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,4,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,4,1,128,1,float16,fp8,0,0.010751999914646149
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,4,128,1,float16,fp8,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,1,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,4,4,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,1,128,1,float16,fp8,0,0.00878399983048439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,2,128,1,float16,float16,0,0.009749333063761393
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,2,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,4,128,1,float16,float16,0,0.011136000355084738
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,4,128,1,float16,float16,0,0.009056000038981438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,4,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,1,128,1,float16,float16,0,0.009103999783595404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,1,128,1,float16,fp8,0,0.00891733355820179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,2,128,1,float16,float16,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,2,128,1,float16,fp8,0,0.008693333094318708
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,4,128,1,float16,float16,0,0.009082666908701261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,4,4,128,1,float16,fp8,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,4,128,1,float16,float16,0,0.00877333308259646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,1,128,1,float16,float16,0,0.008890666688481966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,4,128,1,float16,fp8,0,0.00874133345981439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,1,128,1,float16,fp8,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,2,128,1,float16,float16,0,0.00873066671192646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,2,128,1,float16,fp8,0,0.008842666943868002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,4,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,1,128,1,float16,fp8,0,0.008992000172535578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,4,4,128,1,float16,fp8,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,4,4,128,1,float16,fp8,0,0.009114666531483332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,4,128,1,float16,float16,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,4,128,1,float16,fp8,0,0.008901333436369896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,1,128,1,float16,float16,0,0.008805333326260248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,1,128,1,float16,fp8,0,0.008714666590094566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,2,128,1,float16,float16,0,0.00874133345981439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,4,128,1,float16,float16,0,0.008767999708652496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,4,128,1,float16,fp8,0,0.009205333267649015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,4,128,1,float16,float16,0,0.009152000149091085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,4,128,1,float16,fp8,0,0.008799999952316284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,1,128,1,float16,fp8,0,0.00878399983048439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,4,4,128,1,float16,fp8,0,0.012794667234023413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,2,128,1,float16,fp8,0,0.009061333412925402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,4,128,1,float16,float16,0,0.009072000160813332
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,4,128,1,float16,fp8,0,0.00895999992887179
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,4,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,1,128,1,float16,float16,0,0.008634666601816813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,1,128,1,float16,fp8,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,2,128,1,float16,float16,0,0.009077333534757296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,2,128,1,float16,fp8,0,0.008752000207702318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,4,128,1,float16,float16,0,0.009088000282645226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,4,128,1,float16,fp8,0,0.008703999842206636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,4,128,1,float16,float16,0,0.00855466661353906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,4,128,1,float16,fp8,0,0.009045333291093508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,1,128,1,float16,float16,0,0.008986666798591614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,1,128,1,float16,fp8,0,0.008709333216150602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,2,128,1,float16,float16,0,0.008762666955590248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,2,128,1,float16,fp8,0,0.008736000085870424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,4,4,128,1,float16,fp8,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,4,128,1,float16,float16,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,4,4,128,1,float16,fp8,0,0.00890666681031386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,2,128,1,float16,float16,0,0.6324373483657837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,2,1,128,1,float16,float16,0,1.0454986890157063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,2,128,1,float16,fp8,0,0.5970186789830526
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,1,128,1,float16,float16,0,0.5713066657384237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,1,128,1,float16,fp8,0,0.5202773412068685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,2,128,1,float16,float16,0,0.6389173269271851
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,2,128,1,float16,float16,0,0.37910934289296466
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,2,128,1,float16,fp8,0,0.3478720188140869
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,4,2,128,1,float16,fp8,0,0.008890666688481966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,1,128,1,float16,float16,0,0.3423626820246379
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,1,128,1,float16,fp8,0,0.31964266300201416
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,1,128,1,float16,float16,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,4,2,128,1,float16,float16,0,0.008874666566650072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,2,128,1,float16,float16,0,0.3808906475702922
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,2,2,128,1,float16,fp8,0,0.347978671391805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,2,128,1,float16,fp8,0,0.2208426594734192
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,1,128,1,float16,float16,0,0.22083733479181925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,4,4,128,1,float16,float16,0,0.008938666433095932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,1,128,1,float16,fp8,0,0.2103360096613566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,2,128,1,float16,float16,0,0.23321600755055746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,2,128,1,float16,fp8,0,0.2212053338686625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,2,128,1,float16,float16,0,0.4054773251215617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,2,1,128,1,float16,float16,0,0.6361546516418457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,2,128,1,float16,fp8,0,0.38836268583933514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,1,128,1,float16,fp8,0,0.32569066683451336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,2,128,1,float16,float16,0,0.4064319928487142
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,2,128,1,float16,float16,0,0.24101867278416952
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16384,2,1,128,1,float16,fp8,0,0.9167253176371256
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,1,128,1,float16,fp8,0,0.20982933044433594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,2,128,1,float16,float16,0,0.23492266734441122
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,2,2,128,1,float16,fp8,0,0.5942399899164835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,2,128,1,float16,fp8,0,0.2294879953066508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,2,128,1,float16,float16,0,0.178874671459198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,2,128,1,float16,fp8,0,0.17028266191482544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,1,128,1,float16,float16,0,0.16880534092585245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,2,2,128,1,float16,float16,0,0.2323519984881083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,1,128,1,float16,fp8,0,0.15923200050989786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,2,128,1,float16,float16,0,0.17757334311803183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,2,2,128,1,float16,fp8,0,0.169813334941864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,2,1,128,1,float16,float16,0,0.4675840139389038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,10240,2,1,128,1,float16,fp8,0,0.43457599480946857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,1,128,1,float16,float16,0,0.3558826843897502
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,2,128,1,float16,fp8,0,0.30241600672403973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,1,128,1,float16,float16,0,0.2749226689338684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,2,2,128,1,float16,fp8,0,0.3914080063501994
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,2,128,1,float16,float16,0,0.31759466727574664
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,2,128,1,float16,fp8,0,0.22773333390553793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,2,128,1,float16,fp8,0,0.30157333612442017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,2,128,1,float16,float16,0,0.1774666706720988
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,2,1,128,1,float16,float16,0,0.2239946722984314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,1,128,1,float16,float16,0,0.16497600078582764
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,1,128,1,float16,fp8,0,0.15627200404802957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,2,128,1,float16,float16,0,0.1787253419558207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,2,128,1,float16,float16,0,0.14844800035158792
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,2,128,1,float16,fp8,0,0.14351466298103333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,1,128,1,float16,fp8,0,0.13388267159461975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,2,128,1,float16,float16,0,0.14824000000953674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,2,128,1,float16,fp8,0,0.14380799730618796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,2,1,128,1,float16,float16,0,0.596010684967041
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,12288,2,1,128,1,float16,fp8,0,0.5770933230717977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,8192,2,1,128,1,float16,fp8,0,0.5480480194091797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,1,128,1,float16,fp8,0,0.2539733250935872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,2,128,1,float16,fp8,0,0.3810933430989583
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,2,128,1,float16,float16,0,0.389024019241333
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,1,128,1,float16,float16,0,0.3304640054702759
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,1,128,1,float16,fp8,0,0.30577067534128827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,2,128,1,float16,fp8,0,0.38362133502960205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,2,128,1,float16,float16,0,0.2307093342145284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,2,128,1,float16,fp8,0,0.21450134118398032
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,1,128,1,float16,float16,0,0.1983786622683207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,1,128,1,float16,fp8,0,0.18760534127553305
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,2,128,1,float16,fp8,0,0.17291200160980225
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,2,128,1,float16,float16,0,0.23345067103703818
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,2,2,128,1,float16,fp8,0,0.21547732750574747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,2,128,1,float16,float16,0,0.1360479990641276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,1,128,1,float16,float16,0,0.12452266613642375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,2,128,1,float16,fp8,0,0.13173866271972656
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,2,1,128,1,float16,float16,0,0.1416213313738505
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,1,128,1,float16,fp8,0,0.12029866377512614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,2,128,1,float16,float16,0,0.1360640029112498
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,2,128,1,float16,float16,0,0.12097066640853882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,1,128,1,float16,float16,0,0.11550933122634888
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,1,128,1,float16,fp8,0,0.11134399970372517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,2,128,1,float16,float16,0,0.12125333150227864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,2,128,1,float16,fp8,0,0.11875200271606445
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,2,1,128,1,float16,float16,0,0.3741333484649658
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,6144,2,1,128,1,float16,fp8,0,0.35589333375295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,2,128,1,float16,float16,0,0.2624266743659973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,2,128,1,float16,fp8,0,0.25787200530370075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,1,128,1,float16,float16,0,0.21022933721542358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,1,128,1,float16,fp8,0,0.19744000832239786
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,2,2,128,1,float16,float16,0,0.3168693383534749
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,2,128,1,float16,float16,0,0.14638400077819824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,2,128,1,float16,fp8,0,0.2587040066719055
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,1,128,1,float16,float16,0,0.13129066427548727
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,2,128,1,float16,float16,0,0.1514346698919932
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,2,2,128,1,float16,float16,0,0.3881066640218099
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,2,128,1,float16,fp8,0,0.1453439990679423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,2,2,128,1,float16,fp8,0,0.1723360021909078
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,2,128,1,float16,float16,0,0.10522666573524475
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,1,128,1,float16,float16,0,0.09505066275596619
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,1,128,1,float16,fp8,0,0.09268800417582194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,2,128,1,float16,float16,0,0.10494400064150493
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,2,128,1,float16,fp8,0,0.10300266742706299
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,2,128,1,float16,float16,0,0.09333866834640503
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,2,2,128,1,float16,fp8,0,0.1318826675415039
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,2,2,128,1,float16,fp8,0,0.11807466546694438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,1,128,1,float16,float16,0,0.08914666374524434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,1,128,1,float16,fp8,0,0.08666666348775227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,2,128,1,float16,float16,0,0.0934986670811971
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,2,128,1,float16,fp8,0,0.09327466289202373
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,2,128,1,float16,float16,0,0.2645653287569682
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,2,1,128,1,float16,float16,0,0.3716799815495809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,4096,2,1,128,1,float16,fp8,0,0.3596373399098714
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,2,128,1,float16,fp8,0,0.2709813316663106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,1,128,1,float16,float16,0,0.20914665857950845
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,1,128,1,float16,fp8,0,0.19829867283503214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,2,2,128,1,float16,float16,0,0.2584693431854248
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,2,128,1,float16,float16,0,0.26531734069188434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,2,128,1,float16,float16,0,0.1556053360303243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,2,2,128,1,float16,fp8,0,0.27610133091608685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,1,128,1,float16,float16,0,0.11898666620254517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,1,128,1,float16,fp8,0,0.11750400066375732
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,2,128,1,float16,float16,0,0.1574613352616628
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,2,128,1,float16,fp8,0,0.14381333192189535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,2,128,1,float16,float16,0,0.08705600102742513
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,2,128,1,float16,fp8,0,0.0864586631457011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,1,128,1,float16,float16,0,0.07449600100517273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,1,128,1,float16,fp8,0,0.07438399891058604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,2,128,1,float16,float16,0,0.08668800195058186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,2,2,128,1,float16,fp8,0,0.10187199711799622
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,2,2,128,1,float16,fp8,0,0.08649599552154541
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,2,128,1,float16,float16,0,0.07272000114123027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,2,128,1,float16,fp8,0,0.0724373310804367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,1,128,1,float16,float16,0,0.06639466683069865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,1,128,1,float16,fp8,0,0.06607999900976817
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,2,128,1,float16,float16,0,0.07236800094445546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,2,2,128,1,float16,fp8,0,0.07242133220036824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,2,128,1,float16,float16,0,0.06620266536871593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,1,128,1,float16,float16,0,0.06249066690603892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,1,128,1,float16,fp8,0,0.06215466558933258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,2,2,128,1,float16,fp8,0,0.09285866220792134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,2,128,1,float16,float16,0,0.06613866488138835
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,2,128,1,float16,fp8,0,0.06540800134340923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,2,128,1,float16,float16,0,0.1874506672223409
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,2,1,128,1,float16,float16,0,0.24634667237599692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,3072,2,1,128,1,float16,fp8,0,0.24303466081619263
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,2,128,1,float16,fp8,0,0.18963199853897095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,1,128,1,float16,float16,0,0.13961600263913473
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,1,128,1,float16,fp8,0,0.13013866543769836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,2,128,1,float16,float16,0,0.18546666701634726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,2,128,1,float16,float16,0,0.09824533263842265
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,2,2,128,1,float16,fp8,0,0.19177067279815674
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,2,128,1,float16,fp8,0,0.10150399804115295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,1,128,1,float16,float16,0,0.08265066643555959
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,1,128,1,float16,fp8,0,0.08269333342711131
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,1,128,1,float16,fp8,0,0.12713066736857095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,2,128,1,float16,float16,0,0.1055573324362437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,2,2,128,1,float16,fp8,0,0.10148800412813823
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,2,128,1,float16,fp8,0,0.06806399921576183
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,2,128,1,float16,float16,0,0.06832000116507213
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,1,128,1,float16,float16,0,0.058101331194241844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,1,128,1,float16,fp8,0,0.058176000912984215
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,2,128,1,float16,float16,0,0.06758399804433186
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,2,2,128,1,float16,fp8,0,0.06828799843788147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,2,128,1,float16,float16,0,0.05781333148479462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,2,128,1,float16,fp8,0,0.057999998331069946
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,1,128,1,float16,fp8,0,0.051957334081331887
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,2,128,1,float16,float16,0,0.05717866619427999
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,2,128,1,float16,fp8,0,0.05820266902446747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,2,128,1,float16,float16,0,0.05184000233809153
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,2,128,1,float16,fp8,0,0.05203733344872793
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,1,128,1,float16,float16,0,0.04979733129342397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,1,128,1,float16,fp8,0,0.04940799872080485
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,2,128,1,float16,fp8,0,0.051925331354141235
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,2,2,128,1,float16,fp8,0,0.06603200236956279
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,2,1,128,1,float16,float16,0,0.2635786732037862
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,2,128,1,float16,float16,0,0.2035413384437561
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,2048,2,1,128,1,float16,fp8,0,0.26243199904759723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,2,128,1,float16,fp8,0,0.2136639952659607
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,1,128,1,float16,float16,0,0.14656000336011252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,1,128,1,float16,fp8,0,0.14221866925557455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,2,128,1,float16,fp8,0,0.21384533246358237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,2,128,1,float16,float16,0,0.11176533500353496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,2,2,128,1,float16,float16,0,0.2037386695543925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,2,128,1,float16,fp8,0,0.10893332958221436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,2,2,128,1,float16,fp8,0,0.1446293294429779
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,1,128,1,float16,float16,0,0.08057066798210144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,1,128,1,float16,fp8,0,0.08004266520341237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,2,128,1,float16,float16,0,0.11575466394424438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,2,2,128,1,float16,fp8,0,0.10752532879511516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,2,128,1,float16,fp8,0,0.06354133288065593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,2,128,1,float16,float16,0,0.06102933486302694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,1,128,1,float16,fp8,0,0.05048533280690511
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,2,128,1,float16,float16,0,0.061834668119748436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,2,128,1,float16,fp8,0,0.06221333146095276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,2,128,1,float16,float16,0,0.04772266745567322
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,2,128,1,float16,fp8,0,0.049738665421803795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,2,1,128,1,float16,float16,0,0.053258667389551796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,1,128,1,float16,float16,0,0.042362665136655174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,1,128,1,float16,fp8,0,0.04365866879622141
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,2,128,1,float16,float16,0,0.047914668917655945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,2,2,128,1,float16,fp8,0,0.04980800052483877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,2,128,1,float16,float16,0,0.041477332512537636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,2,128,1,float16,fp8,0,0.041573333243529
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,1,128,1,float16,float16,0,0.039306665460268654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,1,128,1,float16,fp8,0,0.03934400031963984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,2,128,1,float16,float16,0,0.04173333446184794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,2,2,128,1,float16,fp8,0,0.04170133173465729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,2,128,1,float16,float16,0,0.03835733234882355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,1,128,1,float16,float16,0,0.03738666574160258
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,1,128,1,float16,fp8,0,0.03728000074625015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,2,2,128,1,float16,float16,0,0.051813334226608276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,2,128,1,float16,float16,0,0.03766933331886927
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,2,1,128,1,float16,float16,0,0.1833546757698059
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,2,128,1,float16,fp8,0,0.15436800320943198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,2,128,1,float16,float16,0,0.14897599816322327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1536,2,1,128,1,float16,fp8,0,0.18561599651972452
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,1,128,1,float16,float16,0,0.10009599725405376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,1,128,1,float16,fp8,0,0.09444266557693481
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,2,128,1,float16,fp8,0,0.151637335618337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,2,128,1,float16,float16,0,0.07454933226108551
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,2,128,1,float16,fp8,0,0.07835733393828075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,1,128,1,float16,float16,0,0.05849599838256836
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,1,128,1,float16,fp8,0,0.05919999877611796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,2,128,1,float16,float16,0,0.07494933406511943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,2,2,128,1,float16,fp8,0,0.14892799655596414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,2,128,1,float16,float16,0,0.04761599997679392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,2,2,128,1,float16,fp8,0,0.07860800127188365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,1,128,1,float16,float16,0,0.0397119993964831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,2,1,128,1,float16,float16,0,0.049312000473340355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,1,128,1,float16,fp8,0,0.03977066775163015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,2,128,1,float16,float16,0,0.04983466863632202
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,2,128,1,float16,fp8,0,0.04984533290068308
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,2,128,1,float16,float16,0,0.03937600056330363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,1,128,1,float16,float16,0,0.03513599932193756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,1,128,1,float16,fp8,0,0.03532266616821289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,2,128,1,float16,float16,0,0.03896533449490865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,2,128,1,float16,fp8,0,0.03992533435424169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,2,128,1,float16,float16,0,0.03410666684309641
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,2,128,1,float16,fp8,0,0.0352906659245491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,1,128,1,float16,float16,0,0.03145066648721695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,2,128,1,float16,fp8,0,0.039461334546407066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,1,128,1,float16,fp8,0,0.03123733401298523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,2,128,1,float16,float16,0,0.034346667428811394
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,2,2,128,1,float16,fp8,0,0.03555733213822047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,2,2,128,1,float16,fp8,0,0.0383093332250913
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,2,128,1,float16,float16,0,0.031343999008337654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,2,128,1,float16,fp8,0,0.03127466638882955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,1,128,1,float16,float16,0,0.030997333427270252
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,1,128,1,float16,fp8,0,0.02937600016593933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,2,128,1,float16,float16,0,0.031541332602500916
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,2,2,128,1,float16,fp8,0,0.03154666721820831
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,2,128,1,float16,float16,0,0.17225066820780435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,2,2,128,1,float16,float16,0,0.14922666549682617
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,2,128,1,float16,fp8,0,0.18519999583562216
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,2,1,128,1,float16,fp8,0,0.21336533625920615
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,1,128,1,float16,float16,0,0.11600533127784729
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,1,128,1,float16,fp8,0,0.1134986678759257
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,2,128,1,float16,float16,0,0.1709386706352234
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,2,2,128,1,float16,fp8,0,0.18435200055440268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,2,2,128,1,float16,fp8,0,0.05031466484069824
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,2,128,1,float16,fp8,0,0.08756267031033833
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,1,128,1,float16,float16,0,0.05916800101598104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,1,128,1,float16,fp8,0,0.06233066817124685
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,2,2,128,1,float16,fp8,0,0.040949332217375435
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,2,128,1,float16,fp8,0,0.08749333024024963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,2,128,1,float16,float16,0,0.09599467118581136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,2,128,1,float16,float16,0,0.04762133459250132
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,2,128,1,float16,fp8,0,0.05151999990145365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,1,128,1,float16,float16,0,0.035536001125971474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,1,128,1,float16,fp8,0,0.03845866769552231
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,2,128,1,float16,fp8,0,0.05022933085759481
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,2,128,1,float16,float16,0,0.03563733398914337
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,2,128,1,float16,fp8,0,0.037445334096749626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,1,128,1,float16,float16,0,0.029845332105954487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,1,128,1,float16,fp8,0,0.03128000100453695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,2,128,1,float16,float16,0,0.035631999373435974
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,2,2,128,1,float16,fp8,0,0.037717332442601524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,2,128,1,float16,fp8,0,0.03148266673088074
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,1,128,1,float16,float16,0,0.027376001079877216
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,1,128,1,float16,fp8,0,0.027471999327341717
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,2,128,1,float16,float16,0,0.029301332930723827
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,2,128,1,float16,float16,0,0.0269813338915507
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,2,128,1,float16,fp8,0,0.031301334500312805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,2,128,1,float16,fp8,0,0.027082666754722595
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,1,128,1,float16,float16,0,0.02526933451493581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,1,128,1,float16,fp8,0,0.025386666258176167
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,2,128,1,float16,float16,0,0.027290667096773785
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,2,2,128,1,float16,fp8,0,0.027248000105222065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,2,128,1,float16,fp8,0,0.025311999022960663
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,1,128,1,float16,float16,0,0.023210667073726654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,1,128,1,float16,fp8,0,0.02333866556485494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,2,128,1,float16,float16,0,0.025045332809289295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,2,128,1,float16,fp8,0,0.02502399931351344
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1024,2,1,128,1,float16,float16,0,0.20683733622233072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,2,128,1,float16,float16,0,0.15636799732844034
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,2,2,128,1,float16,float16,0,0.09599467118581136
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,2,128,1,float16,fp8,0,0.17071467638015747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,2,1,128,1,float16,float16,0,0.184714674949646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,512,2,1,128,1,float16,fp8,0,0.18885332345962524
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,1,128,1,float16,float16,0,0.10109866658846538
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,1,128,1,float16,fp8,0,0.09846400221188863
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,2,2,128,1,float16,float16,0,0.0491893341143926
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,2,128,1,float16,float16,0,0.15808000167210898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,2,128,1,float16,float16,0,0.08370666702588399
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,2,2,128,1,float16,fp8,0,0.17165867487589517
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,2,128,1,float16,fp8,0,0.07859733204046886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,1,128,1,float16,float16,0,0.04969066878159841
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,1,128,1,float16,fp8,0,0.05180799961090088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,2,2,128,1,float16,float16,0,0.029520000020662945
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,2,128,1,float16,fp8,0,0.07989866534868877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,2,2,128,1,float16,float16,0,0.08532266815503438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,2,128,1,float16,fp8,0,0.0454720010360082
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,1,128,1,float16,float16,0,0.02937600016593933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,1,128,1,float16,fp8,0,0.03252266595760981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,2,128,1,float16,float16,0,0.041877334316571556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,2,128,1,float16,float16,0,0.029215998947620392
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,2,128,1,float16,fp8,0,0.04582933088143667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,2,128,1,float16,fp8,0,0.031727999448776245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,1,128,1,float16,fp8,0,0.025429333249727886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,2,2,128,1,float16,float16,0,0.025397333006064098
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,2,128,1,float16,float16,0,0.0295413335164388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,2,128,1,float16,float16,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,1,128,1,float16,float16,0,0.020901332298914593
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,1,128,1,float16,fp8,0,0.021194666624069214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,2,128,1,float16,float16,0,0.020954666038354237
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,2,128,1,float16,fp8,0,0.0210506667693456
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,1,128,1,float16,float16,0,0.019002666076024372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,1,128,1,float16,fp8,0,0.019098666807015736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,2,128,1,float16,float16,0,0.021157334248224895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,2,2,128,1,float16,fp8,0,0.021007999777793884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,2,128,1,float16,float16,0,0.018906666586796444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,2,128,1,float16,fp8,0,0.0189280000825723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,1,128,1,float16,float16,0,0.018160000443458557
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,1,128,1,float16,fp8,0,0.017242666333913803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,2,128,1,float16,float16,0,0.019248000035683315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,2,2,128,1,float16,fp8,0,0.018922666708628338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,2,128,1,float16,float16,0,0.018911999960740406
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,2,128,1,float16,fp8,0,0.017210666090250015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,1,128,1,float16,float16,0,0.01811733345190684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,1,128,1,float16,fp8,0,0.017103999853134155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,2,128,1,float16,float16,0,0.018794666975736618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,2,2,128,1,float16,fp8,0,0.017312000195185345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,2,2,128,1,float16,float16,0,0.041509332756201424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,1,128,1,float16,float16,0,0.02327466756105423
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,2,1,128,1,float16,float16,0,0.09098133444786072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,2,2,128,1,float16,fp8,0,0.03146666785081228
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,2,128,1,float16,fp8,0,0.02550400048494339
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,2,128,1,float16,float16,0,0.023103999594847362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,2,128,1,float16,float16,0,0.07548800110816956
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,2,128,1,float16,fp8,0,0.07625066737333934
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,1,128,1,float16,float16,0,0.041562666495641075
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,1,128,1,float16,fp8,0,0.046207999189694725
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,2,128,1,float16,float16,0,0.07832533121109009
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,2,128,1,float16,float16,0,0.03944533318281174
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,2,128,1,float16,fp8,0,0.04155199974775314
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,2,2,128,1,float16,fp8,0,0.07380266487598419
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,1,128,1,float16,float16,0,0.026778665681680042
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,2,128,1,float16,float16,0,0.039306665460268654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,2,128,1,float16,fp8,0,0.02942399928967158
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,1,128,1,float16,fp8,0,0.02921066681543986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,2,2,128,1,float16,fp8,0,0.04190933207670847
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,1,128,1,float16,float16,0,0.021029333273569744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,1,128,1,float16,fp8,0,0.02309866746266683
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,2,128,1,float16,float16,0,0.02720533311367035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,2,128,1,float16,fp8,0,0.02940800040960312
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,2,128,1,float16,float16,0,0.021007999777793884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,2,128,1,float16,fp8,0,0.022885332504908245
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,1,128,1,float16,float16,0,0.017029333859682083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,1,128,1,float16,fp8,0,0.019280000279347103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,2,128,1,float16,float16,0,0.02093333254257838
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,2,2,128,1,float16,fp8,0,0.02295999974012375
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,2,128,1,float16,fp8,0,0.0189280000825723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,2,128,1,float16,float16,0,0.017125333348910015
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,1,128,1,float16,float16,0,0.016858667135238647
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,1,128,1,float16,fp8,0,0.016783999900023144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,2,128,1,float16,float16,0,0.017055999487638474
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,2,2,128,1,float16,fp8,0,0.019071999937295914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,2,128,1,float16,float16,0,0.016719999412695568
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,2,128,1,float16,fp8,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,1,128,1,float16,float16,0,0.015087999403476715
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,1,128,1,float16,fp8,0,0.015029333531856537
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,2,128,1,float16,float16,0,0.016890666137139004
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,2,2,128,1,float16,fp8,0,0.017082666357358296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,2,128,1,float16,float16,0,0.01504533365368843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,2,128,1,float16,fp8,0,0.014688000082969666
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,1,128,1,float16,float16,0,0.014896000425020853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,1,128,1,float16,fp8,0,0.014864000181357065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,2,128,1,float16,float16,0,0.015114666273196539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,2,2,128,1,float16,fp8,0,0.015077333897352219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,2,128,1,float16,float16,0,0.014858666807413101
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,2,2,128,1,float16,fp8,0,0.025279998779296875
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,1,128,1,float16,float16,0,0.015061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,1,128,1,float16,fp8,0,0.015034666905800501
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,2,128,1,float16,fp8,0,0.014720000326633453
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,256,2,1,128,1,float16,fp8,0,0.08917867143948872
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,2,128,1,float16,fp8,0,0.04340266684691111
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,2,1,128,1,float16,float16,0,0.04541866481304169
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,1,128,1,float16,float16,0,0.027797333896160126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,128,2,1,128,1,float16,fp8,0,0.048138668139775596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,2,2,128,1,float16,float16,0,0.027024000883102417
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,1,128,1,float16,fp8,0,0.030415999392668407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,2,128,1,float16,fp8,0,0.027066667874654133
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,2,128,1,float16,float16,0,0.03938666731119156
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,2,128,1,float16,float16,0,0.025258667767047882
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,2,128,1,float16,fp8,0,0.04364799956480662
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,1,128,1,float16,float16,0,0.019189332922299702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,1,128,1,float16,fp8,0,0.021007999777793884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,2,128,1,float16,float16,0,0.018250666558742523
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,2,128,1,float16,float16,0,0.025199999411900837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,2,128,1,float16,fp8,0,0.019941333681344986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,1,128,1,float16,float16,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,1,128,1,float16,fp8,0,0.016783999900023144
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,2,128,1,float16,float16,0,0.017157333592573803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,2,2,128,1,float16,fp8,0,0.019813333948453266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,2,128,1,float16,float16,0,0.015077333897352219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,1,128,1,float16,float16,0,0.0129120002190272
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,1,128,1,float16,fp8,0,0.014501333236694336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,2,128,1,float16,float16,0,0.012778667112191519
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,2,128,1,float16,fp8,0,0.01509333277742068
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,2,128,1,float16,fp8,0,0.014671999961137772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,2,2,128,1,float16,float16,0,0.014864000181357065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,1,128,1,float16,float16,0,0.012815999488035837
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,1,128,1,float16,fp8,0,0.012608000387748083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,2,128,1,float16,float16,0,0.012928000340859095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,2,2,128,1,float16,fp8,0,0.014629332969586054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,2,128,1,float16,float16,0,0.012837332983811697
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,2,128,1,float16,fp8,0,0.012005332857370377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,1,128,1,float16,float16,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,1,128,1,float16,fp8,0,0.012624000509579977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,2,2,128,1,float16,float16,0,0.039994666973749794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,2,128,1,float16,fp8,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,2,128,1,float16,float16,0,0.01268799975514412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,1,128,1,float16,float16,0,0.012319999436537424
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,1,128,1,float16,fp8,0,0.012613333761692047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,2,128,1,float16,fp8,0,0.012639999389648438
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,2,128,1,float16,float16,0,0.011744000017642975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,2,128,1,float16,fp8,0,0.011418666690587997
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,1,128,1,float16,float16,0,0.011178666104873022
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,1,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,2,2,128,1,float16,fp8,0,0.02720000098148982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,2,128,1,float16,float16,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,2,2,128,1,float16,fp8,0,0.011306667079528173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,2,128,1,float16,fp8,0,0.014975999792416891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,2,128,1,float16,fp8,0,0.015072000523408255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,2,1,128,1,float16,float16,0,0.03403199960788091
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,2,128,1,float16,float16,0,0.027237333357334137
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,64,2,1,128,1,float16,fp8,0,0.0353973334034284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,2,128,1,float16,fp8,0,0.029498666524887085
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,2,2,128,1,float16,float16,0,0.01179733375708262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,1,128,1,float16,float16,0,0.02125866711139679
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,2,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,1,128,1,float16,fp8,0,0.02332266668478648
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,2,2,128,1,float16,float16,0,0.012821332861979803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,2,128,1,float16,float16,0,0.027104000250498455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,2,128,1,float16,float16,0,0.018768000106016796
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,2,2,128,1,float16,fp8,0,0.029461334149042766
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,2,128,1,float16,fp8,0,0.02107733239730199
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,1,128,1,float16,float16,0,0.014975999792416891
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,1,128,1,float16,fp8,0,0.016832000265518825
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,2,128,1,float16,float16,0,0.01783466711640358
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,2,128,1,float16,fp8,0,0.015061333775520325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,2,2,128,1,float16,fp8,0,0.021002667645613354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,1,128,1,float16,float16,0,0.012831999609867731
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,1,128,1,float16,fp8,0,0.01369599997997284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,2,128,1,float16,float16,0,0.014671999961137772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,2,128,1,float16,float16,0,0.013050666699806849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,2,128,1,float16,fp8,0,0.015168000012636185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,2,128,1,float16,fp8,0,0.01333333303531011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,1,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,2,128,1,float16,float16,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,2,128,1,float16,fp8,0,0.013045333325862885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,2,128,1,float16,float16,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,2,128,1,float16,fp8,0,0.011600000162919363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,1,128,1,float16,float16,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,1,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,2,128,1,float16,float16,0,0.011098666737476984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,2,2,128,1,float16,fp8,0,0.010703999549150467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,2,128,1,float16,float16,0,0.012266666938861212
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,1,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,1,128,1,float16,fp8,0,0.011018666128317514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,2,128,1,float16,float16,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,2,128,1,float16,float16,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,2,2,128,1,float16,float16,0,0.01482133318980535
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,2,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,1,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,1,128,1,float16,fp8,0,0.010703999549150467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,2,128,1,float16,float16,0,0.011402666568756104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,2,2,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,2,128,1,float16,fp8,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,1,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,1,128,1,float16,fp8,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,2,128,1,float16,float16,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,2,128,1,float16,fp8,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,2,128,1,float16,float16,0,0.02143999934196472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,2,1,128,1,float16,float16,0,0.029135999580224354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,32,2,1,128,1,float16,fp8,0,0.02937600016593933
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,2,128,1,float16,fp8,0,0.02292266736427943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,2,2,128,1,float16,float16,0,0.014831999937693277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,1,128,1,float16,fp8,0,0.019248000035683315
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,2,1,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,2,128,1,float16,float16,0,0.021221332252025604
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,2,128,1,float16,float16,0,0.01504533365368843
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,2,128,1,float16,fp8,0,0.023029332359631855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,2,128,1,float16,fp8,0,0.015594666202863058
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,2,128,1,float16,fp8,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,1,128,1,float16,float16,0,0.012981332838535309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,1,128,1,float16,fp8,0,0.014885333677132925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,2,128,1,float16,float16,0,0.014922666052977243
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,2,2,128,1,float16,fp8,0,0.010911999891201654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,2,128,1,float16,float16,0,0.012800000607967377
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,1,128,1,float16,float16,0,0.012618667135636011
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,1,128,1,float16,fp8,0,0.012757333616415659
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,2,128,1,float16,float16,0,0.013023999830087027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,2,128,1,float16,float16,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,2,128,1,float16,fp8,0,0.013754667093356451
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,2,128,1,float16,fp8,0,0.011098666737476984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,1,128,1,float16,float16,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,1,128,1,float16,fp8,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,2,128,1,float16,float16,0,0.011274666835864386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,2,128,1,float16,float16,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,2,2,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,2,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,1,128,1,float16,float16,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,1,128,1,float16,fp8,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,2,128,1,float16,float16,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,2,2,128,1,float16,fp8,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,2,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,2,128,1,float16,fp8,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,1,128,1,float16,float16,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,1,128,1,float16,fp8,0,0.010805333654085795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,2,128,1,float16,float16,0,0.010869332899649939
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,2,2,128,1,float16,fp8,0,0.010944000134865442
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,2,128,1,float16,float16,0,0.011120000233252844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,2,128,1,float16,fp8,0,0.01101333275437355
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,1,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,1,128,1,float16,fp8,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,2,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,2,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,2,128,1,float16,fp8,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,1,128,1,float16,float16,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,1,128,1,float16,fp8,0,0.011120000233252844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,2,128,1,float16,float16,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,2,2,128,1,float16,fp8,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,2,1,128,1,float16,float16,0,0.018944000204404194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,2,128,1,float16,float16,0,0.019018666197856266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,2,1,128,1,float16,float16,0,0.025263999899228413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,16,2,1,128,1,float16,fp8,0,0.02492800106604894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,2,2,128,1,float16,fp8,0,0.012991999586423239
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,2,128,1,float16,fp8,0,0.019050666441520054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,1,128,1,float16,float16,0,0.016869333883126576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,1,128,1,float16,fp8,0,0.016810666769742966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,2,128,1,float16,float16,0,0.019013332823912304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,2,128,1,float16,float16,0,0.013125333935022354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,2,2,128,1,float16,fp8,0,0.018810667097568512
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,2,128,1,float16,fp8,0,0.014730667074521383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,1,128,1,float16,float16,0,0.012928000340859095
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,1,128,1,float16,fp8,0,0.012682666381200155
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,2,128,1,float16,float16,0,0.01303999995191892
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,2,2,128,1,float16,fp8,0,0.014864000181357065
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,2,128,1,float16,float16,0,0.011034666250149408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,2,128,1,float16,fp8,0,0.012725333372751871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,1,128,1,float16,float16,0,0.011936000237862269
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,1,128,1,float16,fp8,0,0.010677333921194077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,2,128,1,float16,float16,0,0.011829332758982977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,2,2,128,1,float16,float16,0,0.010784000158309937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,2,128,1,float16,float16,0,0.011317333827416102
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,2,2,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,1,128,1,float16,float16,0,0.011978667229413986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,1,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,2,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,2,128,1,float16,fp8,0,0.011328000575304031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,2,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,2,128,1,float16,fp8,0,0.011039999624093374
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,1,128,1,float16,float16,0,0.010656000425418219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,1,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,2,128,1,float16,fp8,0,0.011205332974592844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,2,2,128,1,float16,float16,0,0.010890666395425797
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,2,128,1,float16,fp8,0,0.010816000401973724
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,1,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,1,128,1,float16,fp8,0,0.011183999478816986
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,2,128,1,float16,float16,0,0.010714666297038397
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,2,128,1,float16,float16,0,0.010687999427318573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,2,128,1,float16,fp8,0,0.010885333021481832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,1,128,1,float16,float16,0,0.010741333166758219
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,1,128,1,float16,fp8,0,0.01091733326514562
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,2,128,1,float16,float16,0,0.010954666882753372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,2,128,1,float16,float16,0,0.010650667051474253
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,2,128,1,float16,fp8,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,1,128,1,float16,float16,0,0.010640000303586325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,2,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,2,2,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,2,1,128,1,float16,float16,0,0.011055999745925268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,256,1,2,1,128,1,float16,fp8,0,0.010794666906197866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,2,2,128,1,float16,fp8,0,0.011274666835864386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,2,128,1,float16,float16,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,2,128,1,float16,fp8,0,0.010794666906197866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,2,128,1,float16,float16,0,0.01071999967098236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,1,128,1,float16,float16,0,0.010687999427318573
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,1,128,1,float16,fp8,0,0.008986666798591614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,2,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,2,128,1,float16,float16,0,0.00966933307548364
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,2,2,128,1,float16,fp8,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,2,2,128,1,float16,fp8,0,0.010837333897749582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,1,128,1,float16,float16,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,1,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,2,2,128,1,float16,fp8,0,0.01655999943614006
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,2,128,1,float16,float16,0,0.009690666571259499
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,2,128,1,float16,fp8,0,0.009018666421373686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,2,128,1,float16,float16,0,0.008698666468262672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,2,128,1,float16,fp8,0,0.009082666908701261
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,1,128,1,float16,float16,0,0.008986666798591614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,2,128,1,float16,float16,0,0.00884799969693025
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,2,128,1,float16,float16,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,2,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,1,128,1,float16,float16,0,0.008885333314538002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,1,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,2,128,1,float16,float16,0,0.00873066671192646
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,2,2,128,1,float16,fp8,0,0.008901333436369896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,2,128,1,float16,float16,0,0.008687999720374743
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,1,128,1,float16,float16,0,0.008789333204428354
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,1,128,1,float16,fp8,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,2,128,1,float16,float16,0,0.008757333581646284
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,2,128,1,float16,fp8,0,0.008853333070874214
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,2,128,1,float16,float16,0,0.008901333436369896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,2,128,1,float16,fp8,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,1,128,1,float16,float16,0,0.008826666822036108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,1,128,1,float16,fp8,0,0.00897066667675972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,2,128,1,float16,float16,0,0.008879999940594038
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,2,2,128,1,float16,fp8,0,0.008986666798591614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,2,128,1,float16,float16,0,0.008874666566650072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,2,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,2,2,128,1,float16,fp8,0,0.011002667248249054
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,1,128,1,float16,float16,0,0.008869333192706108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,1,128,1,float16,fp8,0,0.008885333314538002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,2,128,1,float16,float16,0,0.008703999842206636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,2,2,128,1,float16,fp8,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,2,128,1,float16,float16,0,0.009061333412925402
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,2,2,128,1,float16,fp8,0,0.009962666779756546
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,2,128,1,float16,fp8,0,0.0086666668454806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,1,128,1,float16,float16,0,0.008650666723648706
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,1,128,1,float16,fp8,0,0.00902399979531765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,2,128,1,float16,fp8,0,0.00898133342464765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,1,1,128,1,float16,float16,0,0.37402133146921795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,1,1,128,1,float16,fp8,0,0.34281599521636963
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,2,2,128,1,float16,fp8,0,0.010928000013033548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,1,128,1,float16,fp8,0,0.008943999807039896
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,2,2,128,1,float16,fp8,0,0.008997333546479544
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,1,1,128,1,float16,float16,0,0.3736533323923747
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,1,1,128,1,float16,float16,0,0.22963200012842813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16384,1,1,128,1,float16,fp8,0,0.34384000301361084
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,1,1,128,1,float16,fp8,0,0.21787200371424356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,1,1,128,1,float16,float16,0,0.23406932751337686
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16384,1,1,128,1,float16,fp8,0,0.21965332825978598
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,1,1,128,1,float16,float16,0,0.2163626750310262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,1,1,128,1,float16,fp8,0,0.2059839963912964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,1,1,128,1,float16,float16,0,0.2167946696281433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16384,1,1,128,1,float16,fp8,0,0.2056480050086975
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,1,1,128,1,float16,float16,0,0.23438932498296103
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,2,2,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,1,1,128,1,float16,fp8,0,0.22645866870880127
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,1,1,128,1,float16,float16,0,0.23484265804290771
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,12288,1,1,128,1,float16,fp8,0,0.22531733910242716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,1,1,128,1,float16,fp8,0,0.16889599959055582
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,1,1,128,1,float16,fp8,0,0.16824533541997275
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,1,1,128,1,float16,float16,0,0.164682666460673
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,1,1,128,1,float16,fp8,0,0.1585919956366221
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,1,1,128,1,float16,float16,0,0.16521599888801575
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,12288,1,1,128,1,float16,fp8,0,0.15843733151753744
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,1,1,128,1,float16,float16,0,0.17631999651590982
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,1,1,128,1,float16,fp8,0,0.17095466454823813
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,1,1,128,1,float16,fp8,0,0.17032533884048462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,1,1,128,1,float16,float16,0,0.1469066639741262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,1,1,128,1,float16,fp8,0,0.14289599657058716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,1,1,128,1,float16,float16,0,0.14748266339302063
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,10240,1,1,128,1,float16,fp8,0,0.14219199617703757
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,1,1,128,1,float16,float16,0,0.139957328637441
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,1,1,128,1,float16,fp8,0,0.13375999530156454
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,1,1,128,1,float16,float16,0,0.13832533359527588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,10240,1,1,128,1,float16,fp8,0,0.13382933537165323
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,1,1,128,1,float16,float16,0,0.2255679965019226
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,1,1,128,1,float16,fp8,0,0.21110934019088745
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,1,1,128,1,float16,float16,0,0.22739734252293906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,8192,1,1,128,1,float16,fp8,0,0.20961066087086996
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,1,1,128,1,float16,float16,0,0.13376532991727194
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,2,2,128,1,float16,float16,0,0.008992000172535578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,1,1,128,1,float16,float16,0,0.13341333468755087
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,1,1,128,1,float16,fp8,0,0.12971199552218118
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,1,1,128,1,float16,fp8,0,0.11760000387827556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,1,1,128,1,float16,float16,0,0.12071466445922852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,1,1,128,1,float16,fp8,0,0.11682132879892985
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,1,1,128,1,float16,float16,0,0.11408533652623494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,1,1,128,1,float16,fp8,0,0.11101333300272624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,1,1,128,1,float16,float16,0,0.11315199732780457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,8192,1,1,128,1,float16,fp8,0,0.11095999677975972
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,1,1,128,1,float16,float16,0,0.14174933234850565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,1,1,128,1,float16,fp8,0,0.14203199744224548
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,1,1,128,1,float16,float16,0,0.14274133245150247
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,6144,1,1,128,1,float16,fp8,0,0.14201066891352335
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,1,1,128,1,float16,float16,0,0.10139200091362
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,1,1,128,1,float16,fp8,0,0.1009173293908437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,1,1,128,1,float16,float16,0,0.10240532954533894
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,1,1,128,1,float16,float16,0,0.1750239928563436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,6144,1,1,128,1,float16,fp8,0,0.10110933581988017
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,1,1,128,1,float16,float16,0,0.09287466605504353
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,12288,1,1,128,1,float16,float16,0,0.17550933361053467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,1,1,128,1,float16,fp8,0,0.09087466200192769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,1,1,128,1,float16,float16,0,0.09275199969609578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,6144,1,1,128,1,float16,fp8,0,0.09091732899347942
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,1,1,128,1,float16,float16,0,0.08849066495895386
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,1,1,128,1,float16,fp8,0,0.0865760048230489
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,1,1,128,1,float16,float16,0,0.08854400118192036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,6144,1,1,128,1,float16,fp8,0,0.08642133076985677
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,10240,1,1,128,1,float16,float16,0,0.1737119952837626
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,1,1,128,1,float16,float16,0,0.15213333566983542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,1,1,128,1,float16,fp8,0,0.1383680005868276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,1,1,128,1,float16,float16,0,0.08390933275222778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,1,1,128,1,float16,float16,0,0.15013333161671957
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,1,1,128,1,float16,fp8,0,0.08474133412043254
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,1,1,128,1,float16,float16,0,0.08306133250395457
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,4096,1,1,128,1,float16,fp8,0,0.08475200335184734
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,1,1,128,1,float16,float16,0,0.07228800157705943
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,1,1,128,1,float16,float16,0,0.07187733550866444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,1,1,128,1,float16,fp8,0,0.07246933380762736
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,4096,1,1,128,1,float16,fp8,0,0.07223999996980031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,1,1,128,1,float16,fp8,0,0.06577066580454509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,1,1,128,1,float16,float16,0,0.06593066453933716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,1,1,128,1,float16,fp8,0,0.06612266600131989
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,1,1,128,1,float16,float16,0,0.06228266656398773
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,1,1,128,1,float16,fp8,0,0.06200533111890157
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,1,1,128,1,float16,float16,0,0.06213866670926412
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,4096,1,1,128,1,float16,fp8,0,0.06185600161552429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,8192,1,1,128,1,float16,fp8,0,0.12994133432706198
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,1,1,128,1,float16,float16,0,0.09770666559537251
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,1,1,128,1,float16,fp8,0,0.09920000036557515
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,1,1,128,1,float16,float16,0,0.09539733330408733
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,3072,1,1,128,1,float16,fp8,0,0.09897599617640178
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,1,1,128,1,float16,float16,0,0.06418133278687795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,1,1,128,1,float16,fp8,0,0.06621333460013072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,1,1,128,1,float16,float16,0,0.06528000036875407
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,3072,1,1,128,1,float16,fp8,0,0.06635733445485432
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,1,1,128,1,float16,fp8,0,0.05738133192062378
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,1,1,128,1,float16,float16,0,0.05605866511662801
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,1,1,128,1,float16,fp8,0,0.05605333546797434
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,1,1,128,1,float16,float16,0,0.051967998345692955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,1,1,128,1,float16,fp8,0,0.0516480008761088
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,1,1,128,1,float16,float16,0,0.05190399785836538
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,3072,1,1,128,1,float16,fp8,0,0.05187733471393585
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,1,1,128,1,float16,float16,0,0.04956800242265066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,1,1,128,1,float16,fp8,0,0.04799999793370565
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,1,1,128,1,float16,float16,0,0.04956266780694326
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,3072,1,1,128,1,float16,fp8,0,0.04923733572165171
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,1,1,128,1,float16,float16,0,0.11202667156855266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,1,1,128,1,float16,fp8,0,0.10309333602587382
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,1,1,128,1,float16,fp8,0,0.10669333736101787
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,1,1,128,1,float16,float16,0,0.05772800246874491
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,1,1,128,1,float16,fp8,0,0.06196266909440359
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,4096,1,1,128,1,float16,fp8,0,0.1400320033232371
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,1,1,128,1,float16,fp8,0,0.06154666841030121
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,1,1,128,1,float16,float16,0,0.04753600060939789
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,4096,1,1,128,1,float16,float16,0,0.06592533489068349
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,1,1,128,1,float16,fp8,0,0.04900800188382467
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,1,1,128,1,float16,float16,0,0.04716266691684723
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,2048,1,1,128,1,float16,fp8,0,0.047594666481018066
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,1,1,128,1,float16,float16,0,0.041749333341916404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,1,1,128,1,float16,fp8,0,0.04269866645336151
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,1,1,128,1,float16,float16,0,0.04141866664091746
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,2048,1,1,128,1,float16,fp8,0,0.0418453315893809
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,8192,1,1,128,1,float16,float16,0,0.12018666664759318
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,1,1,128,1,float16,fp8,0,0.0386559988061587
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,1,1,128,1,float16,float16,0,0.03846933444341024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,1,1,128,1,float16,fp8,0,0.03859733293453852
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,1,1,128,1,float16,float16,0,0.037434667348861694
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,1,1,128,1,float16,fp8,0,0.037477334340413414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,3072,1,1,128,1,float16,float16,0,0.056517332792282104
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,1,1,128,1,float16,float16,0,0.037477334340413414
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,2048,1,1,128,1,float16,fp8,0,0.0373333344856898
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,1,1,128,1,float16,float16,0,0.07223466535409291
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,1,1,128,1,float16,fp8,0,0.07530666887760162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,1,1,128,1,float16,float16,0,0.07211199899514516
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1536,1,1,128,1,float16,fp8,0,0.0753119985262553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,1,1,128,1,float16,float16,0,0.046037331223487854
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,1,1,128,1,float16,fp8,0,0.049466664592425026
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,1,1,128,1,float16,float16,0,0.0459146648645401
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1536,1,1,128,1,float16,fp8,0,0.049738665421803795
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,1,1,128,1,float16,fp8,0,0.03973866750796636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,1,1,128,1,float16,float16,0,0.0376800000667572
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,1,1,128,1,float16,fp8,0,0.03939199944337209
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,1,1,128,1,float16,float16,0,0.033514666060606636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,1,1,128,1,float16,fp8,0,0.03532266616821289
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,2048,1,1,128,1,float16,float16,0,0.1106826663017273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,1,1,128,1,float16,float16,0,0.033887999753157295
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1536,1,1,128,1,float16,fp8,0,0.03533866753180822
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,1,1,128,1,float16,float16,0,0.03151999910672506
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,1,1,128,1,float16,fp8,0,0.03127466638882955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,1,1,128,1,float16,fp8,0,0.03155199935038885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1536,1,1,128,1,float16,float16,0,0.031504000226656594
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,1,1,128,1,float16,float16,0,0.029989334444204967
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,1,1,128,1,float16,fp8,0,0.029279999434947968
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,1,1,128,1,float16,float16,0,0.029472000896930695
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1536,1,1,128,1,float16,fp8,0,0.029445332785447437
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,2048,1,1,128,1,float16,float16,0,0.05793066819508871
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,1,1,128,1,float16,float16,0,0.045381332437197365
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,1,1,128,1,float16,float16,0,0.08989866574605306
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,1,1,128,1,float16,fp8,0,0.049770668148994446
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,1,1,128,1,float16,fp8,0,0.08700799942016602
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,1,1,128,1,float16,float16,0,0.09079999725023906
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,1,1,128,1,float16,float16,0,0.03475733349720637
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,1,1,128,1,float16,fp8,0,0.049653331438700356
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,1,1,128,1,float16,float16,0,0.03555733213822047
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,1,1,128,1,float16,fp8,0,0.03756266583998998
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,1,1,128,1,float16,float16,0,0.02917333443959554
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,1,1,128,1,float16,fp8,0,0.031301334500312805
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,2048,1,1,128,1,float16,float16,0,0.03896533449490865
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,1,1,128,1,float16,float16,0,0.029098667204380035
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,1,1,128,1,float16,float16,0,0.025839999318122864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,1,1,128,1,float16,fp8,0,0.027045334378878277
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,1,1,128,1,float16,float16,0,0.02621866762638092
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1024,1,1,128,1,float16,fp8,0,0.02734400083621343
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,1,1,128,1,float16,float16,0,0.02499733368555705
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,1,1,128,1,float16,fp8,0,0.025050667424996693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,1,1,128,1,float16,float16,0,0.02510933329661687
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1024,1,1,128,1,float16,fp8,0,0.025290665527184803
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,1,1,128,1,float16,float16,0,0.023344000180562336
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,1,1,128,1,float16,fp8,0,0.02333866556485494
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,1,1,128,1,float16,float16,0,0.023290666441122692
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1024,1,1,128,1,float16,fp8,0,0.02317333221435547
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1536,1,1,128,1,float16,float16,0,0.037621334195137024
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,1,1,128,1,float16,float16,0,0.08052266637484233
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,1,1,128,1,float16,fp8,0,0.07457600037256877
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,1,1,128,1,float16,float16,0,0.039706667264302574
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,1,1,128,1,float16,float16,0,0.039733332892258964
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,1,1,128,1,float16,fp8,0,0.043680002291997276
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1024,1,1,128,1,float16,fp8,0,0.08782399694124858
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1024,1,1,128,1,float16,float16,0,0.04584533472855886
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,1,1,128,1,float16,float16,0,0.027690666417280834
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,512,1,1,128,1,float16,fp8,0,0.04398400088151296
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,1,1,128,1,float16,fp8,0,0.031258667508761086
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,1,1,128,1,float16,float16,0,0.029045333464940388
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,512,1,1,128,1,float16,fp8,0,0.03124266614516576
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,1,1,128,1,float16,fp8,0,0.025055999557177227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,1,1,128,1,float16,float16,0,0.023370665808518726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,1,1,128,1,float16,float16,0,0.023013333479563396
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1024,1,1,128,1,float16,fp8,0,0.03143466760714849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,512,1,1,128,1,float16,fp8,0,0.025034666061401367
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,1,1,128,1,float16,float16,0,0.021055998901526134
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,1,1,128,1,float16,fp8,0,0.021210665504137676
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,1,1,128,1,float16,float16,0,0.021157334248224895
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,512,1,1,128,1,float16,fp8,0,0.021295999487241108
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,1,1,128,1,float16,float16,0,0.018933333456516266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,1,1,128,1,float16,fp8,0,0.018917333334684372
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,1,1,128,1,float16,float16,0,0.019306667149066925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,512,1,1,128,1,float16,fp8,0,0.019152000546455383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,1,1,128,1,float16,float16,0,0.01725333308180173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,1,1,128,1,float16,fp8,0,0.017194667210181553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,1,1,128,1,float16,float16,0,0.017008000363906223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,512,1,1,128,1,float16,fp8,0,0.017093333105246227
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,1,1,128,1,float16,float16,0,0.017162666966517765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,1,1,128,1,float16,fp8,0,0.01721599946419398
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,1,1,128,1,float16,float16,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,512,1,1,128,1,float16,fp8,0,0.01720533271630605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,1,1,128,1,float16,float16,0,0.03526400029659271
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,1,1,128,1,float16,fp8,0,0.04004266609748205
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,1,1,128,1,float16,float16,0,0.035205334424972534
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,256,1,1,128,1,float16,fp8,0,0.04144533226887385
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,1,1,128,1,float16,float16,0,0.025061334172884624
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,1,1,128,1,float16,fp8,0,0.029071999092896778
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,1,1,128,1,float16,float16,0,0.025146665672461193
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,256,1,1,128,1,float16,fp8,0,0.02924799919128418
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,1,1,128,1,float16,float16,0,0.019487999379634857
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,1,1,128,1,float16,fp8,0,0.022240000466505688
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,1,1,128,1,float16,float16,0,0.019850666324297588
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,256,1,1,128,1,float16,fp8,0,0.022821334501107533
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,1,1,128,1,float16,float16,0,0.016949333250522614
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,1,1,128,1,float16,float16,0,0.01703466723362605
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,1,1,128,1,float16,fp8,0,0.019205333044131596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,256,1,1,128,1,float16,fp8,0,0.01889066646496455
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,1,1,128,1,float16,float16,0,0.016789333273967106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,1,1,128,1,float16,fp8,0,0.01693333312869072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,1,1,128,1,float16,float16,0,0.01681600014368693
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,1,1,128,1,float16,float16,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,256,1,1,128,1,float16,fp8,0,0.016976000120242436
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,1,1,128,1,float16,fp8,0,0.014869333555301031
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,1,1,128,1,float16,float16,0,0.015103999525308609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,1,1,128,1,float16,float16,0,0.015002666662136713
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,1,1,128,1,float16,fp8,0,0.01515199989080429
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,1,1,128,1,float16,float16,0,0.014933332800865173
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,256,1,1,128,1,float16,fp8,0,0.014906667172908783
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,1,1,128,1,float16,fp8,0,0.014778666198253632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,1,1,128,1,float16,float16,0,0.01469333345691363
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,1,1,128,1,float16,fp8,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,1,1,128,1,float16,fp8,0,0.026373334228992462
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,1,1,128,1,float16,float16,0,0.023269332945346832
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,1,1,128,1,float16,float16,0,0.023525332411130268
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,1,1,128,1,float16,float16,0,0.017114666601022083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,1,1,128,1,float16,fp8,0,0.01971199984351794
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,128,1,1,128,1,float16,fp8,0,0.026288000245889027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,1,1,128,1,float16,float16,0,0.07990399996439616
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,1,1,128,1,float16,float16,0,0.017024000485738117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,128,1,1,128,1,float16,fp8,0,0.01932799940307935
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,1,1,128,1,float16,fp8,0,0.01573866605758667
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,1,1,128,1,float16,float16,0,0.015205333630243937
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,1,1,128,1,float16,fp8,0,0.01523200049996376
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,1,1,128,1,float16,float16,0,0.013050666699806849
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,1,1,128,1,float16,fp8,0,0.013455999394257864
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,1,1,128,1,float16,float16,0,0.012762666990359625
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,128,1,1,128,1,float16,fp8,0,0.014671999961137772
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,1,1,128,1,float16,float16,0,0.011930666863918304
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,1,1,128,1,float16,fp8,0,0.012170666207869848
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,1,1,128,1,float16,float16,0,0.012773333738247553
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,128,1,1,128,1,float16,fp8,0,0.012986666212479273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,1,1,128,1,float16,float16,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,1,1,128,1,float16,fp8,0,0.012741333494583765
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1024,1,1,128,1,float16,fp8,0,0.03743999948104223
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,1,1,128,1,float16,float16,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,128,1,1,128,1,float16,fp8,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,1,1,128,1,float16,float16,0,0.012538666526476542
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,256,1,1,128,1,float16,fp8,0,0.014896000425020853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,1,1,128,1,float16,float16,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,1,1,128,1,float16,fp8,0,0.011711999773979187
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,128,1,1,128,1,float16,fp8,0,0.011098666737476984
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,1,1,128,1,float16,float16,0,0.011722666521867117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,1,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,256,1,1,128,1,float16,float16,0,0.015114666273196539
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,1,1,128,1,float16,fp8,0,0.011157333850860596
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,1,1,128,1,float16,float16,0,0.014885333677132925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,1,1,128,1,float16,float16,0,0.017130666722853977
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,1,1,128,1,float16,fp8,0,0.01956266661485036
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,1,1,128,1,float16,float16,0,0.01820266619324684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,64,1,1,128,1,float16,fp8,0,0.01912533367673556
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,1,1,128,1,float16,fp8,0,0.015130666395028433
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,1,1,128,1,float16,float16,0,0.013034666577974955
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,1,1,128,1,float16,fp8,0,0.013045333325862885
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,1,1,128,1,float16,float16,0,0.014842666685581207
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,64,1,1,128,1,float16,fp8,0,0.016704000532627106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,1,1,128,1,float16,float16,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,64,1,1,128,1,float16,fp8,0,0.012826666235923767
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,1,1,128,1,float16,fp8,0,0.011722666521867117
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,1,1,128,1,float16,float16,0,0.011999999483426413
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,1,1,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,1,1,128,1,float16,fp8,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,1,1,128,1,float16,fp8,0,0.011087999989589056
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,1,1,128,1,float16,fp8,0,0.01102399950226148
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,1,1,128,1,float16,float16,0,0.010832000523805618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,1,1,128,1,float16,fp8,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,1,1,128,1,float16,float16,0,0.011061333119869232
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,64,1,1,128,1,float16,fp8,0,0.01097600037852923
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,1,1,128,1,float16,float16,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,1,1,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,1,1,128,1,float16,float16,0,0.010746666540702185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,64,1,1,128,1,float16,fp8,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,1,1,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,1,1,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,1,1,128,1,float16,float16,0,0.011050666371981302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,64,1,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,128,1,1,128,1,float16,float16,0,0.011205332974592844
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,1,1,128,1,float16,float16,0,0.014970666418472925
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,1,1,128,1,float16,float16,0,0.01505600040157636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,1,1,128,1,float16,fp8,0,0.015861333658297855
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,1,1,128,1,float16,float16,0,0.012975999464591345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,64,1,1,128,1,float16,float16,0,0.011605333536863327
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,1,1,128,1,float16,fp8,0,0.012965332716703415
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,64,1,1,128,1,float16,float16,0,0.010874666273593903
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,1,1,128,1,float16,float16,0,0.012981332838535309
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,1,1,128,1,float16,float16,0,0.011109333485364914
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,32,1,1,128,1,float16,fp8,0,0.013674666484196981
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,128,1,1,128,1,float16,float16,0,0.014730667074521383
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,1,1,128,1,float16,fp8,0,0.011071999867757162
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,1,1,128,1,float16,float16,0,0.010992000500361124
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,1,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,32,1,1,128,1,float16,fp8,0,0.011338666081428528
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,1,1,128,1,float16,float16,0,0.010735999792814255
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,1,1,128,1,float16,float16,0,0.010629333555698395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,32,1,1,128,1,float16,fp8,0,0.010879999647537867
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,1,1,128,1,float16,fp8,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,1,1,128,1,float16,fp8,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,1,1,128,1,float16,float16,0,0.010965333630641302
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,1,1,128,1,float16,fp8,0,0.010832000523805618
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,1,1,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,1,1,128,1,float16,float16,0,0.01090666651725769
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,1,1,128,1,float16,fp8,0,0.010970667004585266
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,1,1,128,1,float16,float16,0,0.010794666906197866
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,32,1,1,128,1,float16,fp8,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,1,1,128,1,float16,fp8,0,0.011045332998037338
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,1,1,128,1,float16,float16,0,0.011007999380429586
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,1,1,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,1,1,128,1,float16,float16,0,0.013376000026861826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,32,1,1,128,1,float16,fp8,0,0.016704000532627106
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,1,1,128,1,float16,fp8,0,0.014778666198253632
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,1,1,128,1,float16,float16,0,0.012693333129088083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,1,1,128,1,float16,float16,0,0.013023999830087027
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,16,1,1,128,1,float16,fp8,0,0.014767999450365702
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,1,1,128,1,float16,fp8,0,0.012693333129088083
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,1,1,128,1,float16,float16,0,0.01184533288081487
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,16,1,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,512,1,1,128,1,float16,fp8,0,0.07480533421039581
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,1,1,128,1,float16,fp8,0,0.010922666639089584
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,1,1,128,1,float16,float16,0,0.010949333508809408
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,1,1,128,1,float16,fp8,0,0.010933333386977514
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,1,1,128,1,float16,float16,0,0.01098666712641716
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,1,1,128,1,float16,fp8,0,0.010746666540702185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,1,1,128,1,float16,float16,0,0.01099733387430509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,1,1,128,1,float16,float16,0,0.01073066641887029
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,16,1,1,128,1,float16,fp8,0,0.010863999525705973
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,1,1,128,1,float16,float16,0,0.010826667149861654
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,32,1,1,128,1,float16,float16,0,0.010666667173306147
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,1,1,128,1,float16,fp8,0,0.010981333752473196
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,1,1,128,1,float16,float16,0,0.010725333044926325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,16,1,1,128,1,float16,fp8,0,0.010629333555698395
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,1,1,128,1,float16,float16,0,0.010677333921194077
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,1,1,128,1,float16,fp8,0,0.011077333241701126
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,32,1,1,128,1,float16,fp8,0,0.01071999967098236
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,1,1,128,1,float16,float16,0,0.01089599976936976
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,16,1,1,128,1,float16,fp8,0,0.010938666760921478
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,1,1,128,1,float16,float16,0,0.010847999403874079
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,1,1,128,1,float16,fp8,0,0.010725333044926325
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,1,1,128,1,float16,float16,0,0.01108266661564509
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,32,1,1,128,1,float16,float16,0,0.010853332777818045
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,16,1,1,128,1,float16,fp8,0,0.010746666540702185
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,1,1,128,1,float16,float16,0,0.010682666053374609
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,1,1,128,1,float16,fp8,0,0.010709332923094431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,1,1,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,16,1,1,128,1,float16,fp8,0,0.010709332923094431
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,1,1,128,1,float16,float16,0,0.008976000050703684
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,1,1,128,1,float16,fp8,0,0.008954666554927826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,1,1,128,1,float16,float16,0,0.010757333288590113
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,1,1,128,1,float16,fp8,0,0.009392000113924345
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,1,1,128,1,float16,float16,0,0.011029332876205444
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,1,1,128,1,float16,float16,0,0.008912000184257826
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,1,1,128,1,float16,fp8,0,0.008965333302815756
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,1,1,128,1,float16,fp8,0,0.008890666688481966
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,1,1,128,1,float16,float16,0,0.009008000294367472
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,1,1,1,128,1,float16,fp8,0,0.00927466650803884
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,1,1,128,1,float16,float16,0,0.008656000097592672
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,1,1,128,1,float16,fp8,0,0.008714666590094566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,1,1,128,1,float16,float16,0,0.008672000219424566
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,16,1,1,1,128,1,float16,fp8,0,0.008725333337982496
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,1,1,128,1,float16,fp8,0,0.008778666456540426
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,1,1,128,1,float16,float16,0,0.00874133345981439
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,1,1,128,1,float16,fp8,0,0.008992000172535578
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,1,1,128,1,float16,fp8,0,0.008618666479984919
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,1,1,128,1,float16,float16,0,0.008581333483258883
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,1,1,128,1,float16,fp8,0,0.009125333279371262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,1,1,128,1,float16,float16,0,0.00871999996403853
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,1,1,128,1,float16,fp8,0,0.009066666786869368
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,1,1,128,1,float16,float16,0,0.0086666668454806
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,2,1,1,1,128,1,float16,fp8,0,0.009002666920423508
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,1,1,128,1,float16,float16,0,0.008703999842206636
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,32,16,1,1,128,1,float16,float16,0,0.010901333143313726
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,1,1,128,1,float16,fp8,0,0.009125333279371262
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,1,1,128,1,float16,float16,0,0.008522666369875273
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,1,1,1,1,128,1,float16,fp8,0,0.008885333314538002
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,64,1,1,1,128,1,float16,float16,0,0.009103999783595404
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,8,1,1,1,128,1,float16,float16,0,0.008874666566650072
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,4,1,1,1,128,1,float16,float16,0,0.008645333349704742
VLLM,0.14.1.dev1+gd68209402,NVIDIA B200,context_attention,vllm_flashinfer,128,1,1,1,128,1,float16,fp8,0,0.010005333150426546
