test_baseline.json 21 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051
  1. [
  2. {
  3. "top": 1,
  4. "dtype": "float16",
  5. "num_local_experts": 1,
  6. "step_time": [
  7. {
  8. "GPU": "V100",
  9. "value": "0.014220118522644043"
  10. },
  11. {
  12. "GPU": "RTX 2080 Ti",
  13. "value": "0.016164422035217285"
  14. }
  15. ],
  16. "losses": [
  17. "30.5",
  18. "29.90625",
  19. "29.34375",
  20. "28.796875",
  21. "28.265625",
  22. "27.765625",
  23. "27.28125",
  24. "26.8125",
  25. "26.359375",
  26. "25.90625",
  27. "25.46875",
  28. "25.03125",
  29. "24.609375",
  30. "24.203125",
  31. "23.796875",
  32. "23.40625",
  33. "23.015625",
  34. "22.625",
  35. "22.25",
  36. "21.875",
  37. "21.515625",
  38. "21.15625",
  39. "20.796875",
  40. "20.453125",
  41. "20.109375",
  42. "19.765625",
  43. "19.4375",
  44. "19.09375",
  45. "18.765625",
  46. "18.4375",
  47. "18.109375",
  48. "17.796875",
  49. "17.46875",
  50. "17.15625",
  51. "16.84375",
  52. "16.515625",
  53. "16.21875",
  54. "15.90625",
  55. "15.59375",
  56. "15.2890625",
  57. "14.984375",
  58. "14.6875",
  59. "14.3828125",
  60. "14.0859375",
  61. "13.7890625",
  62. "13.5",
  63. "13.2109375",
  64. "12.9140625",
  65. "12.625",
  66. "12.3359375",
  67. "12.0546875",
  68. "11.7734375",
  69. "11.4921875",
  70. "11.21875",
  71. "10.9453125",
  72. "10.671875",
  73. "10.40625",
  74. "10.140625",
  75. "9.875",
  76. "9.6171875",
  77. "9.3671875",
  78. "9.1171875",
  79. "8.859375",
  80. "8.6171875",
  81. "8.3671875",
  82. "8.125",
  83. "7.8828125",
  84. "7.640625",
  85. "7.41015625",
  86. "7.16796875",
  87. "6.9375",
  88. "6.70703125",
  89. "6.48046875",
  90. "6.2578125",
  91. "6.0390625",
  92. "5.82421875",
  93. "5.61328125",
  94. "5.40234375",
  95. "5.19921875",
  96. "5.0",
  97. "4.8125",
  98. "4.625",
  99. "4.4453125",
  100. "4.26953125",
  101. "4.09765625",
  102. "3.9375",
  103. "3.78125",
  104. "3.630859375",
  105. "3.486328125",
  106. "3.345703125",
  107. "3.212890625",
  108. "3.083984375",
  109. "2.9609375",
  110. "2.845703125",
  111. "2.734375",
  112. "2.62890625",
  113. "2.525390625",
  114. "2.4296875",
  115. "2.3359375",
  116. "2.24609375"
  117. ]
  118. },
  119. {
  120. "top": 1,
  121. "dtype": "float16",
  122. "num_local_experts": 2,
  123. "step_time": [
  124. {
  125. "GPU": "V100",
  126. "value": "0.015884661674499513"
  127. },
  128. {
  129. "GPU": "RTX 2080 Ti",
  130. "value": "0.016640520095825194"
  131. }
  132. ],
  133. "losses": [
  134. "26.484375",
  135. "26.109375",
  136. "25.75",
  137. "25.390625",
  138. "25.046875",
  139. "24.703125",
  140. "24.375",
  141. "24.046875",
  142. "23.734375",
  143. "23.4375",
  144. "23.125",
  145. "22.828125",
  146. "22.546875",
  147. "22.265625",
  148. "22.0",
  149. "21.734375",
  150. "21.46875",
  151. "21.21875",
  152. "20.984375",
  153. "20.734375",
  154. "20.46875",
  155. "20.234375",
  156. "20.015625",
  157. "19.796875",
  158. "18.6875",
  159. "18.5",
  160. "18.3125",
  161. "18.140625",
  162. "18.84375",
  163. "17.796875",
  164. "17.625",
  165. "17.46875",
  166. "17.296875",
  167. "17.140625",
  168. "17.671875",
  169. "16.828125",
  170. "16.484375",
  171. "16.53125",
  172. "16.171875",
  173. "16.03125",
  174. "16.984375",
  175. "15.7421875",
  176. "15.3203125",
  177. "15.7109375",
  178. "15.0703125",
  179. "15.2265625",
  180. "16.390625",
  181. "14.734375",
  182. "15.3203125",
  183. "14.515625",
  184. "14.9375",
  185. "14.203125",
  186. "14.765625",
  187. "13.890625",
  188. "14.5",
  189. "13.7109375",
  190. "13.6796875",
  191. "14.5078125",
  192. "14.0859375",
  193. "13.3984375",
  194. "14.3671875",
  195. "13.0234375",
  196. "13.2421875",
  197. "13.8359375",
  198. "14.2265625",
  199. "12.7734375",
  200. "13.703125",
  201. "12.8984375",
  202. "13.3671875",
  203. "13.296875",
  204. "12.4765625",
  205. "13.9140625",
  206. "12.6796875",
  207. "12.8984375",
  208. "12.890625",
  209. "12.234375",
  210. "13.203125",
  211. "12.5390625",
  212. "12.8203125",
  213. "12.78125",
  214. "12.3125",
  215. "12.484375",
  216. "12.3359375",
  217. "12.2265625",
  218. "11.921875",
  219. "12.9375",
  220. "12.7109375",
  221. "11.5703125",
  222. "12.203125",
  223. "11.9921875",
  224. "11.5",
  225. "11.984375",
  226. "12.2734375",
  227. "12.3125",
  228. "11.703125",
  229. "11.71875",
  230. "11.421875",
  231. "11.4765625",
  232. "11.3203125",
  233. "11.8828125"
  234. ]
  235. },
  236. {
  237. "top": 1,
  238. "dtype": "float32",
  239. "num_local_experts": 1,
  240. "step_time": [
  241. {
  242. "GPU": "V100",
  243. "value": "0.06863923072814941"
  244. },
  245. {
  246. "GPU": "RTX 2080 Ti",
  247. "value": "0.059459257125854495"
  248. }
  249. ],
  250. "losses": [
  251. "29.98500633239746",
  252. "28.761884689331055",
  253. "27.58047866821289",
  254. "26.43315315246582",
  255. "25.3136043548584",
  256. "24.21868133544922",
  257. "23.148454666137695",
  258. "22.106149673461914",
  259. "21.094993591308594",
  260. "20.11581802368164",
  261. "19.165630340576172",
  262. "18.239715576171875",
  263. "17.333059310913086",
  264. "16.442028045654297",
  265. "15.564196586608887",
  266. "14.698493957519531",
  267. "13.844827651977539",
  268. "13.004915237426758",
  269. "12.182660102844238",
  270. "11.383204460144043",
  271. "10.612395286560059",
  272. "9.876236915588379",
  273. "9.181147575378418",
  274. "8.531209945678711",
  275. "7.925898551940918",
  276. "7.362267017364502",
  277. "6.836770534515381",
  278. "6.344995975494385",
  279. "5.880909442901611",
  280. "5.439388751983643",
  281. "5.016486167907715",
  282. "4.610177040100098",
  283. "4.2208251953125",
  284. "3.8502838611602783",
  285. "3.4999637603759766",
  286. "3.16986083984375",
  287. "2.858599901199341",
  288. "2.564656972885132",
  289. "2.287442207336426",
  290. "2.0266451835632324",
  291. "1.7821216583251953",
  292. "1.5533881187438965",
  293. "1.3392244577407837",
  294. "1.1389491558074951",
  295. "0.9536727070808411",
  296. "0.7866250276565552",
  297. "0.6412524580955505",
  298. "0.5192052125930786",
  299. "0.41924676299095154",
  300. "0.33786702156066895",
  301. "0.27139899134635925",
  302. "0.21766313910484314",
  303. "0.17579089105129242",
  304. "0.14475053548812866",
  305. "0.12243179231882095",
  306. "0.10634990781545639",
  307. "0.0944439247250557",
  308. "0.08532019704580307",
  309. "0.07809687405824661",
  310. "0.07220908999443054",
  311. "0.06729564070701599",
  312. "0.06311818957328796",
  313. "0.05951143428683281",
  314. "0.05635545030236244",
  315. "0.05356447398662567",
  316. "0.051073770970106125",
  317. "0.04883363097906113",
  318. "0.04680565744638443",
  319. "0.04495749622583389",
  320. "0.043265290558338165",
  321. "0.041708823293447495",
  322. "0.04027065634727478",
  323. "0.03893772140145302",
  324. "0.037698108702898026",
  325. "0.03654063865542412",
  326. "0.03545815870165825",
  327. "0.03444304317235947",
  328. "0.03348816931247711",
  329. "0.03258833661675453",
  330. "0.03173903375864029",
  331. "0.030935637652873993",
  332. "0.03017423115670681",
  333. "0.029451580718159676",
  334. "0.02876456454396248",
  335. "0.028110845014452934",
  336. "0.027487512677907944",
  337. "0.026892736554145813",
  338. "0.026324469596147537",
  339. "0.02578059770166874",
  340. "0.025259941816329956",
  341. "0.024760741740465164",
  342. "0.02428179420530796",
  343. "0.02382158301770687",
  344. "0.02337941713631153",
  345. "0.02295416034758091",
  346. "0.02254442311823368",
  347. "0.02214960940182209",
  348. "0.021768951788544655",
  349. "0.021401450037956238",
  350. "0.021046629175543785"
  351. ]
  352. },
  353. {
  354. "top": 1,
  355. "dtype": "float32",
  356. "num_local_experts": 2,
  357. "step_time": [
  358. {
  359. "GPU": "V100",
  360. "value": "0.07085721492767334"
  361. },
  362. {
  363. "GPU": "RTX 2080 Ti",
  364. "value": "0.0697479009628296"
  365. }
  366. ],
  367. "losses": [
  368. "22.07648277282715",
  369. "21.387950897216797",
  370. "20.73025894165039",
  371. "20.102460861206055",
  372. "19.502853393554688",
  373. "18.929244995117188",
  374. "18.37995147705078",
  375. "17.85356330871582",
  376. "17.348365783691406",
  377. "16.86236000061035",
  378. "16.393089294433594",
  379. "15.938311576843262",
  380. "15.495789527893066",
  381. "15.063897132873535",
  382. "14.641019821166992",
  383. "14.226059913635254",
  384. "13.818093299865723",
  385. "13.416620254516602",
  386. "13.02176284790039",
  387. "12.633668899536133",
  388. "12.253036499023438",
  389. "11.880377769470215",
  390. "11.516019821166992",
  391. "11.15967845916748",
  392. "10.810673713684082",
  393. "10.4685697555542",
  394. "10.133264541625977",
  395. "9.805092811584473",
  396. "9.484430313110352",
  397. "9.171442031860352",
  398. "8.86601734161377",
  399. "8.567427635192871",
  400. "8.27490234375",
  401. "7.987741947174072",
  402. "7.705358028411865",
  403. "7.427794456481934",
  404. "7.155190944671631",
  405. "6.887553691864014",
  406. "6.626027584075928",
  407. "6.370643138885498",
  408. "6.1208953857421875",
  409. "5.876578330993652",
  410. "5.637292861938477",
  411. "5.402650833129883",
  412. "5.173330783843994",
  413. "4.949470043182373",
  414. "4.731014251708984",
  415. "4.518482685089111",
  416. "4.311755657196045",
  417. "4.11070442199707",
  418. "3.915586471557617",
  419. "3.726199150085449",
  420. "3.5423967838287354",
  421. "3.3639752864837646",
  422. "3.1907100677490234",
  423. "3.02242112159729",
  424. "2.8590900897979736",
  425. "2.7008259296417236",
  426. "2.547826051712036",
  427. "2.4003584384918213",
  428. "2.2587926387786865",
  429. "2.1234612464904785",
  430. "1.9942450523376465",
  431. "1.8715554475784302",
  432. "1.7550554275512695",
  433. "1.6444458961486816",
  434. "1.539300799369812",
  435. "1.439163088798523",
  436. "1.3436201810836792",
  437. "1.2523038387298584",
  438. "1.1649760007858276",
  439. "1.0814529657363892",
  440. "1.0016427040100098",
  441. "0.9255325198173523",
  442. "0.8531060814857483",
  443. "0.7844582796096802",
  444. "0.7197057008743286",
  445. "0.6593865752220154",
  446. "0.6026942729949951",
  447. "0.5501302480697632",
  448. "0.5016597509384155",
  449. "0.457214891910553",
  450. "0.41650763154029846",
  451. "0.3793560862541199",
  452. "0.34559059143066406",
  453. "0.314958781003952",
  454. "0.287245512008667",
  455. "0.26227840781211853",
  456. "0.23979704082012177",
  457. "0.21966718137264252",
  458. "0.20171628892421722",
  459. "0.18577177822589874",
  460. "0.17165276408195496",
  461. "0.15917447209358215",
  462. "0.14816568791866302",
  463. "0.13842710852622986",
  464. "0.12981705367565155",
  465. "0.1221797838807106",
  466. "0.11538200080394745",
  467. "0.10930977761745453"
  468. ]
  469. },
  470. {
  471. "top": 2,
  472. "dtype": "float16",
  473. "num_local_experts": 1,
  474. "step_time": [
  475. {
  476. "GPU": "V100",
  477. "value": "0.014011073112487792"
  478. },
  479. {
  480. "GPU": "RTX 2080 Ti",
  481. "value": "0.016269850730895995"
  482. }
  483. ],
  484. "losses": [
  485. "30.5",
  486. "29.90625",
  487. "29.34375",
  488. "28.796875",
  489. "28.265625",
  490. "27.765625",
  491. "27.28125",
  492. "26.8125",
  493. "26.359375",
  494. "25.90625",
  495. "25.46875",
  496. "25.03125",
  497. "24.609375",
  498. "24.203125",
  499. "23.796875",
  500. "23.40625",
  501. "23.015625",
  502. "22.625",
  503. "22.25",
  504. "21.875",
  505. "21.515625",
  506. "21.15625",
  507. "20.796875",
  508. "20.453125",
  509. "20.109375",
  510. "19.765625",
  511. "19.4375",
  512. "19.09375",
  513. "18.765625",
  514. "18.4375",
  515. "18.109375",
  516. "17.796875",
  517. "17.46875",
  518. "17.15625",
  519. "16.84375",
  520. "16.515625",
  521. "16.21875",
  522. "15.90625",
  523. "15.59375",
  524. "15.2890625",
  525. "14.984375",
  526. "14.6875",
  527. "14.3828125",
  528. "14.0859375",
  529. "13.7890625",
  530. "13.5",
  531. "13.2109375",
  532. "12.9140625",
  533. "12.625",
  534. "12.3359375",
  535. "12.0546875",
  536. "11.7734375",
  537. "11.4921875",
  538. "11.21875",
  539. "10.9453125",
  540. "10.671875",
  541. "10.40625",
  542. "10.140625",
  543. "9.875",
  544. "9.6171875",
  545. "9.3671875",
  546. "9.1171875",
  547. "8.859375",
  548. "8.6171875",
  549. "8.3671875",
  550. "8.125",
  551. "7.8828125",
  552. "7.640625",
  553. "7.41015625",
  554. "7.16796875",
  555. "6.9375",
  556. "6.70703125",
  557. "6.48046875",
  558. "6.2578125",
  559. "6.0390625",
  560. "5.82421875",
  561. "5.61328125",
  562. "5.40234375",
  563. "5.19921875",
  564. "5.0",
  565. "4.8125",
  566. "4.625",
  567. "4.4453125",
  568. "4.26953125",
  569. "4.09765625",
  570. "3.9375",
  571. "3.78125",
  572. "3.630859375",
  573. "3.486328125",
  574. "3.345703125",
  575. "3.212890625",
  576. "3.083984375",
  577. "2.9609375",
  578. "2.845703125",
  579. "2.734375",
  580. "2.62890625",
  581. "2.525390625",
  582. "2.4296875",
  583. "2.3359375",
  584. "2.24609375"
  585. ]
  586. },
  587. {
  588. "top": 2,
  589. "dtype": "float16",
  590. "num_local_experts": 2,
  591. "step_time": [
  592. {
  593. "GPU": "V100",
  594. "value": "0.030052709579467773"
  595. },
  596. {
  597. "GPU": "RTX 2080 Ti",
  598. "value": "0.03255057334899902"
  599. }
  600. ],
  601. "losses": [
  602. "25.828125",
  603. "25.453125",
  604. "25.09375",
  605. "24.75",
  606. "24.40625",
  607. "24.09375",
  608. "23.78125",
  609. "23.484375",
  610. "23.203125",
  611. "22.921875",
  612. "22.65625",
  613. "22.40625",
  614. "22.15625",
  615. "21.921875",
  616. "21.6875",
  617. "21.46875",
  618. "21.25",
  619. "21.046875",
  620. "20.828125",
  621. "20.625",
  622. "20.4375",
  623. "20.234375",
  624. "20.046875",
  625. "19.859375",
  626. "19.671875",
  627. "19.484375",
  628. "19.3125",
  629. "19.125",
  630. "18.953125",
  631. "18.78125",
  632. "18.609375",
  633. "18.4375",
  634. "18.28125",
  635. "18.109375",
  636. "17.953125",
  637. "17.796875",
  638. "17.640625",
  639. "17.484375",
  640. "17.328125",
  641. "17.171875",
  642. "17.015625",
  643. "16.875",
  644. "16.71875",
  645. "16.578125",
  646. "16.4375",
  647. "16.28125",
  648. "16.140625",
  649. "15.9921875",
  650. "15.859375",
  651. "15.7109375",
  652. "15.578125",
  653. "15.4375",
  654. "15.296875",
  655. "15.1640625",
  656. "15.0234375",
  657. "14.890625",
  658. "14.7578125",
  659. "14.6171875",
  660. "14.484375",
  661. "14.3515625",
  662. "14.21875",
  663. "14.0859375",
  664. "13.953125",
  665. "13.8203125",
  666. "13.6875",
  667. "13.5546875",
  668. "13.4296875",
  669. "13.296875",
  670. "13.1640625",
  671. "13.0390625",
  672. "12.90625",
  673. "12.78125",
  674. "12.6484375",
  675. "12.5234375",
  676. "12.390625",
  677. "12.265625",
  678. "12.140625",
  679. "12.015625",
  680. "11.8828125",
  681. "11.7578125",
  682. "11.6328125",
  683. "11.5",
  684. "11.375",
  685. "11.25",
  686. "11.125",
  687. "11.0",
  688. "10.875",
  689. "10.75",
  690. "10.625",
  691. "10.5",
  692. "10.375",
  693. "10.25",
  694. "10.125",
  695. "10.0",
  696. "9.8828125",
  697. "9.7578125",
  698. "9.6328125",
  699. "9.515625",
  700. "9.390625",
  701. "9.2734375"
  702. ]
  703. },
  704. {
  705. "top": 2,
  706. "dtype": "float32",
  707. "num_local_experts": 1,
  708. "step_time": [
  709. {
  710. "GPU": "V100",
  711. "value": "0.06921641826629639"
  712. },
  713. {
  714. "GPU": "RTX 2080 Ti",
  715. "value": "0.05989260673522949"
  716. }
  717. ],
  718. "losses": [
  719. "29.98500633239746",
  720. "28.761884689331055",
  721. "27.58047866821289",
  722. "26.43315315246582",
  723. "25.3136043548584",
  724. "24.21868133544922",
  725. "23.148454666137695",
  726. "22.106149673461914",
  727. "21.094993591308594",
  728. "20.11581802368164",
  729. "19.165630340576172",
  730. "18.239715576171875",
  731. "17.333059310913086",
  732. "16.442028045654297",
  733. "15.564196586608887",
  734. "14.698493957519531",
  735. "13.844827651977539",
  736. "13.004915237426758",
  737. "12.182660102844238",
  738. "11.383204460144043",
  739. "10.612395286560059",
  740. "9.876236915588379",
  741. "9.181147575378418",
  742. "8.531209945678711",
  743. "7.925898551940918",
  744. "7.362267017364502",
  745. "6.836770534515381",
  746. "6.344995975494385",
  747. "5.880909442901611",
  748. "5.439388751983643",
  749. "5.016486167907715",
  750. "4.610177040100098",
  751. "4.2208251953125",
  752. "3.8502838611602783",
  753. "3.4999637603759766",
  754. "3.16986083984375",
  755. "2.858599901199341",
  756. "2.564656972885132",
  757. "2.287442207336426",
  758. "2.0266451835632324",
  759. "1.7821216583251953",
  760. "1.5533881187438965",
  761. "1.3392244577407837",
  762. "1.1389491558074951",
  763. "0.9536727070808411",
  764. "0.7866250276565552",
  765. "0.6412524580955505",
  766. "0.5192052125930786",
  767. "0.41924676299095154",
  768. "0.33786702156066895",
  769. "0.27139899134635925",
  770. "0.21766313910484314",
  771. "0.17579089105129242",
  772. "0.14475053548812866",
  773. "0.12243179231882095",
  774. "0.10634990781545639",
  775. "0.0944439247250557",
  776. "0.08532019704580307",
  777. "0.07809687405824661",
  778. "0.07220908999443054",
  779. "0.06729564070701599",
  780. "0.06311818957328796",
  781. "0.05951143428683281",
  782. "0.05635545030236244",
  783. "0.05356447398662567",
  784. "0.051073770970106125",
  785. "0.04883363097906113",
  786. "0.04680565744638443",
  787. "0.04495749622583389",
  788. "0.043265290558338165",
  789. "0.041708823293447495",
  790. "0.04027065634727478",
  791. "0.03893772140145302",
  792. "0.037698108702898026",
  793. "0.03654063865542412",
  794. "0.03545815870165825",
  795. "0.03444304317235947",
  796. "0.03348816931247711",
  797. "0.03258833661675453",
  798. "0.03173903375864029",
  799. "0.030935637652873993",
  800. "0.03017423115670681",
  801. "0.029451580718159676",
  802. "0.02876456454396248",
  803. "0.028110845014452934",
  804. "0.027487512677907944",
  805. "0.026892736554145813",
  806. "0.026324469596147537",
  807. "0.02578059770166874",
  808. "0.025259941816329956",
  809. "0.024760741740465164",
  810. "0.02428179420530796",
  811. "0.02382158301770687",
  812. "0.02337941713631153",
  813. "0.02295416034758091",
  814. "0.02254442311823368",
  815. "0.02214960940182209",
  816. "0.021768951788544655",
  817. "0.021401450037956238",
  818. "0.021046629175543785"
  819. ]
  820. },
  821. {
  822. "top": 2,
  823. "dtype": "float32",
  824. "num_local_experts": 2,
  825. "step_time": [
  826. {
  827. "GPU": "V100",
  828. "value": "0.1362759590148926"
  829. },
  830. {
  831. "GPU": "RTX 2080 Ti",
  832. "value": "0.14004385471343994"
  833. }
  834. ],
  835. "losses": [
  836. "24.08013153076172",
  837. "23.29282569885254",
  838. "22.534143447875977",
  839. "21.803869247436523",
  840. "21.100021362304688",
  841. "20.419719696044922",
  842. "19.76018524169922",
  843. "19.118858337402344",
  844. "18.493772506713867",
  845. "17.883146286010742",
  846. "17.285343170166016",
  847. "16.69870376586914",
  848. "16.121618270874023",
  849. "15.552938461303711",
  850. "14.991884231567383",
  851. "14.438118934631348",
  852. "13.891477584838867",
  853. "13.352250099182129",
  854. "12.820815086364746",
  855. "12.297611236572266",
  856. "11.782981872558594",
  857. "11.277116775512695",
  858. "10.780354499816895",
  859. "10.293059349060059",
  860. "9.815596580505371",
  861. "9.3479642868042",
  862. "8.89001750946045",
  863. "8.441652297973633",
  864. "8.003580093383789",
  865. "7.577324390411377",
  866. "7.164087295532227",
  867. "6.763621807098389",
  868. "6.374527931213379",
  869. "5.99547815322876",
  870. "5.625848770141602",
  871. "5.265934944152832",
  872. "4.916399955749512",
  873. "4.578485488891602",
  874. "4.253542900085449",
  875. "3.942594528198242",
  876. "3.6465139389038086",
  877. "3.3662383556365967",
  878. "3.102998971939087",
  879. "2.85768461227417",
  880. "2.6305179595947266",
  881. "2.4205806255340576",
  882. "2.2259671688079834",
  883. "2.0445773601531982",
  884. "1.8744558095932007",
  885. "1.7139946222305298",
  886. "1.5623281002044678",
  887. "1.4191162586212158",
  888. "1.2844462394714355",
  889. "1.1586267948150635",
  890. "1.042003870010376",
  891. "0.9347238540649414",
  892. "0.8366733193397522",
  893. "0.7474625110626221",
  894. "0.6664537787437439",
  895. "0.5928301215171814",
  896. "0.5258411169052124",
  897. "0.4649595022201538",
  898. "0.4099421203136444",
  899. "0.36069005727767944",
  900. "0.317216157913208",
  901. "0.27947312593460083",
  902. "0.24723058938980103",
  903. "0.2200528085231781",
  904. "0.19733883440494537",
  905. "0.17840774357318878",
  906. "0.16259536147117615",
  907. "0.14931270480155945",
  908. "0.13806448876857758",
  909. "0.12845289707183838",
  910. "0.12016049027442932",
  911. "0.11294268816709518",
  912. "0.10660269856452942",
  913. "0.10099020600318909",
  914. "0.09598501771688461",
  915. "0.09149053692817688",
  916. "0.08743057399988174",
  917. "0.08374322205781937",
  918. "0.08037738502025604",
  919. "0.07729102671146393",
  920. "0.07444953173398972",
  921. "0.07182367891073227",
  922. "0.06938879191875458",
  923. "0.06712391972541809",
  924. "0.06501118838787079",
  925. "0.06303507834672928",
  926. "0.06118202582001686",
  927. "0.059440482407808304",
  928. "0.057800669223070145",
  929. "0.056253232061862946",
  930. "0.05479034036397934",
  931. "0.053405433893203735",
  932. "0.05209173634648323",
  933. "0.05084368586540222",
  934. "0.0496567040681839",
  935. "0.048525672405958176"
  936. ]
  937. },
  938. {
  939. "top": 2,
  940. "dtype": "float64",
  941. "num_local_experts": 2,
  942. "step_time": [
  943. {
  944. "GPU": "RTX 2080 Ti",
  945. "value": "0.22079918384552003"
  946. }
  947. ],
  948. "losses": [
  949. "27.448386457158",
  950. "16.62704052924519",
  951. "7.467060806621459",
  952. "0.5762248991293285",
  953. "0.03547520289993352",
  954. "0.02809477135324243",
  955. "0.023334777984187503",
  956. "0.01999105336568195",
  957. "0.017506607273773383",
  958. "0.015584494110390793",
  959. "0.014050133283992646",
  960. "0.01279578691597016",
  961. "0.011750494163433321",
  962. "0.010865584358183304",
  963. "0.010106437103754732",
  964. "0.009447815468700606",
  965. "0.00887086934648234",
  966. "0.008361138633483211",
  967. "0.007907487467681268",
  968. "0.007501064318002476",
  969. "0.007134805202108995",
  970. "0.006803018882902828",
  971. "0.006501021877127516",
  972. "0.006224931701034769",
  973. "0.005971566554243328",
  974. "0.005738199369502883",
  975. "0.005522536404718329",
  976. "0.005322649718323559",
  977. "0.005136823500241541",
  978. "0.004963649647667566",
  979. "0.0048018655041757435",
  980. "0.004650365255696276",
  981. "0.00450820659642881",
  982. "0.004374549145932308",
  983. "0.004248645308252479",
  984. "0.004129832568878503",
  985. "0.004017526471538513",
  986. "0.003911207942749451",
  987. "0.0038104107024693815",
  988. "0.0037147120981585614",
  989. "0.003623730093472561",
  990. "0.0035371256693723942",
  991. "0.003454591252105671",
  992. "0.0033758445997638946",
  993. "0.0033006257584612714",
  994. "0.003228712022183779",
  995. "0.003159875769871621",
  996. "0.003093934997176204",
  997. "0.003030698416011075",
  998. "0.002970019849231812",
  999. "0.0029117267332478655",
  1000. "0.002855690896558411",
  1001. "0.0028017880653116944",
  1002. "0.0027498890149930807",
  1003. "0.002699889628183793",
  1004. "0.002651684023926449",
  1005. "0.0026051776572020113",
  1006. "0.0025602813567713405",
  1007. "0.002516915773172441",
  1008. "0.002475002498674116",
  1009. "0.00243447091402485",
  1010. "0.0023952475300970813",
  1011. "0.0023572766497906923",
  1012. "0.0023204917654825225",
  1013. "0.0022848482980198643",
  1014. "0.0022502838751454696",
  1015. "0.002216755487522313",
  1016. "0.0021842164309187526",
  1017. "0.002152622699402496",
  1018. "0.0021219338370448183",
  1019. "0.0020921096708784856",
  1020. "0.002063120178773446",
  1021. "0.0020349227169327395",
  1022. "0.0020074897880104432",
  1023. "0.0019807943562292814",
  1024. "0.0019547934133035227",
  1025. "0.0019294749544858955",
  1026. "0.0019048060631555329",
  1027. "0.0018807605424784794",
  1028. "0.0018573171489731286",
  1029. "0.0018344699261072263",
  1030. "0.0018121867734525617",
  1031. "0.0017904364642675878",
  1032. "0.0017692078257977402",
  1033. "0.001748478446787041",
  1034. "0.001728227330401437",
  1035. "0.001708446466119141",
  1036. "0.0016891133134218665",
  1037. "0.0016702130731257574",
  1038. "0.0016517332280291494",
  1039. "0.0016336618647047987",
  1040. "0.0016159794684817157",
  1041. "0.0015986795396382262",
  1042. "0.0015817459579443448",
  1043. "0.0015651691352352721",
  1044. "0.0015489365308224445",
  1045. "0.0015330389692089475",
  1046. "0.0015174668485532716",
  1047. "0.0015022072171170233",
  1048. "0.001487254744157283"
  1049. ]
  1050. }
  1051. ]