ege-g commited on
Commit
4b841a9
·
verified ·
1 Parent(s): 3e6ce55

Add BF16 NVFP4 DFlash DGX Spark benchmark

Browse files
MANIFEST.md CHANGED
@@ -18,8 +18,14 @@ e7ffbde19080c85333332bd25f214de07274e172e01fa140aaee806d5c635efc model.safetens
18
  Primary benchmark artifacts copied from `results/`:
19
 
20
  ```text
 
 
 
 
21
  dflash20k-b4-l1-v28k-elyza100-streaming.json
22
  dflash20k-b4-l1-v28k-elyza100-streaming.csv
 
 
23
  elyza100-baseline-vs-dflash-summary.json
24
  elyza100-baseline-vs-dflash-speedup.png
25
  dflash20k-b4-l1-elyza100-tps-histogram.png
 
18
  Primary benchmark artifacts copied from `results/`:
19
 
20
  ```text
21
+ bf16-elyza100-streaming.json
22
+ bf16-elyza100-streaming.csv
23
+ baseline-elyza100-streaming.json
24
+ baseline-elyza100-streaming.csv
25
  dflash20k-b4-l1-v28k-elyza100-streaming.json
26
  dflash20k-b4-l1-v28k-elyza100-streaming.csv
27
+ elyza100-bf16-nvfp4-dflash-speedup-summary.json
28
+ elyza100-bf16-nvfp4-dflash-speedup.png
29
  elyza100-baseline-vs-dflash-summary.json
30
  elyza100-baseline-vs-dflash-speedup.png
31
  dflash20k-b4-l1-elyza100-tps-histogram.png
README.md CHANGED
@@ -50,14 +50,23 @@ If you use an environment wrapper, set the same JSON as `SPECULATIVE_CONFIG`.
50
  The verifier model still needs the serving requirements documented on the NVFP4
51
  model page, including the LLM-jp remote code/plugin setup.
52
 
53
- ## Benchmark
54
 
55
- Representative Japanese streaming decode benchmark on ELYZA-100 prompts:
 
 
 
56
 
57
- | Setup | Mean decode tok/s | Notes |
58
- | --- | ---: | --- |
59
- | Baseline NVFP4 | 29.73 | no speculative decoding |
60
- | NVFP4 + this DFlash drafter | 48.76 paired / 49.04 standalone run | stock vLLM, DFlash |
 
 
 
 
 
 
61
 
62
  Paired speedup summary:
63
 
@@ -71,13 +80,19 @@ Paired speedup summary:
71
 
72
  The raw benchmark artifacts are included under [`benchmark/`](benchmark/):
73
 
 
 
 
 
74
  - `dflash20k-b4-l1-v28k-elyza100-streaming.json`
75
  - `dflash20k-b4-l1-v28k-elyza100-streaming.csv`
 
 
76
  - `elyza100-baseline-vs-dflash-summary.json`
77
  - `elyza100-baseline-vs-dflash-speedup.png`
78
  - `dflash20k-b4-l1-elyza100-tps-histogram.png`
79
 
80
- ![Decode speedup distribution](benchmark/elyza100-baseline-vs-dflash-speedup.png)
81
 
82
  ## Validation Metrics
83
 
 
50
  The verifier model still needs the serving requirements documented on the NVFP4
51
  model page, including the LLM-jp remote code/plugin setup.
52
 
53
+ ## Benchmark (DGX Spark / GB10, SM121)
54
 
55
+ Representative Japanese streaming decode benchmark on ELYZA-100 prompts,
56
+ measured on NVIDIA DGX Spark / GB10 (SM121) with vLLM 0.24.0, stock vLLM
57
+ DFlash, temperature 0, serial streaming requests, and exactly 128 generated
58
+ tokens per request. Decode speed excludes time-to-first-token.
59
 
60
+ | Setup | Mean decode tok/s | Speedup vs BF16 | Notes |
61
+ | --- | ---: | ---: | --- |
62
+ | BF16 base model | 14.42 | 1.00x | `llm-jp/llm-jp-4-8b-instruct`, no speculative decoding |
63
+ | NVFP4 verifier | 29.73 | 2.06x | `kel-jp/llm-jp-4-8b-instruct-NVFP4`, no speculative decoding |
64
+ | NVFP4 + this DFlash drafter | 49.04 | 3.40x | stock vLLM DFlash, `num_speculative_tokens=3` |
65
+
66
+ The DFlash row is 1.65x faster than the NVFP4 verifier baseline by ratio of
67
+ standalone means. In the paired per-prompt run against the same NVFP4 baseline,
68
+ the DFlash mean was 48.76 tok/s and the ratio of mean decode throughput was
69
+ 1.64x.
70
 
71
  Paired speedup summary:
72
 
 
80
 
81
  The raw benchmark artifacts are included under [`benchmark/`](benchmark/):
82
 
83
+ - `bf16-elyza100-streaming.json`
84
+ - `bf16-elyza100-streaming.csv`
85
+ - `baseline-elyza100-streaming.json`
86
+ - `baseline-elyza100-streaming.csv`
87
  - `dflash20k-b4-l1-v28k-elyza100-streaming.json`
88
  - `dflash20k-b4-l1-v28k-elyza100-streaming.csv`
89
+ - `elyza100-bf16-nvfp4-dflash-speedup-summary.json`
90
+ - `elyza100-bf16-nvfp4-dflash-speedup.png`
91
  - `elyza100-baseline-vs-dflash-summary.json`
92
  - `elyza100-baseline-vs-dflash-speedup.png`
93
  - `dflash20k-b4-l1-elyza100-tps-histogram.png`
94
 
95
+ ![BF16, NVFP4, and DFlash speedup distribution](benchmark/elyza100-bf16-nvfp4-dflash-speedup.png)
96
 
97
  ## Validation Metrics
98
 
benchmark/baseline-elyza100-streaming.csv ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ prompt_tokens,completion_tokens,ttft_ms,decode_seconds,decode_tokens_per_second,e2e_seconds,e2e_tokens_per_second,index,prompt_characters
2
+ 89,128,143.46256002318114,4.237766047997866,29.968619919450532,4.381228608021047,29.215549210479615,0,28
3
+ 100,128,149.23513599205762,4.245029520010576,29.91734201171906,4.394264656002633,29.12887821291101,1,50
4
+ 86,128,147.96897600172088,4.250384335988201,29.879650864672428,4.398353311989922,29.101800360392076,2,24
5
+ 223,128,159.77539197774604,4.268197119992692,29.754951898804862,4.427972511970438,28.907135185227308,3,323
6
+ 87,128,147.64036802807823,4.255182991968468,29.845954977661066,4.402823359996546,29.072254218279703,4,15
7
+ 138,128,150.07712005171925,4.252459567971528,29.86506937221286,4.402536688023247,29.0741472633752,5,107
8
+ 197,128,162.3923999723047,4.261921360041015,29.798766629231707,4.4243137600133196,28.931040369888834,6,286
9
+ 129,128,155.38915200158954,4.255304831953254,29.8451004135713,4.410693983954843,29.02037649078274,7,77
10
+ 253,128,157.81886398326606,4.268306992016733,29.754185965895992,4.426125855999999,28.91919574010415,8,359
11
+ 107,128,149.04068800387904,4.26341652800329,29.788316287144156,4.412457216007169,29.00877985528144,9,76
12
+ 115,128,148.95393600454554,4.266124832036439,29.769405490972566,4.415078768040985,28.991555241673503,10,93
13
+ 136,128,186.66489596944302,4.235812655999325,29.98244028099911,4.422477551968768,28.943052507528915,11,126
14
+ 239,128,159.81835196726024,4.27479575999314,29.70902170077099,4.4346141119604,28.86384176128807,12,301
15
+ 113,128,149.9263520236127,4.26866089599207,29.751719120917475,4.418587248015683,28.968535148306227,13,76
16
+ 239,128,159.59620795911178,4.279861872026231,29.673854857347045,4.439458079985343,28.832347933877234,14,323
17
+ 122,128,182.7699519926682,4.240294800023548,29.950747763880642,4.4230647520162165,28.939210067329963,15,120
18
+ 152,128,153.0061599914916,4.274810159986373,29.708921623879746,4.427816319977865,28.90815488946025,16,182
19
+ 141,128,151.20801597367972,4.281187455984764,29.66466694245401,4.4323954719584435,28.878289586250187,17,140
20
+ 196,128,160.8437920222059,4.278787071991246,29.681308712774342,4.439630864013452,28.831225820492485,18,191
21
+ 118,128,149.06441600760445,4.280752464022953,29.667681340454877,4.4298168800305575,28.895099609426076,19,75
22
+ 269,128,159.87711999332532,4.284917648008559,29.638842664578164,4.444794768001884,28.797730082269062,20,368
23
+ 151,128,188.5674080112949,4.241870319994632,29.93962342539522,4.430437728005927,28.891050469094587,21,137
24
+ 83,128,181.51931202737615,4.240379583963659,29.950148915981675,4.421898895991035,28.9468400365388,22,8
25
+ 257,128,158.51934399688616,4.284019791986793,29.645054450390717,4.442539135983679,28.81235169392782,23,379
26
+ 264,128,157.54560002824292,4.294625615992118,29.57184429000833,4.452171216020361,28.75001741609001,24,480
27
+ 115,128,183.9634400093928,4.240709887992125,29.947816133239776,4.424673328001518,28.92868930909606,25,82
28
+ 167,128,151.50518395239487,4.277752512018196,29.688487036872267,4.429257695970591,28.898747552314436,26,145
29
+ 86,128,145.308943989221,4.2739029280492105,29.715228010096187,4.4192118720384315,28.964440652843823,27,11
30
+ 95,128,147.21774403005838,4.27222136000637,29.72692407488236,4.419439104036428,28.962951403288514,28,33
31
+ 138,128,152.53638400463387,4.275733312009834,29.702507320388253,4.428269696014468,28.905195208684464,29,155
32
+ 89,128,147.56200002739206,4.27770995197352,29.688782415322148,4.425271952000912,28.924776011138494,30,23
33
+ 122,128,149.15563201066107,4.275628671981394,29.703234247689284,4.424784303992055,28.927963761876022,31,97
34
+ 94,128,145.9132640156895,4.272694383980706,29.723633049008047,4.418607647996396,28.96840140537058,32,61
35
+ 132,128,153.64075201796368,4.272159695974551,29.72735315106922,4.425800447992515,28.921322030698228,33,136
36
+ 114,128,148.47996801836416,4.275548528006766,29.70379102659995,4.42402849602513,28.93290586057584,34,82
37
+ 97,128,147.8521599783562,4.273406176012941,29.71868218679137,4.421258335991297,28.95103390770332,35,49
38
+ 116,128,147.574128000997,4.273268640041351,29.719638688283133,4.420842768042348,28.953755362053144,36,101
39
+ 127,128,149.55046400427818,4.280436095956247,29.669874085955314,4.4299865599605255,28.893992852461515,37,118
40
+ 99,128,146.9279839657247,4.2737699200515635,29.71615280554638,4.420697904017288,28.95470416191086,38,44
41
+ 89,128,145.21406398853287,4.270830192021094,29.736607237924275,4.416044256009627,28.985216764032575,39,21
42
+ 131,128,153.4373280010186,4.277822831994854,29.68799900971513,4.4312601599958725,28.885688354646103,40,99
43
+ 164,128,152.07355201710016,4.281609760015272,29.66174105496878,4.433683312032372,28.86990138709876,41,255
44
+ 119,128,147.45664002839476,4.278829519986175,29.6810142602761,4.42628616001457,28.918148391828936,42,98
45
+ 109,128,148.86659203330055,4.27835519996006,29.684304847149104,4.427221791993361,28.912036941878142,43,69
46
+ 116,128,149.5863359887153,4.2813182240352035,29.66376086856274,4.430904560023919,28.88800656074367,44,63
47
+ 99,128,148.91756797442213,4.275036848033778,29.70734627899435,4.4239544160082005,28.93339034797205,45,51
48
+ 87,128,181.61635199794546,4.241584288014565,29.94164240914972,4.42320064001251,28.938321007214807,46,16
49
+ 121,128,149.99320002971217,4.277360575972125,29.69120740332639,4.427353776001837,28.911175044066976,47,83
50
+ 103,128,150.74512001592666,4.27451617596671,29.710964883944584,4.4252612959826365,28.92484566192772,48,52
51
+ 128,128,148.76968000316992,4.281864800024778,29.659974317560213,4.430634480027948,28.88976749876072,49,82
52
+ 116,128,148.97512004245073,4.279575983993709,29.675837156531415,4.42855110403616,28.90335845584833,50,89
53
+ 202,128,159.03887996682897,4.285158192040399,29.637178911131013,4.444197072007228,28.801603062617705,51,202
54
+ 133,128,150.73892794316635,4.277238800015766,29.692052732602132,4.4279777279589325,28.9071011337271,52,73
55
+ 97,128,148.48937600618228,4.274834448006004,29.708752828835074,4.423323824012186,28.937515111407173,53,42
56
+ 208,128,160.55590397445485,4.285696112026926,29.633458994817804,4.446252016001381,28.78829169811958,54,254
57
+ 113,128,148.43628800008446,4.276303727994673,29.698545304113676,4.424740015994757,28.92825330692868,55,85
58
+ 127,128,182.12604796281084,4.240071760024875,29.952323259560806,4.422197807987686,28.94488341719978,56,107
59
+ 144,128,185.98363199271262,4.249349439982325,29.88692782124509,4.4353330719750375,28.859162981192316,57,121
60
+ 230,128,162.00067201862112,4.287982367968652,29.617659099695356,4.4499830399872735,28.764154570882603,58,346
61
+ 124,128,183.87990404153243,4.242221055959817,29.937148094057964,4.4261009600013494,28.91935840522738,59,93
62
+ 170,128,152.916015998926,4.2795338880387135,29.67612906512195,4.4324499040376395,28.8779349504664,60,207
63
+ 148,128,153.09798397356644,4.275488400016911,29.70420876349417,4.428586383990478,28.90312819971747,61,122
64
+ 123,128,149.6343039907515,4.274956831999589,29.707902322980047,4.42459113599034,28.929226693699963,62,88
65
+ 98,128,150.04775999113917,4.279320368019398,29.677609778671357,4.4293681280105375,28.898027055044427,63,43
66
+ 196,128,160.95686401240528,4.288259503955487,29.615745008634693,4.449216367967892,28.769111100447994,64,177
67
+ 280,128,194.86894400324672,4.253986447991338,29.85434992628322,4.448855391994584,28.771445399265478,65,494
68
+ 100,128,146.7246399843134,4.280999056005385,29.665972437402065,4.427723695989698,28.908759622000094,66,56
69
+ 111,128,149.47828801814467,4.27679667196935,29.695122247071872,4.426274959987495,28.918221564880284,67,68
70
+ 222,128,161.32521600229666,4.284537935978733,29.64146937141984,4.44586315198103,28.790809708788395,68,333
71
+ 86,128,149.98059201752767,4.274428736011032,29.71157266702229,4.42440932802856,28.93041545435729,69,20
72
+ 116,128,148.25239998754114,4.280142048024572,29.671912421368052,4.428394448012114,28.904380922403746,70,70
73
+ 122,128,151.68241603532806,4.279039520013612,29.679557621752465,4.43072193604894,28.88919725667618,71,100
74
+ 123,128,182.6112479902804,4.241006672033109,29.94572039640699,4.423617920023389,28.93559125452752,72,100
75
+ 304,128,160.21692799404263,4.284923232044093,29.638804039767013,4.445140160038136,28.795492468543863,73,516
76
+ 96,128,145.91110398760065,4.27431379200425,29.712371664797445,4.420224895991851,28.957802603226636,74,51
77
+ 98,128,182.3572000139393,4.2468398079508916,29.904589234148144,4.429197007964831,28.899143517396766,75,38
78
+ 146,128,152.6936799637042,4.277269551996142,29.691839257764542,4.429963231959846,28.894145007016668,76,133
79
+ 140,128,152.17240003403276,4.280287823989056,29.67090186978153,4.432460224023089,28.87786771469813,77,158
80
+ 89,128,147.72603200981393,4.280198383959942,29.671521879904667,4.4279244159697555,28.90744917378334,78,17
81
+ 163,128,186.45403202390298,4.244761504000053,29.919231005162832,4.431215536023956,28.88597924416286,79,193
82
+ 154,128,152.16956799849868,4.277686815999914,29.68894298782685,4.429856383998413,28.894841932655726,80,151
83
+ 119,128,147.88078400306404,4.278590032015927,29.682675612685866,4.426470816018991,28.916942033545045,81,75
84
+ 274,128,158.65448000840843,4.288939487945754,29.611049621226616,4.4475939679541625,28.779605540044024,82,413
85
+ 96,128,151.030623994302,4.277100688021164,29.693011519623024,4.428131312015466,28.906098527990746,83,53
86
+ 187,128,153.6058560013771,4.2891785760293715,29.609399037325215,4.442784432030749,28.810760899666832,84,243
87
+ 175,128,150.53255995735526,4.287062800023705,29.62401203903469,4.43759535998106,28.844450567603424,85,193
88
+ 95,128,146.89548802562058,4.2793887840234675,29.677135312906767,4.426284272049088,28.91816072643345,86,54
89
+ 135,128,152.44775998871773,4.279913456004579,29.673497210982887,4.432361215993296,28.878512775118008,87,153
90
+ 132,128,152.3534239968285,4.280044800019823,29.67258660456353,4.4323982240166515,28.878271655836475,88,83
91
+ 151,128,152.58678403915837,4.286318383994512,29.629156918027626,4.438905168033671,28.8359393036326,89,154
92
+ 98,128,149.5123039931059,4.2784633280243725,29.68355464639302,4.427975632017478,28.90711481663699,90,47
93
+ 120,128,149.92755197454244,4.279246336023789,29.67812320849154,4.429173887998331,28.899294368830212,91,91
94
+ 129,128,149.91504000499845,4.284216400003061,29.643694001990486,4.434131440008059,28.86698369946547,92,149
95
+ 126,128,148.77603197237477,4.277493072033394,29.690287713225434,4.426269104005769,28.9182598238684,93,129
96
+ 176,128,150.4963680054061,4.283495743991807,29.648681261825693,4.433992111997213,28.86789077807914,94,181
97
+ 151,128,151.79175999946892,4.284410799969919,29.642348955168274,4.436202559969388,28.853506635387557,95,182
98
+ 113,128,147.12048001820222,4.276005279971287,29.700618143496026,4.423125759989489,28.93881091011624,96,72
99
+ 159,128,152.09144004620612,4.281069583958015,29.665483708999556,4.433161024004221,28.873302663025065,97,144
100
+ 135,128,156.03772801114246,4.28199811198283,29.659050910041422,4.438035839993972,28.84158772367504,98,146
101
+ 183,128,154.6247520018369,4.2839804160175845,29.64532693127015,4.438605168019421,28.837888290279196,99,214
benchmark/baseline-elyza100-streaming.json ADDED
@@ -0,0 +1,1120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": "training/data/benchmarks/elyza-tasks-100/test.csv",
3
+ "model": "llm-jp-4-8b-instruct-NVFP4",
4
+ "temperature": 0,
5
+ "max_tokens": 128,
6
+ "summary": {
7
+ "count": 100,
8
+ "mean": 29.72816958251134,
9
+ "median": 29.69194599518334,
10
+ "stdev": 0.10065825734656245,
11
+ "min": 29.57184429000833,
12
+ "p05": 29.62401203903469,
13
+ "p25": 29.669874085955314,
14
+ "p75": 29.736607237924275,
15
+ "p95": 29.947816133239776,
16
+ "max": 29.98244028099911
17
+ },
18
+ "records": [
19
+ {
20
+ "prompt_tokens": 89,
21
+ "completion_tokens": 128,
22
+ "ttft_ms": 143.46256002318114,
23
+ "decode_seconds": 4.237766047997866,
24
+ "decode_tokens_per_second": 29.968619919450532,
25
+ "e2e_seconds": 4.381228608021047,
26
+ "e2e_tokens_per_second": 29.215549210479615,
27
+ "index": 0,
28
+ "prompt_characters": 28
29
+ },
30
+ {
31
+ "prompt_tokens": 100,
32
+ "completion_tokens": 128,
33
+ "ttft_ms": 149.23513599205762,
34
+ "decode_seconds": 4.245029520010576,
35
+ "decode_tokens_per_second": 29.91734201171906,
36
+ "e2e_seconds": 4.394264656002633,
37
+ "e2e_tokens_per_second": 29.12887821291101,
38
+ "index": 1,
39
+ "prompt_characters": 50
40
+ },
41
+ {
42
+ "prompt_tokens": 86,
43
+ "completion_tokens": 128,
44
+ "ttft_ms": 147.96897600172088,
45
+ "decode_seconds": 4.250384335988201,
46
+ "decode_tokens_per_second": 29.879650864672428,
47
+ "e2e_seconds": 4.398353311989922,
48
+ "e2e_tokens_per_second": 29.101800360392076,
49
+ "index": 2,
50
+ "prompt_characters": 24
51
+ },
52
+ {
53
+ "prompt_tokens": 223,
54
+ "completion_tokens": 128,
55
+ "ttft_ms": 159.77539197774604,
56
+ "decode_seconds": 4.268197119992692,
57
+ "decode_tokens_per_second": 29.754951898804862,
58
+ "e2e_seconds": 4.427972511970438,
59
+ "e2e_tokens_per_second": 28.907135185227308,
60
+ "index": 3,
61
+ "prompt_characters": 323
62
+ },
63
+ {
64
+ "prompt_tokens": 87,
65
+ "completion_tokens": 128,
66
+ "ttft_ms": 147.64036802807823,
67
+ "decode_seconds": 4.255182991968468,
68
+ "decode_tokens_per_second": 29.845954977661066,
69
+ "e2e_seconds": 4.402823359996546,
70
+ "e2e_tokens_per_second": 29.072254218279703,
71
+ "index": 4,
72
+ "prompt_characters": 15
73
+ },
74
+ {
75
+ "prompt_tokens": 138,
76
+ "completion_tokens": 128,
77
+ "ttft_ms": 150.07712005171925,
78
+ "decode_seconds": 4.252459567971528,
79
+ "decode_tokens_per_second": 29.86506937221286,
80
+ "e2e_seconds": 4.402536688023247,
81
+ "e2e_tokens_per_second": 29.0741472633752,
82
+ "index": 5,
83
+ "prompt_characters": 107
84
+ },
85
+ {
86
+ "prompt_tokens": 197,
87
+ "completion_tokens": 128,
88
+ "ttft_ms": 162.3923999723047,
89
+ "decode_seconds": 4.261921360041015,
90
+ "decode_tokens_per_second": 29.798766629231707,
91
+ "e2e_seconds": 4.4243137600133196,
92
+ "e2e_tokens_per_second": 28.931040369888834,
93
+ "index": 6,
94
+ "prompt_characters": 286
95
+ },
96
+ {
97
+ "prompt_tokens": 129,
98
+ "completion_tokens": 128,
99
+ "ttft_ms": 155.38915200158954,
100
+ "decode_seconds": 4.255304831953254,
101
+ "decode_tokens_per_second": 29.8451004135713,
102
+ "e2e_seconds": 4.410693983954843,
103
+ "e2e_tokens_per_second": 29.02037649078274,
104
+ "index": 7,
105
+ "prompt_characters": 77
106
+ },
107
+ {
108
+ "prompt_tokens": 253,
109
+ "completion_tokens": 128,
110
+ "ttft_ms": 157.81886398326606,
111
+ "decode_seconds": 4.268306992016733,
112
+ "decode_tokens_per_second": 29.754185965895992,
113
+ "e2e_seconds": 4.426125855999999,
114
+ "e2e_tokens_per_second": 28.91919574010415,
115
+ "index": 8,
116
+ "prompt_characters": 359
117
+ },
118
+ {
119
+ "prompt_tokens": 107,
120
+ "completion_tokens": 128,
121
+ "ttft_ms": 149.04068800387904,
122
+ "decode_seconds": 4.26341652800329,
123
+ "decode_tokens_per_second": 29.788316287144156,
124
+ "e2e_seconds": 4.412457216007169,
125
+ "e2e_tokens_per_second": 29.00877985528144,
126
+ "index": 9,
127
+ "prompt_characters": 76
128
+ },
129
+ {
130
+ "prompt_tokens": 115,
131
+ "completion_tokens": 128,
132
+ "ttft_ms": 148.95393600454554,
133
+ "decode_seconds": 4.266124832036439,
134
+ "decode_tokens_per_second": 29.769405490972566,
135
+ "e2e_seconds": 4.415078768040985,
136
+ "e2e_tokens_per_second": 28.991555241673503,
137
+ "index": 10,
138
+ "prompt_characters": 93
139
+ },
140
+ {
141
+ "prompt_tokens": 136,
142
+ "completion_tokens": 128,
143
+ "ttft_ms": 186.66489596944302,
144
+ "decode_seconds": 4.235812655999325,
145
+ "decode_tokens_per_second": 29.98244028099911,
146
+ "e2e_seconds": 4.422477551968768,
147
+ "e2e_tokens_per_second": 28.943052507528915,
148
+ "index": 11,
149
+ "prompt_characters": 126
150
+ },
151
+ {
152
+ "prompt_tokens": 239,
153
+ "completion_tokens": 128,
154
+ "ttft_ms": 159.81835196726024,
155
+ "decode_seconds": 4.27479575999314,
156
+ "decode_tokens_per_second": 29.70902170077099,
157
+ "e2e_seconds": 4.4346141119604,
158
+ "e2e_tokens_per_second": 28.86384176128807,
159
+ "index": 12,
160
+ "prompt_characters": 301
161
+ },
162
+ {
163
+ "prompt_tokens": 113,
164
+ "completion_tokens": 128,
165
+ "ttft_ms": 149.9263520236127,
166
+ "decode_seconds": 4.26866089599207,
167
+ "decode_tokens_per_second": 29.751719120917475,
168
+ "e2e_seconds": 4.418587248015683,
169
+ "e2e_tokens_per_second": 28.968535148306227,
170
+ "index": 13,
171
+ "prompt_characters": 76
172
+ },
173
+ {
174
+ "prompt_tokens": 239,
175
+ "completion_tokens": 128,
176
+ "ttft_ms": 159.59620795911178,
177
+ "decode_seconds": 4.279861872026231,
178
+ "decode_tokens_per_second": 29.673854857347045,
179
+ "e2e_seconds": 4.439458079985343,
180
+ "e2e_tokens_per_second": 28.832347933877234,
181
+ "index": 14,
182
+ "prompt_characters": 323
183
+ },
184
+ {
185
+ "prompt_tokens": 122,
186
+ "completion_tokens": 128,
187
+ "ttft_ms": 182.7699519926682,
188
+ "decode_seconds": 4.240294800023548,
189
+ "decode_tokens_per_second": 29.950747763880642,
190
+ "e2e_seconds": 4.4230647520162165,
191
+ "e2e_tokens_per_second": 28.939210067329963,
192
+ "index": 15,
193
+ "prompt_characters": 120
194
+ },
195
+ {
196
+ "prompt_tokens": 152,
197
+ "completion_tokens": 128,
198
+ "ttft_ms": 153.0061599914916,
199
+ "decode_seconds": 4.274810159986373,
200
+ "decode_tokens_per_second": 29.708921623879746,
201
+ "e2e_seconds": 4.427816319977865,
202
+ "e2e_tokens_per_second": 28.90815488946025,
203
+ "index": 16,
204
+ "prompt_characters": 182
205
+ },
206
+ {
207
+ "prompt_tokens": 141,
208
+ "completion_tokens": 128,
209
+ "ttft_ms": 151.20801597367972,
210
+ "decode_seconds": 4.281187455984764,
211
+ "decode_tokens_per_second": 29.66466694245401,
212
+ "e2e_seconds": 4.4323954719584435,
213
+ "e2e_tokens_per_second": 28.878289586250187,
214
+ "index": 17,
215
+ "prompt_characters": 140
216
+ },
217
+ {
218
+ "prompt_tokens": 196,
219
+ "completion_tokens": 128,
220
+ "ttft_ms": 160.8437920222059,
221
+ "decode_seconds": 4.278787071991246,
222
+ "decode_tokens_per_second": 29.681308712774342,
223
+ "e2e_seconds": 4.439630864013452,
224
+ "e2e_tokens_per_second": 28.831225820492485,
225
+ "index": 18,
226
+ "prompt_characters": 191
227
+ },
228
+ {
229
+ "prompt_tokens": 118,
230
+ "completion_tokens": 128,
231
+ "ttft_ms": 149.06441600760445,
232
+ "decode_seconds": 4.280752464022953,
233
+ "decode_tokens_per_second": 29.667681340454877,
234
+ "e2e_seconds": 4.4298168800305575,
235
+ "e2e_tokens_per_second": 28.895099609426076,
236
+ "index": 19,
237
+ "prompt_characters": 75
238
+ },
239
+ {
240
+ "prompt_tokens": 269,
241
+ "completion_tokens": 128,
242
+ "ttft_ms": 159.87711999332532,
243
+ "decode_seconds": 4.284917648008559,
244
+ "decode_tokens_per_second": 29.638842664578164,
245
+ "e2e_seconds": 4.444794768001884,
246
+ "e2e_tokens_per_second": 28.797730082269062,
247
+ "index": 20,
248
+ "prompt_characters": 368
249
+ },
250
+ {
251
+ "prompt_tokens": 151,
252
+ "completion_tokens": 128,
253
+ "ttft_ms": 188.5674080112949,
254
+ "decode_seconds": 4.241870319994632,
255
+ "decode_tokens_per_second": 29.93962342539522,
256
+ "e2e_seconds": 4.430437728005927,
257
+ "e2e_tokens_per_second": 28.891050469094587,
258
+ "index": 21,
259
+ "prompt_characters": 137
260
+ },
261
+ {
262
+ "prompt_tokens": 83,
263
+ "completion_tokens": 128,
264
+ "ttft_ms": 181.51931202737615,
265
+ "decode_seconds": 4.240379583963659,
266
+ "decode_tokens_per_second": 29.950148915981675,
267
+ "e2e_seconds": 4.421898895991035,
268
+ "e2e_tokens_per_second": 28.9468400365388,
269
+ "index": 22,
270
+ "prompt_characters": 8
271
+ },
272
+ {
273
+ "prompt_tokens": 257,
274
+ "completion_tokens": 128,
275
+ "ttft_ms": 158.51934399688616,
276
+ "decode_seconds": 4.284019791986793,
277
+ "decode_tokens_per_second": 29.645054450390717,
278
+ "e2e_seconds": 4.442539135983679,
279
+ "e2e_tokens_per_second": 28.81235169392782,
280
+ "index": 23,
281
+ "prompt_characters": 379
282
+ },
283
+ {
284
+ "prompt_tokens": 264,
285
+ "completion_tokens": 128,
286
+ "ttft_ms": 157.54560002824292,
287
+ "decode_seconds": 4.294625615992118,
288
+ "decode_tokens_per_second": 29.57184429000833,
289
+ "e2e_seconds": 4.452171216020361,
290
+ "e2e_tokens_per_second": 28.75001741609001,
291
+ "index": 24,
292
+ "prompt_characters": 480
293
+ },
294
+ {
295
+ "prompt_tokens": 115,
296
+ "completion_tokens": 128,
297
+ "ttft_ms": 183.9634400093928,
298
+ "decode_seconds": 4.240709887992125,
299
+ "decode_tokens_per_second": 29.947816133239776,
300
+ "e2e_seconds": 4.424673328001518,
301
+ "e2e_tokens_per_second": 28.92868930909606,
302
+ "index": 25,
303
+ "prompt_characters": 82
304
+ },
305
+ {
306
+ "prompt_tokens": 167,
307
+ "completion_tokens": 128,
308
+ "ttft_ms": 151.50518395239487,
309
+ "decode_seconds": 4.277752512018196,
310
+ "decode_tokens_per_second": 29.688487036872267,
311
+ "e2e_seconds": 4.429257695970591,
312
+ "e2e_tokens_per_second": 28.898747552314436,
313
+ "index": 26,
314
+ "prompt_characters": 145
315
+ },
316
+ {
317
+ "prompt_tokens": 86,
318
+ "completion_tokens": 128,
319
+ "ttft_ms": 145.308943989221,
320
+ "decode_seconds": 4.2739029280492105,
321
+ "decode_tokens_per_second": 29.715228010096187,
322
+ "e2e_seconds": 4.4192118720384315,
323
+ "e2e_tokens_per_second": 28.964440652843823,
324
+ "index": 27,
325
+ "prompt_characters": 11
326
+ },
327
+ {
328
+ "prompt_tokens": 95,
329
+ "completion_tokens": 128,
330
+ "ttft_ms": 147.21774403005838,
331
+ "decode_seconds": 4.27222136000637,
332
+ "decode_tokens_per_second": 29.72692407488236,
333
+ "e2e_seconds": 4.419439104036428,
334
+ "e2e_tokens_per_second": 28.962951403288514,
335
+ "index": 28,
336
+ "prompt_characters": 33
337
+ },
338
+ {
339
+ "prompt_tokens": 138,
340
+ "completion_tokens": 128,
341
+ "ttft_ms": 152.53638400463387,
342
+ "decode_seconds": 4.275733312009834,
343
+ "decode_tokens_per_second": 29.702507320388253,
344
+ "e2e_seconds": 4.428269696014468,
345
+ "e2e_tokens_per_second": 28.905195208684464,
346
+ "index": 29,
347
+ "prompt_characters": 155
348
+ },
349
+ {
350
+ "prompt_tokens": 89,
351
+ "completion_tokens": 128,
352
+ "ttft_ms": 147.56200002739206,
353
+ "decode_seconds": 4.27770995197352,
354
+ "decode_tokens_per_second": 29.688782415322148,
355
+ "e2e_seconds": 4.425271952000912,
356
+ "e2e_tokens_per_second": 28.924776011138494,
357
+ "index": 30,
358
+ "prompt_characters": 23
359
+ },
360
+ {
361
+ "prompt_tokens": 122,
362
+ "completion_tokens": 128,
363
+ "ttft_ms": 149.15563201066107,
364
+ "decode_seconds": 4.275628671981394,
365
+ "decode_tokens_per_second": 29.703234247689284,
366
+ "e2e_seconds": 4.424784303992055,
367
+ "e2e_tokens_per_second": 28.927963761876022,
368
+ "index": 31,
369
+ "prompt_characters": 97
370
+ },
371
+ {
372
+ "prompt_tokens": 94,
373
+ "completion_tokens": 128,
374
+ "ttft_ms": 145.9132640156895,
375
+ "decode_seconds": 4.272694383980706,
376
+ "decode_tokens_per_second": 29.723633049008047,
377
+ "e2e_seconds": 4.418607647996396,
378
+ "e2e_tokens_per_second": 28.96840140537058,
379
+ "index": 32,
380
+ "prompt_characters": 61
381
+ },
382
+ {
383
+ "prompt_tokens": 132,
384
+ "completion_tokens": 128,
385
+ "ttft_ms": 153.64075201796368,
386
+ "decode_seconds": 4.272159695974551,
387
+ "decode_tokens_per_second": 29.72735315106922,
388
+ "e2e_seconds": 4.425800447992515,
389
+ "e2e_tokens_per_second": 28.921322030698228,
390
+ "index": 33,
391
+ "prompt_characters": 136
392
+ },
393
+ {
394
+ "prompt_tokens": 114,
395
+ "completion_tokens": 128,
396
+ "ttft_ms": 148.47996801836416,
397
+ "decode_seconds": 4.275548528006766,
398
+ "decode_tokens_per_second": 29.70379102659995,
399
+ "e2e_seconds": 4.42402849602513,
400
+ "e2e_tokens_per_second": 28.93290586057584,
401
+ "index": 34,
402
+ "prompt_characters": 82
403
+ },
404
+ {
405
+ "prompt_tokens": 97,
406
+ "completion_tokens": 128,
407
+ "ttft_ms": 147.8521599783562,
408
+ "decode_seconds": 4.273406176012941,
409
+ "decode_tokens_per_second": 29.71868218679137,
410
+ "e2e_seconds": 4.421258335991297,
411
+ "e2e_tokens_per_second": 28.95103390770332,
412
+ "index": 35,
413
+ "prompt_characters": 49
414
+ },
415
+ {
416
+ "prompt_tokens": 116,
417
+ "completion_tokens": 128,
418
+ "ttft_ms": 147.574128000997,
419
+ "decode_seconds": 4.273268640041351,
420
+ "decode_tokens_per_second": 29.719638688283133,
421
+ "e2e_seconds": 4.420842768042348,
422
+ "e2e_tokens_per_second": 28.953755362053144,
423
+ "index": 36,
424
+ "prompt_characters": 101
425
+ },
426
+ {
427
+ "prompt_tokens": 127,
428
+ "completion_tokens": 128,
429
+ "ttft_ms": 149.55046400427818,
430
+ "decode_seconds": 4.280436095956247,
431
+ "decode_tokens_per_second": 29.669874085955314,
432
+ "e2e_seconds": 4.4299865599605255,
433
+ "e2e_tokens_per_second": 28.893992852461515,
434
+ "index": 37,
435
+ "prompt_characters": 118
436
+ },
437
+ {
438
+ "prompt_tokens": 99,
439
+ "completion_tokens": 128,
440
+ "ttft_ms": 146.9279839657247,
441
+ "decode_seconds": 4.2737699200515635,
442
+ "decode_tokens_per_second": 29.71615280554638,
443
+ "e2e_seconds": 4.420697904017288,
444
+ "e2e_tokens_per_second": 28.95470416191086,
445
+ "index": 38,
446
+ "prompt_characters": 44
447
+ },
448
+ {
449
+ "prompt_tokens": 89,
450
+ "completion_tokens": 128,
451
+ "ttft_ms": 145.21406398853287,
452
+ "decode_seconds": 4.270830192021094,
453
+ "decode_tokens_per_second": 29.736607237924275,
454
+ "e2e_seconds": 4.416044256009627,
455
+ "e2e_tokens_per_second": 28.985216764032575,
456
+ "index": 39,
457
+ "prompt_characters": 21
458
+ },
459
+ {
460
+ "prompt_tokens": 131,
461
+ "completion_tokens": 128,
462
+ "ttft_ms": 153.4373280010186,
463
+ "decode_seconds": 4.277822831994854,
464
+ "decode_tokens_per_second": 29.68799900971513,
465
+ "e2e_seconds": 4.4312601599958725,
466
+ "e2e_tokens_per_second": 28.885688354646103,
467
+ "index": 40,
468
+ "prompt_characters": 99
469
+ },
470
+ {
471
+ "prompt_tokens": 164,
472
+ "completion_tokens": 128,
473
+ "ttft_ms": 152.07355201710016,
474
+ "decode_seconds": 4.281609760015272,
475
+ "decode_tokens_per_second": 29.66174105496878,
476
+ "e2e_seconds": 4.433683312032372,
477
+ "e2e_tokens_per_second": 28.86990138709876,
478
+ "index": 41,
479
+ "prompt_characters": 255
480
+ },
481
+ {
482
+ "prompt_tokens": 119,
483
+ "completion_tokens": 128,
484
+ "ttft_ms": 147.45664002839476,
485
+ "decode_seconds": 4.278829519986175,
486
+ "decode_tokens_per_second": 29.6810142602761,
487
+ "e2e_seconds": 4.42628616001457,
488
+ "e2e_tokens_per_second": 28.918148391828936,
489
+ "index": 42,
490
+ "prompt_characters": 98
491
+ },
492
+ {
493
+ "prompt_tokens": 109,
494
+ "completion_tokens": 128,
495
+ "ttft_ms": 148.86659203330055,
496
+ "decode_seconds": 4.27835519996006,
497
+ "decode_tokens_per_second": 29.684304847149104,
498
+ "e2e_seconds": 4.427221791993361,
499
+ "e2e_tokens_per_second": 28.912036941878142,
500
+ "index": 43,
501
+ "prompt_characters": 69
502
+ },
503
+ {
504
+ "prompt_tokens": 116,
505
+ "completion_tokens": 128,
506
+ "ttft_ms": 149.5863359887153,
507
+ "decode_seconds": 4.2813182240352035,
508
+ "decode_tokens_per_second": 29.66376086856274,
509
+ "e2e_seconds": 4.430904560023919,
510
+ "e2e_tokens_per_second": 28.88800656074367,
511
+ "index": 44,
512
+ "prompt_characters": 63
513
+ },
514
+ {
515
+ "prompt_tokens": 99,
516
+ "completion_tokens": 128,
517
+ "ttft_ms": 148.91756797442213,
518
+ "decode_seconds": 4.275036848033778,
519
+ "decode_tokens_per_second": 29.70734627899435,
520
+ "e2e_seconds": 4.4239544160082005,
521
+ "e2e_tokens_per_second": 28.93339034797205,
522
+ "index": 45,
523
+ "prompt_characters": 51
524
+ },
525
+ {
526
+ "prompt_tokens": 87,
527
+ "completion_tokens": 128,
528
+ "ttft_ms": 181.61635199794546,
529
+ "decode_seconds": 4.241584288014565,
530
+ "decode_tokens_per_second": 29.94164240914972,
531
+ "e2e_seconds": 4.42320064001251,
532
+ "e2e_tokens_per_second": 28.938321007214807,
533
+ "index": 46,
534
+ "prompt_characters": 16
535
+ },
536
+ {
537
+ "prompt_tokens": 121,
538
+ "completion_tokens": 128,
539
+ "ttft_ms": 149.99320002971217,
540
+ "decode_seconds": 4.277360575972125,
541
+ "decode_tokens_per_second": 29.69120740332639,
542
+ "e2e_seconds": 4.427353776001837,
543
+ "e2e_tokens_per_second": 28.911175044066976,
544
+ "index": 47,
545
+ "prompt_characters": 83
546
+ },
547
+ {
548
+ "prompt_tokens": 103,
549
+ "completion_tokens": 128,
550
+ "ttft_ms": 150.74512001592666,
551
+ "decode_seconds": 4.27451617596671,
552
+ "decode_tokens_per_second": 29.710964883944584,
553
+ "e2e_seconds": 4.4252612959826365,
554
+ "e2e_tokens_per_second": 28.92484566192772,
555
+ "index": 48,
556
+ "prompt_characters": 52
557
+ },
558
+ {
559
+ "prompt_tokens": 128,
560
+ "completion_tokens": 128,
561
+ "ttft_ms": 148.76968000316992,
562
+ "decode_seconds": 4.281864800024778,
563
+ "decode_tokens_per_second": 29.659974317560213,
564
+ "e2e_seconds": 4.430634480027948,
565
+ "e2e_tokens_per_second": 28.88976749876072,
566
+ "index": 49,
567
+ "prompt_characters": 82
568
+ },
569
+ {
570
+ "prompt_tokens": 116,
571
+ "completion_tokens": 128,
572
+ "ttft_ms": 148.97512004245073,
573
+ "decode_seconds": 4.279575983993709,
574
+ "decode_tokens_per_second": 29.675837156531415,
575
+ "e2e_seconds": 4.42855110403616,
576
+ "e2e_tokens_per_second": 28.90335845584833,
577
+ "index": 50,
578
+ "prompt_characters": 89
579
+ },
580
+ {
581
+ "prompt_tokens": 202,
582
+ "completion_tokens": 128,
583
+ "ttft_ms": 159.03887996682897,
584
+ "decode_seconds": 4.285158192040399,
585
+ "decode_tokens_per_second": 29.637178911131013,
586
+ "e2e_seconds": 4.444197072007228,
587
+ "e2e_tokens_per_second": 28.801603062617705,
588
+ "index": 51,
589
+ "prompt_characters": 202
590
+ },
591
+ {
592
+ "prompt_tokens": 133,
593
+ "completion_tokens": 128,
594
+ "ttft_ms": 150.73892794316635,
595
+ "decode_seconds": 4.277238800015766,
596
+ "decode_tokens_per_second": 29.692052732602132,
597
+ "e2e_seconds": 4.4279777279589325,
598
+ "e2e_tokens_per_second": 28.9071011337271,
599
+ "index": 52,
600
+ "prompt_characters": 73
601
+ },
602
+ {
603
+ "prompt_tokens": 97,
604
+ "completion_tokens": 128,
605
+ "ttft_ms": 148.48937600618228,
606
+ "decode_seconds": 4.274834448006004,
607
+ "decode_tokens_per_second": 29.708752828835074,
608
+ "e2e_seconds": 4.423323824012186,
609
+ "e2e_tokens_per_second": 28.937515111407173,
610
+ "index": 53,
611
+ "prompt_characters": 42
612
+ },
613
+ {
614
+ "prompt_tokens": 208,
615
+ "completion_tokens": 128,
616
+ "ttft_ms": 160.55590397445485,
617
+ "decode_seconds": 4.285696112026926,
618
+ "decode_tokens_per_second": 29.633458994817804,
619
+ "e2e_seconds": 4.446252016001381,
620
+ "e2e_tokens_per_second": 28.78829169811958,
621
+ "index": 54,
622
+ "prompt_characters": 254
623
+ },
624
+ {
625
+ "prompt_tokens": 113,
626
+ "completion_tokens": 128,
627
+ "ttft_ms": 148.43628800008446,
628
+ "decode_seconds": 4.276303727994673,
629
+ "decode_tokens_per_second": 29.698545304113676,
630
+ "e2e_seconds": 4.424740015994757,
631
+ "e2e_tokens_per_second": 28.92825330692868,
632
+ "index": 55,
633
+ "prompt_characters": 85
634
+ },
635
+ {
636
+ "prompt_tokens": 127,
637
+ "completion_tokens": 128,
638
+ "ttft_ms": 182.12604796281084,
639
+ "decode_seconds": 4.240071760024875,
640
+ "decode_tokens_per_second": 29.952323259560806,
641
+ "e2e_seconds": 4.422197807987686,
642
+ "e2e_tokens_per_second": 28.94488341719978,
643
+ "index": 56,
644
+ "prompt_characters": 107
645
+ },
646
+ {
647
+ "prompt_tokens": 144,
648
+ "completion_tokens": 128,
649
+ "ttft_ms": 185.98363199271262,
650
+ "decode_seconds": 4.249349439982325,
651
+ "decode_tokens_per_second": 29.88692782124509,
652
+ "e2e_seconds": 4.4353330719750375,
653
+ "e2e_tokens_per_second": 28.859162981192316,
654
+ "index": 57,
655
+ "prompt_characters": 121
656
+ },
657
+ {
658
+ "prompt_tokens": 230,
659
+ "completion_tokens": 128,
660
+ "ttft_ms": 162.00067201862112,
661
+ "decode_seconds": 4.287982367968652,
662
+ "decode_tokens_per_second": 29.617659099695356,
663
+ "e2e_seconds": 4.4499830399872735,
664
+ "e2e_tokens_per_second": 28.764154570882603,
665
+ "index": 58,
666
+ "prompt_characters": 346
667
+ },
668
+ {
669
+ "prompt_tokens": 124,
670
+ "completion_tokens": 128,
671
+ "ttft_ms": 183.87990404153243,
672
+ "decode_seconds": 4.242221055959817,
673
+ "decode_tokens_per_second": 29.937148094057964,
674
+ "e2e_seconds": 4.4261009600013494,
675
+ "e2e_tokens_per_second": 28.91935840522738,
676
+ "index": 59,
677
+ "prompt_characters": 93
678
+ },
679
+ {
680
+ "prompt_tokens": 170,
681
+ "completion_tokens": 128,
682
+ "ttft_ms": 152.916015998926,
683
+ "decode_seconds": 4.2795338880387135,
684
+ "decode_tokens_per_second": 29.67612906512195,
685
+ "e2e_seconds": 4.4324499040376395,
686
+ "e2e_tokens_per_second": 28.8779349504664,
687
+ "index": 60,
688
+ "prompt_characters": 207
689
+ },
690
+ {
691
+ "prompt_tokens": 148,
692
+ "completion_tokens": 128,
693
+ "ttft_ms": 153.09798397356644,
694
+ "decode_seconds": 4.275488400016911,
695
+ "decode_tokens_per_second": 29.70420876349417,
696
+ "e2e_seconds": 4.428586383990478,
697
+ "e2e_tokens_per_second": 28.90312819971747,
698
+ "index": 61,
699
+ "prompt_characters": 122
700
+ },
701
+ {
702
+ "prompt_tokens": 123,
703
+ "completion_tokens": 128,
704
+ "ttft_ms": 149.6343039907515,
705
+ "decode_seconds": 4.274956831999589,
706
+ "decode_tokens_per_second": 29.707902322980047,
707
+ "e2e_seconds": 4.42459113599034,
708
+ "e2e_tokens_per_second": 28.929226693699963,
709
+ "index": 62,
710
+ "prompt_characters": 88
711
+ },
712
+ {
713
+ "prompt_tokens": 98,
714
+ "completion_tokens": 128,
715
+ "ttft_ms": 150.04775999113917,
716
+ "decode_seconds": 4.279320368019398,
717
+ "decode_tokens_per_second": 29.677609778671357,
718
+ "e2e_seconds": 4.4293681280105375,
719
+ "e2e_tokens_per_second": 28.898027055044427,
720
+ "index": 63,
721
+ "prompt_characters": 43
722
+ },
723
+ {
724
+ "prompt_tokens": 196,
725
+ "completion_tokens": 128,
726
+ "ttft_ms": 160.95686401240528,
727
+ "decode_seconds": 4.288259503955487,
728
+ "decode_tokens_per_second": 29.615745008634693,
729
+ "e2e_seconds": 4.449216367967892,
730
+ "e2e_tokens_per_second": 28.769111100447994,
731
+ "index": 64,
732
+ "prompt_characters": 177
733
+ },
734
+ {
735
+ "prompt_tokens": 280,
736
+ "completion_tokens": 128,
737
+ "ttft_ms": 194.86894400324672,
738
+ "decode_seconds": 4.253986447991338,
739
+ "decode_tokens_per_second": 29.85434992628322,
740
+ "e2e_seconds": 4.448855391994584,
741
+ "e2e_tokens_per_second": 28.771445399265478,
742
+ "index": 65,
743
+ "prompt_characters": 494
744
+ },
745
+ {
746
+ "prompt_tokens": 100,
747
+ "completion_tokens": 128,
748
+ "ttft_ms": 146.7246399843134,
749
+ "decode_seconds": 4.280999056005385,
750
+ "decode_tokens_per_second": 29.665972437402065,
751
+ "e2e_seconds": 4.427723695989698,
752
+ "e2e_tokens_per_second": 28.908759622000094,
753
+ "index": 66,
754
+ "prompt_characters": 56
755
+ },
756
+ {
757
+ "prompt_tokens": 111,
758
+ "completion_tokens": 128,
759
+ "ttft_ms": 149.47828801814467,
760
+ "decode_seconds": 4.27679667196935,
761
+ "decode_tokens_per_second": 29.695122247071872,
762
+ "e2e_seconds": 4.426274959987495,
763
+ "e2e_tokens_per_second": 28.918221564880284,
764
+ "index": 67,
765
+ "prompt_characters": 68
766
+ },
767
+ {
768
+ "prompt_tokens": 222,
769
+ "completion_tokens": 128,
770
+ "ttft_ms": 161.32521600229666,
771
+ "decode_seconds": 4.284537935978733,
772
+ "decode_tokens_per_second": 29.64146937141984,
773
+ "e2e_seconds": 4.44586315198103,
774
+ "e2e_tokens_per_second": 28.790809708788395,
775
+ "index": 68,
776
+ "prompt_characters": 333
777
+ },
778
+ {
779
+ "prompt_tokens": 86,
780
+ "completion_tokens": 128,
781
+ "ttft_ms": 149.98059201752767,
782
+ "decode_seconds": 4.274428736011032,
783
+ "decode_tokens_per_second": 29.71157266702229,
784
+ "e2e_seconds": 4.42440932802856,
785
+ "e2e_tokens_per_second": 28.93041545435729,
786
+ "index": 69,
787
+ "prompt_characters": 20
788
+ },
789
+ {
790
+ "prompt_tokens": 116,
791
+ "completion_tokens": 128,
792
+ "ttft_ms": 148.25239998754114,
793
+ "decode_seconds": 4.280142048024572,
794
+ "decode_tokens_per_second": 29.671912421368052,
795
+ "e2e_seconds": 4.428394448012114,
796
+ "e2e_tokens_per_second": 28.904380922403746,
797
+ "index": 70,
798
+ "prompt_characters": 70
799
+ },
800
+ {
801
+ "prompt_tokens": 122,
802
+ "completion_tokens": 128,
803
+ "ttft_ms": 151.68241603532806,
804
+ "decode_seconds": 4.279039520013612,
805
+ "decode_tokens_per_second": 29.679557621752465,
806
+ "e2e_seconds": 4.43072193604894,
807
+ "e2e_tokens_per_second": 28.88919725667618,
808
+ "index": 71,
809
+ "prompt_characters": 100
810
+ },
811
+ {
812
+ "prompt_tokens": 123,
813
+ "completion_tokens": 128,
814
+ "ttft_ms": 182.6112479902804,
815
+ "decode_seconds": 4.241006672033109,
816
+ "decode_tokens_per_second": 29.94572039640699,
817
+ "e2e_seconds": 4.423617920023389,
818
+ "e2e_tokens_per_second": 28.93559125452752,
819
+ "index": 72,
820
+ "prompt_characters": 100
821
+ },
822
+ {
823
+ "prompt_tokens": 304,
824
+ "completion_tokens": 128,
825
+ "ttft_ms": 160.21692799404263,
826
+ "decode_seconds": 4.284923232044093,
827
+ "decode_tokens_per_second": 29.638804039767013,
828
+ "e2e_seconds": 4.445140160038136,
829
+ "e2e_tokens_per_second": 28.795492468543863,
830
+ "index": 73,
831
+ "prompt_characters": 516
832
+ },
833
+ {
834
+ "prompt_tokens": 96,
835
+ "completion_tokens": 128,
836
+ "ttft_ms": 145.91110398760065,
837
+ "decode_seconds": 4.27431379200425,
838
+ "decode_tokens_per_second": 29.712371664797445,
839
+ "e2e_seconds": 4.420224895991851,
840
+ "e2e_tokens_per_second": 28.957802603226636,
841
+ "index": 74,
842
+ "prompt_characters": 51
843
+ },
844
+ {
845
+ "prompt_tokens": 98,
846
+ "completion_tokens": 128,
847
+ "ttft_ms": 182.3572000139393,
848
+ "decode_seconds": 4.2468398079508916,
849
+ "decode_tokens_per_second": 29.904589234148144,
850
+ "e2e_seconds": 4.429197007964831,
851
+ "e2e_tokens_per_second": 28.899143517396766,
852
+ "index": 75,
853
+ "prompt_characters": 38
854
+ },
855
+ {
856
+ "prompt_tokens": 146,
857
+ "completion_tokens": 128,
858
+ "ttft_ms": 152.6936799637042,
859
+ "decode_seconds": 4.277269551996142,
860
+ "decode_tokens_per_second": 29.691839257764542,
861
+ "e2e_seconds": 4.429963231959846,
862
+ "e2e_tokens_per_second": 28.894145007016668,
863
+ "index": 76,
864
+ "prompt_characters": 133
865
+ },
866
+ {
867
+ "prompt_tokens": 140,
868
+ "completion_tokens": 128,
869
+ "ttft_ms": 152.17240003403276,
870
+ "decode_seconds": 4.280287823989056,
871
+ "decode_tokens_per_second": 29.67090186978153,
872
+ "e2e_seconds": 4.432460224023089,
873
+ "e2e_tokens_per_second": 28.87786771469813,
874
+ "index": 77,
875
+ "prompt_characters": 158
876
+ },
877
+ {
878
+ "prompt_tokens": 89,
879
+ "completion_tokens": 128,
880
+ "ttft_ms": 147.72603200981393,
881
+ "decode_seconds": 4.280198383959942,
882
+ "decode_tokens_per_second": 29.671521879904667,
883
+ "e2e_seconds": 4.4279244159697555,
884
+ "e2e_tokens_per_second": 28.90744917378334,
885
+ "index": 78,
886
+ "prompt_characters": 17
887
+ },
888
+ {
889
+ "prompt_tokens": 163,
890
+ "completion_tokens": 128,
891
+ "ttft_ms": 186.45403202390298,
892
+ "decode_seconds": 4.244761504000053,
893
+ "decode_tokens_per_second": 29.919231005162832,
894
+ "e2e_seconds": 4.431215536023956,
895
+ "e2e_tokens_per_second": 28.88597924416286,
896
+ "index": 79,
897
+ "prompt_characters": 193
898
+ },
899
+ {
900
+ "prompt_tokens": 154,
901
+ "completion_tokens": 128,
902
+ "ttft_ms": 152.16956799849868,
903
+ "decode_seconds": 4.277686815999914,
904
+ "decode_tokens_per_second": 29.68894298782685,
905
+ "e2e_seconds": 4.429856383998413,
906
+ "e2e_tokens_per_second": 28.894841932655726,
907
+ "index": 80,
908
+ "prompt_characters": 151
909
+ },
910
+ {
911
+ "prompt_tokens": 119,
912
+ "completion_tokens": 128,
913
+ "ttft_ms": 147.88078400306404,
914
+ "decode_seconds": 4.278590032015927,
915
+ "decode_tokens_per_second": 29.682675612685866,
916
+ "e2e_seconds": 4.426470816018991,
917
+ "e2e_tokens_per_second": 28.916942033545045,
918
+ "index": 81,
919
+ "prompt_characters": 75
920
+ },
921
+ {
922
+ "prompt_tokens": 274,
923
+ "completion_tokens": 128,
924
+ "ttft_ms": 158.65448000840843,
925
+ "decode_seconds": 4.288939487945754,
926
+ "decode_tokens_per_second": 29.611049621226616,
927
+ "e2e_seconds": 4.4475939679541625,
928
+ "e2e_tokens_per_second": 28.779605540044024,
929
+ "index": 82,
930
+ "prompt_characters": 413
931
+ },
932
+ {
933
+ "prompt_tokens": 96,
934
+ "completion_tokens": 128,
935
+ "ttft_ms": 151.030623994302,
936
+ "decode_seconds": 4.277100688021164,
937
+ "decode_tokens_per_second": 29.693011519623024,
938
+ "e2e_seconds": 4.428131312015466,
939
+ "e2e_tokens_per_second": 28.906098527990746,
940
+ "index": 83,
941
+ "prompt_characters": 53
942
+ },
943
+ {
944
+ "prompt_tokens": 187,
945
+ "completion_tokens": 128,
946
+ "ttft_ms": 153.6058560013771,
947
+ "decode_seconds": 4.2891785760293715,
948
+ "decode_tokens_per_second": 29.609399037325215,
949
+ "e2e_seconds": 4.442784432030749,
950
+ "e2e_tokens_per_second": 28.810760899666832,
951
+ "index": 84,
952
+ "prompt_characters": 243
953
+ },
954
+ {
955
+ "prompt_tokens": 175,
956
+ "completion_tokens": 128,
957
+ "ttft_ms": 150.53255995735526,
958
+ "decode_seconds": 4.287062800023705,
959
+ "decode_tokens_per_second": 29.62401203903469,
960
+ "e2e_seconds": 4.43759535998106,
961
+ "e2e_tokens_per_second": 28.844450567603424,
962
+ "index": 85,
963
+ "prompt_characters": 193
964
+ },
965
+ {
966
+ "prompt_tokens": 95,
967
+ "completion_tokens": 128,
968
+ "ttft_ms": 146.89548802562058,
969
+ "decode_seconds": 4.2793887840234675,
970
+ "decode_tokens_per_second": 29.677135312906767,
971
+ "e2e_seconds": 4.426284272049088,
972
+ "e2e_tokens_per_second": 28.91816072643345,
973
+ "index": 86,
974
+ "prompt_characters": 54
975
+ },
976
+ {
977
+ "prompt_tokens": 135,
978
+ "completion_tokens": 128,
979
+ "ttft_ms": 152.44775998871773,
980
+ "decode_seconds": 4.279913456004579,
981
+ "decode_tokens_per_second": 29.673497210982887,
982
+ "e2e_seconds": 4.432361215993296,
983
+ "e2e_tokens_per_second": 28.878512775118008,
984
+ "index": 87,
985
+ "prompt_characters": 153
986
+ },
987
+ {
988
+ "prompt_tokens": 132,
989
+ "completion_tokens": 128,
990
+ "ttft_ms": 152.3534239968285,
991
+ "decode_seconds": 4.280044800019823,
992
+ "decode_tokens_per_second": 29.67258660456353,
993
+ "e2e_seconds": 4.4323982240166515,
994
+ "e2e_tokens_per_second": 28.878271655836475,
995
+ "index": 88,
996
+ "prompt_characters": 83
997
+ },
998
+ {
999
+ "prompt_tokens": 151,
1000
+ "completion_tokens": 128,
1001
+ "ttft_ms": 152.58678403915837,
1002
+ "decode_seconds": 4.286318383994512,
1003
+ "decode_tokens_per_second": 29.629156918027626,
1004
+ "e2e_seconds": 4.438905168033671,
1005
+ "e2e_tokens_per_second": 28.8359393036326,
1006
+ "index": 89,
1007
+ "prompt_characters": 154
1008
+ },
1009
+ {
1010
+ "prompt_tokens": 98,
1011
+ "completion_tokens": 128,
1012
+ "ttft_ms": 149.5123039931059,
1013
+ "decode_seconds": 4.2784633280243725,
1014
+ "decode_tokens_per_second": 29.68355464639302,
1015
+ "e2e_seconds": 4.427975632017478,
1016
+ "e2e_tokens_per_second": 28.90711481663699,
1017
+ "index": 90,
1018
+ "prompt_characters": 47
1019
+ },
1020
+ {
1021
+ "prompt_tokens": 120,
1022
+ "completion_tokens": 128,
1023
+ "ttft_ms": 149.92755197454244,
1024
+ "decode_seconds": 4.279246336023789,
1025
+ "decode_tokens_per_second": 29.67812320849154,
1026
+ "e2e_seconds": 4.429173887998331,
1027
+ "e2e_tokens_per_second": 28.899294368830212,
1028
+ "index": 91,
1029
+ "prompt_characters": 91
1030
+ },
1031
+ {
1032
+ "prompt_tokens": 129,
1033
+ "completion_tokens": 128,
1034
+ "ttft_ms": 149.91504000499845,
1035
+ "decode_seconds": 4.284216400003061,
1036
+ "decode_tokens_per_second": 29.643694001990486,
1037
+ "e2e_seconds": 4.434131440008059,
1038
+ "e2e_tokens_per_second": 28.86698369946547,
1039
+ "index": 92,
1040
+ "prompt_characters": 149
1041
+ },
1042
+ {
1043
+ "prompt_tokens": 126,
1044
+ "completion_tokens": 128,
1045
+ "ttft_ms": 148.77603197237477,
1046
+ "decode_seconds": 4.277493072033394,
1047
+ "decode_tokens_per_second": 29.690287713225434,
1048
+ "e2e_seconds": 4.426269104005769,
1049
+ "e2e_tokens_per_second": 28.9182598238684,
1050
+ "index": 93,
1051
+ "prompt_characters": 129
1052
+ },
1053
+ {
1054
+ "prompt_tokens": 176,
1055
+ "completion_tokens": 128,
1056
+ "ttft_ms": 150.4963680054061,
1057
+ "decode_seconds": 4.283495743991807,
1058
+ "decode_tokens_per_second": 29.648681261825693,
1059
+ "e2e_seconds": 4.433992111997213,
1060
+ "e2e_tokens_per_second": 28.86789077807914,
1061
+ "index": 94,
1062
+ "prompt_characters": 181
1063
+ },
1064
+ {
1065
+ "prompt_tokens": 151,
1066
+ "completion_tokens": 128,
1067
+ "ttft_ms": 151.79175999946892,
1068
+ "decode_seconds": 4.284410799969919,
1069
+ "decode_tokens_per_second": 29.642348955168274,
1070
+ "e2e_seconds": 4.436202559969388,
1071
+ "e2e_tokens_per_second": 28.853506635387557,
1072
+ "index": 95,
1073
+ "prompt_characters": 182
1074
+ },
1075
+ {
1076
+ "prompt_tokens": 113,
1077
+ "completion_tokens": 128,
1078
+ "ttft_ms": 147.12048001820222,
1079
+ "decode_seconds": 4.276005279971287,
1080
+ "decode_tokens_per_second": 29.700618143496026,
1081
+ "e2e_seconds": 4.423125759989489,
1082
+ "e2e_tokens_per_second": 28.93881091011624,
1083
+ "index": 96,
1084
+ "prompt_characters": 72
1085
+ },
1086
+ {
1087
+ "prompt_tokens": 159,
1088
+ "completion_tokens": 128,
1089
+ "ttft_ms": 152.09144004620612,
1090
+ "decode_seconds": 4.281069583958015,
1091
+ "decode_tokens_per_second": 29.665483708999556,
1092
+ "e2e_seconds": 4.433161024004221,
1093
+ "e2e_tokens_per_second": 28.873302663025065,
1094
+ "index": 97,
1095
+ "prompt_characters": 144
1096
+ },
1097
+ {
1098
+ "prompt_tokens": 135,
1099
+ "completion_tokens": 128,
1100
+ "ttft_ms": 156.03772801114246,
1101
+ "decode_seconds": 4.28199811198283,
1102
+ "decode_tokens_per_second": 29.659050910041422,
1103
+ "e2e_seconds": 4.438035839993972,
1104
+ "e2e_tokens_per_second": 28.84158772367504,
1105
+ "index": 98,
1106
+ "prompt_characters": 146
1107
+ },
1108
+ {
1109
+ "prompt_tokens": 183,
1110
+ "completion_tokens": 128,
1111
+ "ttft_ms": 154.6247520018369,
1112
+ "decode_seconds": 4.2839804160175845,
1113
+ "decode_tokens_per_second": 29.64532693127015,
1114
+ "e2e_seconds": 4.438605168019421,
1115
+ "e2e_tokens_per_second": 28.837888290279196,
1116
+ "index": 99,
1117
+ "prompt_characters": 214
1118
+ }
1119
+ ]
1120
+ }
benchmark/bf16-elyza100-streaming.csv ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ prompt_tokens,completion_tokens,ttft_ms,decode_seconds,decode_tokens_per_second,e2e_seconds,e2e_tokens_per_second,index,prompt_characters
2
+ 89,128,287.3321439838037,8.779072656005155,14.466220405765533,9.066404799988959,14.118054821482918,0,28
3
+ 100,128,295.72676797397435,8.767456800036598,14.485386457731947,9.063183568010572,14.123072653166709,1,50
4
+ 86,128,290.60323198791593,8.77216342400061,14.477614456261545,9.062766655988526,14.123722353087368,2,24
5
+ 223,128,304.505567997694,8.789692880003713,14.448741467283934,9.094198448001407,14.074907286428306,3,323
6
+ 87,128,289.189808012452,8.793557535973378,14.442391430369154,9.08274734398583,14.092652272746044,4,15
7
+ 138,128,298.6349919810891,8.804777184035629,14.42398794943612,9.103412176016718,14.06066181834772,5,107
8
+ 197,128,304.21625601593405,8.81909500801703,14.400570566996976,9.123311264032964,14.029993748499768,6,286
9
+ 129,128,299.53299200860783,8.820546608010773,14.398200660791169,9.12007960001938,14.034965221107061,7,77
10
+ 253,128,309.5190239837393,8.817294720036443,14.403510830981398,9.126813744020183,14.024609638151606,8,359
11
+ 107,128,299.14905602345243,8.819510432018433,14.399892259205004,9.118659488041885,14.037150983415694,9,76
12
+ 115,128,296.4645279571414,8.813084144028835,14.410392312666938,9.109548671985976,14.051190087345422,10,93
13
+ 136,128,370.15520001295954,8.748351391986944,14.517020900226493,9.118506591999903,14.037386353627298,11,126
14
+ 239,128,309.37422398710623,8.829718319990207,14.383244787376283,9.139092543977313,14.005766916579956,12,301
15
+ 113,128,297.8560159681365,8.81552940804977,14.406395137656943,9.113385424017906,14.045274510464784,13,76
16
+ 239,128,309.5741919823922,8.828813120024279,14.384719471743757,9.138387312006671,14.00684777628372,14,323
17
+ 122,128,370.8477119798772,8.738631952030119,14.533167284897031,9.109479664009996,14.051296530767418,15,120
18
+ 152,128,300.2828800235875,8.817734479962382,14.402792496031452,9.11801735998597,14.038139536970235,16,182
19
+ 141,128,298.9331199787557,8.812246479967143,14.411762118627612,9.111179599945899,14.04867488297125,17,140
20
+ 196,128,304.14611200103536,8.825933744024951,14.389412348124349,9.130079856025986,14.019592601429233,18,191
21
+ 118,128,296.74635198898613,8.812764559988864,14.410914887775066,9.10951091197785,14.05124833120253,19,75
22
+ 269,128,309.93571202270687,8.82683105597971,14.38794955908488,9.136766768002417,14.009332102934351,20,368
23
+ 151,128,371.7997280182317,8.751005103986245,14.512618663900579,9.122804832004476,14.030772592103743,21,137
24
+ 83,128,368.71777597116306,8.740428032004274,14.53018084869209,9.109145807975437,14.051811519794827,22,8
25
+ 257,128,310.9340959927067,8.822712880035397,14.394665419452082,9.133646976028103,14.014117289177584,23,379
26
+ 264,128,312.00308800907806,8.822620527993422,14.394816097670736,9.1346236160025,14.012618951892343,24,480
27
+ 115,128,368.4656799887307,8.752178527996875,14.5106729248891,9.120644207985606,14.034096395069247,25,82
28
+ 167,128,300.1571359927766,8.814201600034721,14.408565376982041,9.114358736027498,14.043774631564363,26,145
29
+ 86,128,288.03827200317755,8.810084576020017,14.41529861650575,9.098122848023195,14.068836191611915,27,11
30
+ 95,128,290.33766401698813,8.809145903971512,14.416834660752226,9.0994835679885,14.066732363834053,28,33
31
+ 138,128,300.13843200867996,8.82354910398135,14.393301210586026,9.12368753599003,14.029415134514519,29,155
32
+ 89,128,291.2958399974741,8.811235280008987,14.413416049408962,9.10253112000646,14.062022783824236,30,23
33
+ 122,128,298.4383839648217,8.810073536005802,14.415316680498178,9.108511919970624,14.05278942648766,31,97
34
+ 94,128,287.24841598886997,8.81881228799466,14.401032231164425,9.10606070398353,14.05657222820898,32,61
35
+ 132,128,298.8155360217206,8.817851200001314,14.402601849301004,9.116666736023035,14.040219271614777,33,136
36
+ 114,128,297.1212000120431,8.809575375984423,14.416131831530919,9.106696575996466,14.055590732800283,34,82
37
+ 97,128,297.4157760036178,8.811500575975515,14.412982091411816,9.108916351979133,14.05216548861917,35,49
38
+ 116,128,295.1624320121482,8.818597696023062,14.401382666234271,9.11376012803521,14.044697051687148,36,101
39
+ 127,128,294.031776022166,8.825343439995777,14.390374818100085,9.119375216017943,14.036049287145392,37,118
40
+ 99,128,298.44286397565156,8.810942688025534,14.413894687182415,9.109385552001186,14.051441699257142,38,44
41
+ 89,128,290.05577601492405,8.821949823992327,14.395910488473717,9.112005600007251,14.047401375598161,39,21
42
+ 131,128,297.30228800326586,8.816554000019096,14.404720937423502,9.113856288022362,14.044548866567112,40,99
43
+ 164,128,299.2388000129722,8.815770207962487,14.406001631631959,9.115009007975459,14.04277273758067,41,255
44
+ 119,128,297.0329279778525,8.812944128003437,14.410621258388904,9.10997705598129,14.050529349682579,42,98
45
+ 109,128,299.9312960309908,8.82349996798439,14.393381363496673,9.12343126401538,14.029809212774731,43,69
46
+ 116,128,296.43854399910197,8.814559072023258,14.407981041625595,9.11099761602236,14.048955492525053,44,63
47
+ 99,128,296.96737602353096,8.8147550720023,14.407660673792442,9.11172244802583,14.047837906622455,45,51
48
+ 87,128,362.1777279768139,8.752297568018548,14.510475565189427,9.114475295995362,14.043595033522063,46,16
49
+ 121,128,299.46691199438646,8.81550879997667,14.406428815581934,9.114975711971056,14.04282403428597,47,83
50
+ 103,128,297.9078239877708,8.816075871989597,14.405502158109133,9.113983695977367,14.044352532306512,48,52
51
+ 128,128,296.2054240051657,8.818844111985527,14.400980263093285,9.115049535990693,14.042710299553846,49,82
52
+ 116,128,296.5520319994539,8.82329428801313,14.393716887867562,9.119846320012584,14.03532422680379,50,89
53
+ 202,128,303.8067359593697,8.820766672026366,14.397841448721225,9.124573407985736,14.02805306908657,51,202
54
+ 133,128,296.68961599236354,8.814125408011023,14.408689929073581,9.110815024003386,14.049237050996068,52,73
55
+ 97,128,296.0946400417015,8.822594399971422,14.39485872776962,9.118689040013123,14.037105491626216,53,42
56
+ 208,128,304.0414240094833,8.82271115196636,14.394668238877447,9.126752575975843,14.02470363192837,54,254
57
+ 113,128,296.58384004142135,8.816141712013632,14.405394576058015,9.112725552055053,14.04629155885575,55,85
58
+ 127,128,370.9898079978302,8.744085903977975,14.524102507069776,9.115075711975805,14.04266997276037,56,107
59
+ 144,128,370.400496001821,8.756533648003824,14.503455945601425,9.126934144005645,14.024424629388543,57,121
60
+ 230,128,309.8742720321752,8.82315703999484,14.393940788350093,9.133031312027015,14.015061990583634,58,346
61
+ 124,128,369.675631984137,8.867081984004471,14.322637394026373,9.236757615988608,13.85767661353753,59,93
62
+ 170,128,302.19761602347717,8.913596848025918,14.24789590165574,9.215794464049395,13.889198646880102,60,207
63
+ 148,128,298.5740000149235,8.821890159975737,14.396007850583949,9.12046415999066,14.034373443569462,61,122
64
+ 123,128,297.4766080151312,8.81660511996597,14.404637416775923,9.114081727981102,14.04420146980115,62,88
65
+ 98,128,297.258767997846,8.813225760008208,14.410160758196861,9.110484528006054,14.049746707381153,63,43
66
+ 196,128,306.2620800337754,8.828527903999202,14.385184187102217,9.134789984032977,14.012363746045144,64,177
67
+ 280,128,381.4747359720059,8.757706432021223,14.501513722319327,9.139181167993229,14.005631100549252,65,494
68
+ 100,128,296.71494400827214,8.81345068803057,14.40979299657023,9.110165632038843,14.050238510466441,66,56
69
+ 111,128,300.6466400111094,8.820130895997863,14.398879279402337,9.120777536008973,14.03389124388288,67,68
70
+ 222,128,308.8962400215678,8.82139251200715,14.396819983595018,9.130288752028719,14.019271840834042,68,333
71
+ 86,128,288.00887998659164,8.815513295994606,14.406421468129757,9.103522175981198,14.060491920118148,69,20
72
+ 116,128,299.3990720133297,8.814984991971869,14.407284880877684,9.114384063985199,14.043735605325471,70,70
73
+ 122,128,297.36579197924584,8.820748256053776,14.397871508558078,9.118114048033021,14.03799067720725,71,100
74
+ 123,128,366.77401600172743,8.74748774402542,14.518454179800555,9.114261760027148,14.043924057719705,72,100
75
+ 304,128,314.2944479477592,8.825065472046845,14.390828079663438,9.139359919994604,14.005357171673307,73,516
76
+ 96,128,290.49059201497585,8.813162976002786,14.410263414599976,9.103653568017762,14.060288986575621,74,51
77
+ 98,128,368.19483200088143,8.751111984020099,14.51244141680593,9.11930681602098,14.036154565512264,75,38
78
+ 146,128,299.2566720349714,8.823440047970507,14.393479109002556,9.122696720005479,14.030938869128944,76,133
79
+ 140,128,299.837856029626,8.817254160006996,14.403577088209877,9.117092016036622,14.039564345171993,77,158
80
+ 89,128,289.219903992489,8.821538256015629,14.396582128224123,9.110758160008118,14.049324738072727,78,17
81
+ 163,128,367.9596959846094,8.744869392015971,14.522801234281493,9.11282908800058,14.046131971085185,79,193
82
+ 154,128,301.0328320087865,8.822140367992688,14.395599560030176,9.123173200001474,14.030206069087816,80,151
83
+ 119,128,295.11679999995977,8.817949280026369,14.402441652467775,9.113066080026329,14.045766691031192,81,75
84
+ 274,128,308.49830398801714,8.834593360021245,14.37530793151238,9.143091664009262,13.999640898697036,82,413
85
+ 96,128,289.26540800603107,8.812201760010794,14.411835255102517,9.101467168016825,14.063666619575436,83,53
86
+ 187,128,302.0151359960437,8.8206908479915,14.397965214812887,9.122705983987544,14.0309246209041,84,243
87
+ 175,128,299.32001599809155,8.830236239999067,14.382401166654791,9.129556255997159,14.020396655742983,85,193
88
+ 95,128,288.6429119971581,8.814813839970157,14.407564618566008,9.103456751967315,14.060592968966255,86,54
89
+ 135,128,297.78729600366205,8.81484275200637,14.407517362813214,9.112630048010033,14.046438769666937,87,153
90
+ 132,128,300.75819202465937,8.814292319992092,14.40841707869679,9.11505051201675,14.042708795881303,88,83
91
+ 151,128,300.1510240137577,8.82715395197738,14.38742324999901,9.127304975991137,14.02385483301991,89,154
92
+ 98,128,296.32057598792017,8.815685887995642,14.406139421657079,9.112006463983562,14.04740004366078,90,47
93
+ 120,128,293.95171196665615,8.818912415998057,14.400868724993128,9.112864127964713,14.046077962163999,91,91
94
+ 129,128,301.3722080504522,8.82641545595834,14.388627029137595,9.127787664008792,14.023113235281402,92,149
95
+ 126,128,298.66793600376695,8.815982783969957,14.405654265900253,9.114650719973724,14.043324745237083,93,129
96
+ 176,128,298.68044797331095,8.825273936032318,14.390488150342547,9.123954384005629,14.029004816637961,94,181
97
+ 151,128,300.5360160022974,8.817950623983052,14.40243945737069,9.118486639985349,14.037417068607523,95,182
98
+ 113,128,297.2445279592648,8.82403992000036,14.392500617789004,9.121284447959624,14.03311131565827,96,72
99
+ 159,128,301.88958399230614,8.817507632018533,14.403163036551492,9.119397216010839,14.03601542602746,97,144
100
+ 135,128,299.31110399775207,8.813154784031212,14.410276809175603,9.112465888028964,14.046691814578253,98,146
101
+ 183,128,299.85328001203015,8.828092415991705,14.385893805319146,9.127945696003735,14.022870453319975,99,214
benchmark/bf16-elyza100-streaming.json ADDED
@@ -0,0 +1,1120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": "training/data/benchmarks/elyza-tasks-100/test.csv",
3
+ "model": "llm-jp-4-8b-instruct",
4
+ "temperature": 0,
5
+ "max_tokens": 128,
6
+ "summary": {
7
+ "count": 100,
8
+ "mean": 14.415907661667651,
9
+ "median": 14.405057756740758,
10
+ "stdev": 0.04455824876122128,
11
+ "min": 14.24789590165574,
12
+ "p05": 14.384719471743757,
13
+ "p25": 14.395599560030176,
14
+ "p75": 14.411835255102517,
15
+ "p95": 14.517020900226493,
16
+ "max": 14.533167284897031
17
+ },
18
+ "records": [
19
+ {
20
+ "prompt_tokens": 89,
21
+ "completion_tokens": 128,
22
+ "ttft_ms": 287.3321439838037,
23
+ "decode_seconds": 8.779072656005155,
24
+ "decode_tokens_per_second": 14.466220405765533,
25
+ "e2e_seconds": 9.066404799988959,
26
+ "e2e_tokens_per_second": 14.118054821482918,
27
+ "index": 0,
28
+ "prompt_characters": 28
29
+ },
30
+ {
31
+ "prompt_tokens": 100,
32
+ "completion_tokens": 128,
33
+ "ttft_ms": 295.72676797397435,
34
+ "decode_seconds": 8.767456800036598,
35
+ "decode_tokens_per_second": 14.485386457731947,
36
+ "e2e_seconds": 9.063183568010572,
37
+ "e2e_tokens_per_second": 14.123072653166709,
38
+ "index": 1,
39
+ "prompt_characters": 50
40
+ },
41
+ {
42
+ "prompt_tokens": 86,
43
+ "completion_tokens": 128,
44
+ "ttft_ms": 290.60323198791593,
45
+ "decode_seconds": 8.77216342400061,
46
+ "decode_tokens_per_second": 14.477614456261545,
47
+ "e2e_seconds": 9.062766655988526,
48
+ "e2e_tokens_per_second": 14.123722353087368,
49
+ "index": 2,
50
+ "prompt_characters": 24
51
+ },
52
+ {
53
+ "prompt_tokens": 223,
54
+ "completion_tokens": 128,
55
+ "ttft_ms": 304.505567997694,
56
+ "decode_seconds": 8.789692880003713,
57
+ "decode_tokens_per_second": 14.448741467283934,
58
+ "e2e_seconds": 9.094198448001407,
59
+ "e2e_tokens_per_second": 14.074907286428306,
60
+ "index": 3,
61
+ "prompt_characters": 323
62
+ },
63
+ {
64
+ "prompt_tokens": 87,
65
+ "completion_tokens": 128,
66
+ "ttft_ms": 289.189808012452,
67
+ "decode_seconds": 8.793557535973378,
68
+ "decode_tokens_per_second": 14.442391430369154,
69
+ "e2e_seconds": 9.08274734398583,
70
+ "e2e_tokens_per_second": 14.092652272746044,
71
+ "index": 4,
72
+ "prompt_characters": 15
73
+ },
74
+ {
75
+ "prompt_tokens": 138,
76
+ "completion_tokens": 128,
77
+ "ttft_ms": 298.6349919810891,
78
+ "decode_seconds": 8.804777184035629,
79
+ "decode_tokens_per_second": 14.42398794943612,
80
+ "e2e_seconds": 9.103412176016718,
81
+ "e2e_tokens_per_second": 14.06066181834772,
82
+ "index": 5,
83
+ "prompt_characters": 107
84
+ },
85
+ {
86
+ "prompt_tokens": 197,
87
+ "completion_tokens": 128,
88
+ "ttft_ms": 304.21625601593405,
89
+ "decode_seconds": 8.81909500801703,
90
+ "decode_tokens_per_second": 14.400570566996976,
91
+ "e2e_seconds": 9.123311264032964,
92
+ "e2e_tokens_per_second": 14.029993748499768,
93
+ "index": 6,
94
+ "prompt_characters": 286
95
+ },
96
+ {
97
+ "prompt_tokens": 129,
98
+ "completion_tokens": 128,
99
+ "ttft_ms": 299.53299200860783,
100
+ "decode_seconds": 8.820546608010773,
101
+ "decode_tokens_per_second": 14.398200660791169,
102
+ "e2e_seconds": 9.12007960001938,
103
+ "e2e_tokens_per_second": 14.034965221107061,
104
+ "index": 7,
105
+ "prompt_characters": 77
106
+ },
107
+ {
108
+ "prompt_tokens": 253,
109
+ "completion_tokens": 128,
110
+ "ttft_ms": 309.5190239837393,
111
+ "decode_seconds": 8.817294720036443,
112
+ "decode_tokens_per_second": 14.403510830981398,
113
+ "e2e_seconds": 9.126813744020183,
114
+ "e2e_tokens_per_second": 14.024609638151606,
115
+ "index": 8,
116
+ "prompt_characters": 359
117
+ },
118
+ {
119
+ "prompt_tokens": 107,
120
+ "completion_tokens": 128,
121
+ "ttft_ms": 299.14905602345243,
122
+ "decode_seconds": 8.819510432018433,
123
+ "decode_tokens_per_second": 14.399892259205004,
124
+ "e2e_seconds": 9.118659488041885,
125
+ "e2e_tokens_per_second": 14.037150983415694,
126
+ "index": 9,
127
+ "prompt_characters": 76
128
+ },
129
+ {
130
+ "prompt_tokens": 115,
131
+ "completion_tokens": 128,
132
+ "ttft_ms": 296.4645279571414,
133
+ "decode_seconds": 8.813084144028835,
134
+ "decode_tokens_per_second": 14.410392312666938,
135
+ "e2e_seconds": 9.109548671985976,
136
+ "e2e_tokens_per_second": 14.051190087345422,
137
+ "index": 10,
138
+ "prompt_characters": 93
139
+ },
140
+ {
141
+ "prompt_tokens": 136,
142
+ "completion_tokens": 128,
143
+ "ttft_ms": 370.15520001295954,
144
+ "decode_seconds": 8.748351391986944,
145
+ "decode_tokens_per_second": 14.517020900226493,
146
+ "e2e_seconds": 9.118506591999903,
147
+ "e2e_tokens_per_second": 14.037386353627298,
148
+ "index": 11,
149
+ "prompt_characters": 126
150
+ },
151
+ {
152
+ "prompt_tokens": 239,
153
+ "completion_tokens": 128,
154
+ "ttft_ms": 309.37422398710623,
155
+ "decode_seconds": 8.829718319990207,
156
+ "decode_tokens_per_second": 14.383244787376283,
157
+ "e2e_seconds": 9.139092543977313,
158
+ "e2e_tokens_per_second": 14.005766916579956,
159
+ "index": 12,
160
+ "prompt_characters": 301
161
+ },
162
+ {
163
+ "prompt_tokens": 113,
164
+ "completion_tokens": 128,
165
+ "ttft_ms": 297.8560159681365,
166
+ "decode_seconds": 8.81552940804977,
167
+ "decode_tokens_per_second": 14.406395137656943,
168
+ "e2e_seconds": 9.113385424017906,
169
+ "e2e_tokens_per_second": 14.045274510464784,
170
+ "index": 13,
171
+ "prompt_characters": 76
172
+ },
173
+ {
174
+ "prompt_tokens": 239,
175
+ "completion_tokens": 128,
176
+ "ttft_ms": 309.5741919823922,
177
+ "decode_seconds": 8.828813120024279,
178
+ "decode_tokens_per_second": 14.384719471743757,
179
+ "e2e_seconds": 9.138387312006671,
180
+ "e2e_tokens_per_second": 14.00684777628372,
181
+ "index": 14,
182
+ "prompt_characters": 323
183
+ },
184
+ {
185
+ "prompt_tokens": 122,
186
+ "completion_tokens": 128,
187
+ "ttft_ms": 370.8477119798772,
188
+ "decode_seconds": 8.738631952030119,
189
+ "decode_tokens_per_second": 14.533167284897031,
190
+ "e2e_seconds": 9.109479664009996,
191
+ "e2e_tokens_per_second": 14.051296530767418,
192
+ "index": 15,
193
+ "prompt_characters": 120
194
+ },
195
+ {
196
+ "prompt_tokens": 152,
197
+ "completion_tokens": 128,
198
+ "ttft_ms": 300.2828800235875,
199
+ "decode_seconds": 8.817734479962382,
200
+ "decode_tokens_per_second": 14.402792496031452,
201
+ "e2e_seconds": 9.11801735998597,
202
+ "e2e_tokens_per_second": 14.038139536970235,
203
+ "index": 16,
204
+ "prompt_characters": 182
205
+ },
206
+ {
207
+ "prompt_tokens": 141,
208
+ "completion_tokens": 128,
209
+ "ttft_ms": 298.9331199787557,
210
+ "decode_seconds": 8.812246479967143,
211
+ "decode_tokens_per_second": 14.411762118627612,
212
+ "e2e_seconds": 9.111179599945899,
213
+ "e2e_tokens_per_second": 14.04867488297125,
214
+ "index": 17,
215
+ "prompt_characters": 140
216
+ },
217
+ {
218
+ "prompt_tokens": 196,
219
+ "completion_tokens": 128,
220
+ "ttft_ms": 304.14611200103536,
221
+ "decode_seconds": 8.825933744024951,
222
+ "decode_tokens_per_second": 14.389412348124349,
223
+ "e2e_seconds": 9.130079856025986,
224
+ "e2e_tokens_per_second": 14.019592601429233,
225
+ "index": 18,
226
+ "prompt_characters": 191
227
+ },
228
+ {
229
+ "prompt_tokens": 118,
230
+ "completion_tokens": 128,
231
+ "ttft_ms": 296.74635198898613,
232
+ "decode_seconds": 8.812764559988864,
233
+ "decode_tokens_per_second": 14.410914887775066,
234
+ "e2e_seconds": 9.10951091197785,
235
+ "e2e_tokens_per_second": 14.05124833120253,
236
+ "index": 19,
237
+ "prompt_characters": 75
238
+ },
239
+ {
240
+ "prompt_tokens": 269,
241
+ "completion_tokens": 128,
242
+ "ttft_ms": 309.93571202270687,
243
+ "decode_seconds": 8.82683105597971,
244
+ "decode_tokens_per_second": 14.38794955908488,
245
+ "e2e_seconds": 9.136766768002417,
246
+ "e2e_tokens_per_second": 14.009332102934351,
247
+ "index": 20,
248
+ "prompt_characters": 368
249
+ },
250
+ {
251
+ "prompt_tokens": 151,
252
+ "completion_tokens": 128,
253
+ "ttft_ms": 371.7997280182317,
254
+ "decode_seconds": 8.751005103986245,
255
+ "decode_tokens_per_second": 14.512618663900579,
256
+ "e2e_seconds": 9.122804832004476,
257
+ "e2e_tokens_per_second": 14.030772592103743,
258
+ "index": 21,
259
+ "prompt_characters": 137
260
+ },
261
+ {
262
+ "prompt_tokens": 83,
263
+ "completion_tokens": 128,
264
+ "ttft_ms": 368.71777597116306,
265
+ "decode_seconds": 8.740428032004274,
266
+ "decode_tokens_per_second": 14.53018084869209,
267
+ "e2e_seconds": 9.109145807975437,
268
+ "e2e_tokens_per_second": 14.051811519794827,
269
+ "index": 22,
270
+ "prompt_characters": 8
271
+ },
272
+ {
273
+ "prompt_tokens": 257,
274
+ "completion_tokens": 128,
275
+ "ttft_ms": 310.9340959927067,
276
+ "decode_seconds": 8.822712880035397,
277
+ "decode_tokens_per_second": 14.394665419452082,
278
+ "e2e_seconds": 9.133646976028103,
279
+ "e2e_tokens_per_second": 14.014117289177584,
280
+ "index": 23,
281
+ "prompt_characters": 379
282
+ },
283
+ {
284
+ "prompt_tokens": 264,
285
+ "completion_tokens": 128,
286
+ "ttft_ms": 312.00308800907806,
287
+ "decode_seconds": 8.822620527993422,
288
+ "decode_tokens_per_second": 14.394816097670736,
289
+ "e2e_seconds": 9.1346236160025,
290
+ "e2e_tokens_per_second": 14.012618951892343,
291
+ "index": 24,
292
+ "prompt_characters": 480
293
+ },
294
+ {
295
+ "prompt_tokens": 115,
296
+ "completion_tokens": 128,
297
+ "ttft_ms": 368.4656799887307,
298
+ "decode_seconds": 8.752178527996875,
299
+ "decode_tokens_per_second": 14.5106729248891,
300
+ "e2e_seconds": 9.120644207985606,
301
+ "e2e_tokens_per_second": 14.034096395069247,
302
+ "index": 25,
303
+ "prompt_characters": 82
304
+ },
305
+ {
306
+ "prompt_tokens": 167,
307
+ "completion_tokens": 128,
308
+ "ttft_ms": 300.1571359927766,
309
+ "decode_seconds": 8.814201600034721,
310
+ "decode_tokens_per_second": 14.408565376982041,
311
+ "e2e_seconds": 9.114358736027498,
312
+ "e2e_tokens_per_second": 14.043774631564363,
313
+ "index": 26,
314
+ "prompt_characters": 145
315
+ },
316
+ {
317
+ "prompt_tokens": 86,
318
+ "completion_tokens": 128,
319
+ "ttft_ms": 288.03827200317755,
320
+ "decode_seconds": 8.810084576020017,
321
+ "decode_tokens_per_second": 14.41529861650575,
322
+ "e2e_seconds": 9.098122848023195,
323
+ "e2e_tokens_per_second": 14.068836191611915,
324
+ "index": 27,
325
+ "prompt_characters": 11
326
+ },
327
+ {
328
+ "prompt_tokens": 95,
329
+ "completion_tokens": 128,
330
+ "ttft_ms": 290.33766401698813,
331
+ "decode_seconds": 8.809145903971512,
332
+ "decode_tokens_per_second": 14.416834660752226,
333
+ "e2e_seconds": 9.0994835679885,
334
+ "e2e_tokens_per_second": 14.066732363834053,
335
+ "index": 28,
336
+ "prompt_characters": 33
337
+ },
338
+ {
339
+ "prompt_tokens": 138,
340
+ "completion_tokens": 128,
341
+ "ttft_ms": 300.13843200867996,
342
+ "decode_seconds": 8.82354910398135,
343
+ "decode_tokens_per_second": 14.393301210586026,
344
+ "e2e_seconds": 9.12368753599003,
345
+ "e2e_tokens_per_second": 14.029415134514519,
346
+ "index": 29,
347
+ "prompt_characters": 155
348
+ },
349
+ {
350
+ "prompt_tokens": 89,
351
+ "completion_tokens": 128,
352
+ "ttft_ms": 291.2958399974741,
353
+ "decode_seconds": 8.811235280008987,
354
+ "decode_tokens_per_second": 14.413416049408962,
355
+ "e2e_seconds": 9.10253112000646,
356
+ "e2e_tokens_per_second": 14.062022783824236,
357
+ "index": 30,
358
+ "prompt_characters": 23
359
+ },
360
+ {
361
+ "prompt_tokens": 122,
362
+ "completion_tokens": 128,
363
+ "ttft_ms": 298.4383839648217,
364
+ "decode_seconds": 8.810073536005802,
365
+ "decode_tokens_per_second": 14.415316680498178,
366
+ "e2e_seconds": 9.108511919970624,
367
+ "e2e_tokens_per_second": 14.05278942648766,
368
+ "index": 31,
369
+ "prompt_characters": 97
370
+ },
371
+ {
372
+ "prompt_tokens": 94,
373
+ "completion_tokens": 128,
374
+ "ttft_ms": 287.24841598886997,
375
+ "decode_seconds": 8.81881228799466,
376
+ "decode_tokens_per_second": 14.401032231164425,
377
+ "e2e_seconds": 9.10606070398353,
378
+ "e2e_tokens_per_second": 14.05657222820898,
379
+ "index": 32,
380
+ "prompt_characters": 61
381
+ },
382
+ {
383
+ "prompt_tokens": 132,
384
+ "completion_tokens": 128,
385
+ "ttft_ms": 298.8155360217206,
386
+ "decode_seconds": 8.817851200001314,
387
+ "decode_tokens_per_second": 14.402601849301004,
388
+ "e2e_seconds": 9.116666736023035,
389
+ "e2e_tokens_per_second": 14.040219271614777,
390
+ "index": 33,
391
+ "prompt_characters": 136
392
+ },
393
+ {
394
+ "prompt_tokens": 114,
395
+ "completion_tokens": 128,
396
+ "ttft_ms": 297.1212000120431,
397
+ "decode_seconds": 8.809575375984423,
398
+ "decode_tokens_per_second": 14.416131831530919,
399
+ "e2e_seconds": 9.106696575996466,
400
+ "e2e_tokens_per_second": 14.055590732800283,
401
+ "index": 34,
402
+ "prompt_characters": 82
403
+ },
404
+ {
405
+ "prompt_tokens": 97,
406
+ "completion_tokens": 128,
407
+ "ttft_ms": 297.4157760036178,
408
+ "decode_seconds": 8.811500575975515,
409
+ "decode_tokens_per_second": 14.412982091411816,
410
+ "e2e_seconds": 9.108916351979133,
411
+ "e2e_tokens_per_second": 14.05216548861917,
412
+ "index": 35,
413
+ "prompt_characters": 49
414
+ },
415
+ {
416
+ "prompt_tokens": 116,
417
+ "completion_tokens": 128,
418
+ "ttft_ms": 295.1624320121482,
419
+ "decode_seconds": 8.818597696023062,
420
+ "decode_tokens_per_second": 14.401382666234271,
421
+ "e2e_seconds": 9.11376012803521,
422
+ "e2e_tokens_per_second": 14.044697051687148,
423
+ "index": 36,
424
+ "prompt_characters": 101
425
+ },
426
+ {
427
+ "prompt_tokens": 127,
428
+ "completion_tokens": 128,
429
+ "ttft_ms": 294.031776022166,
430
+ "decode_seconds": 8.825343439995777,
431
+ "decode_tokens_per_second": 14.390374818100085,
432
+ "e2e_seconds": 9.119375216017943,
433
+ "e2e_tokens_per_second": 14.036049287145392,
434
+ "index": 37,
435
+ "prompt_characters": 118
436
+ },
437
+ {
438
+ "prompt_tokens": 99,
439
+ "completion_tokens": 128,
440
+ "ttft_ms": 298.44286397565156,
441
+ "decode_seconds": 8.810942688025534,
442
+ "decode_tokens_per_second": 14.413894687182415,
443
+ "e2e_seconds": 9.109385552001186,
444
+ "e2e_tokens_per_second": 14.051441699257142,
445
+ "index": 38,
446
+ "prompt_characters": 44
447
+ },
448
+ {
449
+ "prompt_tokens": 89,
450
+ "completion_tokens": 128,
451
+ "ttft_ms": 290.05577601492405,
452
+ "decode_seconds": 8.821949823992327,
453
+ "decode_tokens_per_second": 14.395910488473717,
454
+ "e2e_seconds": 9.112005600007251,
455
+ "e2e_tokens_per_second": 14.047401375598161,
456
+ "index": 39,
457
+ "prompt_characters": 21
458
+ },
459
+ {
460
+ "prompt_tokens": 131,
461
+ "completion_tokens": 128,
462
+ "ttft_ms": 297.30228800326586,
463
+ "decode_seconds": 8.816554000019096,
464
+ "decode_tokens_per_second": 14.404720937423502,
465
+ "e2e_seconds": 9.113856288022362,
466
+ "e2e_tokens_per_second": 14.044548866567112,
467
+ "index": 40,
468
+ "prompt_characters": 99
469
+ },
470
+ {
471
+ "prompt_tokens": 164,
472
+ "completion_tokens": 128,
473
+ "ttft_ms": 299.2388000129722,
474
+ "decode_seconds": 8.815770207962487,
475
+ "decode_tokens_per_second": 14.406001631631959,
476
+ "e2e_seconds": 9.115009007975459,
477
+ "e2e_tokens_per_second": 14.04277273758067,
478
+ "index": 41,
479
+ "prompt_characters": 255
480
+ },
481
+ {
482
+ "prompt_tokens": 119,
483
+ "completion_tokens": 128,
484
+ "ttft_ms": 297.0329279778525,
485
+ "decode_seconds": 8.812944128003437,
486
+ "decode_tokens_per_second": 14.410621258388904,
487
+ "e2e_seconds": 9.10997705598129,
488
+ "e2e_tokens_per_second": 14.050529349682579,
489
+ "index": 42,
490
+ "prompt_characters": 98
491
+ },
492
+ {
493
+ "prompt_tokens": 109,
494
+ "completion_tokens": 128,
495
+ "ttft_ms": 299.9312960309908,
496
+ "decode_seconds": 8.82349996798439,
497
+ "decode_tokens_per_second": 14.393381363496673,
498
+ "e2e_seconds": 9.12343126401538,
499
+ "e2e_tokens_per_second": 14.029809212774731,
500
+ "index": 43,
501
+ "prompt_characters": 69
502
+ },
503
+ {
504
+ "prompt_tokens": 116,
505
+ "completion_tokens": 128,
506
+ "ttft_ms": 296.43854399910197,
507
+ "decode_seconds": 8.814559072023258,
508
+ "decode_tokens_per_second": 14.407981041625595,
509
+ "e2e_seconds": 9.11099761602236,
510
+ "e2e_tokens_per_second": 14.048955492525053,
511
+ "index": 44,
512
+ "prompt_characters": 63
513
+ },
514
+ {
515
+ "prompt_tokens": 99,
516
+ "completion_tokens": 128,
517
+ "ttft_ms": 296.96737602353096,
518
+ "decode_seconds": 8.8147550720023,
519
+ "decode_tokens_per_second": 14.407660673792442,
520
+ "e2e_seconds": 9.11172244802583,
521
+ "e2e_tokens_per_second": 14.047837906622455,
522
+ "index": 45,
523
+ "prompt_characters": 51
524
+ },
525
+ {
526
+ "prompt_tokens": 87,
527
+ "completion_tokens": 128,
528
+ "ttft_ms": 362.1777279768139,
529
+ "decode_seconds": 8.752297568018548,
530
+ "decode_tokens_per_second": 14.510475565189427,
531
+ "e2e_seconds": 9.114475295995362,
532
+ "e2e_tokens_per_second": 14.043595033522063,
533
+ "index": 46,
534
+ "prompt_characters": 16
535
+ },
536
+ {
537
+ "prompt_tokens": 121,
538
+ "completion_tokens": 128,
539
+ "ttft_ms": 299.46691199438646,
540
+ "decode_seconds": 8.81550879997667,
541
+ "decode_tokens_per_second": 14.406428815581934,
542
+ "e2e_seconds": 9.114975711971056,
543
+ "e2e_tokens_per_second": 14.04282403428597,
544
+ "index": 47,
545
+ "prompt_characters": 83
546
+ },
547
+ {
548
+ "prompt_tokens": 103,
549
+ "completion_tokens": 128,
550
+ "ttft_ms": 297.9078239877708,
551
+ "decode_seconds": 8.816075871989597,
552
+ "decode_tokens_per_second": 14.405502158109133,
553
+ "e2e_seconds": 9.113983695977367,
554
+ "e2e_tokens_per_second": 14.044352532306512,
555
+ "index": 48,
556
+ "prompt_characters": 52
557
+ },
558
+ {
559
+ "prompt_tokens": 128,
560
+ "completion_tokens": 128,
561
+ "ttft_ms": 296.2054240051657,
562
+ "decode_seconds": 8.818844111985527,
563
+ "decode_tokens_per_second": 14.400980263093285,
564
+ "e2e_seconds": 9.115049535990693,
565
+ "e2e_tokens_per_second": 14.042710299553846,
566
+ "index": 49,
567
+ "prompt_characters": 82
568
+ },
569
+ {
570
+ "prompt_tokens": 116,
571
+ "completion_tokens": 128,
572
+ "ttft_ms": 296.5520319994539,
573
+ "decode_seconds": 8.82329428801313,
574
+ "decode_tokens_per_second": 14.393716887867562,
575
+ "e2e_seconds": 9.119846320012584,
576
+ "e2e_tokens_per_second": 14.03532422680379,
577
+ "index": 50,
578
+ "prompt_characters": 89
579
+ },
580
+ {
581
+ "prompt_tokens": 202,
582
+ "completion_tokens": 128,
583
+ "ttft_ms": 303.8067359593697,
584
+ "decode_seconds": 8.820766672026366,
585
+ "decode_tokens_per_second": 14.397841448721225,
586
+ "e2e_seconds": 9.124573407985736,
587
+ "e2e_tokens_per_second": 14.02805306908657,
588
+ "index": 51,
589
+ "prompt_characters": 202
590
+ },
591
+ {
592
+ "prompt_tokens": 133,
593
+ "completion_tokens": 128,
594
+ "ttft_ms": 296.68961599236354,
595
+ "decode_seconds": 8.814125408011023,
596
+ "decode_tokens_per_second": 14.408689929073581,
597
+ "e2e_seconds": 9.110815024003386,
598
+ "e2e_tokens_per_second": 14.049237050996068,
599
+ "index": 52,
600
+ "prompt_characters": 73
601
+ },
602
+ {
603
+ "prompt_tokens": 97,
604
+ "completion_tokens": 128,
605
+ "ttft_ms": 296.0946400417015,
606
+ "decode_seconds": 8.822594399971422,
607
+ "decode_tokens_per_second": 14.39485872776962,
608
+ "e2e_seconds": 9.118689040013123,
609
+ "e2e_tokens_per_second": 14.037105491626216,
610
+ "index": 53,
611
+ "prompt_characters": 42
612
+ },
613
+ {
614
+ "prompt_tokens": 208,
615
+ "completion_tokens": 128,
616
+ "ttft_ms": 304.0414240094833,
617
+ "decode_seconds": 8.82271115196636,
618
+ "decode_tokens_per_second": 14.394668238877447,
619
+ "e2e_seconds": 9.126752575975843,
620
+ "e2e_tokens_per_second": 14.02470363192837,
621
+ "index": 54,
622
+ "prompt_characters": 254
623
+ },
624
+ {
625
+ "prompt_tokens": 113,
626
+ "completion_tokens": 128,
627
+ "ttft_ms": 296.58384004142135,
628
+ "decode_seconds": 8.816141712013632,
629
+ "decode_tokens_per_second": 14.405394576058015,
630
+ "e2e_seconds": 9.112725552055053,
631
+ "e2e_tokens_per_second": 14.04629155885575,
632
+ "index": 55,
633
+ "prompt_characters": 85
634
+ },
635
+ {
636
+ "prompt_tokens": 127,
637
+ "completion_tokens": 128,
638
+ "ttft_ms": 370.9898079978302,
639
+ "decode_seconds": 8.744085903977975,
640
+ "decode_tokens_per_second": 14.524102507069776,
641
+ "e2e_seconds": 9.115075711975805,
642
+ "e2e_tokens_per_second": 14.04266997276037,
643
+ "index": 56,
644
+ "prompt_characters": 107
645
+ },
646
+ {
647
+ "prompt_tokens": 144,
648
+ "completion_tokens": 128,
649
+ "ttft_ms": 370.400496001821,
650
+ "decode_seconds": 8.756533648003824,
651
+ "decode_tokens_per_second": 14.503455945601425,
652
+ "e2e_seconds": 9.126934144005645,
653
+ "e2e_tokens_per_second": 14.024424629388543,
654
+ "index": 57,
655
+ "prompt_characters": 121
656
+ },
657
+ {
658
+ "prompt_tokens": 230,
659
+ "completion_tokens": 128,
660
+ "ttft_ms": 309.8742720321752,
661
+ "decode_seconds": 8.82315703999484,
662
+ "decode_tokens_per_second": 14.393940788350093,
663
+ "e2e_seconds": 9.133031312027015,
664
+ "e2e_tokens_per_second": 14.015061990583634,
665
+ "index": 58,
666
+ "prompt_characters": 346
667
+ },
668
+ {
669
+ "prompt_tokens": 124,
670
+ "completion_tokens": 128,
671
+ "ttft_ms": 369.675631984137,
672
+ "decode_seconds": 8.867081984004471,
673
+ "decode_tokens_per_second": 14.322637394026373,
674
+ "e2e_seconds": 9.236757615988608,
675
+ "e2e_tokens_per_second": 13.85767661353753,
676
+ "index": 59,
677
+ "prompt_characters": 93
678
+ },
679
+ {
680
+ "prompt_tokens": 170,
681
+ "completion_tokens": 128,
682
+ "ttft_ms": 302.19761602347717,
683
+ "decode_seconds": 8.913596848025918,
684
+ "decode_tokens_per_second": 14.24789590165574,
685
+ "e2e_seconds": 9.215794464049395,
686
+ "e2e_tokens_per_second": 13.889198646880102,
687
+ "index": 60,
688
+ "prompt_characters": 207
689
+ },
690
+ {
691
+ "prompt_tokens": 148,
692
+ "completion_tokens": 128,
693
+ "ttft_ms": 298.5740000149235,
694
+ "decode_seconds": 8.821890159975737,
695
+ "decode_tokens_per_second": 14.396007850583949,
696
+ "e2e_seconds": 9.12046415999066,
697
+ "e2e_tokens_per_second": 14.034373443569462,
698
+ "index": 61,
699
+ "prompt_characters": 122
700
+ },
701
+ {
702
+ "prompt_tokens": 123,
703
+ "completion_tokens": 128,
704
+ "ttft_ms": 297.4766080151312,
705
+ "decode_seconds": 8.81660511996597,
706
+ "decode_tokens_per_second": 14.404637416775923,
707
+ "e2e_seconds": 9.114081727981102,
708
+ "e2e_tokens_per_second": 14.04420146980115,
709
+ "index": 62,
710
+ "prompt_characters": 88
711
+ },
712
+ {
713
+ "prompt_tokens": 98,
714
+ "completion_tokens": 128,
715
+ "ttft_ms": 297.258767997846,
716
+ "decode_seconds": 8.813225760008208,
717
+ "decode_tokens_per_second": 14.410160758196861,
718
+ "e2e_seconds": 9.110484528006054,
719
+ "e2e_tokens_per_second": 14.049746707381153,
720
+ "index": 63,
721
+ "prompt_characters": 43
722
+ },
723
+ {
724
+ "prompt_tokens": 196,
725
+ "completion_tokens": 128,
726
+ "ttft_ms": 306.2620800337754,
727
+ "decode_seconds": 8.828527903999202,
728
+ "decode_tokens_per_second": 14.385184187102217,
729
+ "e2e_seconds": 9.134789984032977,
730
+ "e2e_tokens_per_second": 14.012363746045144,
731
+ "index": 64,
732
+ "prompt_characters": 177
733
+ },
734
+ {
735
+ "prompt_tokens": 280,
736
+ "completion_tokens": 128,
737
+ "ttft_ms": 381.4747359720059,
738
+ "decode_seconds": 8.757706432021223,
739
+ "decode_tokens_per_second": 14.501513722319327,
740
+ "e2e_seconds": 9.139181167993229,
741
+ "e2e_tokens_per_second": 14.005631100549252,
742
+ "index": 65,
743
+ "prompt_characters": 494
744
+ },
745
+ {
746
+ "prompt_tokens": 100,
747
+ "completion_tokens": 128,
748
+ "ttft_ms": 296.71494400827214,
749
+ "decode_seconds": 8.81345068803057,
750
+ "decode_tokens_per_second": 14.40979299657023,
751
+ "e2e_seconds": 9.110165632038843,
752
+ "e2e_tokens_per_second": 14.050238510466441,
753
+ "index": 66,
754
+ "prompt_characters": 56
755
+ },
756
+ {
757
+ "prompt_tokens": 111,
758
+ "completion_tokens": 128,
759
+ "ttft_ms": 300.6466400111094,
760
+ "decode_seconds": 8.820130895997863,
761
+ "decode_tokens_per_second": 14.398879279402337,
762
+ "e2e_seconds": 9.120777536008973,
763
+ "e2e_tokens_per_second": 14.03389124388288,
764
+ "index": 67,
765
+ "prompt_characters": 68
766
+ },
767
+ {
768
+ "prompt_tokens": 222,
769
+ "completion_tokens": 128,
770
+ "ttft_ms": 308.8962400215678,
771
+ "decode_seconds": 8.82139251200715,
772
+ "decode_tokens_per_second": 14.396819983595018,
773
+ "e2e_seconds": 9.130288752028719,
774
+ "e2e_tokens_per_second": 14.019271840834042,
775
+ "index": 68,
776
+ "prompt_characters": 333
777
+ },
778
+ {
779
+ "prompt_tokens": 86,
780
+ "completion_tokens": 128,
781
+ "ttft_ms": 288.00887998659164,
782
+ "decode_seconds": 8.815513295994606,
783
+ "decode_tokens_per_second": 14.406421468129757,
784
+ "e2e_seconds": 9.103522175981198,
785
+ "e2e_tokens_per_second": 14.060491920118148,
786
+ "index": 69,
787
+ "prompt_characters": 20
788
+ },
789
+ {
790
+ "prompt_tokens": 116,
791
+ "completion_tokens": 128,
792
+ "ttft_ms": 299.3990720133297,
793
+ "decode_seconds": 8.814984991971869,
794
+ "decode_tokens_per_second": 14.407284880877684,
795
+ "e2e_seconds": 9.114384063985199,
796
+ "e2e_tokens_per_second": 14.043735605325471,
797
+ "index": 70,
798
+ "prompt_characters": 70
799
+ },
800
+ {
801
+ "prompt_tokens": 122,
802
+ "completion_tokens": 128,
803
+ "ttft_ms": 297.36579197924584,
804
+ "decode_seconds": 8.820748256053776,
805
+ "decode_tokens_per_second": 14.397871508558078,
806
+ "e2e_seconds": 9.118114048033021,
807
+ "e2e_tokens_per_second": 14.03799067720725,
808
+ "index": 71,
809
+ "prompt_characters": 100
810
+ },
811
+ {
812
+ "prompt_tokens": 123,
813
+ "completion_tokens": 128,
814
+ "ttft_ms": 366.77401600172743,
815
+ "decode_seconds": 8.74748774402542,
816
+ "decode_tokens_per_second": 14.518454179800555,
817
+ "e2e_seconds": 9.114261760027148,
818
+ "e2e_tokens_per_second": 14.043924057719705,
819
+ "index": 72,
820
+ "prompt_characters": 100
821
+ },
822
+ {
823
+ "prompt_tokens": 304,
824
+ "completion_tokens": 128,
825
+ "ttft_ms": 314.2944479477592,
826
+ "decode_seconds": 8.825065472046845,
827
+ "decode_tokens_per_second": 14.390828079663438,
828
+ "e2e_seconds": 9.139359919994604,
829
+ "e2e_tokens_per_second": 14.005357171673307,
830
+ "index": 73,
831
+ "prompt_characters": 516
832
+ },
833
+ {
834
+ "prompt_tokens": 96,
835
+ "completion_tokens": 128,
836
+ "ttft_ms": 290.49059201497585,
837
+ "decode_seconds": 8.813162976002786,
838
+ "decode_tokens_per_second": 14.410263414599976,
839
+ "e2e_seconds": 9.103653568017762,
840
+ "e2e_tokens_per_second": 14.060288986575621,
841
+ "index": 74,
842
+ "prompt_characters": 51
843
+ },
844
+ {
845
+ "prompt_tokens": 98,
846
+ "completion_tokens": 128,
847
+ "ttft_ms": 368.19483200088143,
848
+ "decode_seconds": 8.751111984020099,
849
+ "decode_tokens_per_second": 14.51244141680593,
850
+ "e2e_seconds": 9.11930681602098,
851
+ "e2e_tokens_per_second": 14.036154565512264,
852
+ "index": 75,
853
+ "prompt_characters": 38
854
+ },
855
+ {
856
+ "prompt_tokens": 146,
857
+ "completion_tokens": 128,
858
+ "ttft_ms": 299.2566720349714,
859
+ "decode_seconds": 8.823440047970507,
860
+ "decode_tokens_per_second": 14.393479109002556,
861
+ "e2e_seconds": 9.122696720005479,
862
+ "e2e_tokens_per_second": 14.030938869128944,
863
+ "index": 76,
864
+ "prompt_characters": 133
865
+ },
866
+ {
867
+ "prompt_tokens": 140,
868
+ "completion_tokens": 128,
869
+ "ttft_ms": 299.837856029626,
870
+ "decode_seconds": 8.817254160006996,
871
+ "decode_tokens_per_second": 14.403577088209877,
872
+ "e2e_seconds": 9.117092016036622,
873
+ "e2e_tokens_per_second": 14.039564345171993,
874
+ "index": 77,
875
+ "prompt_characters": 158
876
+ },
877
+ {
878
+ "prompt_tokens": 89,
879
+ "completion_tokens": 128,
880
+ "ttft_ms": 289.219903992489,
881
+ "decode_seconds": 8.821538256015629,
882
+ "decode_tokens_per_second": 14.396582128224123,
883
+ "e2e_seconds": 9.110758160008118,
884
+ "e2e_tokens_per_second": 14.049324738072727,
885
+ "index": 78,
886
+ "prompt_characters": 17
887
+ },
888
+ {
889
+ "prompt_tokens": 163,
890
+ "completion_tokens": 128,
891
+ "ttft_ms": 367.9596959846094,
892
+ "decode_seconds": 8.744869392015971,
893
+ "decode_tokens_per_second": 14.522801234281493,
894
+ "e2e_seconds": 9.11282908800058,
895
+ "e2e_tokens_per_second": 14.046131971085185,
896
+ "index": 79,
897
+ "prompt_characters": 193
898
+ },
899
+ {
900
+ "prompt_tokens": 154,
901
+ "completion_tokens": 128,
902
+ "ttft_ms": 301.0328320087865,
903
+ "decode_seconds": 8.822140367992688,
904
+ "decode_tokens_per_second": 14.395599560030176,
905
+ "e2e_seconds": 9.123173200001474,
906
+ "e2e_tokens_per_second": 14.030206069087816,
907
+ "index": 80,
908
+ "prompt_characters": 151
909
+ },
910
+ {
911
+ "prompt_tokens": 119,
912
+ "completion_tokens": 128,
913
+ "ttft_ms": 295.11679999995977,
914
+ "decode_seconds": 8.817949280026369,
915
+ "decode_tokens_per_second": 14.402441652467775,
916
+ "e2e_seconds": 9.113066080026329,
917
+ "e2e_tokens_per_second": 14.045766691031192,
918
+ "index": 81,
919
+ "prompt_characters": 75
920
+ },
921
+ {
922
+ "prompt_tokens": 274,
923
+ "completion_tokens": 128,
924
+ "ttft_ms": 308.49830398801714,
925
+ "decode_seconds": 8.834593360021245,
926
+ "decode_tokens_per_second": 14.37530793151238,
927
+ "e2e_seconds": 9.143091664009262,
928
+ "e2e_tokens_per_second": 13.999640898697036,
929
+ "index": 82,
930
+ "prompt_characters": 413
931
+ },
932
+ {
933
+ "prompt_tokens": 96,
934
+ "completion_tokens": 128,
935
+ "ttft_ms": 289.26540800603107,
936
+ "decode_seconds": 8.812201760010794,
937
+ "decode_tokens_per_second": 14.411835255102517,
938
+ "e2e_seconds": 9.101467168016825,
939
+ "e2e_tokens_per_second": 14.063666619575436,
940
+ "index": 83,
941
+ "prompt_characters": 53
942
+ },
943
+ {
944
+ "prompt_tokens": 187,
945
+ "completion_tokens": 128,
946
+ "ttft_ms": 302.0151359960437,
947
+ "decode_seconds": 8.8206908479915,
948
+ "decode_tokens_per_second": 14.397965214812887,
949
+ "e2e_seconds": 9.122705983987544,
950
+ "e2e_tokens_per_second": 14.0309246209041,
951
+ "index": 84,
952
+ "prompt_characters": 243
953
+ },
954
+ {
955
+ "prompt_tokens": 175,
956
+ "completion_tokens": 128,
957
+ "ttft_ms": 299.32001599809155,
958
+ "decode_seconds": 8.830236239999067,
959
+ "decode_tokens_per_second": 14.382401166654791,
960
+ "e2e_seconds": 9.129556255997159,
961
+ "e2e_tokens_per_second": 14.020396655742983,
962
+ "index": 85,
963
+ "prompt_characters": 193
964
+ },
965
+ {
966
+ "prompt_tokens": 95,
967
+ "completion_tokens": 128,
968
+ "ttft_ms": 288.6429119971581,
969
+ "decode_seconds": 8.814813839970157,
970
+ "decode_tokens_per_second": 14.407564618566008,
971
+ "e2e_seconds": 9.103456751967315,
972
+ "e2e_tokens_per_second": 14.060592968966255,
973
+ "index": 86,
974
+ "prompt_characters": 54
975
+ },
976
+ {
977
+ "prompt_tokens": 135,
978
+ "completion_tokens": 128,
979
+ "ttft_ms": 297.78729600366205,
980
+ "decode_seconds": 8.81484275200637,
981
+ "decode_tokens_per_second": 14.407517362813214,
982
+ "e2e_seconds": 9.112630048010033,
983
+ "e2e_tokens_per_second": 14.046438769666937,
984
+ "index": 87,
985
+ "prompt_characters": 153
986
+ },
987
+ {
988
+ "prompt_tokens": 132,
989
+ "completion_tokens": 128,
990
+ "ttft_ms": 300.75819202465937,
991
+ "decode_seconds": 8.814292319992092,
992
+ "decode_tokens_per_second": 14.40841707869679,
993
+ "e2e_seconds": 9.11505051201675,
994
+ "e2e_tokens_per_second": 14.042708795881303,
995
+ "index": 88,
996
+ "prompt_characters": 83
997
+ },
998
+ {
999
+ "prompt_tokens": 151,
1000
+ "completion_tokens": 128,
1001
+ "ttft_ms": 300.1510240137577,
1002
+ "decode_seconds": 8.82715395197738,
1003
+ "decode_tokens_per_second": 14.38742324999901,
1004
+ "e2e_seconds": 9.127304975991137,
1005
+ "e2e_tokens_per_second": 14.02385483301991,
1006
+ "index": 89,
1007
+ "prompt_characters": 154
1008
+ },
1009
+ {
1010
+ "prompt_tokens": 98,
1011
+ "completion_tokens": 128,
1012
+ "ttft_ms": 296.32057598792017,
1013
+ "decode_seconds": 8.815685887995642,
1014
+ "decode_tokens_per_second": 14.406139421657079,
1015
+ "e2e_seconds": 9.112006463983562,
1016
+ "e2e_tokens_per_second": 14.04740004366078,
1017
+ "index": 90,
1018
+ "prompt_characters": 47
1019
+ },
1020
+ {
1021
+ "prompt_tokens": 120,
1022
+ "completion_tokens": 128,
1023
+ "ttft_ms": 293.95171196665615,
1024
+ "decode_seconds": 8.818912415998057,
1025
+ "decode_tokens_per_second": 14.400868724993128,
1026
+ "e2e_seconds": 9.112864127964713,
1027
+ "e2e_tokens_per_second": 14.046077962163999,
1028
+ "index": 91,
1029
+ "prompt_characters": 91
1030
+ },
1031
+ {
1032
+ "prompt_tokens": 129,
1033
+ "completion_tokens": 128,
1034
+ "ttft_ms": 301.3722080504522,
1035
+ "decode_seconds": 8.82641545595834,
1036
+ "decode_tokens_per_second": 14.388627029137595,
1037
+ "e2e_seconds": 9.127787664008792,
1038
+ "e2e_tokens_per_second": 14.023113235281402,
1039
+ "index": 92,
1040
+ "prompt_characters": 149
1041
+ },
1042
+ {
1043
+ "prompt_tokens": 126,
1044
+ "completion_tokens": 128,
1045
+ "ttft_ms": 298.66793600376695,
1046
+ "decode_seconds": 8.815982783969957,
1047
+ "decode_tokens_per_second": 14.405654265900253,
1048
+ "e2e_seconds": 9.114650719973724,
1049
+ "e2e_tokens_per_second": 14.043324745237083,
1050
+ "index": 93,
1051
+ "prompt_characters": 129
1052
+ },
1053
+ {
1054
+ "prompt_tokens": 176,
1055
+ "completion_tokens": 128,
1056
+ "ttft_ms": 298.68044797331095,
1057
+ "decode_seconds": 8.825273936032318,
1058
+ "decode_tokens_per_second": 14.390488150342547,
1059
+ "e2e_seconds": 9.123954384005629,
1060
+ "e2e_tokens_per_second": 14.029004816637961,
1061
+ "index": 94,
1062
+ "prompt_characters": 181
1063
+ },
1064
+ {
1065
+ "prompt_tokens": 151,
1066
+ "completion_tokens": 128,
1067
+ "ttft_ms": 300.5360160022974,
1068
+ "decode_seconds": 8.817950623983052,
1069
+ "decode_tokens_per_second": 14.40243945737069,
1070
+ "e2e_seconds": 9.118486639985349,
1071
+ "e2e_tokens_per_second": 14.037417068607523,
1072
+ "index": 95,
1073
+ "prompt_characters": 182
1074
+ },
1075
+ {
1076
+ "prompt_tokens": 113,
1077
+ "completion_tokens": 128,
1078
+ "ttft_ms": 297.2445279592648,
1079
+ "decode_seconds": 8.82403992000036,
1080
+ "decode_tokens_per_second": 14.392500617789004,
1081
+ "e2e_seconds": 9.121284447959624,
1082
+ "e2e_tokens_per_second": 14.03311131565827,
1083
+ "index": 96,
1084
+ "prompt_characters": 72
1085
+ },
1086
+ {
1087
+ "prompt_tokens": 159,
1088
+ "completion_tokens": 128,
1089
+ "ttft_ms": 301.88958399230614,
1090
+ "decode_seconds": 8.817507632018533,
1091
+ "decode_tokens_per_second": 14.403163036551492,
1092
+ "e2e_seconds": 9.119397216010839,
1093
+ "e2e_tokens_per_second": 14.03601542602746,
1094
+ "index": 97,
1095
+ "prompt_characters": 144
1096
+ },
1097
+ {
1098
+ "prompt_tokens": 135,
1099
+ "completion_tokens": 128,
1100
+ "ttft_ms": 299.31110399775207,
1101
+ "decode_seconds": 8.813154784031212,
1102
+ "decode_tokens_per_second": 14.410276809175603,
1103
+ "e2e_seconds": 9.112465888028964,
1104
+ "e2e_tokens_per_second": 14.046691814578253,
1105
+ "index": 98,
1106
+ "prompt_characters": 146
1107
+ },
1108
+ {
1109
+ "prompt_tokens": 183,
1110
+ "completion_tokens": 128,
1111
+ "ttft_ms": 299.85328001203015,
1112
+ "decode_seconds": 8.828092415991705,
1113
+ "decode_tokens_per_second": 14.385893805319146,
1114
+ "e2e_seconds": 9.127945696003735,
1115
+ "e2e_tokens_per_second": 14.022870453319975,
1116
+ "index": 99,
1117
+ "prompt_characters": 214
1118
+ }
1119
+ ]
1120
+ }
benchmark/dflash20k-b4-l1-elyza100-tps-histogram.png CHANGED

Git LFS Details

  • SHA256: f3d0631ade51a16f62f928bc674ba8a4c6525f7be946fbfd50a4335684cb043e
  • Pointer size: 130 Bytes
  • Size of remote file: 71.5 kB
benchmark/elyza100-baseline-vs-dflash-speedup.png CHANGED

Git LFS Details

  • SHA256: d5178b50be55c5eb1e13e0e733ed3692f6320980550aa6ce8cd59c8852577d5c
  • Pointer size: 130 Bytes
  • Size of remote file: 94 kB
benchmark/elyza100-bf16-nvfp4-dflash-speedup-summary.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "count": 100,
3
+ "bf16": {
4
+ "name": "BF16",
5
+ "mean_tps": 14.415907661667651,
6
+ "median_tps": 14.405057756740758,
7
+ "ratio_of_mean_vs_bf16": 1.0,
8
+ "paired_speedup_mean_vs_bf16": 1.0,
9
+ "paired_speedup_median_vs_bf16": 1.0,
10
+ "paired_speedup_p05_vs_bf16": 1.0,
11
+ "paired_speedup_p95_vs_bf16": 1.0,
12
+ "requests_slower_than_bf16": 0
13
+ },
14
+ "nvfp4": {
15
+ "name": "NVFP4",
16
+ "mean_tps": 29.72816958251134,
17
+ "median_tps": 29.69194599518334,
18
+ "ratio_of_mean_vs_bf16": 2.062178135446821,
19
+ "paired_speedup_mean_vs_bf16": 2.062181336331503,
20
+ "paired_speedup_median_vs_bf16": 2.0610139004279286,
21
+ "paired_speedup_p05_vs_bf16": 2.0583650145121104,
22
+ "paired_speedup_p95_vs_bf16": 2.069276803345841,
23
+ "requests_slower_than_bf16": 0
24
+ },
25
+ "nvfp4_dflash": {
26
+ "name": "NVFP4 + DFlash",
27
+ "mean_tps": 49.038942024401024,
28
+ "median_tps": 48.95244489534201,
29
+ "ratio_of_mean_vs_bf16": 3.401724204629661,
30
+ "paired_speedup_mean_vs_bf16": 3.401837646698169,
31
+ "paired_speedup_median_vs_bf16": 3.3984135188496474,
32
+ "paired_speedup_p05_vs_bf16": 2.8543341194603595,
33
+ "paired_speedup_p95_vs_bf16": 4.120080899996654,
34
+ "requests_slower_than_bf16": 0
35
+ },
36
+ "dflash_paired_speedup_vs_nvfp4": {
37
+ "mean": 1.649690142355652,
38
+ "median": 1.6482948864124465,
39
+ "p05": 1.384932896270244,
40
+ "p95": 1.9981834458804657,
41
+ "requests_slower_than_nvfp4": 0
42
+ }
43
+ }
benchmark/elyza100-bf16-nvfp4-dflash-speedup.png ADDED

Git LFS Details

  • SHA256: 8a1797d2a3bfc3f5d4a60a82490d4ebbba726835cd9ba8b0d64061a990913c7e
  • Pointer size: 131 Bytes
  • Size of remote file: 109 kB