ob-e5-draft-head-fp8.summary.json 4.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217
  1. {
  2. "base": "http://10.0.0.1:8888",
  3. "reps": 5,
  4. "max_tokens": 512,
  5. "phases": {
  6. "c1-prose": {
  7. "start": "2026-09-24T03:33:36.560658Z",
  8. "end": "2026-09-24T03:35:01.787553Z",
  9. "decode_tps": {
  10. "median": 30.498101268011187,
  11. "min": 30.243856550332247,
  12. "max": 30.786130518748354,
  13. "stdev": 0.18168216715658586,
  14. "n": 5
  15. },
  16. "e2e_tps": {
  17. "median": 30.0254319225977,
  18. "min": 29.75342580689415,
  19. "max": 30.29953950058001,
  20. "stdev": 0.18357552959856419,
  21. "n": 5
  22. },
  23. "ttft_s": {
  24. "median": 0.29706955602159724,
  25. "min": 0.29065055400133133,
  26. "max": 0.3121090880013071,
  27. "stdev": 0.0071557501219345525,
  28. "n": 5
  29. },
  30. "completion_tokens": [
  31. 512,
  32. 512,
  33. 512,
  34. 512,
  35. 512
  36. ],
  37. "distinct_outputs": 3,
  38. "errors": []
  39. },
  40. "c1-code": {
  41. "start": "2026-09-24T03:35:01.787910Z",
  42. "end": "2026-09-24T03:35:50.319550Z",
  43. "decode_tps": {
  44. "median": 54.325956174454866,
  45. "min": 54.15575094673564,
  46. "max": 54.61819745104633,
  47. "stdev": 0.20026431060744665,
  48. "n": 5
  49. },
  50. "e2e_tps": {
  51. "median": 52.701175595450266,
  52. "min": 52.56320939517117,
  53. "max": 52.98276508080466,
  54. "stdev": 0.18032406569516368,
  55. "n": 5
  56. },
  57. "ttft_s": {
  58. "median": 0.30766363398288377,
  59. "min": 0.3037464920198545,
  60. "max": 0.3095748559862841,
  61. "stdev": 0.0020789926196880834,
  62. "n": 5
  63. },
  64. "completion_tokens": [
  65. 512,
  66. 512,
  67. 512,
  68. 512,
  69. 512
  70. ],
  71. "distinct_outputs": 1,
  72. "errors": []
  73. },
  74. "c1-chat_sampled": {
  75. "start": "2026-09-24T03:35:50.319819Z",
  76. "end": "2026-09-24T03:37:08.454057Z",
  77. "decode_tps": {
  78. "median": 32.97563454721633,
  79. "min": 32.26059080487023,
  80. "max": 34.60662529795313,
  81. "stdev": 0.8447515063688072,
  82. "n": 5
  83. },
  84. "e2e_tps": {
  85. "median": 32.46911135055481,
  86. "min": 31.79044869550587,
  87. "max": 34.002811961904335,
  88. "stdev": 0.7980424993760676,
  89. "n": 5
  90. },
  91. "ttft_s": {
  92. "median": 0.2725430429854896,
  93. "min": 0.26570678598363884,
  94. "max": 0.29161949298577383,
  95. "stdev": 0.009226482771044918,
  96. "n": 5
  97. },
  98. "completion_tokens": [
  99. 512,
  100. 512,
  101. 512,
  102. 512,
  103. 512
  104. ],
  105. "distinct_outputs": 5,
  106. "errors": []
  107. },
  108. "c1-prose2": {
  109. "start": "2026-09-24T03:37:08.454348Z",
  110. "end": "2026-09-24T03:38:41.014076Z",
  111. "decode_tps": {
  112. "median": 27.930453918121465,
  113. "min": 27.909100079402,
  114. "max": 28.32840390173375,
  115. "stdev": 0.15800339613757916,
  116. "n": 5
  117. },
  118. "e2e_tps": {
  119. "median": 27.567603385073852,
  120. "min": 27.546613983541643,
  121. "max": 27.9441930519772,
  122. "stdev": 0.14886844684677136,
  123. "n": 5
  124. },
  125. "ttft_s": {
  126. "median": 0.280362689984031,
  127. "min": 0.26309179002419114,
  128. "max": 0.287192433985183,
  129. "stdev": 0.008322868765609224,
  130. "n": 5
  131. },
  132. "completion_tokens": [
  133. 512,
  134. 512,
  135. 512,
  136. 512,
  137. 512
  138. ],
  139. "distinct_outputs": 2,
  140. "errors": []
  141. },
  142. "c4": {
  143. "start": "2026-09-24T03:38:41.014255Z",
  144. "end": "2026-09-24T03:41:09.217891Z",
  145. "agg_tps": {
  146. "median": 69.46540319994251,
  147. "min": 68.3257498191875,
  148. "max": 69.57172763665461,
  149. "stdev": 0.5195680406993157,
  150. "n": 5
  151. },
  152. "per_stream_decode_median": 23.291582721256802,
  153. "waves": [
  154. {
  155. "wall_s": 29.974058176012477,
  156. "tokens": 2048,
  157. "agg_tps": 68.3257498191875,
  158. "per_stream_decode": [
  159. 17.27112626772283,
  160. 29.02624189001746,
  161. 17.49654465651642,
  162. 28.211025033862136
  163. ],
  164. "errors": []
  165. },
  166. {
  167. "wall_s": 29.437245121982414,
  168. "tokens": 2048,
  169. "agg_tps": 69.57172763665461,
  170. "per_stream_decode": [
  171. 17.597029985378757,
  172. 29.04793763021858,
  173. 17.643764536075736,
  174. 28.754876881876875
  175. ],
  176. "errors": []
  177. },
  178. {
  179. "wall_s": 29.84434698399855,
  180. "tokens": 2048,
  181. "agg_tps": 68.62271106478097,
  182. "per_stream_decode": [
  183. 17.356541791081067,
  184. 30.79483473243217,
  185. 17.537704299658873,
  186. 28.82988863585415
  187. ],
  188. "errors": []
  189. },
  190. {
  191. "wall_s": 29.482302062009694,
  192. "tokens": 2048,
  193. "agg_tps": 69.46540319994251,
  194. "per_stream_decode": [
  195. 17.578383517708033,
  196. 29.02544550202472,
  197. 17.67299850877779,
  198. 30.794002241591535
  199. ],
  200. "errors": []
  201. },
  202. {
  203. "wall_s": 29.464242398011265,
  204. "tokens": 2048,
  205. "agg_tps": 69.50798097351496,
  206. "per_stream_decode": [
  207. 18.372140408651468,
  208. 30.138107238240536,
  209. 17.58305717872259,
  210. 32.26035773043606
  211. ],
  212. "errors": []
  213. }
  214. ]
  215. }
  216. }
  217. }