diff --git "a/checkpoint-17700/trainer_state.json" "b/checkpoint-17700/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-17700/trainer_state.json" @@ -0,0 +1,21280 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.9979702300405954, + "global_step": 17700, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 0.0, + "loss": 0.0, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 115 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 125 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 135 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 145 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 155 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 0.0, + "loss": 0.0, + "step": 165 + }, + { + "epoch": 0.01, + "learning_rate": 0.00023999999999999998, + "loss": 0.0, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029993232192205736, + "loss": 0.0, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029984772432462914, + "loss": 0.0, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002997631267272009, + "loss": 0.0, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002996785291297727, + "loss": 0.0, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029959393153234444, + "loss": 0.0, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002995093339349162, + "loss": 0.0, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029942473633748796, + "loss": 0.0, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029934013874005974, + "loss": 0.0, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002992555411426315, + "loss": 0.0, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002991709435452033, + "loss": 0.0, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029908634594777504, + "loss": 0.0, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002990017483503468, + "loss": 0.0, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002989171507529186, + "loss": 0.0, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029883255315549034, + "loss": 0.0, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002987479555580621, + "loss": 0.0, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029866335796063385, + "loss": 0.0, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029857876036320564, + "loss": 0.0, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002984941627657774, + "loss": 0.0, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002984095651683492, + "loss": 0.0, + "step": 265 + }, + { + "epoch": 0.02, + "learning_rate": 0.000298324967570921, + "loss": 0.0, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002982403699734927, + "loss": 0.0, + "step": 275 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002981557723760645, + "loss": 0.0, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029807117477863624, + "loss": 0.0, + "step": 285 + }, + { + "epoch": 0.02, + "learning_rate": 0.000297986577181208, + "loss": 0.0, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002979019795837798, + "loss": 0.0, + "step": 295 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002978173819863516, + "loss": 0.0, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002977327843889233, + "loss": 0.0, + "step": 305 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002976481867914951, + "loss": 0.0, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002975635891940669, + "loss": 0.0, + "step": 315 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002974789915966386, + "loss": 0.0, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002973943939992104, + "loss": 0.0, + "step": 325 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029730979640178214, + "loss": 0.0, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002972251988043539, + "loss": 0.0, + "step": 335 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002971406012069257, + "loss": 0.0, + "step": 340 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002970560036094975, + "loss": 0.0, + "step": 345 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002969714060120692, + "loss": 0.0, + "step": 350 + }, + { + "epoch": 0.02, + "learning_rate": 0.000296886808414641, + "loss": 0.0, + "step": 355 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029680221081721274, + "loss": 0.0, + "step": 360 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002967176132197845, + "loss": 0.0, + "step": 365 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002966330156223563, + "loss": 0.0, + "step": 370 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002965484180249281, + "loss": 0.0, + "step": 375 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002964638204274998, + "loss": 0.0, + "step": 380 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002963792228300716, + "loss": 0.0, + "step": 385 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002962946252326434, + "loss": 0.0, + "step": 390 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002962100276352151, + "loss": 0.0, + "step": 395 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002961254300377869, + "loss": 0.0, + "step": 400 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029604083244035863, + "loss": 0.0, + "step": 405 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002959562348429304, + "loss": 0.0, + "step": 410 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002958716372455022, + "loss": 0.0, + "step": 415 + }, + { + "epoch": 0.02, + "learning_rate": 0.000295787039648074, + "loss": 0.0, + "step": 420 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029570244205064577, + "loss": 0.0, + "step": 425 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002956178444532175, + "loss": 0.0, + "step": 430 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002955332468557893, + "loss": 0.0, + "step": 435 + }, + { + "epoch": 0.02, + "learning_rate": 0.000295448649258361, + "loss": 0.0, + "step": 440 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002953640516609328, + "loss": 0.0, + "step": 445 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002952794540635046, + "loss": 0.0, + "step": 450 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002951948564660763, + "loss": 0.0, + "step": 455 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002951102588686481, + "loss": 0.0, + "step": 460 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002950256612712199, + "loss": 0.0, + "step": 465 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029494106367379167, + "loss": 0.0, + "step": 470 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002948564660763634, + "loss": 0.0, + "step": 475 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002947718684789352, + "loss": 0.0, + "step": 480 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002946872708815069, + "loss": 0.0, + "step": 485 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002946026732840787, + "loss": 0.0, + "step": 490 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002945180756866505, + "loss": 0.0, + "step": 495 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029443347808922227, + "loss": 0.0, + "step": 500 + }, + { + "epoch": 0.03, + "learning_rate": 0.000294348880491794, + "loss": 0.0, + "step": 505 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002942642828943658, + "loss": 0.0, + "step": 510 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002941796852969375, + "loss": 0.0, + "step": 515 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002940950876995093, + "loss": 0.0, + "step": 520 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002940104901020811, + "loss": 0.0, + "step": 525 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002939258925046528, + "loss": 0.0, + "step": 530 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002938412949072246, + "loss": 0.0, + "step": 535 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002937566973097964, + "loss": 0.0, + "step": 540 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029367209971236817, + "loss": 0.0, + "step": 545 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002935875021149399, + "loss": 0.0, + "step": 550 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002935029045175117, + "loss": 0.0, + "step": 555 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002934183069200834, + "loss": 0.0, + "step": 560 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002933337093226552, + "loss": 0.0, + "step": 565 + }, + { + "epoch": 0.03, + "learning_rate": 0.000293249111725227, + "loss": 0.0, + "step": 570 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029316451412779877, + "loss": 0.0, + "step": 575 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002930799165303705, + "loss": 0.0, + "step": 580 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002929953189329423, + "loss": 0.0, + "step": 585 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029291072133551407, + "loss": 0.0, + "step": 590 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002928261237380858, + "loss": 0.0, + "step": 595 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002927415261406576, + "loss": 0.0, + "step": 600 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029265692854322936, + "loss": 0.0, + "step": 605 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002925723309458011, + "loss": 0.0, + "step": 610 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002924877333483729, + "loss": 0.0, + "step": 615 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029240313575094466, + "loss": 0.0, + "step": 620 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029231853815351645, + "loss": 0.0, + "step": 625 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002922339405560882, + "loss": 0.0, + "step": 630 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029214934295865996, + "loss": 0.0, + "step": 635 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002920647453612317, + "loss": 0.0, + "step": 640 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002919801477638035, + "loss": 0.0, + "step": 645 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029189555016637526, + "loss": 0.0, + "step": 650 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029181095256894705, + "loss": 0.0, + "step": 655 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002917263549715188, + "loss": 0.0, + "step": 660 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029164175737409056, + "loss": 0.0, + "step": 665 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002915571597766623, + "loss": 0.0, + "step": 670 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002914725621792341, + "loss": 0.0, + "step": 675 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029138796458180586, + "loss": 0.0, + "step": 680 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002913033669843776, + "loss": 0.0, + "step": 685 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002912187693869494, + "loss": 0.0, + "step": 690 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029113417178952116, + "loss": 0.0, + "step": 695 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029104957419209295, + "loss": 0.0, + "step": 700 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002909649765946647, + "loss": 0.0, + "step": 705 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029088037899723646, + "loss": 0.0, + "step": 710 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002907957813998082, + "loss": 0.0, + "step": 715 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029071118380238, + "loss": 0.0, + "step": 720 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029062658620495176, + "loss": 0.0, + "step": 725 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029054198860752354, + "loss": 0.0, + "step": 730 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002904573910100953, + "loss": 0.0, + "step": 735 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029037279341266706, + "loss": 0.0, + "step": 740 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029028819581523884, + "loss": 0.0, + "step": 745 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002902035982178106, + "loss": 0.0, + "step": 750 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029011900062038236, + "loss": 0.0, + "step": 755 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002900344030229541, + "loss": 0.0, + "step": 760 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002899498054255259, + "loss": 0.0, + "step": 765 + }, + { + "epoch": 0.04, + "learning_rate": 0.00028986520782809766, + "loss": 0.0, + "step": 770 + }, + { + "epoch": 0.04, + "learning_rate": 0.00028978061023066944, + "loss": 0.0, + "step": 775 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002896960126332412, + "loss": 0.0, + "step": 780 + }, + { + "epoch": 0.04, + "learning_rate": 0.00028961141503581296, + "loss": 0.0, + "step": 785 + }, + { + "epoch": 0.04, + "learning_rate": 0.00028952681743838474, + "loss": 0.0, + "step": 790 + }, + { + "epoch": 0.04, + "learning_rate": 0.00028944221984095647, + "loss": 0.0, + "step": 795 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028935762224352826, + "loss": 0.0, + "step": 800 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028927302464610004, + "loss": 0.0, + "step": 805 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028918842704867177, + "loss": 0.0, + "step": 810 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028910382945124356, + "loss": 0.0, + "step": 815 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028901923185381534, + "loss": 0.0, + "step": 820 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002889346342563871, + "loss": 0.0, + "step": 825 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028885003665895886, + "loss": 0.0, + "step": 830 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002887654390615306, + "loss": 0.0, + "step": 835 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028868084146410237, + "loss": 0.0, + "step": 840 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028859624386667415, + "loss": 0.0, + "step": 845 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028851164626924594, + "loss": 0.0, + "step": 850 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002884270486718177, + "loss": 0.0, + "step": 855 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028834245107438945, + "loss": 0.0, + "step": 860 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028825785347696124, + "loss": 0.0, + "step": 865 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028817325587953297, + "loss": 0.0, + "step": 870 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028808865828210475, + "loss": 0.0, + "step": 875 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028800406068467654, + "loss": 0.0, + "step": 880 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028791946308724827, + "loss": 0.0, + "step": 885 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028783486548982005, + "loss": 0.0, + "step": 890 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028775026789239184, + "loss": 0.0, + "step": 895 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002876656702949636, + "loss": 0.0, + "step": 900 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028758107269753535, + "loss": 0.0, + "step": 905 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028749647510010714, + "loss": 0.0, + "step": 910 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028741187750267887, + "loss": 0.0, + "step": 915 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028732727990525065, + "loss": 0.0, + "step": 920 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028724268230782244, + "loss": 0.0, + "step": 925 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002871580847103942, + "loss": 0.0, + "step": 930 + }, + { + "epoch": 0.05, + "learning_rate": 0.000287073487112966, + "loss": 0.0, + "step": 935 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028698888951553774, + "loss": 0.0, + "step": 940 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002869042919181095, + "loss": 0.0, + "step": 945 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028681969432068125, + "loss": 0.0, + "step": 950 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028673509672325304, + "loss": 0.0, + "step": 955 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002866504991258248, + "loss": 0.0, + "step": 960 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028656590152839655, + "loss": 0.0, + "step": 965 + }, + { + "epoch": 0.05, + "learning_rate": 0.00028648130393096833, + "loss": 0.0, + "step": 970 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002863967063335401, + "loss": 0.0, + "step": 975 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002863121087361119, + "loss": 0.0, + "step": 980 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028622751113868363, + "loss": 0.0, + "step": 985 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002861429135412554, + "loss": 0.0, + "step": 990 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028605831594382715, + "loss": 0.0, + "step": 995 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028597371834639893, + "loss": 0.0, + "step": 1000 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002858891207489707, + "loss": 0.0, + "step": 1005 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002858045231515425, + "loss": 0.0, + "step": 1010 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028571992555411423, + "loss": 0.0, + "step": 1015 + }, + { + "epoch": 0.06, + "learning_rate": 0.000285635327956686, + "loss": 0.0, + "step": 1020 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028555073035925775, + "loss": 0.0, + "step": 1025 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028546613276182953, + "loss": 0.0, + "step": 1030 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002853815351644013, + "loss": 0.0, + "step": 1035 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028529693756697305, + "loss": 0.0, + "step": 1040 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028521233996954483, + "loss": 0.0, + "step": 1045 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002851277423721166, + "loss": 0.0, + "step": 1050 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002850431447746884, + "loss": 0.0, + "step": 1055 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028495854717726013, + "loss": 0.0, + "step": 1060 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002848739495798319, + "loss": 0.0, + "step": 1065 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028478935198240365, + "loss": 0.0, + "step": 1070 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028470475438497543, + "loss": 0.0, + "step": 1075 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002846201567875472, + "loss": 0.0, + "step": 1080 + }, + { + "epoch": 0.06, + "learning_rate": 0.000284535559190119, + "loss": 0.0, + "step": 1085 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028445096159269073, + "loss": 0.0, + "step": 1090 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002843663639952625, + "loss": 0.0, + "step": 1095 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002842817663978343, + "loss": 0.0, + "step": 1100 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028419716880040603, + "loss": 0.0, + "step": 1105 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002841125712029778, + "loss": 0.0, + "step": 1110 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028402797360554954, + "loss": 0.0, + "step": 1115 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028394337600812133, + "loss": 0.0, + "step": 1120 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002838587784106931, + "loss": 0.0, + "step": 1125 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002837741808132649, + "loss": 0.0, + "step": 1130 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002836895832158367, + "loss": 0.0, + "step": 1135 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002836049856184084, + "loss": 0.0, + "step": 1140 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002835203880209802, + "loss": 0.0, + "step": 1145 + }, + { + "epoch": 0.06, + "learning_rate": 0.00028343579042355193, + "loss": 0.0, + "step": 1150 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002833511928261237, + "loss": 0.0, + "step": 1155 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002832665952286955, + "loss": 0.0, + "step": 1160 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002831819976312672, + "loss": 0.0, + "step": 1165 + }, + { + "epoch": 0.07, + "learning_rate": 0.000283097400033839, + "loss": 0.0, + "step": 1170 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002830128024364108, + "loss": 0.0, + "step": 1175 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002829282048389825, + "loss": 0.0, + "step": 1180 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002828436072415543, + "loss": 0.0, + "step": 1185 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028275900964412604, + "loss": 0.0, + "step": 1190 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002826744120466978, + "loss": 0.0, + "step": 1195 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002825898144492696, + "loss": 0.0, + "step": 1200 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002825052168518414, + "loss": 0.0, + "step": 1205 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002824206192544132, + "loss": 0.0, + "step": 1210 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002823360216569849, + "loss": 0.0, + "step": 1215 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002822514240595567, + "loss": 0.0, + "step": 1220 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002821668264621284, + "loss": 0.0, + "step": 1225 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002820822288647002, + "loss": 0.0, + "step": 1230 + }, + { + "epoch": 0.07, + "learning_rate": 0.000281997631267272, + "loss": 0.0, + "step": 1235 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002819130336698437, + "loss": 0.0, + "step": 1240 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002818284360724155, + "loss": 0.0, + "step": 1245 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002817438384749873, + "loss": 0.0, + "step": 1250 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002816592408775591, + "loss": 0.0, + "step": 1255 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002815746432801308, + "loss": 0.0, + "step": 1260 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002814900456827026, + "loss": 0.0, + "step": 1265 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002814054480852743, + "loss": 0.0, + "step": 1270 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002813208504878461, + "loss": 0.0, + "step": 1275 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002812362528904179, + "loss": 0.0, + "step": 1280 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002811516552929897, + "loss": 0.0, + "step": 1285 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028106705769556146, + "loss": 0.0, + "step": 1290 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002809824600981332, + "loss": 0.0, + "step": 1295 + }, + { + "epoch": 0.07, + "learning_rate": 0.000280897862500705, + "loss": 0.0, + "step": 1300 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002808132649032767, + "loss": 0.0, + "step": 1305 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002807286673058485, + "loss": 0.0, + "step": 1310 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002806440697084203, + "loss": 0.0, + "step": 1315 + }, + { + "epoch": 0.07, + "learning_rate": 0.000280559472110992, + "loss": 0.0, + "step": 1320 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002804748745135638, + "loss": 0.0, + "step": 1325 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002803902769161356, + "loss": 0.0, + "step": 1330 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028030567931870736, + "loss": 0.0, + "step": 1335 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002802210817212791, + "loss": 0.0, + "step": 1340 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002801364841238508, + "loss": 0.0, + "step": 1345 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002800518865264226, + "loss": 0.0, + "step": 1350 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002799672889289944, + "loss": 0.0, + "step": 1355 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002798826913315662, + "loss": 0.0, + "step": 1360 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027979809373413796, + "loss": 0.0, + "step": 1365 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002797134961367097, + "loss": 0.0, + "step": 1370 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027962889853928147, + "loss": 0.0, + "step": 1375 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002795443009418532, + "loss": 0.0, + "step": 1380 + }, + { + "epoch": 0.08, + "learning_rate": 0.000279459703344425, + "loss": 0.0, + "step": 1385 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027937510574699677, + "loss": 0.0, + "step": 1390 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002792905081495685, + "loss": 0.0, + "step": 1395 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002792059105521403, + "loss": 0.0, + "step": 1400 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027912131295471207, + "loss": 0.0, + "step": 1405 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027903671535728386, + "loss": 0.0, + "step": 1410 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002789521177598556, + "loss": 0.0, + "step": 1415 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027886752016242737, + "loss": 0.0, + "step": 1420 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002787829225649991, + "loss": 0.0, + "step": 1425 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002786983249675709, + "loss": 0.0, + "step": 1430 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027861372737014267, + "loss": 0.0, + "step": 1435 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027852912977271445, + "loss": 0.0, + "step": 1440 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002784445321752862, + "loss": 0.0, + "step": 1445 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027835993457785797, + "loss": 0.0, + "step": 1450 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027827533698042975, + "loss": 0.0, + "step": 1455 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002781907393830015, + "loss": 0.0, + "step": 1460 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027810614178557327, + "loss": 0.0, + "step": 1465 + }, + { + "epoch": 0.08, + "learning_rate": 0.000278021544188145, + "loss": 0.0, + "step": 1470 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002779369465907168, + "loss": 0.0, + "step": 1475 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027785234899328857, + "loss": 0.0, + "step": 1480 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027776775139586035, + "loss": 0.0, + "step": 1485 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027768315379843214, + "loss": 0.0, + "step": 1490 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027759855620100387, + "loss": 0.0, + "step": 1495 + }, + { + "epoch": 0.08, + "learning_rate": 0.00027751395860357565, + "loss": 0.0, + "step": 1500 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002774293610061474, + "loss": 0.0, + "step": 1505 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027734476340871917, + "loss": 0.0, + "step": 1510 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027726016581129095, + "loss": 0.0, + "step": 1515 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002771755682138627, + "loss": 0.0, + "step": 1520 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027709097061643447, + "loss": 0.0, + "step": 1525 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027700637301900625, + "loss": 0.0, + "step": 1530 + }, + { + "epoch": 0.09, + "learning_rate": 0.000276921775421578, + "loss": 0.0, + "step": 1535 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027683717782414977, + "loss": 0.0, + "step": 1540 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002767525802267215, + "loss": 0.0, + "step": 1545 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002766679826292933, + "loss": 0.0, + "step": 1550 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027658338503186507, + "loss": 0.0, + "step": 1555 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027649878743443685, + "loss": 0.0, + "step": 1560 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027641418983700863, + "loss": 0.0, + "step": 1565 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027632959223958036, + "loss": 0.0, + "step": 1570 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027624499464215215, + "loss": 0.0, + "step": 1575 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002761603970447239, + "loss": 0.0, + "step": 1580 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027607579944729566, + "loss": 0.0, + "step": 1585 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027599120184986745, + "loss": 0.0, + "step": 1590 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027590660425243923, + "loss": 0.0, + "step": 1595 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027582200665501096, + "loss": 0.0, + "step": 1600 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027573740905758275, + "loss": 0.0, + "step": 1605 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027565281146015453, + "loss": 0.0, + "step": 1610 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027556821386272626, + "loss": 0.0, + "step": 1615 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027548361626529805, + "loss": 0.0, + "step": 1620 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002753990186678698, + "loss": 0.0, + "step": 1625 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027531442107044156, + "loss": 0.0, + "step": 1630 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027522982347301335, + "loss": 0.0, + "step": 1635 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027514522587558513, + "loss": 0.0, + "step": 1640 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002750606282781569, + "loss": 0.0, + "step": 1645 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027497603068072865, + "loss": 0.0, + "step": 1650 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027489143308330043, + "loss": 0.0, + "step": 1655 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027480683548587216, + "loss": 0.0, + "step": 1660 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027472223788844395, + "loss": 0.0, + "step": 1665 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027463764029101573, + "loss": 0.0, + "step": 1670 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027455304269358746, + "loss": 0.0, + "step": 1675 + }, + { + "epoch": 0.09, + "learning_rate": 0.00027446844509615925, + "loss": 0.0, + "step": 1680 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027438384749873103, + "loss": 0.0, + "step": 1685 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027429924990130276, + "loss": 0.0, + "step": 1690 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027421465230387454, + "loss": 0.0, + "step": 1695 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002741300547064463, + "loss": 0.0, + "step": 1700 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027404545710901806, + "loss": 0.0, + "step": 1705 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027396085951158984, + "loss": 0.0, + "step": 1710 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027387626191416163, + "loss": 0.0, + "step": 1715 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002737916643167334, + "loss": 0.0, + "step": 1720 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027370706671930514, + "loss": 0.0, + "step": 1725 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027362246912187693, + "loss": 0.0, + "step": 1730 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027353787152444866, + "loss": 0.0, + "step": 1735 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027345327392702044, + "loss": 0.0, + "step": 1740 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027336867632959223, + "loss": 0.0, + "step": 1745 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027328407873216396, + "loss": 0.0, + "step": 1750 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027319948113473574, + "loss": 0.0, + "step": 1755 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002731148835373075, + "loss": 0.0, + "step": 1760 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002730302859398793, + "loss": 0.0, + "step": 1765 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027294568834245104, + "loss": 0.0, + "step": 1770 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002728610907450228, + "loss": 0.0, + "step": 1775 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027277649314759456, + "loss": 0.0, + "step": 1780 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027269189555016634, + "loss": 0.0, + "step": 1785 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002726072979527381, + "loss": 0.0, + "step": 1790 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002725227003553099, + "loss": 0.0, + "step": 1795 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027243810275788164, + "loss": 0.0, + "step": 1800 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002723535051604534, + "loss": 0.0, + "step": 1805 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002722689075630252, + "loss": 0.0, + "step": 1810 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027218430996559694, + "loss": 0.0, + "step": 1815 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002720997123681687, + "loss": 0.0, + "step": 1820 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027201511477074045, + "loss": 0.0, + "step": 1825 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027193051717331224, + "loss": 0.0, + "step": 1830 + }, + { + "epoch": 0.1, + "learning_rate": 0.000271845919575884, + "loss": 0.0, + "step": 1835 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002717613219784558, + "loss": 0.0, + "step": 1840 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002716767243810276, + "loss": 0.0, + "step": 1845 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002715921267835993, + "loss": 0.0, + "step": 1850 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027150752918617105, + "loss": 0.0, + "step": 1855 + }, + { + "epoch": 0.1, + "learning_rate": 0.00027142293158874284, + "loss": 0.0, + "step": 1860 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002713383339913146, + "loss": 0.0, + "step": 1865 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002712537363938864, + "loss": 0.0, + "step": 1870 + }, + { + "epoch": 0.11, + "learning_rate": 0.00027116913879645814, + "loss": 0.0, + "step": 1875 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002710845411990299, + "loss": 0.0, + "step": 1880 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002709999436016017, + "loss": 0.0, + "step": 1885 + }, + { + "epoch": 0.11, + "learning_rate": 0.00027091534600417344, + "loss": 0.0, + "step": 1890 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002708307484067452, + "loss": 0.0, + "step": 1895 + }, + { + "epoch": 0.11, + "learning_rate": 0.00027074615080931695, + "loss": 0.0, + "step": 1900 + }, + { + "epoch": 0.11, + "learning_rate": 0.00027066155321188874, + "loss": 0.0, + "step": 1905 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002705769556144605, + "loss": 0.0, + "step": 1910 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002704923580170323, + "loss": 0.0, + "step": 1915 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002704077604196041, + "loss": 0.0, + "step": 1920 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002703231628221758, + "loss": 0.0, + "step": 1925 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002702385652247476, + "loss": 0.0, + "step": 1930 + }, + { + "epoch": 0.11, + "learning_rate": 0.00027015396762731933, + "loss": 0.0, + "step": 1935 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002700693700298911, + "loss": 0.0, + "step": 1940 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002699847724324629, + "loss": 0.0, + "step": 1945 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002699001748350347, + "loss": 0.0, + "step": 1950 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002698155772376064, + "loss": 0.0, + "step": 1955 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002697309796401782, + "loss": 0.0, + "step": 1960 + }, + { + "epoch": 0.11, + "learning_rate": 0.00026964638204275, + "loss": 0.0, + "step": 1965 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002695617844453217, + "loss": 0.0, + "step": 1970 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002694771868478935, + "loss": 0.0, + "step": 1975 + }, + { + "epoch": 0.11, + "learning_rate": 0.00026939258925046523, + "loss": 0.0, + "step": 1980 + }, + { + "epoch": 0.11, + "learning_rate": 0.000269307991653037, + "loss": 0.0, + "step": 1985 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002692233940556088, + "loss": 0.0, + "step": 1990 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002691387964581806, + "loss": 0.0, + "step": 1995 + }, + { + "epoch": 0.11, + "learning_rate": 0.00026905419886075237, + "loss": 0.0, + "step": 2000 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002689696012633241, + "loss": 0.0, + "step": 2005 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002688850036658959, + "loss": 0.0, + "step": 2010 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002688004060684676, + "loss": 0.0, + "step": 2015 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002687158084710394, + "loss": 0.0, + "step": 2020 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002686312108736112, + "loss": 0.0, + "step": 2025 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002685466132761829, + "loss": 0.0, + "step": 2030 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002684620156787547, + "loss": 0.0, + "step": 2035 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002683774180813265, + "loss": 0.0, + "step": 2040 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002682928204838982, + "loss": 0.0, + "step": 2045 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026820822288647, + "loss": 0.0, + "step": 2050 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026812362528904173, + "loss": 0.0, + "step": 2055 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002680390276916135, + "loss": 0.0, + "step": 2060 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002679544300941853, + "loss": 0.0, + "step": 2065 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002678698324967571, + "loss": 0.0, + "step": 2070 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026778523489932887, + "loss": 0.0, + "step": 2075 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002677006373019006, + "loss": 0.0, + "step": 2080 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002676160397044724, + "loss": 0.0, + "step": 2085 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002675314421070441, + "loss": 0.0, + "step": 2090 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002674468445096159, + "loss": 0.0, + "step": 2095 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002673622469121877, + "loss": 0.0, + "step": 2100 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002672776493147594, + "loss": 0.0, + "step": 2105 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002671930517173312, + "loss": 0.0, + "step": 2110 + }, + { + "epoch": 0.12, + "learning_rate": 0.000267108454119903, + "loss": 0.0, + "step": 2115 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026702385652247477, + "loss": 0.0, + "step": 2120 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002669392589250465, + "loss": 0.0, + "step": 2125 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002668546613276183, + "loss": 0.0, + "step": 2130 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026677006373019, + "loss": 0.0, + "step": 2135 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002666854661327618, + "loss": 0.0, + "step": 2140 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002666008685353336, + "loss": 0.0, + "step": 2145 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026651627093790537, + "loss": 0.0, + "step": 2150 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002664316733404771, + "loss": 0.0, + "step": 2155 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002663470757430489, + "loss": 0.0, + "step": 2160 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026626247814562066, + "loss": 0.0, + "step": 2165 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002661778805481924, + "loss": 0.0, + "step": 2170 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002660932829507642, + "loss": 0.0, + "step": 2175 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002660086853533359, + "loss": 0.0, + "step": 2180 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002659240877559077, + "loss": 0.0, + "step": 2185 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002658394901584795, + "loss": 0.0, + "step": 2190 + }, + { + "epoch": 0.12, + "learning_rate": 0.00026575489256105126, + "loss": 0.0, + "step": 2195 + }, + { + "epoch": 0.12, + "learning_rate": 0.000265670294963623, + "loss": 0.0, + "step": 2200 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002655856973661948, + "loss": 0.0, + "step": 2205 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002655010997687665, + "loss": 0.0, + "step": 2210 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002654165021713383, + "loss": 0.0, + "step": 2215 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002653319045739101, + "loss": 0.0, + "step": 2220 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026524730697648186, + "loss": 0.0, + "step": 2225 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002651627093790536, + "loss": 0.0, + "step": 2230 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002650781117816254, + "loss": 0.0, + "step": 2235 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026499351418419716, + "loss": 0.0, + "step": 2240 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002649089165867689, + "loss": 0.0, + "step": 2245 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002648243189893407, + "loss": 0.0, + "step": 2250 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026473972139191246, + "loss": 0.0, + "step": 2255 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002646551237944842, + "loss": 0.0, + "step": 2260 + }, + { + "epoch": 0.13, + "learning_rate": 0.000264570526197056, + "loss": 0.0, + "step": 2265 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026448592859962776, + "loss": 0.0, + "step": 2270 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026440133100219955, + "loss": 0.0, + "step": 2275 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002643167334047713, + "loss": 0.0, + "step": 2280 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026423213580734306, + "loss": 0.0, + "step": 2285 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002641475382099148, + "loss": 0.0, + "step": 2290 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002640629406124866, + "loss": 0.0, + "step": 2295 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026397834301505836, + "loss": 0.0, + "step": 2300 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026389374541763014, + "loss": 0.0, + "step": 2305 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002638091478202019, + "loss": 0.0, + "step": 2310 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026372455022277366, + "loss": 0.0, + "step": 2315 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026363995262534544, + "loss": 0.0, + "step": 2320 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002635553550279172, + "loss": 0.0, + "step": 2325 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026347075743048896, + "loss": 0.0, + "step": 2330 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002633861598330607, + "loss": 0.0, + "step": 2335 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002633015622356325, + "loss": 0.0, + "step": 2340 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026321696463820426, + "loss": 0.0, + "step": 2345 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026313236704077604, + "loss": 0.0, + "step": 2350 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002630477694433478, + "loss": 0.0, + "step": 2355 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026296317184591956, + "loss": 0.0, + "step": 2360 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002628785742484913, + "loss": 0.0, + "step": 2365 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026279397665106307, + "loss": 0.0, + "step": 2370 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026270937905363486, + "loss": 0.0, + "step": 2375 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026262478145620664, + "loss": 0.0, + "step": 2380 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026254018385877837, + "loss": 0.0, + "step": 2385 + }, + { + "epoch": 0.13, + "learning_rate": 0.00026245558626135016, + "loss": 0.0, + "step": 2390 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026237098866392194, + "loss": 0.0, + "step": 2395 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026228639106649367, + "loss": 0.0, + "step": 2400 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026220179346906546, + "loss": 0.0, + "step": 2405 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002621171958716372, + "loss": 0.0, + "step": 2410 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026203259827420897, + "loss": 0.0, + "step": 2415 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026194800067678075, + "loss": 0.0, + "step": 2420 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026186340307935254, + "loss": 0.0, + "step": 2425 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002617788054819243, + "loss": 0.0, + "step": 2430 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026169420788449605, + "loss": 0.0, + "step": 2435 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026160961028706784, + "loss": 0.0, + "step": 2440 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026152501268963957, + "loss": 0.0, + "step": 2445 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026144041509221135, + "loss": 0.0, + "step": 2450 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026135581749478314, + "loss": 0.0, + "step": 2455 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026127121989735487, + "loss": 0.0, + "step": 2460 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026118662229992665, + "loss": 0.0, + "step": 2465 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026110202470249844, + "loss": 0.0, + "step": 2470 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002610174271050702, + "loss": 0.0, + "step": 2475 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026093282950764195, + "loss": 0.0, + "step": 2480 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026084823191021374, + "loss": 0.0, + "step": 2485 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026076363431278547, + "loss": 0.0, + "step": 2490 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026067903671535725, + "loss": 0.0, + "step": 2495 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026059443911792904, + "loss": 0.0, + "step": 2500 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002605098415205008, + "loss": 0.0, + "step": 2505 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026042524392307255, + "loss": 0.0, + "step": 2510 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026034064632564434, + "loss": 0.0, + "step": 2515 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002602560487282161, + "loss": 0.0, + "step": 2520 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026017145113078785, + "loss": 0.0, + "step": 2525 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026008685353335963, + "loss": 0.0, + "step": 2530 + }, + { + "epoch": 0.14, + "learning_rate": 0.00026000225593593137, + "loss": 0.0, + "step": 2535 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025991765833850315, + "loss": 0.0, + "step": 2540 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025983306074107493, + "loss": 0.0, + "step": 2545 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002597484631436467, + "loss": 0.0, + "step": 2550 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025966386554621845, + "loss": 0.0, + "step": 2555 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025957926794879023, + "loss": 0.0, + "step": 2560 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025949467035136196, + "loss": 0.0, + "step": 2565 + }, + { + "epoch": 0.14, + "learning_rate": 0.00025941007275393375, + "loss": 0.0, + "step": 2570 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025932547515650553, + "loss": 0.0, + "step": 2575 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002592408775590773, + "loss": 0.0, + "step": 2580 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002591562799616491, + "loss": 0.0, + "step": 2585 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025907168236422083, + "loss": 0.0, + "step": 2590 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002589870847667926, + "loss": 0.0, + "step": 2595 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025890248716936435, + "loss": 0.0, + "step": 2600 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025881788957193613, + "loss": 0.0, + "step": 2605 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002587332919745079, + "loss": 0.0, + "step": 2610 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025864869437707965, + "loss": 0.0, + "step": 2615 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025856409677965143, + "loss": 0.0, + "step": 2620 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002584794991822232, + "loss": 0.0, + "step": 2625 + }, + { + "epoch": 0.15, + "learning_rate": 0.000258394901584795, + "loss": 0.0, + "step": 2630 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025831030398736673, + "loss": 0.0, + "step": 2635 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002582257063899385, + "loss": 0.0, + "step": 2640 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025814110879251025, + "loss": 0.0, + "step": 2645 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025805651119508203, + "loss": 0.0, + "step": 2650 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002579719135976538, + "loss": 0.0, + "step": 2655 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002578873160002256, + "loss": 0.0, + "step": 2660 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025780271840279733, + "loss": 0.0, + "step": 2665 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002577181208053691, + "loss": 0.0, + "step": 2670 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002576335232079409, + "loss": 0.0, + "step": 2675 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025754892561051263, + "loss": 0.0, + "step": 2680 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002574643280130844, + "loss": 0.0, + "step": 2685 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025737973041565614, + "loss": 0.0, + "step": 2690 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025729513281822793, + "loss": 0.0, + "step": 2695 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002572105352207997, + "loss": 0.0, + "step": 2700 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002571259376233715, + "loss": 0.0, + "step": 2705 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025704134002594323, + "loss": 0.0, + "step": 2710 + }, + { + "epoch": 0.15, + "learning_rate": 0.000256956742428515, + "loss": 0.0, + "step": 2715 + }, + { + "epoch": 0.15, + "learning_rate": 0.00025687214483108674, + "loss": 0.0, + "step": 2720 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002567875472336585, + "loss": 0.0, + "step": 2725 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002567029496362303, + "loss": 0.0, + "step": 2730 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002566183520388021, + "loss": 0.0, + "step": 2735 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002565337544413738, + "loss": 0.0, + "step": 2740 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002564491568439456, + "loss": 0.0, + "step": 2745 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002563645592465174, + "loss": 0.0, + "step": 2750 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002562799616490891, + "loss": 0.0, + "step": 2755 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002561953640516609, + "loss": 0.0, + "step": 2760 + }, + { + "epoch": 0.16, + "learning_rate": 0.00025611076645423264, + "loss": 0.0, + "step": 2765 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002560261688568044, + "loss": 0.0, + "step": 2770 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002559415712593762, + "loss": 0.0, + "step": 2775 + }, + { + "epoch": 0.16, + "learning_rate": 0.000255856973661948, + "loss": 0.0, + "step": 2780 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002557723760645198, + "loss": 0.0, + "step": 2785 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002556877784670915, + "loss": 0.0, + "step": 2790 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002556031808696633, + "loss": 0.0, + "step": 2795 + }, + { + "epoch": 0.16, + "learning_rate": 0.000255518583272235, + "loss": 0.0, + "step": 2800 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002554339856748068, + "loss": 0.0, + "step": 2805 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002553493880773786, + "loss": 0.0, + "step": 2810 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002552647904799503, + "loss": 0.0, + "step": 2815 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002551801928825221, + "loss": 0.0, + "step": 2820 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002550955952850939, + "loss": 0.0, + "step": 2825 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002550109976876657, + "loss": 0.0, + "step": 2830 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002549264000902374, + "loss": 0.0, + "step": 2835 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002548418024928092, + "loss": 0.0, + "step": 2840 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002547572048953809, + "loss": 0.0, + "step": 2845 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002546726072979527, + "loss": 0.0, + "step": 2850 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002545880097005245, + "loss": 0.0, + "step": 2855 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002545034121030963, + "loss": 0.0, + "step": 2860 + }, + { + "epoch": 0.16, + "learning_rate": 0.000254418814505668, + "loss": 0.0, + "step": 2865 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002543342169082398, + "loss": 0.0, + "step": 2870 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002542496193108115, + "loss": 0.0, + "step": 2875 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002541650217133833, + "loss": 0.0, + "step": 2880 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002540804241159551, + "loss": 0.0, + "step": 2885 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002539958265185269, + "loss": 0.0, + "step": 2890 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002539112289210986, + "loss": 0.0, + "step": 2895 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002538266313236704, + "loss": 0.0, + "step": 2900 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002537420337262422, + "loss": 0.0, + "step": 2905 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002536574361288139, + "loss": 0.0, + "step": 2910 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002535728385313857, + "loss": 0.0, + "step": 2915 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002534882409339574, + "loss": 0.0, + "step": 2920 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002534036433365292, + "loss": 0.0, + "step": 2925 + }, + { + "epoch": 0.17, + "learning_rate": 0.000253319045739101, + "loss": 0.0, + "step": 2930 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002532344481416728, + "loss": 0.0, + "step": 2935 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025314985054424456, + "loss": 0.0, + "step": 2940 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002530652529468163, + "loss": 0.0, + "step": 2945 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025298065534938807, + "loss": 0.0, + "step": 2950 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002528960577519598, + "loss": 0.0, + "step": 2955 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002528114601545316, + "loss": 0.0, + "step": 2960 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025272686255710337, + "loss": 0.0, + "step": 2965 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002526422649596751, + "loss": 0.0, + "step": 2970 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002525576673622469, + "loss": 0.0, + "step": 2975 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025247306976481867, + "loss": 0.0, + "step": 2980 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025238847216739046, + "loss": 0.0, + "step": 2985 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002523038745699622, + "loss": 0.0, + "step": 2990 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025221927697253397, + "loss": 0.0, + "step": 2995 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002521346793751057, + "loss": 0.0, + "step": 3000 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002520500817776775, + "loss": 0.0, + "step": 3005 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025196548418024927, + "loss": 0.0, + "step": 3010 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025188088658282105, + "loss": 0.0, + "step": 3015 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002517962889853928, + "loss": 0.0, + "step": 3020 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025171169138796457, + "loss": 0.0, + "step": 3025 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025162709379053635, + "loss": 0.0, + "step": 3030 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002515424961931081, + "loss": 0.0, + "step": 3035 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025145789859567987, + "loss": 0.0, + "step": 3040 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002513733009982516, + "loss": 0.0, + "step": 3045 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002512887034008234, + "loss": 0.0, + "step": 3050 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025120410580339517, + "loss": 0.0, + "step": 3055 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025111950820596695, + "loss": 0.0, + "step": 3060 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002510349106085387, + "loss": 0.0, + "step": 3065 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025095031301111047, + "loss": 0.0, + "step": 3070 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002508657154136822, + "loss": 0.0, + "step": 3075 + }, + { + "epoch": 0.17, + "learning_rate": 0.000250781117816254, + "loss": 0.0, + "step": 3080 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025069652021882577, + "loss": 0.0, + "step": 3085 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025061192262139755, + "loss": 0.0, + "step": 3090 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002505273250239693, + "loss": 0.0, + "step": 3095 + }, + { + "epoch": 0.17, + "learning_rate": 0.00025044272742654107, + "loss": 0.0, + "step": 3100 + }, + { + "epoch": 0.18, + "learning_rate": 0.00025035812982911285, + "loss": 0.0, + "step": 3105 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002502735322316846, + "loss": 0.0, + "step": 3110 + }, + { + "epoch": 0.18, + "learning_rate": 0.00025018893463425637, + "loss": 0.0, + "step": 3115 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002501043370368281, + "loss": 0.0, + "step": 3120 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002500197394393999, + "loss": 0.0, + "step": 3125 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024993514184197167, + "loss": 0.0, + "step": 3130 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024985054424454345, + "loss": 0.0, + "step": 3135 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024976594664711523, + "loss": 0.0, + "step": 3140 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024968134904968696, + "loss": 0.0, + "step": 3145 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024959675145225875, + "loss": 0.0, + "step": 3150 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002495121538548305, + "loss": 0.0, + "step": 3155 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024942755625740226, + "loss": 0.0, + "step": 3160 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024934295865997405, + "loss": 0.0, + "step": 3165 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002492583610625458, + "loss": 0.0, + "step": 3170 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024917376346511756, + "loss": 0.0, + "step": 3175 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024908916586768935, + "loss": 0.0, + "step": 3180 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024900456827026113, + "loss": 0.0, + "step": 3185 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024891997067283286, + "loss": 0.0, + "step": 3190 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024883537307540465, + "loss": 0.0, + "step": 3195 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002487507754779764, + "loss": 0.0, + "step": 3200 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024866617788054816, + "loss": 0.0, + "step": 3205 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024858158028311995, + "loss": 0.0, + "step": 3210 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024849698268569173, + "loss": 0.0, + "step": 3215 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024841238508826346, + "loss": 0.0, + "step": 3220 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024832778749083525, + "loss": 0.0, + "step": 3225 + }, + { + "epoch": 0.18, + "learning_rate": 0.000248243189893407, + "loss": 0.0, + "step": 3230 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024815859229597876, + "loss": 0.0, + "step": 3235 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024807399469855055, + "loss": 0.0, + "step": 3240 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024798939710112233, + "loss": 0.0, + "step": 3245 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024790479950369406, + "loss": 0.0, + "step": 3250 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024782020190626584, + "loss": 0.0, + "step": 3255 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024773560430883763, + "loss": 0.0, + "step": 3260 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024765100671140936, + "loss": 0.0, + "step": 3265 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024756640911398114, + "loss": 0.0, + "step": 3270 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002474818115165529, + "loss": 0.0, + "step": 3275 + }, + { + "epoch": 0.18, + "learning_rate": 0.00024739721391912466, + "loss": 0.0, + "step": 3280 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024731261632169644, + "loss": 0.0, + "step": 3285 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024722801872426823, + "loss": 0.0, + "step": 3290 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024714342112684, + "loss": 0.0, + "step": 3295 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024705882352941174, + "loss": 0.0, + "step": 3300 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024697422593198353, + "loss": 0.0, + "step": 3305 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024688962833455526, + "loss": 0.0, + "step": 3310 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024680503073712704, + "loss": 0.0, + "step": 3315 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002467204331396988, + "loss": 0.0, + "step": 3320 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024663583554227056, + "loss": 0.0, + "step": 3325 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024655123794484234, + "loss": 0.0, + "step": 3330 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002464666403474141, + "loss": 0.0, + "step": 3335 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002463820427499859, + "loss": 0.0, + "step": 3340 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024629744515255764, + "loss": 0.0, + "step": 3345 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002462128475551294, + "loss": 0.0, + "step": 3350 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024612824995770116, + "loss": 0.0, + "step": 3355 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024604365236027294, + "loss": 0.0, + "step": 3360 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002459590547628447, + "loss": 0.0, + "step": 3365 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002458744571654165, + "loss": 0.0, + "step": 3370 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024578985956798824, + "loss": 0.0, + "step": 3375 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024570526197056, + "loss": 0.0, + "step": 3380 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024562066437313175, + "loss": 0.0, + "step": 3385 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024553606677570354, + "loss": 0.0, + "step": 3390 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002454514691782753, + "loss": 0.0, + "step": 3395 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024536687158084705, + "loss": 0.0, + "step": 3400 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024528227398341884, + "loss": 0.0, + "step": 3405 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002451976763859906, + "loss": 0.0, + "step": 3410 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002451130787885624, + "loss": 0.0, + "step": 3415 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024502848119113414, + "loss": 0.0, + "step": 3420 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002449438835937059, + "loss": 0.0, + "step": 3425 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024485928599627765, + "loss": 0.0, + "step": 3430 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024477468839884944, + "loss": 0.0, + "step": 3435 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002446900908014212, + "loss": 0.0, + "step": 3440 + }, + { + "epoch": 0.19, + "learning_rate": 0.000244605493203993, + "loss": 0.0, + "step": 3445 + }, + { + "epoch": 0.19, + "learning_rate": 0.00024452089560656474, + "loss": 0.0, + "step": 3450 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002444362980091365, + "loss": 0.0, + "step": 3455 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002443517004117083, + "loss": 0.0, + "step": 3460 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024426710281428004, + "loss": 0.0, + "step": 3465 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002441825052168518, + "loss": 0.0, + "step": 3470 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024409790761942358, + "loss": 0.0, + "step": 3475 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024401331002199534, + "loss": 0.0, + "step": 3480 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024392871242456712, + "loss": 0.0, + "step": 3485 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002438441148271389, + "loss": 0.0, + "step": 3490 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024375951722971066, + "loss": 0.0, + "step": 3495 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002436749196322824, + "loss": 0.0, + "step": 3500 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024359032203485418, + "loss": 0.0, + "step": 3505 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024350572443742596, + "loss": 0.0, + "step": 3510 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024342112683999772, + "loss": 0.0, + "step": 3515 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002433365292425695, + "loss": 0.0, + "step": 3520 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024325193164514123, + "loss": 0.0, + "step": 3525 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024316733404771302, + "loss": 0.0, + "step": 3530 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024308273645028478, + "loss": 0.0, + "step": 3535 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024299813885285656, + "loss": 0.0, + "step": 3540 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024291354125542835, + "loss": 0.0, + "step": 3545 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002428289436580001, + "loss": 0.0, + "step": 3550 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024274434606057186, + "loss": 0.0, + "step": 3555 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024265974846314362, + "loss": 0.0, + "step": 3560 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002425751508657154, + "loss": 0.0, + "step": 3565 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024249055326828716, + "loss": 0.0, + "step": 3570 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024240595567085894, + "loss": 0.0, + "step": 3575 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024232135807343067, + "loss": 0.0, + "step": 3580 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024223676047600246, + "loss": 0.0, + "step": 3585 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024215216287857422, + "loss": 0.0, + "step": 3590 + }, + { + "epoch": 0.2, + "learning_rate": 0.000242067565281146, + "loss": 0.0, + "step": 3595 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024198296768371779, + "loss": 0.0, + "step": 3600 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024189837008628952, + "loss": 0.0, + "step": 3605 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002418137724888613, + "loss": 0.0, + "step": 3610 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024172917489143306, + "loss": 0.0, + "step": 3615 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024164457729400484, + "loss": 0.0, + "step": 3620 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002415599796965766, + "loss": 0.0, + "step": 3625 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024147538209914836, + "loss": 0.0, + "step": 3630 + }, + { + "epoch": 0.2, + "learning_rate": 0.00024139078450172011, + "loss": 0.0, + "step": 3635 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002413061869042919, + "loss": 0.0, + "step": 3640 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024122158930686368, + "loss": 0.0, + "step": 3645 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024113699170943544, + "loss": 0.0, + "step": 3650 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002410523941120072, + "loss": 0.0, + "step": 3655 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024096779651457896, + "loss": 0.0, + "step": 3660 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024088319891715074, + "loss": 0.0, + "step": 3665 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002407986013197225, + "loss": 0.0, + "step": 3670 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024071400372229428, + "loss": 0.0, + "step": 3675 + }, + { + "epoch": 0.21, + "learning_rate": 0.000240629406124866, + "loss": 0.0, + "step": 3680 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002405448085274378, + "loss": 0.0, + "step": 3685 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024046021093000955, + "loss": 0.0, + "step": 3690 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024037561333258134, + "loss": 0.0, + "step": 3695 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024029101573515312, + "loss": 0.0, + "step": 3700 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024020641813772485, + "loss": 0.0, + "step": 3705 + }, + { + "epoch": 0.21, + "learning_rate": 0.00024012182054029664, + "loss": 0.0, + "step": 3710 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002400372229428684, + "loss": 0.0, + "step": 3715 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023995262534544018, + "loss": 0.0, + "step": 3720 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023986802774801194, + "loss": 0.0, + "step": 3725 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002397834301505837, + "loss": 0.0, + "step": 3730 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023969883255315545, + "loss": 0.0, + "step": 3735 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023961423495572724, + "loss": 0.0, + "step": 3740 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023952963735829902, + "loss": 0.0, + "step": 3745 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023944503976087078, + "loss": 0.0, + "step": 3750 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002393604421634425, + "loss": 0.0, + "step": 3755 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002392758445660143, + "loss": 0.0, + "step": 3760 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023919124696858608, + "loss": 0.0, + "step": 3765 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023910664937115784, + "loss": 0.0, + "step": 3770 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023902205177372962, + "loss": 0.0, + "step": 3775 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023893745417630135, + "loss": 0.0, + "step": 3780 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023885285657887314, + "loss": 0.0, + "step": 3785 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002387682589814449, + "loss": 0.0, + "step": 3790 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023868366138401668, + "loss": 0.0, + "step": 3795 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023859906378658846, + "loss": 0.0, + "step": 3800 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002385144661891602, + "loss": 0.0, + "step": 3805 + }, + { + "epoch": 0.21, + "learning_rate": 0.00023842986859173198, + "loss": 0.0, + "step": 3810 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023834527099430373, + "loss": 0.0, + "step": 3815 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023826067339687552, + "loss": 0.0, + "step": 3820 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023817607579944728, + "loss": 0.0, + "step": 3825 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023809147820201903, + "loss": 0.0, + "step": 3830 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002380068806045908, + "loss": 0.0, + "step": 3835 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023792228300716258, + "loss": 0.0, + "step": 3840 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023783768540973433, + "loss": 0.0, + "step": 3845 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023775308781230612, + "loss": 0.0, + "step": 3850 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023766849021487785, + "loss": 0.0, + "step": 3855 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023758389261744963, + "loss": 0.0, + "step": 3860 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023749929502002142, + "loss": 0.0, + "step": 3865 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023741469742259317, + "loss": 0.0, + "step": 3870 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023733009982516496, + "loss": 0.0, + "step": 3875 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023724550222773672, + "loss": 0.0, + "step": 3880 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023716090463030847, + "loss": 0.0, + "step": 3885 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023707630703288023, + "loss": 0.0, + "step": 3890 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023699170943545202, + "loss": 0.0, + "step": 3895 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002369071118380238, + "loss": 0.0, + "step": 3900 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023682251424059556, + "loss": 0.0, + "step": 3905 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023673791664316732, + "loss": 0.0, + "step": 3910 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023665331904573907, + "loss": 0.0, + "step": 3915 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023656872144831086, + "loss": 0.0, + "step": 3920 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023648412385088261, + "loss": 0.0, + "step": 3925 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002363995262534544, + "loss": 0.0, + "step": 3930 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023631492865602613, + "loss": 0.0, + "step": 3935 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023623033105859791, + "loss": 0.0, + "step": 3940 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023614573346116967, + "loss": 0.0, + "step": 3945 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023606113586374146, + "loss": 0.0, + "step": 3950 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023597653826631324, + "loss": 0.0, + "step": 3955 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023589194066888497, + "loss": 0.0, + "step": 3960 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023580734307145676, + "loss": 0.0, + "step": 3965 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002357227454740285, + "loss": 0.0, + "step": 3970 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002356381478766003, + "loss": 0.0, + "step": 3975 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023555355027917205, + "loss": 0.0, + "step": 3980 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002354689526817438, + "loss": 0.0, + "step": 3985 + }, + { + "epoch": 0.22, + "learning_rate": 0.00023538435508431557, + "loss": 0.0, + "step": 3990 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023529975748688735, + "loss": 0.0, + "step": 3995 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023521515988945914, + "loss": 0.0, + "step": 4000 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002351305622920309, + "loss": 0.0, + "step": 4005 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023504596469460263, + "loss": 0.0, + "step": 4010 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002349613670971744, + "loss": 0.0, + "step": 4015 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002348767694997462, + "loss": 0.0, + "step": 4020 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023479217190231795, + "loss": 0.0, + "step": 4025 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023470757430488974, + "loss": 0.0, + "step": 4030 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023462297670746147, + "loss": 0.0, + "step": 4035 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023453837911003325, + "loss": 0.0, + "step": 4040 + }, + { + "epoch": 0.23, + "learning_rate": 0.000234453781512605, + "loss": 0.0, + "step": 4045 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002343691839151768, + "loss": 0.0, + "step": 4050 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023428458631774858, + "loss": 0.0, + "step": 4055 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002341999887203203, + "loss": 0.0, + "step": 4060 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002341153911228921, + "loss": 0.0, + "step": 4065 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023403079352546385, + "loss": 0.0, + "step": 4070 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023394619592803564, + "loss": 0.0, + "step": 4075 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002338615983306074, + "loss": 0.0, + "step": 4080 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023377700073317915, + "loss": 0.0, + "step": 4085 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002336924031357509, + "loss": 0.0, + "step": 4090 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002336078055383227, + "loss": 0.0, + "step": 4095 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023352320794089445, + "loss": 0.0, + "step": 4100 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023343861034346623, + "loss": 0.0, + "step": 4105 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023335401274603796, + "loss": 0.0, + "step": 4110 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023326941514860975, + "loss": 0.0, + "step": 4115 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023318481755118153, + "loss": 0.0, + "step": 4120 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002331002199537533, + "loss": 0.0, + "step": 4125 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023301562235632508, + "loss": 0.0, + "step": 4130 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002329310247588968, + "loss": 0.0, + "step": 4135 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002328464271614686, + "loss": 0.0, + "step": 4140 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023276182956404035, + "loss": 0.0, + "step": 4145 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023267723196661213, + "loss": 0.0, + "step": 4150 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023259263436918392, + "loss": 0.0, + "step": 4155 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023250803677175565, + "loss": 0.0, + "step": 4160 + }, + { + "epoch": 0.23, + "learning_rate": 0.00023242343917432743, + "loss": 0.0, + "step": 4165 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002323388415768992, + "loss": 0.0, + "step": 4170 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023225424397947097, + "loss": 0.0, + "step": 4175 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023216964638204273, + "loss": 0.0, + "step": 4180 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002320850487846145, + "loss": 0.0, + "step": 4185 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023200045118718625, + "loss": 0.0, + "step": 4190 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023191585358975803, + "loss": 0.0, + "step": 4195 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002318312559923298, + "loss": 0.0, + "step": 4200 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023174665839490157, + "loss": 0.0, + "step": 4205 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023166206079747336, + "loss": 0.0, + "step": 4210 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002315774632000451, + "loss": 0.0, + "step": 4215 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023149286560261687, + "loss": 0.0, + "step": 4220 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023140826800518863, + "loss": 0.0, + "step": 4225 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023132367040776041, + "loss": 0.0, + "step": 4230 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023123907281033217, + "loss": 0.0, + "step": 4235 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023115447521290393, + "loss": 0.0, + "step": 4240 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002310698776154757, + "loss": 0.0, + "step": 4245 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023098528001804747, + "loss": 0.0, + "step": 4250 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023090068242061926, + "loss": 0.0, + "step": 4255 + }, + { + "epoch": 0.24, + "learning_rate": 0.000230816084823191, + "loss": 0.0, + "step": 4260 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023073148722576274, + "loss": 0.0, + "step": 4265 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023064688962833453, + "loss": 0.0, + "step": 4270 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002305622920309063, + "loss": 0.0, + "step": 4275 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023047769443347807, + "loss": 0.0, + "step": 4280 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023039309683604985, + "loss": 0.0, + "step": 4285 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023030849923862158, + "loss": 0.0, + "step": 4290 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023022390164119337, + "loss": 0.0, + "step": 4295 + }, + { + "epoch": 0.24, + "learning_rate": 0.00023013930404376513, + "loss": 0.0, + "step": 4300 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002300547064463369, + "loss": 0.0, + "step": 4305 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002299701088489087, + "loss": 0.0, + "step": 4310 + }, + { + "epoch": 0.24, + "learning_rate": 0.00022988551125148043, + "loss": 0.0, + "step": 4315 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002298009136540522, + "loss": 0.0, + "step": 4320 + }, + { + "epoch": 0.24, + "learning_rate": 0.00022971631605662397, + "loss": 0.0, + "step": 4325 + }, + { + "epoch": 0.24, + "learning_rate": 0.00022963171845919575, + "loss": 0.0, + "step": 4330 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002295471208617675, + "loss": 0.0, + "step": 4335 + }, + { + "epoch": 0.24, + "learning_rate": 0.00022946252326433927, + "loss": 0.0, + "step": 4340 + }, + { + "epoch": 0.24, + "learning_rate": 0.00022937792566691102, + "loss": 0.0, + "step": 4345 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002292933280694828, + "loss": 0.0, + "step": 4350 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022920873047205457, + "loss": 0.0, + "step": 4355 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022912413287462635, + "loss": 0.0, + "step": 4360 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022903953527719808, + "loss": 0.0, + "step": 4365 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022895493767976987, + "loss": 0.0, + "step": 4370 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022887034008234165, + "loss": 0.0, + "step": 4375 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002287857424849134, + "loss": 0.0, + "step": 4380 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002287011448874852, + "loss": 0.0, + "step": 4385 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022861654729005692, + "loss": 0.0, + "step": 4390 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002285319496926287, + "loss": 0.0, + "step": 4395 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022844735209520047, + "loss": 0.0, + "step": 4400 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022836275449777225, + "loss": 0.0, + "step": 4405 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022827815690034403, + "loss": 0.0, + "step": 4410 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022819355930291576, + "loss": 0.0, + "step": 4415 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022810896170548755, + "loss": 0.0, + "step": 4420 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002280243641080593, + "loss": 0.0, + "step": 4425 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002279397665106311, + "loss": 0.0, + "step": 4430 + }, + { + "epoch": 0.25, + "eval_loss": NaN, + "eval_runtime": 4258.2924, + "eval_samples_per_second": 2.083, + "eval_steps_per_second": 0.26, + "step": 4434 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022785516891320285, + "loss": 0.0, + "step": 4435 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002277705713157746, + "loss": 0.0, + "step": 4440 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022768597371834636, + "loss": 0.0, + "step": 4445 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022760137612091815, + "loss": 0.0, + "step": 4450 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002275167785234899, + "loss": 0.0, + "step": 4455 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002274321809260617, + "loss": 0.0, + "step": 4460 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022734758332863342, + "loss": 0.0, + "step": 4465 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002272629857312052, + "loss": 0.0, + "step": 4470 + }, + { + "epoch": 0.25, + "learning_rate": 0.000227178388133777, + "loss": 0.0, + "step": 4475 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022709379053634875, + "loss": 0.0, + "step": 4480 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022700919293892053, + "loss": 0.0, + "step": 4485 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022692459534149226, + "loss": 0.0, + "step": 4490 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022683999774406405, + "loss": 0.0, + "step": 4495 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002267554001466358, + "loss": 0.0, + "step": 4500 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002266708025492076, + "loss": 0.0, + "step": 4505 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022658620495177937, + "loss": 0.0, + "step": 4510 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002265016073543511, + "loss": 0.0, + "step": 4515 + }, + { + "epoch": 0.25, + "learning_rate": 0.00022641700975692286, + "loss": 0.0, + "step": 4520 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022633241215949464, + "loss": 0.0, + "step": 4525 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022624781456206643, + "loss": 0.0, + "step": 4530 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002261632169646382, + "loss": 0.0, + "step": 4535 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022607861936720997, + "loss": 0.0, + "step": 4540 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002259940217697817, + "loss": 0.0, + "step": 4545 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022590942417235349, + "loss": 0.0, + "step": 4550 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022582482657492524, + "loss": 0.0, + "step": 4555 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022574022897749703, + "loss": 0.0, + "step": 4560 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002256556313800688, + "loss": 0.0, + "step": 4565 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022557103378264054, + "loss": 0.0, + "step": 4570 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022548643618521233, + "loss": 0.0, + "step": 4575 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022540183858778409, + "loss": 0.0, + "step": 4580 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022531724099035587, + "loss": 0.0, + "step": 4585 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022523264339292763, + "loss": 0.0, + "step": 4590 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022514804579549938, + "loss": 0.0, + "step": 4595 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022506344819807114, + "loss": 0.0, + "step": 4600 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022497885060064293, + "loss": 0.0, + "step": 4605 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022489425300321468, + "loss": 0.0, + "step": 4610 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022480965540578647, + "loss": 0.0, + "step": 4615 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002247250578083582, + "loss": 0.0, + "step": 4620 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022464046021092998, + "loss": 0.0, + "step": 4625 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022455586261350177, + "loss": 0.0, + "step": 4630 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022447126501607353, + "loss": 0.0, + "step": 4635 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002243866674186453, + "loss": 0.0, + "step": 4640 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022430206982121704, + "loss": 0.0, + "step": 4645 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022421747222378882, + "loss": 0.0, + "step": 4650 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022413287462636058, + "loss": 0.0, + "step": 4655 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022404827702893237, + "loss": 0.0, + "step": 4660 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022396367943150415, + "loss": 0.0, + "step": 4665 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022387908183407588, + "loss": 0.0, + "step": 4670 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022379448423664767, + "loss": 0.0, + "step": 4675 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022370988663921942, + "loss": 0.0, + "step": 4680 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002236252890417912, + "loss": 0.0, + "step": 4685 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022354069144436297, + "loss": 0.0, + "step": 4690 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022345609384693472, + "loss": 0.0, + "step": 4695 + }, + { + "epoch": 0.26, + "learning_rate": 0.00022337149624950648, + "loss": 0.0, + "step": 4700 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022328689865207826, + "loss": 0.0, + "step": 4705 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022320230105465002, + "loss": 0.0, + "step": 4710 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002231177034572218, + "loss": 0.0, + "step": 4715 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022303310585979354, + "loss": 0.0, + "step": 4720 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022294850826236532, + "loss": 0.0, + "step": 4725 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002228639106649371, + "loss": 0.0, + "step": 4730 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022277931306750886, + "loss": 0.0, + "step": 4735 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022269471547008065, + "loss": 0.0, + "step": 4740 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022261011787265238, + "loss": 0.0, + "step": 4745 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022252552027522416, + "loss": 0.0, + "step": 4750 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022244092267779592, + "loss": 0.0, + "step": 4755 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002223563250803677, + "loss": 0.0, + "step": 4760 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002222717274829395, + "loss": 0.0, + "step": 4765 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022218712988551122, + "loss": 0.0, + "step": 4770 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022210253228808298, + "loss": 0.0, + "step": 4775 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022201793469065476, + "loss": 0.0, + "step": 4780 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022193333709322655, + "loss": 0.0, + "step": 4785 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002218487394957983, + "loss": 0.0, + "step": 4790 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022176414189837006, + "loss": 0.0, + "step": 4795 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022167954430094182, + "loss": 0.0, + "step": 4800 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002215949467035136, + "loss": 0.0, + "step": 4805 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022151034910608536, + "loss": 0.0, + "step": 4810 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022142575150865715, + "loss": 0.0, + "step": 4815 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022134115391122888, + "loss": 0.0, + "step": 4820 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022125655631380066, + "loss": 0.0, + "step": 4825 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022117195871637244, + "loss": 0.0, + "step": 4830 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002210873611189442, + "loss": 0.0, + "step": 4835 + }, + { + "epoch": 0.27, + "learning_rate": 0.000221002763521516, + "loss": 0.0, + "step": 4840 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022091816592408772, + "loss": 0.0, + "step": 4845 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002208335683266595, + "loss": 0.0, + "step": 4850 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022074897072923126, + "loss": 0.0, + "step": 4855 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022066437313180304, + "loss": 0.0, + "step": 4860 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002205797755343748, + "loss": 0.0, + "step": 4865 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022049517793694659, + "loss": 0.0, + "step": 4870 + }, + { + "epoch": 0.27, + "learning_rate": 0.00022041058033951832, + "loss": 0.0, + "step": 4875 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002203259827420901, + "loss": 0.0, + "step": 4880 + }, + { + "epoch": 0.28, + "learning_rate": 0.00022024138514466188, + "loss": 0.0, + "step": 4885 + }, + { + "epoch": 0.28, + "learning_rate": 0.00022015678754723364, + "loss": 0.0, + "step": 4890 + }, + { + "epoch": 0.28, + "learning_rate": 0.00022007218994980543, + "loss": 0.0, + "step": 4895 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021998759235237716, + "loss": 0.0, + "step": 4900 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021990299475494894, + "loss": 0.0, + "step": 4905 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002198183971575207, + "loss": 0.0, + "step": 4910 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021973379956009248, + "loss": 0.0, + "step": 4915 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021964920196266427, + "loss": 0.0, + "step": 4920 + }, + { + "epoch": 0.28, + "learning_rate": 0.000219564604365236, + "loss": 0.0, + "step": 4925 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021948000676780778, + "loss": 0.0, + "step": 4930 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021939540917037954, + "loss": 0.0, + "step": 4935 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021931081157295132, + "loss": 0.0, + "step": 4940 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021922621397552308, + "loss": 0.0, + "step": 4945 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021914161637809484, + "loss": 0.0, + "step": 4950 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002190570187806666, + "loss": 0.0, + "step": 4955 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021897242118323838, + "loss": 0.0, + "step": 4960 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021888782358581014, + "loss": 0.0, + "step": 4965 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021880322598838192, + "loss": 0.0, + "step": 4970 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021871862839095365, + "loss": 0.0, + "step": 4975 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021863403079352544, + "loss": 0.0, + "step": 4980 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021854943319609722, + "loss": 0.0, + "step": 4985 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021846483559866898, + "loss": 0.0, + "step": 4990 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021838023800124077, + "loss": 0.0, + "step": 4995 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002182956404038125, + "loss": 0.0, + "step": 5000 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021821104280638428, + "loss": 0.0, + "step": 5005 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021812644520895604, + "loss": 0.0, + "step": 5010 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021804184761152782, + "loss": 0.0, + "step": 5015 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002179572500140996, + "loss": 0.0, + "step": 5020 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021787265241667134, + "loss": 0.0, + "step": 5025 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002177880548192431, + "loss": 0.0, + "step": 5030 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021770345722181488, + "loss": 0.0, + "step": 5035 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021761885962438666, + "loss": 0.0, + "step": 5040 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021753426202695842, + "loss": 0.0, + "step": 5045 + }, + { + "epoch": 0.28, + "learning_rate": 0.00021744966442953018, + "loss": 0.0, + "step": 5050 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021736506683210194, + "loss": 0.0, + "step": 5055 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021728046923467372, + "loss": 0.0, + "step": 5060 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021719587163724548, + "loss": 0.0, + "step": 5065 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021711127403981726, + "loss": 0.0, + "step": 5070 + }, + { + "epoch": 0.29, + "learning_rate": 0.000217026676442389, + "loss": 0.0, + "step": 5075 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021694207884496078, + "loss": 0.0, + "step": 5080 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021685748124753256, + "loss": 0.0, + "step": 5085 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021677288365010432, + "loss": 0.0, + "step": 5090 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002166882860526761, + "loss": 0.0, + "step": 5095 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021660368845524783, + "loss": 0.0, + "step": 5100 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021651909085781962, + "loss": 0.0, + "step": 5105 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021643449326039138, + "loss": 0.0, + "step": 5110 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021634989566296316, + "loss": 0.0, + "step": 5115 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021626529806553492, + "loss": 0.0, + "step": 5120 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021618070046810668, + "loss": 0.0, + "step": 5125 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021609610287067843, + "loss": 0.0, + "step": 5130 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021601150527325022, + "loss": 0.0, + "step": 5135 + }, + { + "epoch": 0.29, + "learning_rate": 0.000215926907675822, + "loss": 0.0, + "step": 5140 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021584231007839376, + "loss": 0.0, + "step": 5145 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021575771248096552, + "loss": 0.0, + "step": 5150 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021567311488353727, + "loss": 0.0, + "step": 5155 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021558851728610906, + "loss": 0.0, + "step": 5160 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021550391968868082, + "loss": 0.0, + "step": 5165 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002154193220912526, + "loss": 0.0, + "step": 5170 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021533472449382433, + "loss": 0.0, + "step": 5175 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021525012689639612, + "loss": 0.0, + "step": 5180 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002151655292989679, + "loss": 0.0, + "step": 5185 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021508093170153966, + "loss": 0.0, + "step": 5190 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021499633410411144, + "loss": 0.0, + "step": 5195 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002149117365066832, + "loss": 0.0, + "step": 5200 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021482713890925496, + "loss": 0.0, + "step": 5205 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021474254131182671, + "loss": 0.0, + "step": 5210 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002146579437143985, + "loss": 0.0, + "step": 5215 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021457334611697026, + "loss": 0.0, + "step": 5220 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021448874851954204, + "loss": 0.0, + "step": 5225 + }, + { + "epoch": 0.29, + "learning_rate": 0.00021440415092211377, + "loss": 0.0, + "step": 5230 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021431955332468556, + "loss": 0.0, + "step": 5235 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021423495572725734, + "loss": 0.0, + "step": 5240 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002141503581298291, + "loss": 0.0, + "step": 5245 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021406576053240088, + "loss": 0.0, + "step": 5250 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002139811629349726, + "loss": 0.0, + "step": 5255 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002138965653375444, + "loss": 0.0, + "step": 5260 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021381196774011615, + "loss": 0.0, + "step": 5265 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021372737014268794, + "loss": 0.0, + "step": 5270 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021364277254525972, + "loss": 0.0, + "step": 5275 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021355817494783145, + "loss": 0.0, + "step": 5280 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002134735773504032, + "loss": 0.0, + "step": 5285 + }, + { + "epoch": 0.3, + "learning_rate": 0.000213388979752975, + "loss": 0.0, + "step": 5290 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021330438215554678, + "loss": 0.0, + "step": 5295 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021321978455811854, + "loss": 0.0, + "step": 5300 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002131351869606903, + "loss": 0.0, + "step": 5305 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021305058936326205, + "loss": 0.0, + "step": 5310 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021296599176583384, + "loss": 0.0, + "step": 5315 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002128813941684056, + "loss": 0.0, + "step": 5320 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021279679657097738, + "loss": 0.0, + "step": 5325 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002127121989735491, + "loss": 0.0, + "step": 5330 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002126276013761209, + "loss": 0.0, + "step": 5335 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021254300377869268, + "loss": 0.0, + "step": 5340 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021245840618126444, + "loss": 0.0, + "step": 5345 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021237380858383622, + "loss": 0.0, + "step": 5350 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021228921098640795, + "loss": 0.0, + "step": 5355 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021220461338897974, + "loss": 0.0, + "step": 5360 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002121200157915515, + "loss": 0.0, + "step": 5365 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021203541819412328, + "loss": 0.0, + "step": 5370 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021195082059669503, + "loss": 0.0, + "step": 5375 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002118662229992668, + "loss": 0.0, + "step": 5380 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021178162540183855, + "loss": 0.0, + "step": 5385 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021169702780441033, + "loss": 0.0, + "step": 5390 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021161243020698212, + "loss": 0.0, + "step": 5395 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021152783260955388, + "loss": 0.0, + "step": 5400 + }, + { + "epoch": 0.3, + "learning_rate": 0.00021144323501212563, + "loss": 0.0, + "step": 5405 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002113586374146974, + "loss": 0.0, + "step": 5410 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021127403981726918, + "loss": 0.0, + "step": 5415 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021118944221984093, + "loss": 0.0, + "step": 5420 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021110484462241272, + "loss": 0.0, + "step": 5425 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021102024702498445, + "loss": 0.0, + "step": 5430 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021093564942755623, + "loss": 0.0, + "step": 5435 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021085105183012802, + "loss": 0.0, + "step": 5440 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021076645423269977, + "loss": 0.0, + "step": 5445 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021068185663527156, + "loss": 0.0, + "step": 5450 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002105972590378433, + "loss": 0.0, + "step": 5455 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021051266144041507, + "loss": 0.0, + "step": 5460 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021042806384298683, + "loss": 0.0, + "step": 5465 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021034346624555862, + "loss": 0.0, + "step": 5470 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021025886864813037, + "loss": 0.0, + "step": 5475 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021017427105070213, + "loss": 0.0, + "step": 5480 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002100896734532739, + "loss": 0.0, + "step": 5485 + }, + { + "epoch": 0.31, + "learning_rate": 0.00021000507585584567, + "loss": 0.0, + "step": 5490 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020992047825841746, + "loss": 0.0, + "step": 5495 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020983588066098921, + "loss": 0.0, + "step": 5500 + }, + { + "epoch": 0.31, + "learning_rate": 0.000209751283063561, + "loss": 0.0, + "step": 5505 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020966668546613273, + "loss": 0.0, + "step": 5510 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020958208786870451, + "loss": 0.0, + "step": 5515 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020949749027127627, + "loss": 0.0, + "step": 5520 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020941289267384806, + "loss": 0.0, + "step": 5525 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020932829507641984, + "loss": 0.0, + "step": 5530 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020924369747899157, + "loss": 0.0, + "step": 5535 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020915909988156333, + "loss": 0.0, + "step": 5540 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002090745022841351, + "loss": 0.0, + "step": 5545 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002089899046867069, + "loss": 0.0, + "step": 5550 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020890530708927865, + "loss": 0.0, + "step": 5555 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002088207094918504, + "loss": 0.0, + "step": 5560 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020873611189442217, + "loss": 0.0, + "step": 5565 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020865151429699395, + "loss": 0.0, + "step": 5570 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002085669166995657, + "loss": 0.0, + "step": 5575 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002084823191021375, + "loss": 0.0, + "step": 5580 + }, + { + "epoch": 0.31, + "learning_rate": 0.00020839772150470923, + "loss": 0.0, + "step": 5585 + }, + { + "epoch": 0.32, + "learning_rate": 0.000208313123907281, + "loss": 0.0, + "step": 5590 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002082285263098528, + "loss": 0.0, + "step": 5595 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020814392871242455, + "loss": 0.0, + "step": 5600 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020805933111499634, + "loss": 0.0, + "step": 5605 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020797473351756807, + "loss": 0.0, + "step": 5610 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020789013592013985, + "loss": 0.0, + "step": 5615 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002078055383227116, + "loss": 0.0, + "step": 5620 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002077209407252834, + "loss": 0.0, + "step": 5625 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020763634312785515, + "loss": 0.0, + "step": 5630 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002075517455304269, + "loss": 0.0, + "step": 5635 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020746714793299867, + "loss": 0.0, + "step": 5640 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020738255033557045, + "loss": 0.0, + "step": 5645 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020729795273814224, + "loss": 0.0, + "step": 5650 + }, + { + "epoch": 0.32, + "learning_rate": 0.000207213355140714, + "loss": 0.0, + "step": 5655 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020712875754328575, + "loss": 0.0, + "step": 5660 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002070441599458575, + "loss": 0.0, + "step": 5665 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002069595623484293, + "loss": 0.0, + "step": 5670 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020687496475100105, + "loss": 0.0, + "step": 5675 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020679036715357283, + "loss": 0.0, + "step": 5680 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020670576955614456, + "loss": 0.0, + "step": 5685 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020662117195871635, + "loss": 0.0, + "step": 5690 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020653657436128813, + "loss": 0.0, + "step": 5695 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002064519767638599, + "loss": 0.0, + "step": 5700 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020636737916643168, + "loss": 0.0, + "step": 5705 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002062827815690034, + "loss": 0.0, + "step": 5710 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002061981839715752, + "loss": 0.0, + "step": 5715 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020611358637414695, + "loss": 0.0, + "step": 5720 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020602898877671873, + "loss": 0.0, + "step": 5725 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002059443911792905, + "loss": 0.0, + "step": 5730 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020585979358186225, + "loss": 0.0, + "step": 5735 + }, + { + "epoch": 0.32, + "learning_rate": 0.000205775195984434, + "loss": 0.0, + "step": 5740 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002056905983870058, + "loss": 0.0, + "step": 5745 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020560600078957757, + "loss": 0.0, + "step": 5750 + }, + { + "epoch": 0.32, + "learning_rate": 0.00020552140319214933, + "loss": 0.0, + "step": 5755 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002054368055947211, + "loss": 0.0, + "step": 5760 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020535220799729285, + "loss": 0.0, + "step": 5765 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020526761039986463, + "loss": 0.0, + "step": 5770 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002051830128024364, + "loss": 0.0, + "step": 5775 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020509841520500817, + "loss": 0.0, + "step": 5780 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002050138176075799, + "loss": 0.0, + "step": 5785 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002049292200101517, + "loss": 0.0, + "step": 5790 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020484462241272344, + "loss": 0.0, + "step": 5795 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020476002481529523, + "loss": 0.0, + "step": 5800 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020467542721786701, + "loss": 0.0, + "step": 5805 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020459082962043874, + "loss": 0.0, + "step": 5810 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020450623202301053, + "loss": 0.0, + "step": 5815 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020442163442558229, + "loss": 0.0, + "step": 5820 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020433703682815407, + "loss": 0.0, + "step": 5825 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020425243923072583, + "loss": 0.0, + "step": 5830 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002041678416332976, + "loss": 0.0, + "step": 5835 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020408324403586934, + "loss": 0.0, + "step": 5840 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020399864643844113, + "loss": 0.0, + "step": 5845 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002039140488410129, + "loss": 0.0, + "step": 5850 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020382945124358467, + "loss": 0.0, + "step": 5855 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020374485364615645, + "loss": 0.0, + "step": 5860 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020366025604872818, + "loss": 0.0, + "step": 5865 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020357565845129997, + "loss": 0.0, + "step": 5870 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020349106085387173, + "loss": 0.0, + "step": 5875 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002034064632564435, + "loss": 0.0, + "step": 5880 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020332186565901527, + "loss": 0.0, + "step": 5885 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020323726806158703, + "loss": 0.0, + "step": 5890 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020315267046415878, + "loss": 0.0, + "step": 5895 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020306807286673057, + "loss": 0.0, + "step": 5900 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020298347526930235, + "loss": 0.0, + "step": 5905 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002028988776718741, + "loss": 0.0, + "step": 5910 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020281428007444587, + "loss": 0.0, + "step": 5915 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020272968247701762, + "loss": 0.0, + "step": 5920 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002026450848795894, + "loss": 0.0, + "step": 5925 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020256048728216117, + "loss": 0.0, + "step": 5930 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020247588968473295, + "loss": 0.0, + "step": 5935 + }, + { + "epoch": 0.33, + "learning_rate": 0.00020239129208730468, + "loss": 0.0, + "step": 5940 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020230669448987647, + "loss": 0.0, + "step": 5945 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020222209689244825, + "loss": 0.0, + "step": 5950 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020213749929502, + "loss": 0.0, + "step": 5955 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002020529016975918, + "loss": 0.0, + "step": 5960 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020196830410016352, + "loss": 0.0, + "step": 5965 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002018837065027353, + "loss": 0.0, + "step": 5970 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020179910890530706, + "loss": 0.0, + "step": 5975 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020171451130787885, + "loss": 0.0, + "step": 5980 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002016299137104506, + "loss": 0.0, + "step": 5985 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020154531611302236, + "loss": 0.0, + "step": 5990 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020146071851559412, + "loss": 0.0, + "step": 5995 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002013761209181659, + "loss": 0.0, + "step": 6000 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002012915233207377, + "loss": 0.0, + "step": 6005 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020120692572330945, + "loss": 0.0, + "step": 6010 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002011223281258812, + "loss": 0.0, + "step": 6015 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020103773052845296, + "loss": 0.0, + "step": 6020 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020095313293102475, + "loss": 0.0, + "step": 6025 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002008685353335965, + "loss": 0.0, + "step": 6030 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002007839377361683, + "loss": 0.0, + "step": 6035 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020069934013874002, + "loss": 0.0, + "step": 6040 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002006147425413118, + "loss": 0.0, + "step": 6045 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020053014494388356, + "loss": 0.0, + "step": 6050 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020044554734645535, + "loss": 0.0, + "step": 6055 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020036094974902713, + "loss": 0.0, + "step": 6060 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020027635215159886, + "loss": 0.0, + "step": 6065 + }, + { + "epoch": 0.34, + "learning_rate": 0.00020019175455417065, + "loss": 0.0, + "step": 6070 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002001071569567424, + "loss": 0.0, + "step": 6075 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002000225593593142, + "loss": 0.0, + "step": 6080 + }, + { + "epoch": 0.34, + "learning_rate": 0.00019993796176188595, + "loss": 0.0, + "step": 6085 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001998533641644577, + "loss": 0.0, + "step": 6090 + }, + { + "epoch": 0.34, + "learning_rate": 0.00019976876656702946, + "loss": 0.0, + "step": 6095 + }, + { + "epoch": 0.34, + "learning_rate": 0.00019968416896960124, + "loss": 0.0, + "step": 6100 + }, + { + "epoch": 0.34, + "learning_rate": 0.00019959957137217303, + "loss": 0.0, + "step": 6105 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001995149737747448, + "loss": 0.0, + "step": 6110 + }, + { + "epoch": 0.34, + "learning_rate": 0.00019943037617731654, + "loss": 0.0, + "step": 6115 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001993457785798883, + "loss": 0.0, + "step": 6120 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019926118098246009, + "loss": 0.0, + "step": 6125 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019917658338503184, + "loss": 0.0, + "step": 6130 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019909198578760363, + "loss": 0.0, + "step": 6135 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019900738819017536, + "loss": 0.0, + "step": 6140 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019892279059274714, + "loss": 0.0, + "step": 6145 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001988381929953189, + "loss": 0.0, + "step": 6150 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019875359539789068, + "loss": 0.0, + "step": 6155 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019866899780046247, + "loss": 0.0, + "step": 6160 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019858440020303423, + "loss": 0.0, + "step": 6165 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019849980260560598, + "loss": 0.0, + "step": 6170 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019841520500817774, + "loss": 0.0, + "step": 6175 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019833060741074953, + "loss": 0.0, + "step": 6180 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019824600981332128, + "loss": 0.0, + "step": 6185 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019816141221589307, + "loss": 0.0, + "step": 6190 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001980768146184648, + "loss": 0.0, + "step": 6195 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019799221702103658, + "loss": 0.0, + "step": 6200 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019790761942360837, + "loss": 0.0, + "step": 6205 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019782302182618012, + "loss": 0.0, + "step": 6210 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001977384242287519, + "loss": 0.0, + "step": 6215 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019765382663132364, + "loss": 0.0, + "step": 6220 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019756922903389542, + "loss": 0.0, + "step": 6225 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019748463143646718, + "loss": 0.0, + "step": 6230 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019740003383903897, + "loss": 0.0, + "step": 6235 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019731543624161072, + "loss": 0.0, + "step": 6240 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019723083864418248, + "loss": 0.0, + "step": 6245 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019714624104675424, + "loss": 0.0, + "step": 6250 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019706164344932602, + "loss": 0.0, + "step": 6255 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001969770458518978, + "loss": 0.0, + "step": 6260 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019689244825446957, + "loss": 0.0, + "step": 6265 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019680785065704132, + "loss": 0.0, + "step": 6270 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019672325305961308, + "loss": 0.0, + "step": 6275 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019663865546218486, + "loss": 0.0, + "step": 6280 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019655405786475662, + "loss": 0.0, + "step": 6285 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001964694602673284, + "loss": 0.0, + "step": 6290 + }, + { + "epoch": 0.35, + "learning_rate": 0.00019638486266990014, + "loss": 0.0, + "step": 6295 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019630026507247192, + "loss": 0.0, + "step": 6300 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019621566747504368, + "loss": 0.0, + "step": 6305 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019613106987761546, + "loss": 0.0, + "step": 6310 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019604647228018725, + "loss": 0.0, + "step": 6315 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019596187468275898, + "loss": 0.0, + "step": 6320 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019587727708533076, + "loss": 0.0, + "step": 6325 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019579267948790252, + "loss": 0.0, + "step": 6330 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001957080818904743, + "loss": 0.0, + "step": 6335 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019562348429304606, + "loss": 0.0, + "step": 6340 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019553888669561782, + "loss": 0.0, + "step": 6345 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019545428909818958, + "loss": 0.0, + "step": 6350 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019536969150076136, + "loss": 0.0, + "step": 6355 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019528509390333315, + "loss": 0.0, + "step": 6360 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001952004963059049, + "loss": 0.0, + "step": 6365 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019511589870847666, + "loss": 0.0, + "step": 6370 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019503130111104842, + "loss": 0.0, + "step": 6375 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001949467035136202, + "loss": 0.0, + "step": 6380 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019486210591619196, + "loss": 0.0, + "step": 6385 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019477750831876374, + "loss": 0.0, + "step": 6390 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019469291072133548, + "loss": 0.0, + "step": 6395 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019460831312390726, + "loss": 0.0, + "step": 6400 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019452371552647902, + "loss": 0.0, + "step": 6405 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001944391179290508, + "loss": 0.0, + "step": 6410 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019435452033162259, + "loss": 0.0, + "step": 6415 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019426992273419432, + "loss": 0.0, + "step": 6420 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001941853251367661, + "loss": 0.0, + "step": 6425 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019410072753933786, + "loss": 0.0, + "step": 6430 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019401612994190964, + "loss": 0.0, + "step": 6435 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001939315323444814, + "loss": 0.0, + "step": 6440 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019384693474705316, + "loss": 0.0, + "step": 6445 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019376233714962492, + "loss": 0.0, + "step": 6450 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001936777395521967, + "loss": 0.0, + "step": 6455 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019359314195476848, + "loss": 0.0, + "step": 6460 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019350854435734024, + "loss": 0.0, + "step": 6465 + }, + { + "epoch": 0.36, + "learning_rate": 0.00019342394675991197, + "loss": 0.0, + "step": 6470 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019333934916248376, + "loss": 0.0, + "step": 6475 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019325475156505554, + "loss": 0.0, + "step": 6480 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001931701539676273, + "loss": 0.0, + "step": 6485 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019308555637019908, + "loss": 0.0, + "step": 6490 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019300095877277084, + "loss": 0.0, + "step": 6495 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001929163611753426, + "loss": 0.0, + "step": 6500 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019283176357791436, + "loss": 0.0, + "step": 6505 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019274716598048614, + "loss": 0.0, + "step": 6510 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019266256838305792, + "loss": 0.0, + "step": 6515 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019257797078562968, + "loss": 0.0, + "step": 6520 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019249337318820144, + "loss": 0.0, + "step": 6525 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001924087755907732, + "loss": 0.0, + "step": 6530 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019232417799334498, + "loss": 0.0, + "step": 6535 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019223958039591674, + "loss": 0.0, + "step": 6540 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019215498279848852, + "loss": 0.0, + "step": 6545 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019207038520106025, + "loss": 0.0, + "step": 6550 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019198578760363204, + "loss": 0.0, + "step": 6555 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001919011900062038, + "loss": 0.0, + "step": 6560 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019181659240877558, + "loss": 0.0, + "step": 6565 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019173199481134736, + "loss": 0.0, + "step": 6570 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001916473972139191, + "loss": 0.0, + "step": 6575 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019156279961649088, + "loss": 0.0, + "step": 6580 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019147820201906264, + "loss": 0.0, + "step": 6585 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019139360442163442, + "loss": 0.0, + "step": 6590 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019130900682420618, + "loss": 0.0, + "step": 6595 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019122440922677794, + "loss": 0.0, + "step": 6600 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001911398116293497, + "loss": 0.0, + "step": 6605 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019105521403192148, + "loss": 0.0, + "step": 6610 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019097061643449326, + "loss": 0.0, + "step": 6615 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019088601883706502, + "loss": 0.0, + "step": 6620 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019080142123963678, + "loss": 0.0, + "step": 6625 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019071682364220854, + "loss": 0.0, + "step": 6630 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019063222604478032, + "loss": 0.0, + "step": 6635 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019054762844735208, + "loss": 0.0, + "step": 6640 + }, + { + "epoch": 0.37, + "learning_rate": 0.00019046303084992386, + "loss": 0.0, + "step": 6645 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001903784332524956, + "loss": 0.0, + "step": 6650 + }, + { + "epoch": 0.38, + "learning_rate": 0.00019029383565506738, + "loss": 0.0, + "step": 6655 + }, + { + "epoch": 0.38, + "learning_rate": 0.00019020923805763913, + "loss": 0.0, + "step": 6660 + }, + { + "epoch": 0.38, + "learning_rate": 0.00019012464046021092, + "loss": 0.0, + "step": 6665 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001900400428627827, + "loss": 0.0, + "step": 6670 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018995544526535443, + "loss": 0.0, + "step": 6675 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018987084766792622, + "loss": 0.0, + "step": 6680 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018978625007049798, + "loss": 0.0, + "step": 6685 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018970165247306976, + "loss": 0.0, + "step": 6690 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018961705487564152, + "loss": 0.0, + "step": 6695 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018953245727821327, + "loss": 0.0, + "step": 6700 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018944785968078503, + "loss": 0.0, + "step": 6705 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018936326208335682, + "loss": 0.0, + "step": 6710 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001892786644859286, + "loss": 0.0, + "step": 6715 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018919406688850036, + "loss": 0.0, + "step": 6720 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001891094692910721, + "loss": 0.0, + "step": 6725 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018902487169364387, + "loss": 0.0, + "step": 6730 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018894027409621566, + "loss": 0.0, + "step": 6735 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018885567649878742, + "loss": 0.0, + "step": 6740 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001887710789013592, + "loss": 0.0, + "step": 6745 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018868648130393093, + "loss": 0.0, + "step": 6750 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018860188370650271, + "loss": 0.0, + "step": 6755 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018851728610907447, + "loss": 0.0, + "step": 6760 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018843268851164626, + "loss": 0.0, + "step": 6765 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018834809091421804, + "loss": 0.0, + "step": 6770 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018826349331678977, + "loss": 0.0, + "step": 6775 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018817889571936156, + "loss": 0.0, + "step": 6780 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018809429812193331, + "loss": 0.0, + "step": 6785 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001880097005245051, + "loss": 0.0, + "step": 6790 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018792510292707686, + "loss": 0.0, + "step": 6795 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001878405053296486, + "loss": 0.0, + "step": 6800 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018775590773222037, + "loss": 0.0, + "step": 6805 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018767131013479216, + "loss": 0.0, + "step": 6810 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001875867125373639, + "loss": 0.0, + "step": 6815 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001875021149399357, + "loss": 0.0, + "step": 6820 + }, + { + "epoch": 0.38, + "learning_rate": 0.00018741751734250748, + "loss": 0.0, + "step": 6825 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001873329197450792, + "loss": 0.0, + "step": 6830 + }, + { + "epoch": 0.39, + "learning_rate": 0.000187248322147651, + "loss": 0.0, + "step": 6835 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018716372455022275, + "loss": 0.0, + "step": 6840 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018707912695279454, + "loss": 0.0, + "step": 6845 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001869945293553663, + "loss": 0.0, + "step": 6850 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018690993175793805, + "loss": 0.0, + "step": 6855 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001868253341605098, + "loss": 0.0, + "step": 6860 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001867407365630816, + "loss": 0.0, + "step": 6865 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018665613896565338, + "loss": 0.0, + "step": 6870 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018657154136822514, + "loss": 0.0, + "step": 6875 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001864869437707969, + "loss": 0.0, + "step": 6880 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018640234617336865, + "loss": 0.0, + "step": 6885 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018631774857594044, + "loss": 0.0, + "step": 6890 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001862331509785122, + "loss": 0.0, + "step": 6895 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018614855338108398, + "loss": 0.0, + "step": 6900 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001860639557836557, + "loss": 0.0, + "step": 6905 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001859793581862275, + "loss": 0.0, + "step": 6910 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018589476058879925, + "loss": 0.0, + "step": 6915 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018581016299137104, + "loss": 0.0, + "step": 6920 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018572556539394282, + "loss": 0.0, + "step": 6925 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018564096779651455, + "loss": 0.0, + "step": 6930 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018555637019908633, + "loss": 0.0, + "step": 6935 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001854717726016581, + "loss": 0.0, + "step": 6940 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018538717500422988, + "loss": 0.0, + "step": 6945 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018530257740680163, + "loss": 0.0, + "step": 6950 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001852179798093734, + "loss": 0.0, + "step": 6955 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018513338221194515, + "loss": 0.0, + "step": 6960 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018504878461451693, + "loss": 0.0, + "step": 6965 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018496418701708872, + "loss": 0.0, + "step": 6970 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018487958941966048, + "loss": 0.0, + "step": 6975 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001847949918222322, + "loss": 0.0, + "step": 6980 + }, + { + "epoch": 0.39, + "learning_rate": 0.000184710394224804, + "loss": 0.0, + "step": 6985 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018462579662737578, + "loss": 0.0, + "step": 6990 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018454119902994753, + "loss": 0.0, + "step": 6995 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018445660143251932, + "loss": 0.0, + "step": 7000 + }, + { + "epoch": 0.39, + "learning_rate": 0.00018437200383509105, + "loss": 0.0, + "step": 7005 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018428740623766283, + "loss": 0.0, + "step": 7010 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001842028086402346, + "loss": 0.0, + "step": 7015 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018411821104280637, + "loss": 0.0, + "step": 7020 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018403361344537816, + "loss": 0.0, + "step": 7025 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001839490158479499, + "loss": 0.0, + "step": 7030 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018386441825052167, + "loss": 0.0, + "step": 7035 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018377982065309343, + "loss": 0.0, + "step": 7040 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018369522305566522, + "loss": 0.0, + "step": 7045 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018361062545823697, + "loss": 0.0, + "step": 7050 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018352602786080873, + "loss": 0.0, + "step": 7055 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001834414302633805, + "loss": 0.0, + "step": 7060 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018335683266595227, + "loss": 0.0, + "step": 7065 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018327223506852403, + "loss": 0.0, + "step": 7070 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018318763747109581, + "loss": 0.0, + "step": 7075 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018310303987366754, + "loss": 0.0, + "step": 7080 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018301844227623933, + "loss": 0.0, + "step": 7085 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001829338446788111, + "loss": 0.0, + "step": 7090 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018284924708138287, + "loss": 0.0, + "step": 7095 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018276464948395466, + "loss": 0.0, + "step": 7100 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018268005188652639, + "loss": 0.0, + "step": 7105 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018259545428909817, + "loss": 0.0, + "step": 7110 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018251085669166993, + "loss": 0.0, + "step": 7115 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001824262590942417, + "loss": 0.0, + "step": 7120 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001823416614968135, + "loss": 0.0, + "step": 7125 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018225706389938523, + "loss": 0.0, + "step": 7130 + }, + { + "epoch": 0.4, + "learning_rate": 0.000182172466301957, + "loss": 0.0, + "step": 7135 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018208786870452877, + "loss": 0.0, + "step": 7140 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018200327110710055, + "loss": 0.0, + "step": 7145 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001819186735096723, + "loss": 0.0, + "step": 7150 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001818340759122441, + "loss": 0.0, + "step": 7155 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018174947831481583, + "loss": 0.0, + "step": 7160 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001816648807173876, + "loss": 0.0, + "step": 7165 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018158028311995937, + "loss": 0.0, + "step": 7170 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018149568552253115, + "loss": 0.0, + "step": 7175 + }, + { + "epoch": 0.4, + "learning_rate": 0.00018141108792510294, + "loss": 0.0, + "step": 7180 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018132649032767467, + "loss": 0.0, + "step": 7185 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018124189273024645, + "loss": 0.0, + "step": 7190 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001811572951328182, + "loss": 0.0, + "step": 7195 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018107269753539, + "loss": 0.0, + "step": 7200 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018098809993796175, + "loss": 0.0, + "step": 7205 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001809035023405335, + "loss": 0.0, + "step": 7210 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018081890474310527, + "loss": 0.0, + "step": 7215 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018073430714567705, + "loss": 0.0, + "step": 7220 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018064970954824884, + "loss": 0.0, + "step": 7225 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001805651119508206, + "loss": 0.0, + "step": 7230 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018048051435339232, + "loss": 0.0, + "step": 7235 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001803959167559641, + "loss": 0.0, + "step": 7240 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001803113191585359, + "loss": 0.0, + "step": 7245 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018022672156110765, + "loss": 0.0, + "step": 7250 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018014212396367943, + "loss": 0.0, + "step": 7255 + }, + { + "epoch": 0.41, + "learning_rate": 0.00018005752636625116, + "loss": 0.0, + "step": 7260 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017997292876882295, + "loss": 0.0, + "step": 7265 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001798883311713947, + "loss": 0.0, + "step": 7270 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001798037335739665, + "loss": 0.0, + "step": 7275 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017971913597653828, + "loss": 0.0, + "step": 7280 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017963453837911, + "loss": 0.0, + "step": 7285 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001795499407816818, + "loss": 0.0, + "step": 7290 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017946534318425355, + "loss": 0.0, + "step": 7295 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017938074558682533, + "loss": 0.0, + "step": 7300 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001792961479893971, + "loss": 0.0, + "step": 7305 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017921155039196885, + "loss": 0.0, + "step": 7310 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001791269527945406, + "loss": 0.0, + "step": 7315 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001790423551971124, + "loss": 0.0, + "step": 7320 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017895775759968415, + "loss": 0.0, + "step": 7325 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017887316000225593, + "loss": 0.0, + "step": 7330 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017878856240482766, + "loss": 0.0, + "step": 7335 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017870396480739945, + "loss": 0.0, + "step": 7340 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017861936720997123, + "loss": 0.0, + "step": 7345 + }, + { + "epoch": 0.41, + "learning_rate": 0.000178534769612543, + "loss": 0.0, + "step": 7350 + }, + { + "epoch": 0.41, + "learning_rate": 0.00017845017201511477, + "loss": 0.0, + "step": 7355 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001783655744176865, + "loss": 0.0, + "step": 7360 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001782809768202583, + "loss": 0.0, + "step": 7365 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017819637922283004, + "loss": 0.0, + "step": 7370 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017811178162540183, + "loss": 0.0, + "step": 7375 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017802718402797361, + "loss": 0.0, + "step": 7380 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017794258643054534, + "loss": 0.0, + "step": 7385 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017785798883311713, + "loss": 0.0, + "step": 7390 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017777339123568889, + "loss": 0.0, + "step": 7395 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017768879363826067, + "loss": 0.0, + "step": 7400 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017760419604083243, + "loss": 0.0, + "step": 7405 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017751959844340419, + "loss": 0.0, + "step": 7410 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017743500084597594, + "loss": 0.0, + "step": 7415 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017735040324854773, + "loss": 0.0, + "step": 7420 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017726580565111948, + "loss": 0.0, + "step": 7425 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017718120805369127, + "loss": 0.0, + "step": 7430 + }, + { + "epoch": 0.42, + "learning_rate": 0.000177096610456263, + "loss": 0.0, + "step": 7435 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017701201285883478, + "loss": 0.0, + "step": 7440 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017692741526140657, + "loss": 0.0, + "step": 7445 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017684281766397833, + "loss": 0.0, + "step": 7450 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001767582200665501, + "loss": 0.0, + "step": 7455 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017667362246912184, + "loss": 0.0, + "step": 7460 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017658902487169363, + "loss": 0.0, + "step": 7465 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017650442727426538, + "loss": 0.0, + "step": 7470 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017641982967683717, + "loss": 0.0, + "step": 7475 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017633523207940895, + "loss": 0.0, + "step": 7480 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001762506344819807, + "loss": 0.0, + "step": 7485 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017616603688455244, + "loss": 0.0, + "step": 7490 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017608143928712422, + "loss": 0.0, + "step": 7495 + }, + { + "epoch": 0.42, + "learning_rate": 0.000175996841689696, + "loss": 0.0, + "step": 7500 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017591224409226777, + "loss": 0.0, + "step": 7505 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017582764649483955, + "loss": 0.0, + "step": 7510 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017574304889741128, + "loss": 0.0, + "step": 7515 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017565845129998307, + "loss": 0.0, + "step": 7520 + }, + { + "epoch": 0.42, + "learning_rate": 0.00017557385370255482, + "loss": 0.0, + "step": 7525 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001754892561051266, + "loss": 0.0, + "step": 7530 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001754046585076984, + "loss": 0.0, + "step": 7535 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017532006091027012, + "loss": 0.0, + "step": 7540 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001752354633128419, + "loss": 0.0, + "step": 7545 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017515086571541366, + "loss": 0.0, + "step": 7550 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017506626811798545, + "loss": 0.0, + "step": 7555 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001749816705205572, + "loss": 0.0, + "step": 7560 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017489707292312896, + "loss": 0.0, + "step": 7565 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017481247532570072, + "loss": 0.0, + "step": 7570 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001747278777282725, + "loss": 0.0, + "step": 7575 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017464328013084426, + "loss": 0.0, + "step": 7580 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017455868253341605, + "loss": 0.0, + "step": 7585 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017447408493598778, + "loss": 0.0, + "step": 7590 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017438948733855956, + "loss": 0.0, + "step": 7595 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017430488974113135, + "loss": 0.0, + "step": 7600 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001742202921437031, + "loss": 0.0, + "step": 7605 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001741356945462749, + "loss": 0.0, + "step": 7610 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017405109694884662, + "loss": 0.0, + "step": 7615 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001739664993514184, + "loss": 0.0, + "step": 7620 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017388190175399016, + "loss": 0.0, + "step": 7625 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017379730415656195, + "loss": 0.0, + "step": 7630 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017371270655913373, + "loss": 0.0, + "step": 7635 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017362810896170546, + "loss": 0.0, + "step": 7640 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017354351136427725, + "loss": 0.0, + "step": 7645 + }, + { + "epoch": 0.43, + "learning_rate": 0.000173458913766849, + "loss": 0.0, + "step": 7650 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001733743161694208, + "loss": 0.0, + "step": 7655 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017328971857199254, + "loss": 0.0, + "step": 7660 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001732051209745643, + "loss": 0.0, + "step": 7665 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017312052337713606, + "loss": 0.0, + "step": 7670 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017303592577970784, + "loss": 0.0, + "step": 7675 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001729513281822796, + "loss": 0.0, + "step": 7680 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017286673058485139, + "loss": 0.0, + "step": 7685 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017278213298742312, + "loss": 0.0, + "step": 7690 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001726975353899949, + "loss": 0.0, + "step": 7695 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017261293779256669, + "loss": 0.0, + "step": 7700 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017252834019513844, + "loss": 0.0, + "step": 7705 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017244374259771023, + "loss": 0.0, + "step": 7710 + }, + { + "epoch": 0.43, + "learning_rate": 0.00017235914500028196, + "loss": 0.0, + "step": 7715 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017227454740285374, + "loss": 0.0, + "step": 7720 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001721899498054255, + "loss": 0.0, + "step": 7725 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017210535220799728, + "loss": 0.0, + "step": 7730 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017202075461056907, + "loss": 0.0, + "step": 7735 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001719361570131408, + "loss": 0.0, + "step": 7740 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017185155941571256, + "loss": 0.0, + "step": 7745 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017176696181828434, + "loss": 0.0, + "step": 7750 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017168236422085613, + "loss": 0.0, + "step": 7755 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017159776662342788, + "loss": 0.0, + "step": 7760 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017151316902599964, + "loss": 0.0, + "step": 7765 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001714285714285714, + "loss": 0.0, + "step": 7770 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017134397383114318, + "loss": 0.0, + "step": 7775 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017125937623371494, + "loss": 0.0, + "step": 7780 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017117477863628672, + "loss": 0.0, + "step": 7785 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017109018103885845, + "loss": 0.0, + "step": 7790 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017100558344143024, + "loss": 0.0, + "step": 7795 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017092098584400202, + "loss": 0.0, + "step": 7800 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017083638824657378, + "loss": 0.0, + "step": 7805 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017075179064914557, + "loss": 0.0, + "step": 7810 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017066719305171732, + "loss": 0.0, + "step": 7815 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017058259545428908, + "loss": 0.0, + "step": 7820 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017049799785686084, + "loss": 0.0, + "step": 7825 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017041340025943262, + "loss": 0.0, + "step": 7830 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017032880266200438, + "loss": 0.0, + "step": 7835 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017024420506457616, + "loss": 0.0, + "step": 7840 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001701596074671479, + "loss": 0.0, + "step": 7845 + }, + { + "epoch": 0.44, + "learning_rate": 0.00017007500986971968, + "loss": 0.0, + "step": 7850 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016999041227229146, + "loss": 0.0, + "step": 7855 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016990581467486322, + "loss": 0.0, + "step": 7860 + }, + { + "epoch": 0.44, + "learning_rate": 0.000169821217077435, + "loss": 0.0, + "step": 7865 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016973661948000674, + "loss": 0.0, + "step": 7870 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016965202188257852, + "loss": 0.0, + "step": 7875 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016956742428515028, + "loss": 0.0, + "step": 7880 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016948282668772206, + "loss": 0.0, + "step": 7885 + }, + { + "epoch": 0.44, + "learning_rate": 0.00016939822909029385, + "loss": 0.0, + "step": 7890 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016931363149286558, + "loss": 0.0, + "step": 7895 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016922903389543736, + "loss": 0.0, + "step": 7900 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016914443629800912, + "loss": 0.0, + "step": 7905 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001690598387005809, + "loss": 0.0, + "step": 7910 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016897524110315266, + "loss": 0.0, + "step": 7915 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016889064350572442, + "loss": 0.0, + "step": 7920 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016880604590829618, + "loss": 0.0, + "step": 7925 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016872144831086796, + "loss": 0.0, + "step": 7930 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016863685071343972, + "loss": 0.0, + "step": 7935 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001685522531160115, + "loss": 0.0, + "step": 7940 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016846765551858323, + "loss": 0.0, + "step": 7945 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016838305792115502, + "loss": 0.0, + "step": 7950 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001682984603237268, + "loss": 0.0, + "step": 7955 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016821386272629856, + "loss": 0.0, + "step": 7960 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016812926512887034, + "loss": 0.0, + "step": 7965 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016804466753144207, + "loss": 0.0, + "step": 7970 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016796006993401386, + "loss": 0.0, + "step": 7975 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016787547233658562, + "loss": 0.0, + "step": 7980 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001677908747391574, + "loss": 0.0, + "step": 7985 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016770627714172919, + "loss": 0.0, + "step": 7990 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016762167954430092, + "loss": 0.0, + "step": 7995 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016753708194687267, + "loss": 0.0, + "step": 8000 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016745248434944446, + "loss": 0.0, + "step": 8005 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016736788675201624, + "loss": 0.0, + "step": 8010 + }, + { + "epoch": 0.45, + "learning_rate": 0.000167283289154588, + "loss": 0.0, + "step": 8015 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016719869155715976, + "loss": 0.0, + "step": 8020 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016711409395973151, + "loss": 0.0, + "step": 8025 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001670294963623033, + "loss": 0.0, + "step": 8030 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016694489876487506, + "loss": 0.0, + "step": 8035 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016686030116744684, + "loss": 0.0, + "step": 8040 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016677570357001857, + "loss": 0.0, + "step": 8045 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016669110597259036, + "loss": 0.0, + "step": 8050 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016660650837516214, + "loss": 0.0, + "step": 8055 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001665219107777339, + "loss": 0.0, + "step": 8060 + }, + { + "epoch": 0.45, + "learning_rate": 0.00016643731318030568, + "loss": 0.0, + "step": 8065 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001663527155828774, + "loss": 0.0, + "step": 8070 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001662681179854492, + "loss": 0.0, + "step": 8075 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016618352038802096, + "loss": 0.0, + "step": 8080 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016609892279059274, + "loss": 0.0, + "step": 8085 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001660143251931645, + "loss": 0.0, + "step": 8090 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016592972759573625, + "loss": 0.0, + "step": 8095 + }, + { + "epoch": 0.46, + "learning_rate": 0.000165845129998308, + "loss": 0.0, + "step": 8100 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001657605324008798, + "loss": 0.0, + "step": 8105 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016567593480345158, + "loss": 0.0, + "step": 8110 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016559133720602334, + "loss": 0.0, + "step": 8115 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001655067396085951, + "loss": 0.0, + "step": 8120 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016542214201116685, + "loss": 0.0, + "step": 8125 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016533754441373864, + "loss": 0.0, + "step": 8130 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001652529468163104, + "loss": 0.0, + "step": 8135 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016516834921888218, + "loss": 0.0, + "step": 8140 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016508375162145396, + "loss": 0.0, + "step": 8145 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001649991540240257, + "loss": 0.0, + "step": 8150 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016491455642659748, + "loss": 0.0, + "step": 8155 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016482995882916924, + "loss": 0.0, + "step": 8160 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016474536123174102, + "loss": 0.0, + "step": 8165 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016466076363431278, + "loss": 0.0, + "step": 8170 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016457616603688454, + "loss": 0.0, + "step": 8175 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001644915684394563, + "loss": 0.0, + "step": 8180 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016440697084202808, + "loss": 0.0, + "step": 8185 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016432237324459984, + "loss": 0.0, + "step": 8190 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016423777564717162, + "loss": 0.0, + "step": 8195 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016415317804974335, + "loss": 0.0, + "step": 8200 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016406858045231513, + "loss": 0.0, + "step": 8205 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016398398285488692, + "loss": 0.0, + "step": 8210 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016389938525745868, + "loss": 0.0, + "step": 8215 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016381478766003046, + "loss": 0.0, + "step": 8220 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001637301900626022, + "loss": 0.0, + "step": 8225 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016364559246517398, + "loss": 0.0, + "step": 8230 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016356099486774573, + "loss": 0.0, + "step": 8235 + }, + { + "epoch": 0.46, + "learning_rate": 0.00016347639727031752, + "loss": 0.0, + "step": 8240 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001633917996728893, + "loss": 0.0, + "step": 8245 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016330720207546103, + "loss": 0.0, + "step": 8250 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001632226044780328, + "loss": 0.0, + "step": 8255 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016313800688060458, + "loss": 0.0, + "step": 8260 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016305340928317636, + "loss": 0.0, + "step": 8265 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016296881168574812, + "loss": 0.0, + "step": 8270 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016288421408831987, + "loss": 0.0, + "step": 8275 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016279961649089163, + "loss": 0.0, + "step": 8280 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016271501889346342, + "loss": 0.0, + "step": 8285 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016263042129603517, + "loss": 0.0, + "step": 8290 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016254582369860696, + "loss": 0.0, + "step": 8295 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001624612261011787, + "loss": 0.0, + "step": 8300 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016237662850375047, + "loss": 0.0, + "step": 8305 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016229203090632226, + "loss": 0.0, + "step": 8310 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016220743330889402, + "loss": 0.0, + "step": 8315 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001621228357114658, + "loss": 0.0, + "step": 8320 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016203823811403753, + "loss": 0.0, + "step": 8325 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016195364051660931, + "loss": 0.0, + "step": 8330 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016186904291918107, + "loss": 0.0, + "step": 8335 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016178444532175286, + "loss": 0.0, + "step": 8340 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016169984772432461, + "loss": 0.0, + "step": 8345 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016161525012689637, + "loss": 0.0, + "step": 8350 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016153065252946813, + "loss": 0.0, + "step": 8355 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001614460549320399, + "loss": 0.0, + "step": 8360 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001613614573346117, + "loss": 0.0, + "step": 8365 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016127685973718346, + "loss": 0.0, + "step": 8370 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001611922621397552, + "loss": 0.0, + "step": 8375 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016110766454232697, + "loss": 0.0, + "step": 8380 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016102306694489875, + "loss": 0.0, + "step": 8385 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001609384693474705, + "loss": 0.0, + "step": 8390 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001608538717500423, + "loss": 0.0, + "step": 8395 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016076927415261403, + "loss": 0.0, + "step": 8400 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001606846765551858, + "loss": 0.0, + "step": 8405 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001606000789577576, + "loss": 0.0, + "step": 8410 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016051548136032935, + "loss": 0.0, + "step": 8415 + }, + { + "epoch": 0.47, + "learning_rate": 0.00016043088376290114, + "loss": 0.0, + "step": 8420 + }, + { + "epoch": 0.48, + "learning_rate": 0.00016034628616547287, + "loss": 0.0, + "step": 8425 + }, + { + "epoch": 0.48, + "learning_rate": 0.00016026168856804465, + "loss": 0.0, + "step": 8430 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001601770909706164, + "loss": 0.0, + "step": 8435 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001600924933731882, + "loss": 0.0, + "step": 8440 + }, + { + "epoch": 0.48, + "learning_rate": 0.00016000789577575995, + "loss": 0.0, + "step": 8445 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015992329817833174, + "loss": 0.0, + "step": 8450 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015983870058090347, + "loss": 0.0, + "step": 8455 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015975410298347525, + "loss": 0.0, + "step": 8460 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015966950538604704, + "loss": 0.0, + "step": 8465 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001595849077886188, + "loss": 0.0, + "step": 8470 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015950031019119058, + "loss": 0.0, + "step": 8475 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001594157125937623, + "loss": 0.0, + "step": 8480 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001593311149963341, + "loss": 0.0, + "step": 8485 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015924651739890585, + "loss": 0.0, + "step": 8490 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015916191980147764, + "loss": 0.0, + "step": 8495 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015907732220404942, + "loss": 0.0, + "step": 8500 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015899272460662115, + "loss": 0.0, + "step": 8505 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001589081270091929, + "loss": 0.0, + "step": 8510 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001588235294117647, + "loss": 0.0, + "step": 8515 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015873893181433648, + "loss": 0.0, + "step": 8520 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015865433421690823, + "loss": 0.0, + "step": 8525 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015856973661948, + "loss": 0.0, + "step": 8530 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015848513902205175, + "loss": 0.0, + "step": 8535 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015840054142462353, + "loss": 0.0, + "step": 8540 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001583159438271953, + "loss": 0.0, + "step": 8545 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015823134622976708, + "loss": 0.0, + "step": 8550 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001581467486323388, + "loss": 0.0, + "step": 8555 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001580621510349106, + "loss": 0.0, + "step": 8560 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015797755343748237, + "loss": 0.0, + "step": 8565 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015789295584005413, + "loss": 0.0, + "step": 8570 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015780835824262592, + "loss": 0.0, + "step": 8575 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015772376064519765, + "loss": 0.0, + "step": 8580 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015763916304776943, + "loss": 0.0, + "step": 8585 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001575545654503412, + "loss": 0.0, + "step": 8590 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015746996785291297, + "loss": 0.0, + "step": 8595 + }, + { + "epoch": 0.48, + "learning_rate": 0.00015738537025548473, + "loss": 0.0, + "step": 8600 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001573007726580565, + "loss": 0.0, + "step": 8605 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015721617506062825, + "loss": 0.0, + "step": 8610 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015713157746320003, + "loss": 0.0, + "step": 8615 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015704697986577181, + "loss": 0.0, + "step": 8620 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015696238226834357, + "loss": 0.0, + "step": 8625 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015687778467091533, + "loss": 0.0, + "step": 8630 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001567931870734871, + "loss": 0.0, + "step": 8635 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015670858947605887, + "loss": 0.0, + "step": 8640 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015662399187863063, + "loss": 0.0, + "step": 8645 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015653939428120241, + "loss": 0.0, + "step": 8650 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015645479668377414, + "loss": 0.0, + "step": 8655 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015637019908634593, + "loss": 0.0, + "step": 8660 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001562856014889177, + "loss": 0.0, + "step": 8665 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015620100389148947, + "loss": 0.0, + "step": 8670 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015611640629406126, + "loss": 0.0, + "step": 8675 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015603180869663299, + "loss": 0.0, + "step": 8680 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015594721109920477, + "loss": 0.0, + "step": 8685 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015586261350177653, + "loss": 0.0, + "step": 8690 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001557780159043483, + "loss": 0.0, + "step": 8695 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015569341830692007, + "loss": 0.0, + "step": 8700 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015560882070949183, + "loss": 0.0, + "step": 8705 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015552422311206358, + "loss": 0.0, + "step": 8710 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015543962551463537, + "loss": 0.0, + "step": 8715 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015535502791720715, + "loss": 0.0, + "step": 8720 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001552704303197789, + "loss": 0.0, + "step": 8725 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015518583272235067, + "loss": 0.0, + "step": 8730 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015510123512492243, + "loss": 0.0, + "step": 8735 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001550166375274942, + "loss": 0.0, + "step": 8740 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015493203993006597, + "loss": 0.0, + "step": 8745 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015484744233263775, + "loss": 0.0, + "step": 8750 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015476284473520948, + "loss": 0.0, + "step": 8755 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015467824713778127, + "loss": 0.0, + "step": 8760 + }, + { + "epoch": 0.49, + "learning_rate": 0.00015459364954035302, + "loss": 0.0, + "step": 8765 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001545090519429248, + "loss": 0.0, + "step": 8770 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001544244543454966, + "loss": 0.0, + "step": 8775 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015433985674806835, + "loss": 0.0, + "step": 8780 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001542552591506401, + "loss": 0.0, + "step": 8785 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015417066155321187, + "loss": 0.0, + "step": 8790 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015408606395578365, + "loss": 0.0, + "step": 8795 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001540014663583554, + "loss": 0.0, + "step": 8800 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001539168687609272, + "loss": 0.0, + "step": 8805 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015383227116349892, + "loss": 0.0, + "step": 8810 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001537476735660707, + "loss": 0.0, + "step": 8815 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001536630759686425, + "loss": 0.0, + "step": 8820 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015357847837121425, + "loss": 0.0, + "step": 8825 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015349388077378603, + "loss": 0.0, + "step": 8830 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015340928317635776, + "loss": 0.0, + "step": 8835 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015332468557892955, + "loss": 0.0, + "step": 8840 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001532400879815013, + "loss": 0.0, + "step": 8845 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001531554903840731, + "loss": 0.0, + "step": 8850 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015307089278664485, + "loss": 0.0, + "step": 8855 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001529862951892166, + "loss": 0.0, + "step": 8860 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015290169759178836, + "loss": 0.0, + "step": 8865 + }, + { + "epoch": 0.5, + "eval_loss": NaN, + "eval_runtime": 4250.8174, + "eval_samples_per_second": 2.086, + "eval_steps_per_second": 0.261, + "step": 8868 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015281709999436015, + "loss": 0.0, + "step": 8870 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015273250239693193, + "loss": 0.0, + "step": 8875 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001526479047995037, + "loss": 0.0, + "step": 8880 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015256330720207545, + "loss": 0.0, + "step": 8885 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001524787096046472, + "loss": 0.0, + "step": 8890 + }, + { + "epoch": 0.5, + "learning_rate": 0.000152394112007219, + "loss": 0.0, + "step": 8895 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015230951440979075, + "loss": 0.0, + "step": 8900 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015222491681236253, + "loss": 0.0, + "step": 8905 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015214031921493426, + "loss": 0.0, + "step": 8910 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015205572161750605, + "loss": 0.0, + "step": 8915 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015197112402007783, + "loss": 0.0, + "step": 8920 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001518865264226496, + "loss": 0.0, + "step": 8925 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015180192882522137, + "loss": 0.0, + "step": 8930 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001517173312277931, + "loss": 0.0, + "step": 8935 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001516327336303649, + "loss": 0.0, + "step": 8940 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015154813603293664, + "loss": 0.0, + "step": 8945 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015146353843550843, + "loss": 0.0, + "step": 8950 + }, + { + "epoch": 0.5, + "learning_rate": 0.00015137894083808019, + "loss": 0.0, + "step": 8955 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015129434324065194, + "loss": 0.0, + "step": 8960 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001512097456432237, + "loss": 0.0, + "step": 8965 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015112514804579549, + "loss": 0.0, + "step": 8970 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015104055044836727, + "loss": 0.0, + "step": 8975 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015095595285093903, + "loss": 0.0, + "step": 8980 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015087135525351079, + "loss": 0.0, + "step": 8985 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015078675765608254, + "loss": 0.0, + "step": 8990 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015070216005865433, + "loss": 0.0, + "step": 8995 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015061756246122608, + "loss": 0.0, + "step": 9000 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015053296486379787, + "loss": 0.0, + "step": 9005 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001504483672663696, + "loss": 0.0, + "step": 9010 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015036376966894138, + "loss": 0.0, + "step": 9015 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015027917207151314, + "loss": 0.0, + "step": 9020 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015019457447408493, + "loss": 0.0, + "step": 9025 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001501099768766567, + "loss": 0.0, + "step": 9030 + }, + { + "epoch": 0.51, + "learning_rate": 0.00015002537927922844, + "loss": 0.0, + "step": 9035 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014994078168180023, + "loss": 0.0, + "step": 9040 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014985618408437198, + "loss": 0.0, + "step": 9045 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014977158648694377, + "loss": 0.0, + "step": 9050 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014968698888951552, + "loss": 0.0, + "step": 9055 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014960239129208728, + "loss": 0.0, + "step": 9060 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014951779369465907, + "loss": 0.0, + "step": 9065 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014943319609723082, + "loss": 0.0, + "step": 9070 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001493485984998026, + "loss": 0.0, + "step": 9075 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014926400090237437, + "loss": 0.0, + "step": 9080 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014917940330494612, + "loss": 0.0, + "step": 9085 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001490948057075179, + "loss": 0.0, + "step": 9090 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014901020811008967, + "loss": 0.0, + "step": 9095 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014892561051266142, + "loss": 0.0, + "step": 9100 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014884101291523318, + "loss": 0.0, + "step": 9105 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014875641531780496, + "loss": 0.0, + "step": 9110 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014867181772037672, + "loss": 0.0, + "step": 9115 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014858722012294848, + "loss": 0.0, + "step": 9120 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014850262252552026, + "loss": 0.0, + "step": 9125 + }, + { + "epoch": 0.51, + "learning_rate": 0.00014841802492809202, + "loss": 0.0, + "step": 9130 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001483334273306638, + "loss": 0.0, + "step": 9135 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014824882973323556, + "loss": 0.0, + "step": 9140 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014816423213580732, + "loss": 0.0, + "step": 9145 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001480796345383791, + "loss": 0.0, + "step": 9150 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014799503694095086, + "loss": 0.0, + "step": 9155 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014791043934352262, + "loss": 0.0, + "step": 9160 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001478258417460944, + "loss": 0.0, + "step": 9165 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014774124414866616, + "loss": 0.0, + "step": 9170 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014765664655123795, + "loss": 0.0, + "step": 9175 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014757204895380968, + "loss": 0.0, + "step": 9180 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014748745135638146, + "loss": 0.0, + "step": 9185 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014740285375895325, + "loss": 0.0, + "step": 9190 + }, + { + "epoch": 0.52, + "learning_rate": 0.000147318256161525, + "loss": 0.0, + "step": 9195 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014723365856409676, + "loss": 0.0, + "step": 9200 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014714906096666855, + "loss": 0.0, + "step": 9205 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001470644633692403, + "loss": 0.0, + "step": 9210 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014697986577181206, + "loss": 0.0, + "step": 9215 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014689526817438382, + "loss": 0.0, + "step": 9220 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001468106705769556, + "loss": 0.0, + "step": 9225 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001467260729795274, + "loss": 0.0, + "step": 9230 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014664147538209914, + "loss": 0.0, + "step": 9235 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001465568777846709, + "loss": 0.0, + "step": 9240 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014647228018724266, + "loss": 0.0, + "step": 9245 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014638768258981444, + "loss": 0.0, + "step": 9250 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001463030849923862, + "loss": 0.0, + "step": 9255 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014621848739495796, + "loss": 0.0, + "step": 9260 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014613388979752974, + "loss": 0.0, + "step": 9265 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001460492922001015, + "loss": 0.0, + "step": 9270 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014596469460267326, + "loss": 0.0, + "step": 9275 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014588009700524504, + "loss": 0.0, + "step": 9280 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001457954994078168, + "loss": 0.0, + "step": 9285 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014571090181038858, + "loss": 0.0, + "step": 9290 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014562630421296034, + "loss": 0.0, + "step": 9295 + }, + { + "epoch": 0.52, + "learning_rate": 0.0001455417066155321, + "loss": 0.0, + "step": 9300 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014545710901810388, + "loss": 0.0, + "step": 9305 + }, + { + "epoch": 0.52, + "learning_rate": 0.00014537251142067564, + "loss": 0.0, + "step": 9310 + }, + { + "epoch": 0.53, + "learning_rate": 0.0001452879138232474, + "loss": 0.0, + "step": 9315 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014520331622581916, + "loss": 0.0, + "step": 9320 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014511871862839094, + "loss": 0.0, + "step": 9325 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014503412103096273, + "loss": 0.0, + "step": 9330 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014494952343353448, + "loss": 0.0, + "step": 9335 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014486492583610624, + "loss": 0.0, + "step": 9340 + }, + { + "epoch": 0.53, + "learning_rate": 0.000144780328238678, + "loss": 0.0, + "step": 9345 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014469573064124978, + "loss": 0.0, + "step": 9350 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014461113304382154, + "loss": 0.0, + "step": 9355 + }, + { + "epoch": 0.53, + "learning_rate": 0.0001445265354463933, + "loss": 0.0, + "step": 9360 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014444193784896508, + "loss": 0.0, + "step": 9365 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014435734025153684, + "loss": 0.0, + "step": 9370 + }, + { + "epoch": 0.53, + "learning_rate": 0.0001442727426541086, + "loss": 0.0, + "step": 9375 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014418814505668038, + "loss": 0.0, + "step": 9380 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014410354745925214, + "loss": 0.0, + "step": 9385 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014401894986182392, + "loss": 0.0, + "step": 9390 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014393435226439568, + "loss": 0.0, + "step": 9395 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014384975466696744, + "loss": 0.0, + "step": 9400 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014376515706953922, + "loss": 0.0, + "step": 9405 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014368055947211098, + "loss": 0.0, + "step": 9410 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014359596187468274, + "loss": 0.0, + "step": 9415 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014351136427725452, + "loss": 0.0, + "step": 9420 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014342676667982628, + "loss": 0.0, + "step": 9425 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014334216908239806, + "loss": 0.0, + "step": 9430 + }, + { + "epoch": 0.53, + "learning_rate": 0.0001432575714849698, + "loss": 0.0, + "step": 9435 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014317297388754158, + "loss": 0.0, + "step": 9440 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014308837629011336, + "loss": 0.0, + "step": 9445 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014300377869268512, + "loss": 0.0, + "step": 9450 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014291918109525688, + "loss": 0.0, + "step": 9455 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014283458349782864, + "loss": 0.0, + "step": 9460 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014274998590040042, + "loss": 0.0, + "step": 9465 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014266538830297218, + "loss": 0.0, + "step": 9470 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014258079070554393, + "loss": 0.0, + "step": 9475 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014249619310811572, + "loss": 0.0, + "step": 9480 + }, + { + "epoch": 0.53, + "learning_rate": 0.00014241159551068748, + "loss": 0.0, + "step": 9485 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014232699791325926, + "loss": 0.0, + "step": 9490 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014224240031583102, + "loss": 0.0, + "step": 9495 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014215780271840278, + "loss": 0.0, + "step": 9500 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014207320512097456, + "loss": 0.0, + "step": 9505 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014198860752354632, + "loss": 0.0, + "step": 9510 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014190400992611808, + "loss": 0.0, + "step": 9515 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014181941232868986, + "loss": 0.0, + "step": 9520 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014173481473126162, + "loss": 0.0, + "step": 9525 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014165021713383338, + "loss": 0.0, + "step": 9530 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014156561953640516, + "loss": 0.0, + "step": 9535 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014148102193897692, + "loss": 0.0, + "step": 9540 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001413964243415487, + "loss": 0.0, + "step": 9545 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014131182674412046, + "loss": 0.0, + "step": 9550 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014122722914669222, + "loss": 0.0, + "step": 9555 + }, + { + "epoch": 0.54, + "learning_rate": 0.000141142631549264, + "loss": 0.0, + "step": 9560 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014105803395183576, + "loss": 0.0, + "step": 9565 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014097343635440752, + "loss": 0.0, + "step": 9570 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014088883875697927, + "loss": 0.0, + "step": 9575 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014080424115955106, + "loss": 0.0, + "step": 9580 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014071964356212284, + "loss": 0.0, + "step": 9585 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001406350459646946, + "loss": 0.0, + "step": 9590 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014055044836726636, + "loss": 0.0, + "step": 9595 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014046585076983811, + "loss": 0.0, + "step": 9600 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001403812531724099, + "loss": 0.0, + "step": 9605 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014029665557498166, + "loss": 0.0, + "step": 9610 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014021205797755341, + "loss": 0.0, + "step": 9615 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001401274603801252, + "loss": 0.0, + "step": 9620 + }, + { + "epoch": 0.54, + "learning_rate": 0.00014004286278269696, + "loss": 0.0, + "step": 9625 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001399582651852687, + "loss": 0.0, + "step": 9630 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001398736675878405, + "loss": 0.0, + "step": 9635 + }, + { + "epoch": 0.54, + "learning_rate": 0.00013978906999041226, + "loss": 0.0, + "step": 9640 + }, + { + "epoch": 0.54, + "learning_rate": 0.00013970447239298404, + "loss": 0.0, + "step": 9645 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001396198747955558, + "loss": 0.0, + "step": 9650 + }, + { + "epoch": 0.54, + "learning_rate": 0.00013953527719812755, + "loss": 0.0, + "step": 9655 + }, + { + "epoch": 0.54, + "learning_rate": 0.00013945067960069934, + "loss": 0.0, + "step": 9660 + }, + { + "epoch": 0.54, + "learning_rate": 0.0001393660820032711, + "loss": 0.0, + "step": 9665 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013928148440584285, + "loss": 0.0, + "step": 9670 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013919688680841464, + "loss": 0.0, + "step": 9675 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001391122892109864, + "loss": 0.0, + "step": 9680 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013902769161355818, + "loss": 0.0, + "step": 9685 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001389430940161299, + "loss": 0.0, + "step": 9690 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001388584964187017, + "loss": 0.0, + "step": 9695 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013877389882127348, + "loss": 0.0, + "step": 9700 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013868930122384524, + "loss": 0.0, + "step": 9705 + }, + { + "epoch": 0.55, + "learning_rate": 0.000138604703626417, + "loss": 0.0, + "step": 9710 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013852010602898875, + "loss": 0.0, + "step": 9715 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013843550843156054, + "loss": 0.0, + "step": 9720 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001383509108341323, + "loss": 0.0, + "step": 9725 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013826631323670405, + "loss": 0.0, + "step": 9730 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013818171563927584, + "loss": 0.0, + "step": 9735 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001380971180418476, + "loss": 0.0, + "step": 9740 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013801252044441938, + "loss": 0.0, + "step": 9745 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013792792284699114, + "loss": 0.0, + "step": 9750 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001378433252495629, + "loss": 0.0, + "step": 9755 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013775872765213468, + "loss": 0.0, + "step": 9760 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013767413005470644, + "loss": 0.0, + "step": 9765 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001375895324572782, + "loss": 0.0, + "step": 9770 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013750493485984998, + "loss": 0.0, + "step": 9775 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013742033726242173, + "loss": 0.0, + "step": 9780 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001373357396649935, + "loss": 0.0, + "step": 9785 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013725114206756525, + "loss": 0.0, + "step": 9790 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013716654447013703, + "loss": 0.0, + "step": 9795 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013708194687270882, + "loss": 0.0, + "step": 9800 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013699734927528058, + "loss": 0.0, + "step": 9805 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013691275167785233, + "loss": 0.0, + "step": 9810 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001368281540804241, + "loss": 0.0, + "step": 9815 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013674355648299588, + "loss": 0.0, + "step": 9820 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013665895888556763, + "loss": 0.0, + "step": 9825 + }, + { + "epoch": 0.55, + "learning_rate": 0.0001365743612881394, + "loss": 0.0, + "step": 9830 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013648976369071117, + "loss": 0.0, + "step": 9835 + }, + { + "epoch": 0.55, + "learning_rate": 0.00013640516609328296, + "loss": 0.0, + "step": 9840 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013632056849585472, + "loss": 0.0, + "step": 9845 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013623597089842647, + "loss": 0.0, + "step": 9850 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013615137330099823, + "loss": 0.0, + "step": 9855 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013606677570357002, + "loss": 0.0, + "step": 9860 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013598217810614177, + "loss": 0.0, + "step": 9865 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013589758050871353, + "loss": 0.0, + "step": 9870 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013581298291128532, + "loss": 0.0, + "step": 9875 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013572838531385707, + "loss": 0.0, + "step": 9880 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013564378771642883, + "loss": 0.0, + "step": 9885 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013555919011900061, + "loss": 0.0, + "step": 9890 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013547459252157237, + "loss": 0.0, + "step": 9895 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013538999492414416, + "loss": 0.0, + "step": 9900 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013530539732671591, + "loss": 0.0, + "step": 9905 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013522079972928767, + "loss": 0.0, + "step": 9910 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013513620213185946, + "loss": 0.0, + "step": 9915 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013505160453443121, + "loss": 0.0, + "step": 9920 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013496700693700297, + "loss": 0.0, + "step": 9925 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013488240933957473, + "loss": 0.0, + "step": 9930 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001347978117421465, + "loss": 0.0, + "step": 9935 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001347132141447183, + "loss": 0.0, + "step": 9940 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013462861654729003, + "loss": 0.0, + "step": 9945 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001345440189498618, + "loss": 0.0, + "step": 9950 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013445942135243357, + "loss": 0.0, + "step": 9955 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013437482375500535, + "loss": 0.0, + "step": 9960 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001342902261575771, + "loss": 0.0, + "step": 9965 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013420562856014887, + "loss": 0.0, + "step": 9970 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013412103096272065, + "loss": 0.0, + "step": 9975 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001340364333652924, + "loss": 0.0, + "step": 9980 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013395183576786417, + "loss": 0.0, + "step": 9985 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013386723817043595, + "loss": 0.0, + "step": 9990 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001337826405730077, + "loss": 0.0, + "step": 9995 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001336980429755795, + "loss": 0.0, + "step": 10000 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013361344537815125, + "loss": 0.0, + "step": 10005 + }, + { + "epoch": 0.56, + "learning_rate": 0.000133528847780723, + "loss": 0.0, + "step": 10010 + }, + { + "epoch": 0.56, + "learning_rate": 0.0001334442501832948, + "loss": 0.0, + "step": 10015 + }, + { + "epoch": 0.56, + "learning_rate": 0.00013335965258586655, + "loss": 0.0, + "step": 10020 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001332750549884383, + "loss": 0.0, + "step": 10025 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001331904573910101, + "loss": 0.0, + "step": 10030 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013310585979358185, + "loss": 0.0, + "step": 10035 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001330212621961536, + "loss": 0.0, + "step": 10040 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013293666459872537, + "loss": 0.0, + "step": 10045 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013285206700129715, + "loss": 0.0, + "step": 10050 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013276746940386894, + "loss": 0.0, + "step": 10055 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001326828718064407, + "loss": 0.0, + "step": 10060 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013259827420901245, + "loss": 0.0, + "step": 10065 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001325136766115842, + "loss": 0.0, + "step": 10070 + }, + { + "epoch": 0.57, + "learning_rate": 0.000132429079014156, + "loss": 0.0, + "step": 10075 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013234448141672775, + "loss": 0.0, + "step": 10080 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001322598838192995, + "loss": 0.0, + "step": 10085 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001321752862218713, + "loss": 0.0, + "step": 10090 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013209068862444305, + "loss": 0.0, + "step": 10095 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013200609102701483, + "loss": 0.0, + "step": 10100 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001319214934295866, + "loss": 0.0, + "step": 10105 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013183689583215835, + "loss": 0.0, + "step": 10110 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013175229823473013, + "loss": 0.0, + "step": 10115 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001316677006373019, + "loss": 0.0, + "step": 10120 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013158310303987365, + "loss": 0.0, + "step": 10125 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013149850544244543, + "loss": 0.0, + "step": 10130 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001314139078450172, + "loss": 0.0, + "step": 10135 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013132931024758895, + "loss": 0.0, + "step": 10140 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001312447126501607, + "loss": 0.0, + "step": 10145 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001311601150527325, + "loss": 0.0, + "step": 10150 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013107551745530427, + "loss": 0.0, + "step": 10155 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013099091985787603, + "loss": 0.0, + "step": 10160 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001309063222604478, + "loss": 0.0, + "step": 10165 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013082172466301957, + "loss": 0.0, + "step": 10170 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013073712706559133, + "loss": 0.0, + "step": 10175 + }, + { + "epoch": 0.57, + "learning_rate": 0.0001306525294681631, + "loss": 0.0, + "step": 10180 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013056793187073485, + "loss": 0.0, + "step": 10185 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013048333427330663, + "loss": 0.0, + "step": 10190 + }, + { + "epoch": 0.57, + "learning_rate": 0.00013039873667587841, + "loss": 0.0, + "step": 10195 + }, + { + "epoch": 0.58, + "learning_rate": 0.00013031413907845014, + "loss": 0.0, + "step": 10200 + }, + { + "epoch": 0.58, + "learning_rate": 0.00013022954148102193, + "loss": 0.0, + "step": 10205 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001301449438835937, + "loss": 0.0, + "step": 10210 + }, + { + "epoch": 0.58, + "learning_rate": 0.00013006034628616547, + "loss": 0.0, + "step": 10215 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012997574868873723, + "loss": 0.0, + "step": 10220 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012989115109130899, + "loss": 0.0, + "step": 10225 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012980655349388077, + "loss": 0.0, + "step": 10230 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012972195589645253, + "loss": 0.0, + "step": 10235 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012963735829902429, + "loss": 0.0, + "step": 10240 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012955276070159607, + "loss": 0.0, + "step": 10245 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012946816310416783, + "loss": 0.0, + "step": 10250 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001293835655067396, + "loss": 0.0, + "step": 10255 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012929896790931137, + "loss": 0.0, + "step": 10260 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012921437031188313, + "loss": 0.0, + "step": 10265 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001291297727144549, + "loss": 0.0, + "step": 10270 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012904517511702667, + "loss": 0.0, + "step": 10275 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012896057751959843, + "loss": 0.0, + "step": 10280 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012887597992217018, + "loss": 0.0, + "step": 10285 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012879138232474197, + "loss": 0.0, + "step": 10290 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012870678472731373, + "loss": 0.0, + "step": 10295 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012862218712988548, + "loss": 0.0, + "step": 10300 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012853758953245727, + "loss": 0.0, + "step": 10305 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012845299193502903, + "loss": 0.0, + "step": 10310 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001283683943376008, + "loss": 0.0, + "step": 10315 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012828379674017257, + "loss": 0.0, + "step": 10320 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012819919914274432, + "loss": 0.0, + "step": 10325 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001281146015453161, + "loss": 0.0, + "step": 10330 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012803000394788787, + "loss": 0.0, + "step": 10335 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012794540635045962, + "loss": 0.0, + "step": 10340 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001278608087530314, + "loss": 0.0, + "step": 10345 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012777621115560317, + "loss": 0.0, + "step": 10350 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012769161355817495, + "loss": 0.0, + "step": 10355 + }, + { + "epoch": 0.58, + "learning_rate": 0.0001276070159607467, + "loss": 0.0, + "step": 10360 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012752241836331847, + "loss": 0.0, + "step": 10365 + }, + { + "epoch": 0.58, + "learning_rate": 0.00012743782076589025, + "loss": 0.0, + "step": 10370 + }, + { + "epoch": 0.58, + "learning_rate": 0.000127353223168462, + "loss": 0.0, + "step": 10375 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012726862557103376, + "loss": 0.0, + "step": 10380 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012718402797360555, + "loss": 0.0, + "step": 10385 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001270994303761773, + "loss": 0.0, + "step": 10390 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012701483277874906, + "loss": 0.0, + "step": 10395 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012693023518132082, + "loss": 0.0, + "step": 10400 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001268456375838926, + "loss": 0.0, + "step": 10405 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001267610399864644, + "loss": 0.0, + "step": 10410 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012667644238903615, + "loss": 0.0, + "step": 10415 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001265918447916079, + "loss": 0.0, + "step": 10420 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012650724719417966, + "loss": 0.0, + "step": 10425 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012642264959675145, + "loss": 0.0, + "step": 10430 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001263380519993232, + "loss": 0.0, + "step": 10435 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012625345440189496, + "loss": 0.0, + "step": 10440 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012616885680446675, + "loss": 0.0, + "step": 10445 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001260842592070385, + "loss": 0.0, + "step": 10450 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012599966160961026, + "loss": 0.0, + "step": 10455 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012591506401218205, + "loss": 0.0, + "step": 10460 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001258304664147538, + "loss": 0.0, + "step": 10465 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001257458688173256, + "loss": 0.0, + "step": 10470 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012566127121989735, + "loss": 0.0, + "step": 10475 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001255766736224691, + "loss": 0.0, + "step": 10480 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001254920760250409, + "loss": 0.0, + "step": 10485 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012540747842761265, + "loss": 0.0, + "step": 10490 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001253228808301844, + "loss": 0.0, + "step": 10495 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001252382832327562, + "loss": 0.0, + "step": 10500 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012515368563532794, + "loss": 0.0, + "step": 10505 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012506908803789973, + "loss": 0.0, + "step": 10510 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001249844904404715, + "loss": 0.0, + "step": 10515 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012489989284304324, + "loss": 0.0, + "step": 10520 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012481529524561503, + "loss": 0.0, + "step": 10525 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012473069764818679, + "loss": 0.0, + "step": 10530 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012464610005075854, + "loss": 0.0, + "step": 10535 + }, + { + "epoch": 0.59, + "learning_rate": 0.0001245615024533303, + "loss": 0.0, + "step": 10540 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012447690485590209, + "loss": 0.0, + "step": 10545 + }, + { + "epoch": 0.59, + "learning_rate": 0.00012439230725847384, + "loss": 0.0, + "step": 10550 + }, + { + "epoch": 0.6, + "learning_rate": 0.0001243077096610456, + "loss": 0.0, + "step": 10555 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012422311206361738, + "loss": 0.0, + "step": 10560 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012413851446618914, + "loss": 0.0, + "step": 10565 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012405391686876093, + "loss": 0.0, + "step": 10570 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012396931927133268, + "loss": 0.0, + "step": 10575 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012388472167390444, + "loss": 0.0, + "step": 10580 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012380012407647623, + "loss": 0.0, + "step": 10585 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012371552647904798, + "loss": 0.0, + "step": 10590 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012363092888161974, + "loss": 0.0, + "step": 10595 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012354633128419153, + "loss": 0.0, + "step": 10600 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012346173368676328, + "loss": 0.0, + "step": 10605 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012337713608933507, + "loss": 0.0, + "step": 10610 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012329253849190682, + "loss": 0.0, + "step": 10615 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012320794089447858, + "loss": 0.0, + "step": 10620 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012312334329705037, + "loss": 0.0, + "step": 10625 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012303874569962212, + "loss": 0.0, + "step": 10630 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012295414810219388, + "loss": 0.0, + "step": 10635 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012286955050476564, + "loss": 0.0, + "step": 10640 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012278495290733742, + "loss": 0.0, + "step": 10645 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012270035530990918, + "loss": 0.0, + "step": 10650 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012261575771248094, + "loss": 0.0, + "step": 10655 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012253116011505272, + "loss": 0.0, + "step": 10660 + }, + { + "epoch": 0.6, + "learning_rate": 0.0001224465625176245, + "loss": 0.0, + "step": 10665 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012236196492019627, + "loss": 0.0, + "step": 10670 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012227736732276802, + "loss": 0.0, + "step": 10675 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012219276972533978, + "loss": 0.0, + "step": 10680 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012210817212791156, + "loss": 0.0, + "step": 10685 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012202357453048334, + "loss": 0.0, + "step": 10690 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012193897693305509, + "loss": 0.0, + "step": 10695 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012185437933562686, + "loss": 0.0, + "step": 10700 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012176978173819862, + "loss": 0.0, + "step": 10705 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012168518414077039, + "loss": 0.0, + "step": 10710 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012160058654334216, + "loss": 0.0, + "step": 10715 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012151598894591392, + "loss": 0.0, + "step": 10720 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012143139134848569, + "loss": 0.0, + "step": 10725 + }, + { + "epoch": 0.6, + "learning_rate": 0.00012134679375105745, + "loss": 0.0, + "step": 10730 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012126219615362922, + "loss": 0.0, + "step": 10735 + }, + { + "epoch": 0.61, + "learning_rate": 0.000121177598556201, + "loss": 0.0, + "step": 10740 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012109300095877276, + "loss": 0.0, + "step": 10745 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012100840336134453, + "loss": 0.0, + "step": 10750 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012092380576391629, + "loss": 0.0, + "step": 10755 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012083920816648806, + "loss": 0.0, + "step": 10760 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012075461056905983, + "loss": 0.0, + "step": 10765 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012067001297163159, + "loss": 0.0, + "step": 10770 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012058541537420336, + "loss": 0.0, + "step": 10775 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012050081777677512, + "loss": 0.0, + "step": 10780 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012041622017934689, + "loss": 0.0, + "step": 10785 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012033162258191867, + "loss": 0.0, + "step": 10790 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012024702498449043, + "loss": 0.0, + "step": 10795 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001201624273870622, + "loss": 0.0, + "step": 10800 + }, + { + "epoch": 0.61, + "learning_rate": 0.00012007782978963396, + "loss": 0.0, + "step": 10805 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011999323219220573, + "loss": 0.0, + "step": 10810 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001199086345947775, + "loss": 0.0, + "step": 10815 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011982403699734926, + "loss": 0.0, + "step": 10820 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011973943939992103, + "loss": 0.0, + "step": 10825 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001196548418024928, + "loss": 0.0, + "step": 10830 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011957024420506456, + "loss": 0.0, + "step": 10835 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011948564660763634, + "loss": 0.0, + "step": 10840 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001194010490102081, + "loss": 0.0, + "step": 10845 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011931645141277987, + "loss": 0.0, + "step": 10850 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011923185381535164, + "loss": 0.0, + "step": 10855 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001191472562179234, + "loss": 0.0, + "step": 10860 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011906265862049517, + "loss": 0.0, + "step": 10865 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011897806102306693, + "loss": 0.0, + "step": 10870 + }, + { + "epoch": 0.61, + "learning_rate": 0.0001188934634256387, + "loss": 0.0, + "step": 10875 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011880886582821047, + "loss": 0.0, + "step": 10880 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011872426823078223, + "loss": 0.0, + "step": 10885 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011863967063335401, + "loss": 0.0, + "step": 10890 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011855507303592576, + "loss": 0.0, + "step": 10895 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011847047543849754, + "loss": 0.0, + "step": 10900 + }, + { + "epoch": 0.61, + "learning_rate": 0.00011838587784106931, + "loss": 0.0, + "step": 10905 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011830128024364107, + "loss": 0.0, + "step": 10910 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011821668264621284, + "loss": 0.0, + "step": 10915 + }, + { + "epoch": 0.62, + "learning_rate": 0.0001181320850487846, + "loss": 0.0, + "step": 10920 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011804748745135637, + "loss": 0.0, + "step": 10925 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011796288985392814, + "loss": 0.0, + "step": 10930 + }, + { + "epoch": 0.62, + "learning_rate": 0.0001178782922564999, + "loss": 0.0, + "step": 10935 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011779369465907168, + "loss": 0.0, + "step": 10940 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011770909706164343, + "loss": 0.0, + "step": 10945 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011762449946421521, + "loss": 0.0, + "step": 10950 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011753990186678698, + "loss": 0.0, + "step": 10955 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011745530426935874, + "loss": 0.0, + "step": 10960 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011737070667193051, + "loss": 0.0, + "step": 10965 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011728610907450227, + "loss": 0.0, + "step": 10970 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011720151147707404, + "loss": 0.0, + "step": 10975 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011711691387964581, + "loss": 0.0, + "step": 10980 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011703231628221757, + "loss": 0.0, + "step": 10985 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011694771868478934, + "loss": 0.0, + "step": 10990 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011686312108736112, + "loss": 0.0, + "step": 10995 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011677852348993288, + "loss": 0.0, + "step": 11000 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011669392589250465, + "loss": 0.0, + "step": 11005 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011660932829507641, + "loss": 0.0, + "step": 11010 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011652473069764818, + "loss": 0.0, + "step": 11015 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011644013310021995, + "loss": 0.0, + "step": 11020 + }, + { + "epoch": 0.62, + "learning_rate": 0.0001163555355027917, + "loss": 0.0, + "step": 11025 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011627093790536348, + "loss": 0.0, + "step": 11030 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011618634030793524, + "loss": 0.0, + "step": 11035 + }, + { + "epoch": 0.62, + "learning_rate": 0.000116101742710507, + "loss": 0.0, + "step": 11040 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011601714511307879, + "loss": 0.0, + "step": 11045 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011593254751565055, + "loss": 0.0, + "step": 11050 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011584794991822232, + "loss": 0.0, + "step": 11055 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011576335232079408, + "loss": 0.0, + "step": 11060 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011567875472336585, + "loss": 0.0, + "step": 11065 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011559415712593762, + "loss": 0.0, + "step": 11070 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011550955952850938, + "loss": 0.0, + "step": 11075 + }, + { + "epoch": 0.62, + "learning_rate": 0.00011542496193108115, + "loss": 0.0, + "step": 11080 + }, + { + "epoch": 0.62, + "learning_rate": 0.0001153403643336529, + "loss": 0.0, + "step": 11085 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011525576673622468, + "loss": 0.0, + "step": 11090 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011517116913879646, + "loss": 0.0, + "step": 11095 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011508657154136822, + "loss": 0.0, + "step": 11100 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011500197394393999, + "loss": 0.0, + "step": 11105 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011491737634651175, + "loss": 0.0, + "step": 11110 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011483277874908352, + "loss": 0.0, + "step": 11115 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011474818115165529, + "loss": 0.0, + "step": 11120 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011466358355422705, + "loss": 0.0, + "step": 11125 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011457898595679882, + "loss": 0.0, + "step": 11130 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011449438835937057, + "loss": 0.0, + "step": 11135 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011440979076194234, + "loss": 0.0, + "step": 11140 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011432519316451413, + "loss": 0.0, + "step": 11145 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011424059556708587, + "loss": 0.0, + "step": 11150 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011415599796965766, + "loss": 0.0, + "step": 11155 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011407140037222943, + "loss": 0.0, + "step": 11160 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011398680277480119, + "loss": 0.0, + "step": 11165 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011390220517737296, + "loss": 0.0, + "step": 11170 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011381760757994471, + "loss": 0.0, + "step": 11175 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011373300998251649, + "loss": 0.0, + "step": 11180 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011364841238508826, + "loss": 0.0, + "step": 11185 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011356381478766001, + "loss": 0.0, + "step": 11190 + }, + { + "epoch": 0.63, + "learning_rate": 0.0001134792171902318, + "loss": 0.0, + "step": 11195 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011339461959280354, + "loss": 0.0, + "step": 11200 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011331002199537533, + "loss": 0.0, + "step": 11205 + }, + { + "epoch": 0.63, + "learning_rate": 0.0001132254243979471, + "loss": 0.0, + "step": 11210 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011314082680051886, + "loss": 0.0, + "step": 11215 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011305622920309063, + "loss": 0.0, + "step": 11220 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011297163160566238, + "loss": 0.0, + "step": 11225 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011288703400823415, + "loss": 0.0, + "step": 11230 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011280243641080593, + "loss": 0.0, + "step": 11235 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011271783881337768, + "loss": 0.0, + "step": 11240 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011263324121594945, + "loss": 0.0, + "step": 11245 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011254864361852121, + "loss": 0.0, + "step": 11250 + }, + { + "epoch": 0.63, + "learning_rate": 0.000112464046021093, + "loss": 0.0, + "step": 11255 + }, + { + "epoch": 0.63, + "learning_rate": 0.00011237944842366477, + "loss": 0.0, + "step": 11260 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011229485082623652, + "loss": 0.0, + "step": 11265 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001122102532288083, + "loss": 0.0, + "step": 11270 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011212565563138005, + "loss": 0.0, + "step": 11275 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011204105803395182, + "loss": 0.0, + "step": 11280 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001119564604365236, + "loss": 0.0, + "step": 11285 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011187186283909535, + "loss": 0.0, + "step": 11290 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011178726524166712, + "loss": 0.0, + "step": 11295 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011170266764423888, + "loss": 0.0, + "step": 11300 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011161807004681067, + "loss": 0.0, + "step": 11305 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011153347244938244, + "loss": 0.0, + "step": 11310 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001114488748519542, + "loss": 0.0, + "step": 11315 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011136427725452596, + "loss": 0.0, + "step": 11320 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011127967965709774, + "loss": 0.0, + "step": 11325 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011119508205966949, + "loss": 0.0, + "step": 11330 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011111048446224126, + "loss": 0.0, + "step": 11335 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011102588686481302, + "loss": 0.0, + "step": 11340 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011094128926738479, + "loss": 0.0, + "step": 11345 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011085669166995658, + "loss": 0.0, + "step": 11350 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011077209407252833, + "loss": 0.0, + "step": 11355 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001106874964751001, + "loss": 0.0, + "step": 11360 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011060289887767186, + "loss": 0.0, + "step": 11365 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011051830128024363, + "loss": 0.0, + "step": 11370 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001104337036828154, + "loss": 0.0, + "step": 11375 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011034910608538716, + "loss": 0.0, + "step": 11380 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011026450848795893, + "loss": 0.0, + "step": 11385 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011017991089053069, + "loss": 0.0, + "step": 11390 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011009531329310246, + "loss": 0.0, + "step": 11395 + }, + { + "epoch": 0.64, + "learning_rate": 0.00011001071569567425, + "loss": 0.0, + "step": 11400 + }, + { + "epoch": 0.64, + "learning_rate": 0.00010992611809824599, + "loss": 0.0, + "step": 11405 + }, + { + "epoch": 0.64, + "learning_rate": 0.00010984152050081777, + "loss": 0.0, + "step": 11410 + }, + { + "epoch": 0.64, + "learning_rate": 0.00010975692290338953, + "loss": 0.0, + "step": 11415 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001096723253059613, + "loss": 0.0, + "step": 11420 + }, + { + "epoch": 0.64, + "learning_rate": 0.00010958772770853307, + "loss": 0.0, + "step": 11425 + }, + { + "epoch": 0.64, + "learning_rate": 0.00010950313011110483, + "loss": 0.0, + "step": 11430 + }, + { + "epoch": 0.64, + "learning_rate": 0.0001094185325136766, + "loss": 0.0, + "step": 11435 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010933393491624836, + "loss": 0.0, + "step": 11440 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010924933731882013, + "loss": 0.0, + "step": 11445 + }, + { + "epoch": 0.65, + "learning_rate": 0.0001091647397213919, + "loss": 0.0, + "step": 11450 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010908014212396366, + "loss": 0.0, + "step": 11455 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010899554452653544, + "loss": 0.0, + "step": 11460 + }, + { + "epoch": 0.65, + "learning_rate": 0.0001089109469291072, + "loss": 0.0, + "step": 11465 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010882634933167897, + "loss": 0.0, + "step": 11470 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010874175173425074, + "loss": 0.0, + "step": 11475 + }, + { + "epoch": 0.65, + "learning_rate": 0.0001086571541368225, + "loss": 0.0, + "step": 11480 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010857255653939427, + "loss": 0.0, + "step": 11485 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010848795894196604, + "loss": 0.0, + "step": 11490 + }, + { + "epoch": 0.65, + "learning_rate": 0.0001084033613445378, + "loss": 0.0, + "step": 11495 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010831876374710957, + "loss": 0.0, + "step": 11500 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010823416614968133, + "loss": 0.0, + "step": 11505 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010814956855225311, + "loss": 0.0, + "step": 11510 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010806497095482488, + "loss": 0.0, + "step": 11515 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010798037335739664, + "loss": 0.0, + "step": 11520 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010789577575996841, + "loss": 0.0, + "step": 11525 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010781117816254017, + "loss": 0.0, + "step": 11530 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010772658056511194, + "loss": 0.0, + "step": 11535 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010764198296768371, + "loss": 0.0, + "step": 11540 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010755738537025547, + "loss": 0.0, + "step": 11545 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010747278777282724, + "loss": 0.0, + "step": 11550 + }, + { + "epoch": 0.65, + "learning_rate": 0.000107388190175399, + "loss": 0.0, + "step": 11555 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010730359257797078, + "loss": 0.0, + "step": 11560 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010721899498054255, + "loss": 0.0, + "step": 11565 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010713439738311431, + "loss": 0.0, + "step": 11570 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010704979978568608, + "loss": 0.0, + "step": 11575 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010696520218825784, + "loss": 0.0, + "step": 11580 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010688060459082961, + "loss": 0.0, + "step": 11585 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010679600699340138, + "loss": 0.0, + "step": 11590 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010671140939597314, + "loss": 0.0, + "step": 11595 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010662681179854491, + "loss": 0.0, + "step": 11600 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010654221420111667, + "loss": 0.0, + "step": 11605 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010645761660368845, + "loss": 0.0, + "step": 11610 + }, + { + "epoch": 0.65, + "learning_rate": 0.00010637301900626022, + "loss": 0.0, + "step": 11615 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010628842140883198, + "loss": 0.0, + "step": 11620 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010620382381140375, + "loss": 0.0, + "step": 11625 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010611922621397551, + "loss": 0.0, + "step": 11630 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010603462861654728, + "loss": 0.0, + "step": 11635 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010595003101911905, + "loss": 0.0, + "step": 11640 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010586543342169081, + "loss": 0.0, + "step": 11645 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010578083582426258, + "loss": 0.0, + "step": 11650 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010569623822683436, + "loss": 0.0, + "step": 11655 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010561164062940611, + "loss": 0.0, + "step": 11660 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010552704303197789, + "loss": 0.0, + "step": 11665 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010544244543454965, + "loss": 0.0, + "step": 11670 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010535784783712142, + "loss": 0.0, + "step": 11675 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010527325023969319, + "loss": 0.0, + "step": 11680 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010518865264226495, + "loss": 0.0, + "step": 11685 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010510405504483672, + "loss": 0.0, + "step": 11690 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010501945744740848, + "loss": 0.0, + "step": 11695 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010493485984998025, + "loss": 0.0, + "step": 11700 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010485026225255202, + "loss": 0.0, + "step": 11705 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010476566465512378, + "loss": 0.0, + "step": 11710 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010468106705769556, + "loss": 0.0, + "step": 11715 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010459646946026732, + "loss": 0.0, + "step": 11720 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010451187186283909, + "loss": 0.0, + "step": 11725 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010442727426541086, + "loss": 0.0, + "step": 11730 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010434267666798262, + "loss": 0.0, + "step": 11735 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010425807907055439, + "loss": 0.0, + "step": 11740 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010417348147312615, + "loss": 0.0, + "step": 11745 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010408888387569792, + "loss": 0.0, + "step": 11750 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010400428627826969, + "loss": 0.0, + "step": 11755 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010391968868084145, + "loss": 0.0, + "step": 11760 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010383509108341323, + "loss": 0.0, + "step": 11765 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010375049348598499, + "loss": 0.0, + "step": 11770 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010366589588855676, + "loss": 0.0, + "step": 11775 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010358129829112853, + "loss": 0.0, + "step": 11780 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010349670069370029, + "loss": 0.0, + "step": 11785 + }, + { + "epoch": 0.66, + "learning_rate": 0.00010341210309627206, + "loss": 0.0, + "step": 11790 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010332750549884381, + "loss": 0.0, + "step": 11795 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010324290790141559, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010315831030398736, + "loss": 0.0, + "step": 11805 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010307371270655911, + "loss": 0.0, + "step": 11810 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001029891151091309, + "loss": 0.0, + "step": 11815 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010290451751170267, + "loss": 0.0, + "step": 11820 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010281991991427443, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001027353223168462, + "loss": 0.0, + "step": 11830 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010265072471941796, + "loss": 0.0, + "step": 11835 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010256612712198973, + "loss": 0.0, + "step": 11840 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001024815295245615, + "loss": 0.0, + "step": 11845 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010239693192713326, + "loss": 0.0, + "step": 11850 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010231233432970503, + "loss": 0.0, + "step": 11855 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010222773673227678, + "loss": 0.0, + "step": 11860 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010214313913484857, + "loss": 0.0, + "step": 11865 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010205854153742034, + "loss": 0.0, + "step": 11870 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001019739439399921, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010188934634256387, + "loss": 0.0, + "step": 11880 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010180474874513562, + "loss": 0.0, + "step": 11885 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001017201511477074, + "loss": 0.0, + "step": 11890 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010163555355027917, + "loss": 0.0, + "step": 11895 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010155095595285092, + "loss": 0.0, + "step": 11900 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001014663583554227, + "loss": 0.0, + "step": 11905 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010138176075799445, + "loss": 0.0, + "step": 11910 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010129716316056622, + "loss": 0.0, + "step": 11915 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010121256556313801, + "loss": 0.0, + "step": 11920 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010112796796570977, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010104337036828154, + "loss": 0.0, + "step": 11930 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001009587727708533, + "loss": 0.0, + "step": 11935 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010087417517342507, + "loss": 0.0, + "step": 11940 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010078957757599684, + "loss": 0.0, + "step": 11945 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001007049799785686, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010062038238114036, + "loss": 0.0, + "step": 11955 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010053578478371212, + "loss": 0.0, + "step": 11960 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010045118718628389, + "loss": 0.0, + "step": 11965 + }, + { + "epoch": 0.67, + "learning_rate": 0.00010036658958885568, + "loss": 0.0, + "step": 11970 + }, + { + "epoch": 0.68, + "learning_rate": 0.00010028199199142743, + "loss": 0.0, + "step": 11975 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001001973943939992, + "loss": 0.0, + "step": 11980 + }, + { + "epoch": 0.68, + "learning_rate": 0.00010011279679657098, + "loss": 0.0, + "step": 11985 + }, + { + "epoch": 0.68, + "learning_rate": 0.00010002819919914273, + "loss": 0.0, + "step": 11990 + }, + { + "epoch": 0.68, + "learning_rate": 9.99436016017145e-05, + "loss": 0.0, + "step": 11995 + }, + { + "epoch": 0.68, + "learning_rate": 9.985900400428626e-05, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 0.68, + "learning_rate": 9.977440640685803e-05, + "loss": 0.0, + "step": 12005 + }, + { + "epoch": 0.68, + "learning_rate": 9.96898088094298e-05, + "loss": 0.0, + "step": 12010 + }, + { + "epoch": 0.68, + "learning_rate": 9.960521121200156e-05, + "loss": 0.0, + "step": 12015 + }, + { + "epoch": 0.68, + "learning_rate": 9.952061361457335e-05, + "loss": 0.0, + "step": 12020 + }, + { + "epoch": 0.68, + "learning_rate": 9.94360160171451e-05, + "loss": 0.0, + "step": 12025 + }, + { + "epoch": 0.68, + "learning_rate": 9.935141841971688e-05, + "loss": 0.0, + "step": 12030 + }, + { + "epoch": 0.68, + "learning_rate": 9.926682082228865e-05, + "loss": 0.0, + "step": 12035 + }, + { + "epoch": 0.68, + "learning_rate": 9.91822232248604e-05, + "loss": 0.0, + "step": 12040 + }, + { + "epoch": 0.68, + "learning_rate": 9.909762562743217e-05, + "loss": 0.0, + "step": 12045 + }, + { + "epoch": 0.68, + "learning_rate": 9.901302803000393e-05, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 0.68, + "learning_rate": 9.89284304325757e-05, + "loss": 0.0, + "step": 12055 + }, + { + "epoch": 0.68, + "learning_rate": 9.884383283514747e-05, + "loss": 0.0, + "step": 12060 + }, + { + "epoch": 0.68, + "learning_rate": 9.875923523771923e-05, + "loss": 0.0, + "step": 12065 + }, + { + "epoch": 0.68, + "learning_rate": 9.867463764029102e-05, + "loss": 0.0, + "step": 12070 + }, + { + "epoch": 0.68, + "learning_rate": 9.859004004286277e-05, + "loss": 0.0, + "step": 12075 + }, + { + "epoch": 0.68, + "learning_rate": 9.850544244543454e-05, + "loss": 0.0, + "step": 12080 + }, + { + "epoch": 0.68, + "learning_rate": 9.842084484800632e-05, + "loss": 0.0, + "step": 12085 + }, + { + "epoch": 0.68, + "learning_rate": 9.833624725057807e-05, + "loss": 0.0, + "step": 12090 + }, + { + "epoch": 0.68, + "learning_rate": 9.825164965314984e-05, + "loss": 0.0, + "step": 12095 + }, + { + "epoch": 0.68, + "learning_rate": 9.81670520557216e-05, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 0.68, + "learning_rate": 9.808245445829337e-05, + "loss": 0.0, + "step": 12105 + }, + { + "epoch": 0.68, + "learning_rate": 9.799785686086514e-05, + "loss": 0.0, + "step": 12110 + }, + { + "epoch": 0.68, + "learning_rate": 9.79132592634369e-05, + "loss": 0.0, + "step": 12115 + }, + { + "epoch": 0.68, + "learning_rate": 9.782866166600868e-05, + "loss": 0.0, + "step": 12120 + }, + { + "epoch": 0.68, + "learning_rate": 9.774406406858043e-05, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 0.68, + "learning_rate": 9.765946647115221e-05, + "loss": 0.0, + "step": 12130 + }, + { + "epoch": 0.68, + "learning_rate": 9.757486887372398e-05, + "loss": 0.0, + "step": 12135 + }, + { + "epoch": 0.68, + "learning_rate": 9.749027127629574e-05, + "loss": 0.0, + "step": 12140 + }, + { + "epoch": 0.68, + "learning_rate": 9.740567367886751e-05, + "loss": 0.0, + "step": 12145 + }, + { + "epoch": 0.69, + "learning_rate": 9.732107608143928e-05, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 0.69, + "learning_rate": 9.723647848401104e-05, + "loss": 0.0, + "step": 12155 + }, + { + "epoch": 0.69, + "learning_rate": 9.715188088658281e-05, + "loss": 0.0, + "step": 12160 + }, + { + "epoch": 0.69, + "learning_rate": 9.706728328915457e-05, + "loss": 0.0, + "step": 12165 + }, + { + "epoch": 0.69, + "learning_rate": 9.698268569172634e-05, + "loss": 0.0, + "step": 12170 + }, + { + "epoch": 0.69, + "learning_rate": 9.689808809429813e-05, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 0.69, + "learning_rate": 9.681349049686988e-05, + "loss": 0.0, + "step": 12180 + }, + { + "epoch": 0.69, + "learning_rate": 9.672889289944165e-05, + "loss": 0.0, + "step": 12185 + }, + { + "epoch": 0.69, + "learning_rate": 9.664429530201341e-05, + "loss": 0.0, + "step": 12190 + }, + { + "epoch": 0.69, + "learning_rate": 9.655969770458518e-05, + "loss": 0.0, + "step": 12195 + }, + { + "epoch": 0.69, + "learning_rate": 9.647510010715695e-05, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 0.69, + "learning_rate": 9.639050250972871e-05, + "loss": 0.0, + "step": 12205 + }, + { + "epoch": 0.69, + "learning_rate": 9.630590491230048e-05, + "loss": 0.0, + "step": 12210 + }, + { + "epoch": 0.69, + "learning_rate": 9.622130731487224e-05, + "loss": 0.0, + "step": 12215 + }, + { + "epoch": 0.69, + "learning_rate": 9.613670971744401e-05, + "loss": 0.0, + "step": 12220 + }, + { + "epoch": 0.69, + "learning_rate": 9.60521121200158e-05, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 0.69, + "learning_rate": 9.596751452258755e-05, + "loss": 0.0, + "step": 12230 + }, + { + "epoch": 0.69, + "learning_rate": 9.588291692515932e-05, + "loss": 0.0, + "step": 12235 + }, + { + "epoch": 0.69, + "learning_rate": 9.579831932773108e-05, + "loss": 0.0, + "step": 12240 + }, + { + "epoch": 0.69, + "learning_rate": 9.571372173030285e-05, + "loss": 0.0, + "step": 12245 + }, + { + "epoch": 0.69, + "learning_rate": 9.562912413287462e-05, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 0.69, + "learning_rate": 9.554452653544638e-05, + "loss": 0.0, + "step": 12255 + }, + { + "epoch": 0.69, + "learning_rate": 9.545992893801815e-05, + "loss": 0.0, + "step": 12260 + }, + { + "epoch": 0.69, + "learning_rate": 9.537533134058991e-05, + "loss": 0.0, + "step": 12265 + }, + { + "epoch": 0.69, + "learning_rate": 9.529073374316168e-05, + "loss": 0.0, + "step": 12270 + }, + { + "epoch": 0.69, + "learning_rate": 9.520613614573346e-05, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 0.69, + "learning_rate": 9.512153854830522e-05, + "loss": 0.0, + "step": 12280 + }, + { + "epoch": 0.69, + "learning_rate": 9.503694095087699e-05, + "loss": 0.0, + "step": 12285 + }, + { + "epoch": 0.69, + "learning_rate": 9.495234335344875e-05, + "loss": 0.0, + "step": 12290 + }, + { + "epoch": 0.69, + "learning_rate": 9.486774575602052e-05, + "loss": 0.0, + "step": 12295 + }, + { + "epoch": 0.69, + "learning_rate": 9.478314815859229e-05, + "loss": 0.0, + "step": 12300 + }, + { + "epoch": 0.69, + "learning_rate": 9.469855056116405e-05, + "loss": 0.0, + "step": 12305 + }, + { + "epoch": 0.69, + "learning_rate": 9.461395296373582e-05, + "loss": 0.0, + "step": 12310 + }, + { + "epoch": 0.69, + "learning_rate": 9.452935536630759e-05, + "loss": 0.0, + "step": 12315 + }, + { + "epoch": 0.69, + "learning_rate": 9.444475776887935e-05, + "loss": 0.0, + "step": 12320 + }, + { + "epoch": 0.69, + "learning_rate": 9.436016017145113e-05, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 0.7, + "learning_rate": 9.427556257402289e-05, + "loss": 0.0, + "step": 12330 + }, + { + "epoch": 0.7, + "learning_rate": 9.419096497659466e-05, + "loss": 0.0, + "step": 12335 + }, + { + "epoch": 0.7, + "learning_rate": 9.410636737916643e-05, + "loss": 0.0, + "step": 12340 + }, + { + "epoch": 0.7, + "learning_rate": 9.402176978173819e-05, + "loss": 0.0, + "step": 12345 + }, + { + "epoch": 0.7, + "learning_rate": 9.393717218430996e-05, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 0.7, + "learning_rate": 9.385257458688172e-05, + "loss": 0.0, + "step": 12355 + }, + { + "epoch": 0.7, + "learning_rate": 9.376797698945349e-05, + "loss": 0.0, + "step": 12360 + }, + { + "epoch": 0.7, + "learning_rate": 9.368337939202526e-05, + "loss": 0.0, + "step": 12365 + }, + { + "epoch": 0.7, + "learning_rate": 9.359878179459702e-05, + "loss": 0.0, + "step": 12370 + }, + { + "epoch": 0.7, + "learning_rate": 9.35141841971688e-05, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 0.7, + "learning_rate": 9.342958659974055e-05, + "loss": 0.0, + "step": 12380 + }, + { + "epoch": 0.7, + "learning_rate": 9.334498900231233e-05, + "loss": 0.0, + "step": 12385 + }, + { + "epoch": 0.7, + "learning_rate": 9.32603914048841e-05, + "loss": 0.0, + "step": 12390 + }, + { + "epoch": 0.7, + "learning_rate": 9.317579380745586e-05, + "loss": 0.0, + "step": 12395 + }, + { + "epoch": 0.7, + "learning_rate": 9.309119621002763e-05, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 0.7, + "learning_rate": 9.300659861259939e-05, + "loss": 0.0, + "step": 12405 + }, + { + "epoch": 0.7, + "learning_rate": 9.292200101517116e-05, + "loss": 0.0, + "step": 12410 + }, + { + "epoch": 0.7, + "learning_rate": 9.283740341774293e-05, + "loss": 0.0, + "step": 12415 + }, + { + "epoch": 0.7, + "learning_rate": 9.275280582031469e-05, + "loss": 0.0, + "step": 12420 + }, + { + "epoch": 0.7, + "learning_rate": 9.266820822288646e-05, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 0.7, + "learning_rate": 9.258361062545821e-05, + "loss": 0.0, + "step": 12430 + }, + { + "epoch": 0.7, + "learning_rate": 9.249901302803e-05, + "loss": 0.0, + "step": 12435 + }, + { + "epoch": 0.7, + "learning_rate": 9.241441543060177e-05, + "loss": 0.0, + "step": 12440 + }, + { + "epoch": 0.7, + "learning_rate": 9.232981783317353e-05, + "loss": 0.0, + "step": 12445 + }, + { + "epoch": 0.7, + "learning_rate": 9.22452202357453e-05, + "loss": 0.0, + "step": 12450 + }, + { + "epoch": 0.7, + "learning_rate": 9.216062263831706e-05, + "loss": 0.0, + "step": 12455 + }, + { + "epoch": 0.7, + "learning_rate": 9.207602504088883e-05, + "loss": 0.0, + "step": 12460 + }, + { + "epoch": 0.7, + "learning_rate": 9.19914274434606e-05, + "loss": 0.0, + "step": 12465 + }, + { + "epoch": 0.7, + "learning_rate": 9.190682984603236e-05, + "loss": 0.0, + "step": 12470 + }, + { + "epoch": 0.7, + "learning_rate": 9.182223224860413e-05, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 0.7, + "learning_rate": 9.173763465117591e-05, + "loss": 0.0, + "step": 12480 + }, + { + "epoch": 0.7, + "learning_rate": 9.165303705374767e-05, + "loss": 0.0, + "step": 12485 + }, + { + "epoch": 0.7, + "learning_rate": 9.156843945631944e-05, + "loss": 0.0, + "step": 12490 + }, + { + "epoch": 0.7, + "learning_rate": 9.14838418588912e-05, + "loss": 0.0, + "step": 12495 + }, + { + "epoch": 0.7, + "learning_rate": 9.139924426146297e-05, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 0.71, + "learning_rate": 9.131464666403474e-05, + "loss": 0.0, + "step": 12505 + }, + { + "epoch": 0.71, + "learning_rate": 9.12300490666065e-05, + "loss": 0.0, + "step": 12510 + }, + { + "epoch": 0.71, + "learning_rate": 9.114545146917827e-05, + "loss": 0.0, + "step": 12515 + }, + { + "epoch": 0.71, + "learning_rate": 9.106085387175002e-05, + "loss": 0.0, + "step": 12520 + }, + { + "epoch": 0.71, + "learning_rate": 9.09762562743218e-05, + "loss": 0.0, + "step": 12525 + }, + { + "epoch": 0.71, + "learning_rate": 9.089165867689358e-05, + "loss": 0.0, + "step": 12530 + }, + { + "epoch": 0.71, + "learning_rate": 9.080706107946534e-05, + "loss": 0.0, + "step": 12535 + }, + { + "epoch": 0.71, + "learning_rate": 9.072246348203711e-05, + "loss": 0.0, + "step": 12540 + }, + { + "epoch": 0.71, + "learning_rate": 9.063786588460887e-05, + "loss": 0.0, + "step": 12545 + }, + { + "epoch": 0.71, + "learning_rate": 9.055326828718064e-05, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 0.71, + "learning_rate": 9.046867068975241e-05, + "loss": 0.0, + "step": 12555 + }, + { + "epoch": 0.71, + "learning_rate": 9.038407309232417e-05, + "loss": 0.0, + "step": 12560 + }, + { + "epoch": 0.71, + "learning_rate": 9.029947549489594e-05, + "loss": 0.0, + "step": 12565 + }, + { + "epoch": 0.71, + "learning_rate": 9.02148778974677e-05, + "loss": 0.0, + "step": 12570 + }, + { + "epoch": 0.71, + "learning_rate": 9.013028030003947e-05, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 0.71, + "learning_rate": 9.004568270261125e-05, + "loss": 0.0, + "step": 12580 + }, + { + "epoch": 0.71, + "learning_rate": 8.996108510518301e-05, + "loss": 0.0, + "step": 12585 + }, + { + "epoch": 0.71, + "learning_rate": 8.987648750775478e-05, + "loss": 0.0, + "step": 12590 + }, + { + "epoch": 0.71, + "learning_rate": 8.979188991032654e-05, + "loss": 0.0, + "step": 12595 + }, + { + "epoch": 0.71, + "learning_rate": 8.97072923128983e-05, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 0.71, + "learning_rate": 8.962269471547008e-05, + "loss": 0.0, + "step": 12605 + }, + { + "epoch": 0.71, + "learning_rate": 8.953809711804183e-05, + "loss": 0.0, + "step": 12610 + }, + { + "epoch": 0.71, + "learning_rate": 8.94534995206136e-05, + "loss": 0.0, + "step": 12615 + }, + { + "epoch": 0.71, + "learning_rate": 8.936890192318538e-05, + "loss": 0.0, + "step": 12620 + }, + { + "epoch": 0.71, + "learning_rate": 8.928430432575713e-05, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 0.71, + "learning_rate": 8.919970672832892e-05, + "loss": 0.0, + "step": 12630 + }, + { + "epoch": 0.71, + "learning_rate": 8.911510913090066e-05, + "loss": 0.0, + "step": 12635 + }, + { + "epoch": 0.71, + "learning_rate": 8.903051153347245e-05, + "loss": 0.0, + "step": 12640 + }, + { + "epoch": 0.71, + "learning_rate": 8.894591393604422e-05, + "loss": 0.0, + "step": 12645 + }, + { + "epoch": 0.71, + "learning_rate": 8.886131633861598e-05, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 0.71, + "learning_rate": 8.877671874118775e-05, + "loss": 0.0, + "step": 12655 + }, + { + "epoch": 0.71, + "learning_rate": 8.86921211437595e-05, + "loss": 0.0, + "step": 12660 + }, + { + "epoch": 0.71, + "learning_rate": 8.860752354633128e-05, + "loss": 0.0, + "step": 12665 + }, + { + "epoch": 0.71, + "learning_rate": 8.852292594890305e-05, + "loss": 0.0, + "step": 12670 + }, + { + "epoch": 0.71, + "learning_rate": 8.84383283514748e-05, + "loss": 0.0, + "step": 12675 + }, + { + "epoch": 0.71, + "learning_rate": 8.835373075404657e-05, + "loss": 0.0, + "step": 12680 + }, + { + "epoch": 0.72, + "learning_rate": 8.826913315661833e-05, + "loss": 0.0, + "step": 12685 + }, + { + "epoch": 0.72, + "learning_rate": 8.818453555919012e-05, + "loss": 0.0, + "step": 12690 + }, + { + "epoch": 0.72, + "learning_rate": 8.809993796176189e-05, + "loss": 0.0, + "step": 12695 + }, + { + "epoch": 0.72, + "learning_rate": 8.801534036433364e-05, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 0.72, + "learning_rate": 8.793074276690542e-05, + "loss": 0.0, + "step": 12705 + }, + { + "epoch": 0.72, + "learning_rate": 8.784614516947717e-05, + "loss": 0.0, + "step": 12710 + }, + { + "epoch": 0.72, + "learning_rate": 8.776154757204894e-05, + "loss": 0.0, + "step": 12715 + }, + { + "epoch": 0.72, + "learning_rate": 8.767694997462072e-05, + "loss": 0.0, + "step": 12720 + }, + { + "epoch": 0.72, + "learning_rate": 8.759235237719247e-05, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 0.72, + "learning_rate": 8.750775477976424e-05, + "loss": 0.0, + "step": 12730 + }, + { + "epoch": 0.72, + "learning_rate": 8.7423157182336e-05, + "loss": 0.0, + "step": 12735 + }, + { + "epoch": 0.72, + "learning_rate": 8.733855958490779e-05, + "loss": 0.0, + "step": 12740 + }, + { + "epoch": 0.72, + "learning_rate": 8.725396198747956e-05, + "loss": 0.0, + "step": 12745 + }, + { + "epoch": 0.72, + "learning_rate": 8.716936439005131e-05, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 0.72, + "learning_rate": 8.708476679262308e-05, + "loss": 0.0, + "step": 12755 + }, + { + "epoch": 0.72, + "learning_rate": 8.700016919519484e-05, + "loss": 0.0, + "step": 12760 + }, + { + "epoch": 0.72, + "learning_rate": 8.691557159776661e-05, + "loss": 0.0, + "step": 12765 + }, + { + "epoch": 0.72, + "learning_rate": 8.683097400033838e-05, + "loss": 0.0, + "step": 12770 + }, + { + "epoch": 0.72, + "learning_rate": 8.674637640291014e-05, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 0.72, + "learning_rate": 8.666177880548191e-05, + "loss": 0.0, + "step": 12780 + }, + { + "epoch": 0.72, + "learning_rate": 8.65771812080537e-05, + "loss": 0.0, + "step": 12785 + }, + { + "epoch": 0.72, + "learning_rate": 8.649258361062545e-05, + "loss": 0.0, + "step": 12790 + }, + { + "epoch": 0.72, + "learning_rate": 8.640798601319723e-05, + "loss": 0.0, + "step": 12795 + }, + { + "epoch": 0.72, + "learning_rate": 8.632338841576898e-05, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 0.72, + "learning_rate": 8.623879081834075e-05, + "loss": 0.0, + "step": 12805 + }, + { + "epoch": 0.72, + "learning_rate": 8.615419322091253e-05, + "loss": 0.0, + "step": 12810 + }, + { + "epoch": 0.72, + "learning_rate": 8.606959562348428e-05, + "loss": 0.0, + "step": 12815 + }, + { + "epoch": 0.72, + "learning_rate": 8.598499802605605e-05, + "loss": 0.0, + "step": 12820 + }, + { + "epoch": 0.72, + "learning_rate": 8.590040042862781e-05, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 0.72, + "learning_rate": 8.581580283119958e-05, + "loss": 0.0, + "step": 12830 + }, + { + "epoch": 0.72, + "learning_rate": 8.573120523377137e-05, + "loss": 0.0, + "step": 12835 + }, + { + "epoch": 0.72, + "learning_rate": 8.564660763634312e-05, + "loss": 0.0, + "step": 12840 + }, + { + "epoch": 0.72, + "learning_rate": 8.55620100389149e-05, + "loss": 0.0, + "step": 12845 + }, + { + "epoch": 0.72, + "learning_rate": 8.547741244148665e-05, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 0.72, + "learning_rate": 8.539281484405842e-05, + "loss": 0.0, + "step": 12855 + }, + { + "epoch": 0.73, + "learning_rate": 8.53082172466302e-05, + "loss": 0.0, + "step": 12860 + }, + { + "epoch": 0.73, + "learning_rate": 8.522361964920195e-05, + "loss": 0.0, + "step": 12865 + }, + { + "epoch": 0.73, + "learning_rate": 8.513902205177372e-05, + "loss": 0.0, + "step": 12870 + }, + { + "epoch": 0.73, + "learning_rate": 8.505442445434548e-05, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 0.73, + "learning_rate": 8.496982685691725e-05, + "loss": 0.0, + "step": 12880 + }, + { + "epoch": 0.73, + "learning_rate": 8.488522925948904e-05, + "loss": 0.0, + "step": 12885 + }, + { + "epoch": 0.73, + "learning_rate": 8.480063166206078e-05, + "loss": 0.0, + "step": 12890 + }, + { + "epoch": 0.73, + "learning_rate": 8.471603406463256e-05, + "loss": 0.0, + "step": 12895 + }, + { + "epoch": 0.73, + "learning_rate": 8.463143646720432e-05, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 0.73, + "learning_rate": 8.454683886977609e-05, + "loss": 0.0, + "step": 12905 + }, + { + "epoch": 0.73, + "learning_rate": 8.446224127234786e-05, + "loss": 0.0, + "step": 12910 + }, + { + "epoch": 0.73, + "learning_rate": 8.437764367491962e-05, + "loss": 0.0, + "step": 12915 + }, + { + "epoch": 0.73, + "learning_rate": 8.429304607749139e-05, + "loss": 0.0, + "step": 12920 + }, + { + "epoch": 0.73, + "learning_rate": 8.420844848006315e-05, + "loss": 0.0, + "step": 12925 + }, + { + "epoch": 0.73, + "learning_rate": 8.412385088263492e-05, + "loss": 0.0, + "step": 12930 + }, + { + "epoch": 0.73, + "learning_rate": 8.403925328520669e-05, + "loss": 0.0, + "step": 12935 + }, + { + "epoch": 0.73, + "learning_rate": 8.395465568777845e-05, + "loss": 0.0, + "step": 12940 + }, + { + "epoch": 0.73, + "learning_rate": 8.387005809035023e-05, + "loss": 0.0, + "step": 12945 + }, + { + "epoch": 0.73, + "learning_rate": 8.3785460492922e-05, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 0.73, + "learning_rate": 8.370086289549376e-05, + "loss": 0.0, + "step": 12955 + }, + { + "epoch": 0.73, + "learning_rate": 8.361626529806553e-05, + "loss": 0.0, + "step": 12960 + }, + { + "epoch": 0.73, + "learning_rate": 8.353166770063729e-05, + "loss": 0.0, + "step": 12965 + }, + { + "epoch": 0.73, + "learning_rate": 8.344707010320906e-05, + "loss": 0.0, + "step": 12970 + }, + { + "epoch": 0.73, + "learning_rate": 8.336247250578083e-05, + "loss": 0.0, + "step": 12975 + }, + { + "epoch": 0.73, + "learning_rate": 8.327787490835259e-05, + "loss": 0.0, + "step": 12980 + }, + { + "epoch": 0.73, + "learning_rate": 8.319327731092436e-05, + "loss": 0.0, + "step": 12985 + }, + { + "epoch": 0.73, + "learning_rate": 8.310867971349612e-05, + "loss": 0.0, + "step": 12990 + }, + { + "epoch": 0.73, + "learning_rate": 8.30240821160679e-05, + "loss": 0.0, + "step": 12995 + }, + { + "epoch": 0.73, + "learning_rate": 8.293948451863967e-05, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 0.73, + "learning_rate": 8.285488692121143e-05, + "loss": 0.0, + "step": 13005 + }, + { + "epoch": 0.73, + "learning_rate": 8.27702893237832e-05, + "loss": 0.0, + "step": 13010 + }, + { + "epoch": 0.73, + "learning_rate": 8.268569172635496e-05, + "loss": 0.0, + "step": 13015 + }, + { + "epoch": 0.73, + "learning_rate": 8.260109412892673e-05, + "loss": 0.0, + "step": 13020 + }, + { + "epoch": 0.73, + "learning_rate": 8.25164965314985e-05, + "loss": 0.0, + "step": 13025 + }, + { + "epoch": 0.73, + "learning_rate": 8.243189893407026e-05, + "loss": 0.0, + "step": 13030 + }, + { + "epoch": 0.73, + "learning_rate": 8.234730133664203e-05, + "loss": 0.0, + "step": 13035 + }, + { + "epoch": 0.74, + "learning_rate": 8.226270373921379e-05, + "loss": 0.0, + "step": 13040 + }, + { + "epoch": 0.74, + "learning_rate": 8.217810614178557e-05, + "loss": 0.0, + "step": 13045 + }, + { + "epoch": 0.74, + "learning_rate": 8.209350854435734e-05, + "loss": 0.0, + "step": 13050 + }, + { + "epoch": 0.74, + "learning_rate": 8.20089109469291e-05, + "loss": 0.0, + "step": 13055 + }, + { + "epoch": 0.74, + "learning_rate": 8.192431334950087e-05, + "loss": 0.0, + "step": 13060 + }, + { + "epoch": 0.74, + "learning_rate": 8.183971575207263e-05, + "loss": 0.0, + "step": 13065 + }, + { + "epoch": 0.74, + "learning_rate": 8.17551181546444e-05, + "loss": 0.0, + "step": 13070 + }, + { + "epoch": 0.74, + "learning_rate": 8.167052055721617e-05, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 0.74, + "learning_rate": 8.158592295978793e-05, + "loss": 0.0, + "step": 13080 + }, + { + "epoch": 0.74, + "learning_rate": 8.15013253623597e-05, + "loss": 0.0, + "step": 13085 + }, + { + "epoch": 0.74, + "learning_rate": 8.141672776493146e-05, + "loss": 0.0, + "step": 13090 + }, + { + "epoch": 0.74, + "learning_rate": 8.133213016750324e-05, + "loss": 0.0, + "step": 13095 + }, + { + "epoch": 0.74, + "learning_rate": 8.124753257007501e-05, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 0.74, + "learning_rate": 8.116293497264677e-05, + "loss": 0.0, + "step": 13105 + }, + { + "epoch": 0.74, + "learning_rate": 8.107833737521854e-05, + "loss": 0.0, + "step": 13110 + }, + { + "epoch": 0.74, + "learning_rate": 8.099373977779031e-05, + "loss": 0.0, + "step": 13115 + }, + { + "epoch": 0.74, + "learning_rate": 8.090914218036207e-05, + "loss": 0.0, + "step": 13120 + }, + { + "epoch": 0.74, + "learning_rate": 8.082454458293384e-05, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 0.74, + "learning_rate": 8.07399469855056e-05, + "loss": 0.0, + "step": 13130 + }, + { + "epoch": 0.74, + "learning_rate": 8.065534938807737e-05, + "loss": 0.0, + "step": 13135 + }, + { + "epoch": 0.74, + "learning_rate": 8.057075179064915e-05, + "loss": 0.0, + "step": 13140 + }, + { + "epoch": 0.74, + "learning_rate": 8.04861541932209e-05, + "loss": 0.0, + "step": 13145 + }, + { + "epoch": 0.74, + "learning_rate": 8.040155659579268e-05, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 0.74, + "learning_rate": 8.031695899836444e-05, + "loss": 0.0, + "step": 13155 + }, + { + "epoch": 0.74, + "learning_rate": 8.023236140093621e-05, + "loss": 0.0, + "step": 13160 + }, + { + "epoch": 0.74, + "learning_rate": 8.014776380350798e-05, + "loss": 0.0, + "step": 13165 + }, + { + "epoch": 0.74, + "learning_rate": 8.006316620607974e-05, + "loss": 0.0, + "step": 13170 + }, + { + "epoch": 0.74, + "learning_rate": 7.997856860865151e-05, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 0.74, + "learning_rate": 7.989397101122327e-05, + "loss": 0.0, + "step": 13180 + }, + { + "epoch": 0.74, + "learning_rate": 7.980937341379504e-05, + "loss": 0.0, + "step": 13185 + }, + { + "epoch": 0.74, + "learning_rate": 7.972477581636681e-05, + "loss": 0.0, + "step": 13190 + }, + { + "epoch": 0.74, + "learning_rate": 7.964017821893857e-05, + "loss": 0.0, + "step": 13195 + }, + { + "epoch": 0.74, + "learning_rate": 7.955558062151035e-05, + "loss": 0.0, + "step": 13200 + }, + { + "epoch": 0.74, + "learning_rate": 7.947098302408211e-05, + "loss": 0.0, + "step": 13205 + }, + { + "epoch": 0.74, + "learning_rate": 7.938638542665388e-05, + "loss": 0.0, + "step": 13210 + }, + { + "epoch": 0.75, + "learning_rate": 7.930178782922565e-05, + "loss": 0.0, + "step": 13215 + }, + { + "epoch": 0.75, + "learning_rate": 7.921719023179741e-05, + "loss": 0.0, + "step": 13220 + }, + { + "epoch": 0.75, + "learning_rate": 7.913259263436918e-05, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 0.75, + "learning_rate": 7.904799503694094e-05, + "loss": 0.0, + "step": 13230 + }, + { + "epoch": 0.75, + "learning_rate": 7.89633974395127e-05, + "loss": 0.0, + "step": 13235 + }, + { + "epoch": 0.75, + "learning_rate": 7.887879984208448e-05, + "loss": 0.0, + "step": 13240 + }, + { + "epoch": 0.75, + "learning_rate": 7.879420224465623e-05, + "loss": 0.0, + "step": 13245 + }, + { + "epoch": 0.75, + "learning_rate": 7.870960464722802e-05, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 0.75, + "learning_rate": 7.862500704979978e-05, + "loss": 0.0, + "step": 13255 + }, + { + "epoch": 0.75, + "learning_rate": 7.854040945237155e-05, + "loss": 0.0, + "step": 13260 + }, + { + "epoch": 0.75, + "learning_rate": 7.845581185494332e-05, + "loss": 0.0, + "step": 13265 + }, + { + "epoch": 0.75, + "learning_rate": 7.837121425751508e-05, + "loss": 0.0, + "step": 13270 + }, + { + "epoch": 0.75, + "learning_rate": 7.828661666008685e-05, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 0.75, + "learning_rate": 7.820201906265862e-05, + "loss": 0.0, + "step": 13280 + }, + { + "epoch": 0.75, + "learning_rate": 7.811742146523038e-05, + "loss": 0.0, + "step": 13285 + }, + { + "epoch": 0.75, + "learning_rate": 7.803282386780215e-05, + "loss": 0.0, + "step": 13290 + }, + { + "epoch": 0.75, + "learning_rate": 7.79482262703739e-05, + "loss": 0.0, + "step": 13295 + }, + { + "epoch": 0.75, + "learning_rate": 7.786362867294569e-05, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 0.75, + "eval_loss": NaN, + "eval_runtime": 4256.2357, + "eval_samples_per_second": 2.084, + "eval_steps_per_second": 0.261, + "step": 13302 + }, + { + "epoch": 0.75, + "learning_rate": 7.777903107551746e-05, + "loss": 0.0, + "step": 13305 + }, + { + "epoch": 0.75, + "learning_rate": 7.769443347808922e-05, + "loss": 0.0, + "step": 13310 + }, + { + "epoch": 0.75, + "learning_rate": 7.760983588066099e-05, + "loss": 0.0, + "step": 13315 + }, + { + "epoch": 0.75, + "learning_rate": 7.752523828323275e-05, + "loss": 0.0, + "step": 13320 + }, + { + "epoch": 0.75, + "learning_rate": 7.744064068580452e-05, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 0.75, + "learning_rate": 7.735604308837629e-05, + "loss": 0.0, + "step": 13330 + }, + { + "epoch": 0.75, + "learning_rate": 7.727144549094804e-05, + "loss": 0.0, + "step": 13335 + }, + { + "epoch": 0.75, + "learning_rate": 7.718684789351982e-05, + "loss": 0.0, + "step": 13340 + }, + { + "epoch": 0.75, + "learning_rate": 7.710225029609157e-05, + "loss": 0.0, + "step": 13345 + }, + { + "epoch": 0.75, + "learning_rate": 7.701765269866336e-05, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 0.75, + "learning_rate": 7.693305510123513e-05, + "loss": 0.0, + "step": 13355 + }, + { + "epoch": 0.75, + "learning_rate": 7.684845750380689e-05, + "loss": 0.0, + "step": 13360 + }, + { + "epoch": 0.75, + "learning_rate": 7.676385990637866e-05, + "loss": 0.0, + "step": 13365 + }, + { + "epoch": 0.75, + "learning_rate": 7.667926230895041e-05, + "loss": 0.0, + "step": 13370 + }, + { + "epoch": 0.75, + "learning_rate": 7.659466471152219e-05, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 0.75, + "learning_rate": 7.651006711409396e-05, + "loss": 0.0, + "step": 13380 + }, + { + "epoch": 0.75, + "learning_rate": 7.642546951666571e-05, + "loss": 0.0, + "step": 13385 + }, + { + "epoch": 0.75, + "learning_rate": 7.634087191923748e-05, + "loss": 0.0, + "step": 13390 + }, + { + "epoch": 0.76, + "learning_rate": 7.625627432180924e-05, + "loss": 0.0, + "step": 13395 + }, + { + "epoch": 0.76, + "learning_rate": 7.617167672438101e-05, + "loss": 0.0, + "step": 13400 + }, + { + "epoch": 0.76, + "learning_rate": 7.60870791269528e-05, + "loss": 0.0, + "step": 13405 + }, + { + "epoch": 0.76, + "learning_rate": 7.600248152952456e-05, + "loss": 0.0, + "step": 13410 + }, + { + "epoch": 0.76, + "learning_rate": 7.591788393209633e-05, + "loss": 0.0, + "step": 13415 + }, + { + "epoch": 0.76, + "learning_rate": 7.583328633466808e-05, + "loss": 0.0, + "step": 13420 + }, + { + "epoch": 0.76, + "learning_rate": 7.574868873723985e-05, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 0.76, + "learning_rate": 7.566409113981163e-05, + "loss": 0.0, + "step": 13430 + }, + { + "epoch": 0.76, + "learning_rate": 7.557949354238338e-05, + "loss": 0.0, + "step": 13435 + }, + { + "epoch": 0.76, + "learning_rate": 7.549489594495515e-05, + "loss": 0.0, + "step": 13440 + }, + { + "epoch": 0.76, + "learning_rate": 7.541029834752693e-05, + "loss": 0.0, + "step": 13445 + }, + { + "epoch": 0.76, + "learning_rate": 7.532570075009868e-05, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 0.76, + "learning_rate": 7.524110315267047e-05, + "loss": 0.0, + "step": 13455 + }, + { + "epoch": 0.76, + "learning_rate": 7.515650555524222e-05, + "loss": 0.0, + "step": 13460 + }, + { + "epoch": 0.76, + "learning_rate": 7.5071907957814e-05, + "loss": 0.0, + "step": 13465 + }, + { + "epoch": 0.76, + "learning_rate": 7.498731036038575e-05, + "loss": 0.0, + "step": 13470 + }, + { + "epoch": 0.76, + "learning_rate": 7.490271276295752e-05, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 0.76, + "learning_rate": 7.48181151655293e-05, + "loss": 0.0, + "step": 13480 + }, + { + "epoch": 0.76, + "learning_rate": 7.473351756810107e-05, + "loss": 0.0, + "step": 13485 + }, + { + "epoch": 0.76, + "learning_rate": 7.464891997067282e-05, + "loss": 0.0, + "step": 13490 + }, + { + "epoch": 0.76, + "learning_rate": 7.45643223732446e-05, + "loss": 0.0, + "step": 13495 + }, + { + "epoch": 0.76, + "learning_rate": 7.447972477581635e-05, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 0.76, + "learning_rate": 7.439512717838814e-05, + "loss": 0.0, + "step": 13505 + }, + { + "epoch": 0.76, + "learning_rate": 7.43105295809599e-05, + "loss": 0.0, + "step": 13510 + }, + { + "epoch": 0.76, + "learning_rate": 7.422593198353166e-05, + "loss": 0.0, + "step": 13515 + }, + { + "epoch": 0.76, + "learning_rate": 7.414133438610342e-05, + "loss": 0.0, + "step": 13520 + }, + { + "epoch": 0.76, + "learning_rate": 7.405673678867519e-05, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 0.76, + "learning_rate": 7.397213919124696e-05, + "loss": 0.0, + "step": 13530 + }, + { + "epoch": 0.76, + "learning_rate": 7.388754159381874e-05, + "loss": 0.0, + "step": 13535 + }, + { + "epoch": 0.76, + "learning_rate": 7.380294399639049e-05, + "loss": 0.0, + "step": 13540 + }, + { + "epoch": 0.76, + "learning_rate": 7.371834639896226e-05, + "loss": 0.0, + "step": 13545 + }, + { + "epoch": 0.76, + "learning_rate": 7.363374880153402e-05, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 0.76, + "learning_rate": 7.35491512041058e-05, + "loss": 0.0, + "step": 13555 + }, + { + "epoch": 0.76, + "learning_rate": 7.346455360667756e-05, + "loss": 0.0, + "step": 13560 + }, + { + "epoch": 0.76, + "learning_rate": 7.337995600924933e-05, + "loss": 0.0, + "step": 13565 + }, + { + "epoch": 0.77, + "learning_rate": 7.329535841182109e-05, + "loss": 0.0, + "step": 13570 + }, + { + "epoch": 0.77, + "learning_rate": 7.321076081439286e-05, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 0.77, + "learning_rate": 7.312616321696463e-05, + "loss": 0.0, + "step": 13580 + }, + { + "epoch": 0.77, + "learning_rate": 7.30415656195364e-05, + "loss": 0.0, + "step": 13585 + }, + { + "epoch": 0.77, + "learning_rate": 7.295696802210816e-05, + "loss": 0.0, + "step": 13590 + }, + { + "epoch": 0.77, + "learning_rate": 7.287237042467993e-05, + "loss": 0.0, + "step": 13595 + }, + { + "epoch": 0.77, + "learning_rate": 7.27877728272517e-05, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 0.77, + "learning_rate": 7.270317522982347e-05, + "loss": 0.0, + "step": 13605 + }, + { + "epoch": 0.77, + "learning_rate": 7.261857763239523e-05, + "loss": 0.0, + "step": 13610 + }, + { + "epoch": 0.77, + "learning_rate": 7.2533980034967e-05, + "loss": 0.0, + "step": 13615 + }, + { + "epoch": 0.77, + "learning_rate": 7.244938243753876e-05, + "loss": 0.0, + "step": 13620 + }, + { + "epoch": 0.77, + "learning_rate": 7.236478484011053e-05, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 0.77, + "learning_rate": 7.22801872426823e-05, + "loss": 0.0, + "step": 13630 + }, + { + "epoch": 0.77, + "learning_rate": 7.219558964525407e-05, + "loss": 0.0, + "step": 13635 + }, + { + "epoch": 0.77, + "learning_rate": 7.211099204782583e-05, + "loss": 0.0, + "step": 13640 + }, + { + "epoch": 0.77, + "learning_rate": 7.20263944503976e-05, + "loss": 0.0, + "step": 13645 + }, + { + "epoch": 0.77, + "learning_rate": 7.194179685296937e-05, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 0.77, + "learning_rate": 7.185719925554113e-05, + "loss": 0.0, + "step": 13655 + }, + { + "epoch": 0.77, + "learning_rate": 7.17726016581129e-05, + "loss": 0.0, + "step": 13660 + }, + { + "epoch": 0.77, + "learning_rate": 7.168800406068467e-05, + "loss": 0.0, + "step": 13665 + }, + { + "epoch": 0.77, + "learning_rate": 7.160340646325644e-05, + "loss": 0.0, + "step": 13670 + }, + { + "epoch": 0.77, + "learning_rate": 7.15188088658282e-05, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 0.77, + "learning_rate": 7.143421126839997e-05, + "loss": 0.0, + "step": 13680 + }, + { + "epoch": 0.77, + "learning_rate": 7.134961367097174e-05, + "loss": 0.0, + "step": 13685 + }, + { + "epoch": 0.77, + "learning_rate": 7.12650160735435e-05, + "loss": 0.0, + "step": 13690 + }, + { + "epoch": 0.77, + "learning_rate": 7.118041847611527e-05, + "loss": 0.0, + "step": 13695 + }, + { + "epoch": 0.77, + "learning_rate": 7.109582087868704e-05, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 0.77, + "learning_rate": 7.10112232812588e-05, + "loss": 0.0, + "step": 13705 + }, + { + "epoch": 0.77, + "learning_rate": 7.092662568383057e-05, + "loss": 0.0, + "step": 13710 + }, + { + "epoch": 0.77, + "learning_rate": 7.084202808640234e-05, + "loss": 0.0, + "step": 13715 + }, + { + "epoch": 0.77, + "learning_rate": 7.075743048897411e-05, + "loss": 0.0, + "step": 13720 + }, + { + "epoch": 0.77, + "learning_rate": 7.067283289154587e-05, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 0.77, + "learning_rate": 7.058823529411764e-05, + "loss": 0.0, + "step": 13730 + }, + { + "epoch": 0.77, + "learning_rate": 7.050363769668941e-05, + "loss": 0.0, + "step": 13735 + }, + { + "epoch": 0.77, + "learning_rate": 7.041904009926118e-05, + "loss": 0.0, + "step": 13740 + }, + { + "epoch": 0.77, + "learning_rate": 7.033444250183294e-05, + "loss": 0.0, + "step": 13745 + }, + { + "epoch": 0.78, + "learning_rate": 7.024984490440471e-05, + "loss": 0.0, + "step": 13750 + }, + { + "epoch": 0.78, + "learning_rate": 7.016524730697647e-05, + "loss": 0.0, + "step": 13755 + }, + { + "epoch": 0.78, + "learning_rate": 7.008064970954824e-05, + "loss": 0.0, + "step": 13760 + }, + { + "epoch": 0.78, + "learning_rate": 6.999605211212001e-05, + "loss": 0.0, + "step": 13765 + }, + { + "epoch": 0.78, + "learning_rate": 6.991145451469178e-05, + "loss": 0.0, + "step": 13770 + }, + { + "epoch": 0.78, + "learning_rate": 6.982685691726354e-05, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 0.78, + "learning_rate": 6.974225931983531e-05, + "loss": 0.0, + "step": 13780 + }, + { + "epoch": 0.78, + "learning_rate": 6.965766172240707e-05, + "loss": 0.0, + "step": 13785 + }, + { + "epoch": 0.78, + "learning_rate": 6.957306412497885e-05, + "loss": 0.0, + "step": 13790 + }, + { + "epoch": 0.78, + "learning_rate": 6.948846652755061e-05, + "loss": 0.0, + "step": 13795 + }, + { + "epoch": 0.78, + "learning_rate": 6.940386893012238e-05, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 0.78, + "learning_rate": 6.931927133269414e-05, + "loss": 0.0, + "step": 13805 + }, + { + "epoch": 0.78, + "learning_rate": 6.923467373526592e-05, + "loss": 0.0, + "step": 13810 + }, + { + "epoch": 0.78, + "learning_rate": 6.915007613783768e-05, + "loss": 0.0, + "step": 13815 + }, + { + "epoch": 0.78, + "learning_rate": 6.906547854040945e-05, + "loss": 0.0, + "step": 13820 + }, + { + "epoch": 0.78, + "learning_rate": 6.898088094298121e-05, + "loss": 0.0, + "step": 13825 + }, + { + "epoch": 0.78, + "learning_rate": 6.889628334555298e-05, + "loss": 0.0, + "step": 13830 + }, + { + "epoch": 0.78, + "learning_rate": 6.881168574812475e-05, + "loss": 0.0, + "step": 13835 + }, + { + "epoch": 0.78, + "learning_rate": 6.872708815069652e-05, + "loss": 0.0, + "step": 13840 + }, + { + "epoch": 0.78, + "learning_rate": 6.864249055326828e-05, + "loss": 0.0, + "step": 13845 + }, + { + "epoch": 0.78, + "learning_rate": 6.855789295584005e-05, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 0.78, + "learning_rate": 6.847329535841181e-05, + "loss": 0.0, + "step": 13855 + }, + { + "epoch": 0.78, + "learning_rate": 6.838869776098359e-05, + "loss": 0.0, + "step": 13860 + }, + { + "epoch": 0.78, + "learning_rate": 6.830410016355535e-05, + "loss": 0.0, + "step": 13865 + }, + { + "epoch": 0.78, + "learning_rate": 6.821950256612712e-05, + "loss": 0.0, + "step": 13870 + }, + { + "epoch": 0.78, + "learning_rate": 6.813490496869888e-05, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 0.78, + "learning_rate": 6.805030737127065e-05, + "loss": 0.0, + "step": 13880 + }, + { + "epoch": 0.78, + "learning_rate": 6.796570977384242e-05, + "loss": 0.0, + "step": 13885 + }, + { + "epoch": 0.78, + "learning_rate": 6.788111217641419e-05, + "loss": 0.0, + "step": 13890 + }, + { + "epoch": 0.78, + "learning_rate": 6.779651457898595e-05, + "loss": 0.0, + "step": 13895 + }, + { + "epoch": 0.78, + "learning_rate": 6.771191698155772e-05, + "loss": 0.0, + "step": 13900 + }, + { + "epoch": 0.78, + "learning_rate": 6.762731938412949e-05, + "loss": 0.0, + "step": 13905 + }, + { + "epoch": 0.78, + "learning_rate": 6.754272178670125e-05, + "loss": 0.0, + "step": 13910 + }, + { + "epoch": 0.78, + "learning_rate": 6.745812418927302e-05, + "loss": 0.0, + "step": 13915 + }, + { + "epoch": 0.78, + "learning_rate": 6.737352659184479e-05, + "loss": 0.0, + "step": 13920 + }, + { + "epoch": 0.79, + "learning_rate": 6.728892899441655e-05, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 0.79, + "learning_rate": 6.720433139698832e-05, + "loss": 0.0, + "step": 13930 + }, + { + "epoch": 0.79, + "learning_rate": 6.711973379956009e-05, + "loss": 0.0, + "step": 13935 + }, + { + "epoch": 0.79, + "learning_rate": 6.703513620213186e-05, + "loss": 0.0, + "step": 13940 + }, + { + "epoch": 0.79, + "learning_rate": 6.695053860470362e-05, + "loss": 0.0, + "step": 13945 + }, + { + "epoch": 0.79, + "learning_rate": 6.686594100727539e-05, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 0.79, + "learning_rate": 6.678134340984716e-05, + "loss": 0.0, + "step": 13955 + }, + { + "epoch": 0.79, + "learning_rate": 6.669674581241892e-05, + "loss": 0.0, + "step": 13960 + }, + { + "epoch": 0.79, + "learning_rate": 6.661214821499069e-05, + "loss": 0.0, + "step": 13965 + }, + { + "epoch": 0.79, + "learning_rate": 6.652755061756246e-05, + "loss": 0.0, + "step": 13970 + }, + { + "epoch": 0.79, + "learning_rate": 6.644295302013423e-05, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 0.79, + "learning_rate": 6.635835542270599e-05, + "loss": 0.0, + "step": 13980 + }, + { + "epoch": 0.79, + "learning_rate": 6.627375782527776e-05, + "loss": 0.0, + "step": 13985 + }, + { + "epoch": 0.79, + "learning_rate": 6.618916022784953e-05, + "loss": 0.0, + "step": 13990 + }, + { + "epoch": 0.79, + "learning_rate": 6.610456263042129e-05, + "loss": 0.0, + "step": 13995 + }, + { + "epoch": 0.79, + "learning_rate": 6.601996503299306e-05, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 0.79, + "learning_rate": 6.593536743556483e-05, + "loss": 0.0, + "step": 14005 + }, + { + "epoch": 0.79, + "learning_rate": 6.585076983813659e-05, + "loss": 0.0, + "step": 14010 + }, + { + "epoch": 0.79, + "learning_rate": 6.576617224070836e-05, + "loss": 0.0, + "step": 14015 + }, + { + "epoch": 0.79, + "learning_rate": 6.568157464328013e-05, + "loss": 0.0, + "step": 14020 + }, + { + "epoch": 0.79, + "learning_rate": 6.55969770458519e-05, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 0.79, + "learning_rate": 6.551237944842366e-05, + "loss": 0.0, + "step": 14030 + }, + { + "epoch": 0.79, + "learning_rate": 6.542778185099543e-05, + "loss": 0.0, + "step": 14035 + }, + { + "epoch": 0.79, + "learning_rate": 6.534318425356718e-05, + "loss": 0.0, + "step": 14040 + }, + { + "epoch": 0.79, + "learning_rate": 6.525858665613896e-05, + "loss": 0.0, + "step": 14045 + }, + { + "epoch": 0.79, + "learning_rate": 6.517398905871073e-05, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 0.79, + "learning_rate": 6.50893914612825e-05, + "loss": 0.0, + "step": 14055 + }, + { + "epoch": 0.79, + "learning_rate": 6.500479386385425e-05, + "loss": 0.0, + "step": 14060 + }, + { + "epoch": 0.79, + "learning_rate": 6.492019626642603e-05, + "loss": 0.0, + "step": 14065 + }, + { + "epoch": 0.79, + "learning_rate": 6.48355986689978e-05, + "loss": 0.0, + "step": 14070 + }, + { + "epoch": 0.79, + "learning_rate": 6.475100107156957e-05, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 0.79, + "learning_rate": 6.466640347414133e-05, + "loss": 0.0, + "step": 14080 + }, + { + "epoch": 0.79, + "learning_rate": 6.45818058767131e-05, + "loss": 0.0, + "step": 14085 + }, + { + "epoch": 0.79, + "learning_rate": 6.449720827928485e-05, + "loss": 0.0, + "step": 14090 + }, + { + "epoch": 0.79, + "learning_rate": 6.441261068185664e-05, + "loss": 0.0, + "step": 14095 + }, + { + "epoch": 0.79, + "learning_rate": 6.43280130844284e-05, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 0.8, + "learning_rate": 6.424341548700017e-05, + "loss": 0.0, + "step": 14105 + }, + { + "epoch": 0.8, + "learning_rate": 6.415881788957192e-05, + "loss": 0.0, + "step": 14110 + }, + { + "epoch": 0.8, + "learning_rate": 6.40742202921437e-05, + "loss": 0.0, + "step": 14115 + }, + { + "epoch": 0.8, + "learning_rate": 6.398962269471547e-05, + "loss": 0.0, + "step": 14120 + }, + { + "epoch": 0.8, + "learning_rate": 6.390502509728724e-05, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 0.8, + "learning_rate": 6.3820427499859e-05, + "loss": 0.0, + "step": 14130 + }, + { + "epoch": 0.8, + "learning_rate": 6.373582990243077e-05, + "loss": 0.0, + "step": 14135 + }, + { + "epoch": 0.8, + "learning_rate": 6.365123230500254e-05, + "loss": 0.0, + "step": 14140 + }, + { + "epoch": 0.8, + "learning_rate": 6.356663470757431e-05, + "loss": 0.0, + "step": 14145 + }, + { + "epoch": 0.8, + "learning_rate": 6.348203711014606e-05, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 0.8, + "learning_rate": 6.339743951271784e-05, + "loss": 0.0, + "step": 14155 + }, + { + "epoch": 0.8, + "learning_rate": 6.331284191528959e-05, + "loss": 0.0, + "step": 14160 + }, + { + "epoch": 0.8, + "learning_rate": 6.322824431786136e-05, + "loss": 0.0, + "step": 14165 + }, + { + "epoch": 0.8, + "learning_rate": 6.314364672043314e-05, + "loss": 0.0, + "step": 14170 + }, + { + "epoch": 0.8, + "learning_rate": 6.30590491230049e-05, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 0.8, + "learning_rate": 6.297445152557666e-05, + "loss": 0.0, + "step": 14180 + }, + { + "epoch": 0.8, + "learning_rate": 6.288985392814843e-05, + "loss": 0.0, + "step": 14185 + }, + { + "epoch": 0.8, + "learning_rate": 6.28052563307202e-05, + "loss": 0.0, + "step": 14190 + }, + { + "epoch": 0.8, + "learning_rate": 6.272065873329198e-05, + "loss": 0.0, + "step": 14195 + }, + { + "epoch": 0.8, + "learning_rate": 6.263606113586373e-05, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 0.8, + "learning_rate": 6.25514635384355e-05, + "loss": 0.0, + "step": 14205 + }, + { + "epoch": 0.8, + "learning_rate": 6.246686594100726e-05, + "loss": 0.0, + "step": 14210 + }, + { + "epoch": 0.8, + "learning_rate": 6.238226834357903e-05, + "loss": 0.0, + "step": 14215 + }, + { + "epoch": 0.8, + "learning_rate": 6.22976707461508e-05, + "loss": 0.0, + "step": 14220 + }, + { + "epoch": 0.8, + "learning_rate": 6.221307314872258e-05, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 0.8, + "learning_rate": 6.212847555129433e-05, + "loss": 0.0, + "step": 14230 + }, + { + "epoch": 0.8, + "learning_rate": 6.20438779538661e-05, + "loss": 0.0, + "step": 14235 + }, + { + "epoch": 0.8, + "learning_rate": 6.195928035643787e-05, + "loss": 0.0, + "step": 14240 + }, + { + "epoch": 0.8, + "learning_rate": 6.187468275900963e-05, + "loss": 0.0, + "step": 14245 + }, + { + "epoch": 0.8, + "learning_rate": 6.17900851615814e-05, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 0.8, + "learning_rate": 6.170548756415317e-05, + "loss": 0.0, + "step": 14255 + }, + { + "epoch": 0.8, + "learning_rate": 6.162088996672495e-05, + "loss": 0.0, + "step": 14260 + }, + { + "epoch": 0.8, + "learning_rate": 6.15362923692967e-05, + "loss": 0.0, + "step": 14265 + }, + { + "epoch": 0.8, + "learning_rate": 6.145169477186847e-05, + "loss": 0.0, + "step": 14270 + }, + { + "epoch": 0.8, + "learning_rate": 6.136709717444024e-05, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 0.81, + "learning_rate": 6.1282499577012e-05, + "loss": 0.0, + "step": 14280 + }, + { + "epoch": 0.81, + "learning_rate": 6.119790197958377e-05, + "loss": 0.0, + "step": 14285 + }, + { + "epoch": 0.81, + "learning_rate": 6.111330438215554e-05, + "loss": 0.0, + "step": 14290 + }, + { + "epoch": 0.81, + "learning_rate": 6.102870678472731e-05, + "loss": 0.0, + "step": 14295 + }, + { + "epoch": 0.81, + "learning_rate": 6.094410918729907e-05, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 0.81, + "learning_rate": 6.085951158987084e-05, + "loss": 0.0, + "step": 14305 + }, + { + "epoch": 0.81, + "learning_rate": 6.0774913992442614e-05, + "loss": 0.0, + "step": 14310 + }, + { + "epoch": 0.81, + "learning_rate": 6.069031639501438e-05, + "loss": 0.0, + "step": 14315 + }, + { + "epoch": 0.81, + "learning_rate": 6.060571879758614e-05, + "loss": 0.0, + "step": 14320 + }, + { + "epoch": 0.81, + "learning_rate": 6.052112120015791e-05, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 0.81, + "learning_rate": 6.043652360272968e-05, + "loss": 0.0, + "step": 14330 + }, + { + "epoch": 0.81, + "learning_rate": 6.035192600530145e-05, + "loss": 0.0, + "step": 14335 + }, + { + "epoch": 0.81, + "learning_rate": 6.026732840787321e-05, + "loss": 0.0, + "step": 14340 + }, + { + "epoch": 0.81, + "learning_rate": 6.018273081044498e-05, + "loss": 0.0, + "step": 14345 + }, + { + "epoch": 0.81, + "learning_rate": 6.009813321301674e-05, + "loss": 0.0, + "step": 14350 + }, + { + "epoch": 0.81, + "learning_rate": 6.001353561558851e-05, + "loss": 0.0, + "step": 14355 + }, + { + "epoch": 0.81, + "learning_rate": 5.9928938018160283e-05, + "loss": 0.0, + "step": 14360 + }, + { + "epoch": 0.81, + "learning_rate": 5.984434042073205e-05, + "loss": 0.0, + "step": 14365 + }, + { + "epoch": 0.81, + "learning_rate": 5.975974282330381e-05, + "loss": 0.0, + "step": 14370 + }, + { + "epoch": 0.81, + "learning_rate": 5.9675145225875576e-05, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 0.81, + "learning_rate": 5.959054762844735e-05, + "loss": 0.0, + "step": 14380 + }, + { + "epoch": 0.81, + "learning_rate": 5.950595003101911e-05, + "loss": 0.0, + "step": 14385 + }, + { + "epoch": 0.81, + "learning_rate": 5.942135243359088e-05, + "loss": 0.0, + "step": 14390 + }, + { + "epoch": 0.81, + "learning_rate": 5.9336754836162646e-05, + "loss": 0.0, + "step": 14395 + }, + { + "epoch": 0.81, + "learning_rate": 5.925215723873442e-05, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 0.81, + "learning_rate": 5.916755964130618e-05, + "loss": 0.0, + "step": 14405 + }, + { + "epoch": 0.81, + "learning_rate": 5.9082962043877946e-05, + "loss": 0.0, + "step": 14410 + }, + { + "epoch": 0.81, + "learning_rate": 5.899836444644972e-05, + "loss": 0.0, + "step": 14415 + }, + { + "epoch": 0.81, + "learning_rate": 5.891376684902148e-05, + "loss": 0.0, + "step": 14420 + }, + { + "epoch": 0.81, + "learning_rate": 5.882916925159325e-05, + "loss": 0.0, + "step": 14425 + }, + { + "epoch": 0.81, + "learning_rate": 5.8744571654165016e-05, + "loss": 0.0, + "step": 14430 + }, + { + "epoch": 0.81, + "learning_rate": 5.865997405673678e-05, + "loss": 0.0, + "step": 14435 + }, + { + "epoch": 0.81, + "learning_rate": 5.857537645930855e-05, + "loss": 0.0, + "step": 14440 + }, + { + "epoch": 0.81, + "learning_rate": 5.8490778861880316e-05, + "loss": 0.0, + "step": 14445 + }, + { + "epoch": 0.81, + "learning_rate": 5.840618126445209e-05, + "loss": 0.0, + "step": 14450 + }, + { + "epoch": 0.82, + "learning_rate": 5.832158366702385e-05, + "loss": 0.0, + "step": 14455 + }, + { + "epoch": 0.82, + "learning_rate": 5.8236986069595615e-05, + "loss": 0.0, + "step": 14460 + }, + { + "epoch": 0.82, + "learning_rate": 5.8152388472167386e-05, + "loss": 0.0, + "step": 14465 + }, + { + "epoch": 0.82, + "learning_rate": 5.806779087473916e-05, + "loss": 0.0, + "step": 14470 + }, + { + "epoch": 0.82, + "learning_rate": 5.798319327731092e-05, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 0.82, + "learning_rate": 5.7898595679882685e-05, + "loss": 0.0, + "step": 14480 + }, + { + "epoch": 0.82, + "learning_rate": 5.781399808245445e-05, + "loss": 0.0, + "step": 14485 + }, + { + "epoch": 0.82, + "learning_rate": 5.7729400485026214e-05, + "loss": 0.0, + "step": 14490 + }, + { + "epoch": 0.82, + "learning_rate": 5.764480288759799e-05, + "loss": 0.0, + "step": 14495 + }, + { + "epoch": 0.82, + "learning_rate": 5.7560205290169756e-05, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 0.82, + "learning_rate": 5.747560769274152e-05, + "loss": 0.0, + "step": 14505 + }, + { + "epoch": 0.82, + "learning_rate": 5.7391010095313284e-05, + "loss": 0.0, + "step": 14510 + }, + { + "epoch": 0.82, + "learning_rate": 5.730641249788505e-05, + "loss": 0.0, + "step": 14515 + }, + { + "epoch": 0.82, + "learning_rate": 5.7221814900456826e-05, + "loss": 0.0, + "step": 14520 + }, + { + "epoch": 0.82, + "learning_rate": 5.713721730302859e-05, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 0.82, + "learning_rate": 5.7052619705600355e-05, + "loss": 0.0, + "step": 14530 + }, + { + "epoch": 0.82, + "learning_rate": 5.696802210817212e-05, + "loss": 0.0, + "step": 14535 + }, + { + "epoch": 0.82, + "learning_rate": 5.688342451074388e-05, + "loss": 0.0, + "step": 14540 + }, + { + "epoch": 0.82, + "learning_rate": 5.679882691331566e-05, + "loss": 0.0, + "step": 14545 + }, + { + "epoch": 0.82, + "learning_rate": 5.6714229315887425e-05, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 0.82, + "learning_rate": 5.662963171845919e-05, + "loss": 0.0, + "step": 14555 + }, + { + "epoch": 0.82, + "learning_rate": 5.6545034121030953e-05, + "loss": 0.0, + "step": 14560 + }, + { + "epoch": 0.82, + "learning_rate": 5.646043652360273e-05, + "loss": 0.0, + "step": 14565 + }, + { + "epoch": 0.82, + "learning_rate": 5.6375838926174495e-05, + "loss": 0.0, + "step": 14570 + }, + { + "epoch": 0.82, + "learning_rate": 5.629124132874626e-05, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 0.82, + "learning_rate": 5.6206643731318024e-05, + "loss": 0.0, + "step": 14580 + }, + { + "epoch": 0.82, + "learning_rate": 5.612204613388979e-05, + "loss": 0.0, + "step": 14585 + }, + { + "epoch": 0.82, + "learning_rate": 5.6037448536461566e-05, + "loss": 0.0, + "step": 14590 + }, + { + "epoch": 0.82, + "learning_rate": 5.595285093903333e-05, + "loss": 0.0, + "step": 14595 + }, + { + "epoch": 0.82, + "learning_rate": 5.5868253341605094e-05, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 0.82, + "learning_rate": 5.578365574417686e-05, + "loss": 0.0, + "step": 14605 + }, + { + "epoch": 0.82, + "learning_rate": 5.569905814674862e-05, + "loss": 0.0, + "step": 14610 + }, + { + "epoch": 0.82, + "learning_rate": 5.56144605493204e-05, + "loss": 0.0, + "step": 14615 + }, + { + "epoch": 0.82, + "learning_rate": 5.5529862951892165e-05, + "loss": 0.0, + "step": 14620 + }, + { + "epoch": 0.82, + "learning_rate": 5.544526535446393e-05, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 0.82, + "learning_rate": 5.536066775703569e-05, + "loss": 0.0, + "step": 14630 + }, + { + "epoch": 0.83, + "learning_rate": 5.5276070159607464e-05, + "loss": 0.0, + "step": 14635 + }, + { + "epoch": 0.83, + "learning_rate": 5.519147256217923e-05, + "loss": 0.0, + "step": 14640 + }, + { + "epoch": 0.83, + "learning_rate": 5.5106874964751e-05, + "loss": 0.0, + "step": 14645 + }, + { + "epoch": 0.83, + "learning_rate": 5.502227736732276e-05, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 0.83, + "learning_rate": 5.493767976989453e-05, + "loss": 0.0, + "step": 14655 + }, + { + "epoch": 0.83, + "learning_rate": 5.48530821724663e-05, + "loss": 0.0, + "step": 14660 + }, + { + "epoch": 0.83, + "learning_rate": 5.476848457503806e-05, + "loss": 0.0, + "step": 14665 + }, + { + "epoch": 0.83, + "learning_rate": 5.4683886977609834e-05, + "loss": 0.0, + "step": 14670 + }, + { + "epoch": 0.83, + "learning_rate": 5.45992893801816e-05, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 0.83, + "learning_rate": 5.451469178275336e-05, + "loss": 0.0, + "step": 14680 + }, + { + "epoch": 0.83, + "learning_rate": 5.443009418532513e-05, + "loss": 0.0, + "step": 14685 + }, + { + "epoch": 0.83, + "learning_rate": 5.43454965878969e-05, + "loss": 0.0, + "step": 14690 + }, + { + "epoch": 0.83, + "learning_rate": 5.426089899046867e-05, + "loss": 0.0, + "step": 14695 + }, + { + "epoch": 0.83, + "learning_rate": 5.417630139304043e-05, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 0.83, + "learning_rate": 5.4091703795612204e-05, + "loss": 0.0, + "step": 14705 + }, + { + "epoch": 0.83, + "learning_rate": 5.400710619818397e-05, + "loss": 0.0, + "step": 14710 + }, + { + "epoch": 0.83, + "learning_rate": 5.392250860075573e-05, + "loss": 0.0, + "step": 14715 + }, + { + "epoch": 0.83, + "learning_rate": 5.38379110033275e-05, + "loss": 0.0, + "step": 14720 + }, + { + "epoch": 0.83, + "learning_rate": 5.375331340589927e-05, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 0.83, + "learning_rate": 5.366871580847104e-05, + "loss": 0.0, + "step": 14730 + }, + { + "epoch": 0.83, + "learning_rate": 5.35841182110428e-05, + "loss": 0.0, + "step": 14735 + }, + { + "epoch": 0.83, + "learning_rate": 5.3499520613614567e-05, + "loss": 0.0, + "step": 14740 + }, + { + "epoch": 0.83, + "learning_rate": 5.341492301618633e-05, + "loss": 0.0, + "step": 14745 + }, + { + "epoch": 0.83, + "learning_rate": 5.33303254187581e-05, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 0.83, + "learning_rate": 5.324572782132987e-05, + "loss": 0.0, + "step": 14755 + }, + { + "epoch": 0.83, + "learning_rate": 5.316113022390164e-05, + "loss": 0.0, + "step": 14760 + }, + { + "epoch": 0.83, + "learning_rate": 5.30765326264734e-05, + "loss": 0.0, + "step": 14765 + }, + { + "epoch": 0.83, + "learning_rate": 5.2991935029045165e-05, + "loss": 0.0, + "step": 14770 + }, + { + "epoch": 0.83, + "learning_rate": 5.2907337431616936e-05, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 0.83, + "learning_rate": 5.282273983418871e-05, + "loss": 0.0, + "step": 14780 + }, + { + "epoch": 0.83, + "learning_rate": 5.273814223676047e-05, + "loss": 0.0, + "step": 14785 + }, + { + "epoch": 0.83, + "learning_rate": 5.2653544639332236e-05, + "loss": 0.0, + "step": 14790 + }, + { + "epoch": 0.83, + "learning_rate": 5.2568947041904e-05, + "loss": 0.0, + "step": 14795 + }, + { + "epoch": 0.83, + "learning_rate": 5.248434944447578e-05, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 0.83, + "learning_rate": 5.239975184704754e-05, + "loss": 0.0, + "step": 14805 + }, + { + "epoch": 0.84, + "learning_rate": 5.2315154249619306e-05, + "loss": 0.0, + "step": 14810 + }, + { + "epoch": 0.84, + "learning_rate": 5.223055665219107e-05, + "loss": 0.0, + "step": 14815 + }, + { + "epoch": 0.84, + "learning_rate": 5.2145959054762835e-05, + "loss": 0.0, + "step": 14820 + }, + { + "epoch": 0.84, + "learning_rate": 5.206136145733461e-05, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 0.84, + "learning_rate": 5.1976763859906377e-05, + "loss": 0.0, + "step": 14830 + }, + { + "epoch": 0.84, + "learning_rate": 5.189216626247814e-05, + "loss": 0.0, + "step": 14835 + }, + { + "epoch": 0.84, + "learning_rate": 5.1807568665049905e-05, + "loss": 0.0, + "step": 14840 + }, + { + "epoch": 0.84, + "learning_rate": 5.172297106762167e-05, + "loss": 0.0, + "step": 14845 + }, + { + "epoch": 0.84, + "learning_rate": 5.163837347019345e-05, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 0.84, + "learning_rate": 5.155377587276521e-05, + "loss": 0.0, + "step": 14855 + }, + { + "epoch": 0.84, + "learning_rate": 5.1469178275336975e-05, + "loss": 0.0, + "step": 14860 + }, + { + "epoch": 0.84, + "learning_rate": 5.138458067790874e-05, + "loss": 0.0, + "step": 14865 + }, + { + "epoch": 0.84, + "learning_rate": 5.129998308048052e-05, + "loss": 0.0, + "step": 14870 + }, + { + "epoch": 0.84, + "learning_rate": 5.121538548305228e-05, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 0.84, + "learning_rate": 5.1130787885624046e-05, + "loss": 0.0, + "step": 14880 + }, + { + "epoch": 0.84, + "learning_rate": 5.104619028819581e-05, + "loss": 0.0, + "step": 14885 + }, + { + "epoch": 0.84, + "learning_rate": 5.0961592690767574e-05, + "loss": 0.0, + "step": 14890 + }, + { + "epoch": 0.84, + "learning_rate": 5.0876995093339345e-05, + "loss": 0.0, + "step": 14895 + }, + { + "epoch": 0.84, + "learning_rate": 5.0792397495911116e-05, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 0.84, + "learning_rate": 5.070779989848288e-05, + "loss": 0.0, + "step": 14905 + }, + { + "epoch": 0.84, + "learning_rate": 5.0623202301054644e-05, + "loss": 0.0, + "step": 14910 + }, + { + "epoch": 0.84, + "learning_rate": 5.053860470362641e-05, + "loss": 0.0, + "step": 14915 + }, + { + "epoch": 0.84, + "learning_rate": 5.045400710619818e-05, + "loss": 0.0, + "step": 14920 + }, + { + "epoch": 0.84, + "learning_rate": 5.036940950876995e-05, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 0.84, + "learning_rate": 5.0284811911341715e-05, + "loss": 0.0, + "step": 14930 + }, + { + "epoch": 0.84, + "learning_rate": 5.020021431391348e-05, + "loss": 0.0, + "step": 14935 + }, + { + "epoch": 0.84, + "learning_rate": 5.011561671648524e-05, + "loss": 0.0, + "step": 14940 + }, + { + "epoch": 0.84, + "learning_rate": 5.0031019119057014e-05, + "loss": 0.0, + "step": 14945 + }, + { + "epoch": 0.84, + "learning_rate": 4.9946421521628785e-05, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 0.84, + "learning_rate": 4.986182392420055e-05, + "loss": 0.0, + "step": 14955 + }, + { + "epoch": 0.84, + "learning_rate": 4.9777226326772314e-05, + "loss": 0.0, + "step": 14960 + }, + { + "epoch": 0.84, + "learning_rate": 4.9692628729344085e-05, + "loss": 0.0, + "step": 14965 + }, + { + "epoch": 0.84, + "learning_rate": 4.960803113191585e-05, + "loss": 0.0, + "step": 14970 + }, + { + "epoch": 0.84, + "learning_rate": 4.952343353448762e-05, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 0.84, + "learning_rate": 4.9438835937059384e-05, + "loss": 0.0, + "step": 14980 + }, + { + "epoch": 0.84, + "learning_rate": 4.935423833963115e-05, + "loss": 0.0, + "step": 14985 + }, + { + "epoch": 0.85, + "learning_rate": 4.926964074220292e-05, + "loss": 0.0, + "step": 14990 + }, + { + "epoch": 0.85, + "learning_rate": 4.9185043144774683e-05, + "loss": 0.0, + "step": 14995 + }, + { + "epoch": 0.85, + "learning_rate": 4.910044554734645e-05, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 0.85, + "learning_rate": 4.901584794991822e-05, + "loss": 0.0, + "step": 15005 + }, + { + "epoch": 0.85, + "learning_rate": 4.893125035248998e-05, + "loss": 0.0, + "step": 15010 + }, + { + "epoch": 0.85, + "learning_rate": 4.8846652755061754e-05, + "loss": 0.0, + "step": 15015 + }, + { + "epoch": 0.85, + "learning_rate": 4.876205515763352e-05, + "loss": 0.0, + "step": 15020 + }, + { + "epoch": 0.85, + "learning_rate": 4.867745756020528e-05, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 0.85, + "learning_rate": 4.859285996277705e-05, + "loss": 0.0, + "step": 15030 + }, + { + "epoch": 0.85, + "learning_rate": 4.8508262365348824e-05, + "loss": 0.0, + "step": 15035 + }, + { + "epoch": 0.85, + "learning_rate": 4.842366476792059e-05, + "loss": 0.0, + "step": 15040 + }, + { + "epoch": 0.85, + "learning_rate": 4.833906717049235e-05, + "loss": 0.0, + "step": 15045 + }, + { + "epoch": 0.85, + "learning_rate": 4.825446957306412e-05, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 0.85, + "learning_rate": 4.816987197563589e-05, + "loss": 0.0, + "step": 15055 + }, + { + "epoch": 0.85, + "learning_rate": 4.808527437820766e-05, + "loss": 0.0, + "step": 15060 + }, + { + "epoch": 0.85, + "learning_rate": 4.800067678077942e-05, + "loss": 0.0, + "step": 15065 + }, + { + "epoch": 0.85, + "learning_rate": 4.791607918335119e-05, + "loss": 0.0, + "step": 15070 + }, + { + "epoch": 0.85, + "learning_rate": 4.783148158592295e-05, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 0.85, + "learning_rate": 4.774688398849472e-05, + "loss": 0.0, + "step": 15080 + }, + { + "epoch": 0.85, + "learning_rate": 4.7662286391066493e-05, + "loss": 0.0, + "step": 15085 + }, + { + "epoch": 0.85, + "learning_rate": 4.757768879363826e-05, + "loss": 0.0, + "step": 15090 + }, + { + "epoch": 0.85, + "learning_rate": 4.749309119621002e-05, + "loss": 0.0, + "step": 15095 + }, + { + "epoch": 0.85, + "learning_rate": 4.7408493598781786e-05, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 0.85, + "learning_rate": 4.732389600135355e-05, + "loss": 0.0, + "step": 15105 + }, + { + "epoch": 0.85, + "learning_rate": 4.723929840392533e-05, + "loss": 0.0, + "step": 15110 + }, + { + "epoch": 0.85, + "learning_rate": 4.715470080649709e-05, + "loss": 0.0, + "step": 15115 + }, + { + "epoch": 0.85, + "learning_rate": 4.7070103209068856e-05, + "loss": 0.0, + "step": 15120 + }, + { + "epoch": 0.85, + "learning_rate": 4.698550561164062e-05, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 0.85, + "learning_rate": 4.69009080142124e-05, + "loss": 0.0, + "step": 15130 + }, + { + "epoch": 0.85, + "learning_rate": 4.681631041678416e-05, + "loss": 0.0, + "step": 15135 + }, + { + "epoch": 0.85, + "learning_rate": 4.673171281935593e-05, + "loss": 0.0, + "step": 15140 + }, + { + "epoch": 0.85, + "learning_rate": 4.664711522192769e-05, + "loss": 0.0, + "step": 15145 + }, + { + "epoch": 0.85, + "learning_rate": 4.6562517624499455e-05, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 0.85, + "learning_rate": 4.647792002707123e-05, + "loss": 0.0, + "step": 15155 + }, + { + "epoch": 0.85, + "learning_rate": 4.6393322429643e-05, + "loss": 0.0, + "step": 15160 + }, + { + "epoch": 0.86, + "learning_rate": 4.630872483221476e-05, + "loss": 0.0, + "step": 15165 + }, + { + "epoch": 0.86, + "learning_rate": 4.6224127234786526e-05, + "loss": 0.0, + "step": 15170 + }, + { + "epoch": 0.86, + "learning_rate": 4.613952963735829e-05, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 0.86, + "learning_rate": 4.605493203993007e-05, + "loss": 0.0, + "step": 15180 + }, + { + "epoch": 0.86, + "learning_rate": 4.597033444250183e-05, + "loss": 0.0, + "step": 15185 + }, + { + "epoch": 0.86, + "learning_rate": 4.5885736845073596e-05, + "loss": 0.0, + "step": 15190 + }, + { + "epoch": 0.86, + "learning_rate": 4.580113924764536e-05, + "loss": 0.0, + "step": 15195 + }, + { + "epoch": 0.86, + "learning_rate": 4.571654165021713e-05, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 0.86, + "learning_rate": 4.56319440527889e-05, + "loss": 0.0, + "step": 15205 + }, + { + "epoch": 0.86, + "learning_rate": 4.5547346455360666e-05, + "loss": 0.0, + "step": 15210 + }, + { + "epoch": 0.86, + "learning_rate": 4.546274885793243e-05, + "loss": 0.0, + "step": 15215 + }, + { + "epoch": 0.86, + "learning_rate": 4.5378151260504195e-05, + "loss": 0.0, + "step": 15220 + }, + { + "epoch": 0.86, + "learning_rate": 4.5293553663075966e-05, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 0.86, + "learning_rate": 4.520895606564774e-05, + "loss": 0.0, + "step": 15230 + }, + { + "epoch": 0.86, + "learning_rate": 4.51243584682195e-05, + "loss": 0.0, + "step": 15235 + }, + { + "epoch": 0.86, + "learning_rate": 4.5039760870791265e-05, + "loss": 0.0, + "step": 15240 + }, + { + "epoch": 0.86, + "learning_rate": 4.495516327336303e-05, + "loss": 0.0, + "step": 15245 + }, + { + "epoch": 0.86, + "learning_rate": 4.48705656759348e-05, + "loss": 0.0, + "step": 15250 + }, + { + "epoch": 0.86, + "learning_rate": 4.4785968078506565e-05, + "loss": 0.0, + "step": 15255 + }, + { + "epoch": 0.86, + "learning_rate": 4.4701370481078336e-05, + "loss": 0.0, + "step": 15260 + }, + { + "epoch": 0.86, + "learning_rate": 4.46167728836501e-05, + "loss": 0.0, + "step": 15265 + }, + { + "epoch": 0.86, + "learning_rate": 4.4532175286221864e-05, + "loss": 0.0, + "step": 15270 + }, + { + "epoch": 0.86, + "learning_rate": 4.4447577688793635e-05, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 0.86, + "learning_rate": 4.43629800913654e-05, + "loss": 0.0, + "step": 15280 + }, + { + "epoch": 0.86, + "learning_rate": 4.427838249393717e-05, + "loss": 0.0, + "step": 15285 + }, + { + "epoch": 0.86, + "learning_rate": 4.4193784896508934e-05, + "loss": 0.0, + "step": 15290 + }, + { + "epoch": 0.86, + "learning_rate": 4.4109187299080705e-05, + "loss": 0.0, + "step": 15295 + }, + { + "epoch": 0.86, + "learning_rate": 4.402458970165247e-05, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 0.86, + "learning_rate": 4.3939992104224234e-05, + "loss": 0.0, + "step": 15305 + }, + { + "epoch": 0.86, + "learning_rate": 4.3855394506796005e-05, + "loss": 0.0, + "step": 15310 + }, + { + "epoch": 0.86, + "learning_rate": 4.377079690936777e-05, + "loss": 0.0, + "step": 15315 + }, + { + "epoch": 0.86, + "learning_rate": 4.368619931193954e-05, + "loss": 0.0, + "step": 15320 + }, + { + "epoch": 0.86, + "learning_rate": 4.3601601714511304e-05, + "loss": 0.0, + "step": 15325 + }, + { + "epoch": 0.86, + "learning_rate": 4.351700411708307e-05, + "loss": 0.0, + "step": 15330 + }, + { + "epoch": 0.86, + "learning_rate": 4.343240651965484e-05, + "loss": 0.0, + "step": 15335 + }, + { + "epoch": 0.86, + "learning_rate": 4.3347808922226604e-05, + "loss": 0.0, + "step": 15340 + }, + { + "epoch": 0.87, + "learning_rate": 4.3263211324798375e-05, + "loss": 0.0, + "step": 15345 + }, + { + "epoch": 0.87, + "learning_rate": 4.317861372737014e-05, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 0.87, + "learning_rate": 4.30940161299419e-05, + "loss": 0.0, + "step": 15355 + }, + { + "epoch": 0.87, + "learning_rate": 4.300941853251367e-05, + "loss": 0.0, + "step": 15360 + }, + { + "epoch": 0.87, + "learning_rate": 4.2924820935085445e-05, + "loss": 0.0, + "step": 15365 + }, + { + "epoch": 0.87, + "learning_rate": 4.284022333765721e-05, + "loss": 0.0, + "step": 15370 + }, + { + "epoch": 0.87, + "learning_rate": 4.275562574022897e-05, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 0.87, + "learning_rate": 4.267102814280074e-05, + "loss": 0.0, + "step": 15380 + }, + { + "epoch": 0.87, + "learning_rate": 4.25864305453725e-05, + "loss": 0.0, + "step": 15385 + }, + { + "epoch": 0.87, + "learning_rate": 4.250183294794428e-05, + "loss": 0.0, + "step": 15390 + }, + { + "epoch": 0.87, + "learning_rate": 4.2417235350516044e-05, + "loss": 0.0, + "step": 15395 + }, + { + "epoch": 0.87, + "learning_rate": 4.233263775308781e-05, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 0.87, + "learning_rate": 4.224804015565957e-05, + "loss": 0.0, + "step": 15405 + }, + { + "epoch": 0.87, + "learning_rate": 4.2163442558231336e-05, + "loss": 0.0, + "step": 15410 + }, + { + "epoch": 0.87, + "learning_rate": 4.2078844960803114e-05, + "loss": 0.0, + "step": 15415 + }, + { + "epoch": 0.87, + "learning_rate": 4.199424736337488e-05, + "loss": 0.0, + "step": 15420 + }, + { + "epoch": 0.87, + "learning_rate": 4.190964976594664e-05, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 0.87, + "learning_rate": 4.182505216851841e-05, + "loss": 0.0, + "step": 15430 + }, + { + "epoch": 0.87, + "learning_rate": 4.174045457109017e-05, + "loss": 0.0, + "step": 15435 + }, + { + "epoch": 0.87, + "learning_rate": 4.165585697366195e-05, + "loss": 0.0, + "step": 15440 + }, + { + "epoch": 0.87, + "learning_rate": 4.157125937623371e-05, + "loss": 0.0, + "step": 15445 + }, + { + "epoch": 0.87, + "learning_rate": 4.148666177880548e-05, + "loss": 0.0, + "step": 15450 + }, + { + "epoch": 0.87, + "learning_rate": 4.140206418137724e-05, + "loss": 0.0, + "step": 15455 + }, + { + "epoch": 0.87, + "learning_rate": 4.131746658394902e-05, + "loss": 0.0, + "step": 15460 + }, + { + "epoch": 0.87, + "learning_rate": 4.123286898652078e-05, + "loss": 0.0, + "step": 15465 + }, + { + "epoch": 0.87, + "learning_rate": 4.114827138909255e-05, + "loss": 0.0, + "step": 15470 + }, + { + "epoch": 0.87, + "learning_rate": 4.106367379166431e-05, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 0.87, + "learning_rate": 4.0979076194236076e-05, + "loss": 0.0, + "step": 15480 + }, + { + "epoch": 0.87, + "learning_rate": 4.0894478596807854e-05, + "loss": 0.0, + "step": 15485 + }, + { + "epoch": 0.87, + "learning_rate": 4.080988099937962e-05, + "loss": 0.0, + "step": 15490 + }, + { + "epoch": 0.87, + "learning_rate": 4.072528340195138e-05, + "loss": 0.0, + "step": 15495 + }, + { + "epoch": 0.87, + "learning_rate": 4.0640685804523146e-05, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 0.87, + "learning_rate": 4.055608820709491e-05, + "loss": 0.0, + "step": 15505 + }, + { + "epoch": 0.87, + "learning_rate": 4.047149060966668e-05, + "loss": 0.0, + "step": 15510 + }, + { + "epoch": 0.87, + "learning_rate": 4.038689301223845e-05, + "loss": 0.0, + "step": 15515 + }, + { + "epoch": 0.88, + "learning_rate": 4.030229541481022e-05, + "loss": 0.0, + "step": 15520 + }, + { + "epoch": 0.88, + "learning_rate": 4.021769781738198e-05, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 0.88, + "learning_rate": 4.013310021995375e-05, + "loss": 0.0, + "step": 15530 + }, + { + "epoch": 0.88, + "learning_rate": 4.0048502622525516e-05, + "loss": 0.0, + "step": 15535 + }, + { + "epoch": 0.88, + "learning_rate": 3.996390502509729e-05, + "loss": 0.0, + "step": 15540 + }, + { + "epoch": 0.88, + "learning_rate": 3.987930742766905e-05, + "loss": 0.0, + "step": 15545 + }, + { + "epoch": 0.88, + "learning_rate": 3.9794709830240815e-05, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 0.88, + "learning_rate": 3.9710112232812586e-05, + "loss": 0.0, + "step": 15555 + }, + { + "epoch": 0.88, + "learning_rate": 3.962551463538435e-05, + "loss": 0.0, + "step": 15560 + }, + { + "epoch": 0.88, + "learning_rate": 3.954091703795612e-05, + "loss": 0.0, + "step": 15565 + }, + { + "epoch": 0.88, + "learning_rate": 3.9456319440527886e-05, + "loss": 0.0, + "step": 15570 + }, + { + "epoch": 0.88, + "learning_rate": 3.937172184309965e-05, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 0.88, + "learning_rate": 3.928712424567142e-05, + "loss": 0.0, + "step": 15580 + }, + { + "epoch": 0.88, + "learning_rate": 3.9202526648243185e-05, + "loss": 0.0, + "step": 15585 + }, + { + "epoch": 0.88, + "learning_rate": 3.9117929050814956e-05, + "loss": 0.0, + "step": 15590 + }, + { + "epoch": 0.88, + "learning_rate": 3.903333145338672e-05, + "loss": 0.0, + "step": 15595 + }, + { + "epoch": 0.88, + "learning_rate": 3.8948733855958485e-05, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 0.88, + "learning_rate": 3.8864136258530256e-05, + "loss": 0.0, + "step": 15605 + }, + { + "epoch": 0.88, + "learning_rate": 3.877953866110202e-05, + "loss": 0.0, + "step": 15610 + }, + { + "epoch": 0.88, + "learning_rate": 3.8694941063673784e-05, + "loss": 0.0, + "step": 15615 + }, + { + "epoch": 0.88, + "learning_rate": 3.8610343466245555e-05, + "loss": 0.0, + "step": 15620 + }, + { + "epoch": 0.88, + "learning_rate": 3.8525745868817326e-05, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 0.88, + "learning_rate": 3.844114827138909e-05, + "loss": 0.0, + "step": 15630 + }, + { + "epoch": 0.88, + "learning_rate": 3.8356550673960854e-05, + "loss": 0.0, + "step": 15635 + }, + { + "epoch": 0.88, + "learning_rate": 3.827195307653262e-05, + "loss": 0.0, + "step": 15640 + }, + { + "epoch": 0.88, + "learning_rate": 3.818735547910439e-05, + "loss": 0.0, + "step": 15645 + }, + { + "epoch": 0.88, + "learning_rate": 3.810275788167616e-05, + "loss": 0.0, + "step": 15650 + }, + { + "epoch": 0.88, + "learning_rate": 3.8018160284247925e-05, + "loss": 0.0, + "step": 15655 + }, + { + "epoch": 0.88, + "learning_rate": 3.793356268681969e-05, + "loss": 0.0, + "step": 15660 + }, + { + "epoch": 0.88, + "learning_rate": 3.784896508939145e-05, + "loss": 0.0, + "step": 15665 + }, + { + "epoch": 0.88, + "learning_rate": 3.7764367491963224e-05, + "loss": 0.0, + "step": 15670 + }, + { + "epoch": 0.88, + "learning_rate": 3.7679769894534995e-05, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 0.88, + "learning_rate": 3.759517229710676e-05, + "loss": 0.0, + "step": 15680 + }, + { + "epoch": 0.88, + "learning_rate": 3.7510574699678524e-05, + "loss": 0.0, + "step": 15685 + }, + { + "epoch": 0.88, + "learning_rate": 3.7425977102250295e-05, + "loss": 0.0, + "step": 15690 + }, + { + "epoch": 0.88, + "learning_rate": 3.734137950482206e-05, + "loss": 0.0, + "step": 15695 + }, + { + "epoch": 0.89, + "learning_rate": 3.725678190739382e-05, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 0.89, + "learning_rate": 3.7172184309965594e-05, + "loss": 0.0, + "step": 15705 + }, + { + "epoch": 0.89, + "learning_rate": 3.708758671253736e-05, + "loss": 0.0, + "step": 15710 + }, + { + "epoch": 0.89, + "learning_rate": 3.700298911510913e-05, + "loss": 0.0, + "step": 15715 + }, + { + "epoch": 0.89, + "learning_rate": 3.6918391517680893e-05, + "loss": 0.0, + "step": 15720 + }, + { + "epoch": 0.89, + "learning_rate": 3.6833793920252664e-05, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 0.89, + "learning_rate": 3.674919632282443e-05, + "loss": 0.0, + "step": 15730 + }, + { + "epoch": 0.89, + "learning_rate": 3.666459872539619e-05, + "loss": 0.0, + "step": 15735 + }, + { + "epoch": 0.89, + "learning_rate": 3.6580001127967964e-05, + "loss": 0.0, + "step": 15740 + }, + { + "epoch": 0.89, + "learning_rate": 3.649540353053973e-05, + "loss": 0.0, + "step": 15745 + }, + { + "epoch": 0.89, + "learning_rate": 3.64108059331115e-05, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 0.89, + "learning_rate": 3.632620833568326e-05, + "loss": 0.0, + "step": 15755 + }, + { + "epoch": 0.89, + "learning_rate": 3.6241610738255034e-05, + "loss": 0.0, + "step": 15760 + }, + { + "epoch": 0.89, + "learning_rate": 3.61570131408268e-05, + "loss": 0.0, + "step": 15765 + }, + { + "epoch": 0.89, + "learning_rate": 3.607241554339856e-05, + "loss": 0.0, + "step": 15770 + }, + { + "epoch": 0.89, + "learning_rate": 3.5987817945970334e-05, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 0.89, + "learning_rate": 3.59032203485421e-05, + "loss": 0.0, + "step": 15780 + }, + { + "epoch": 0.89, + "learning_rate": 3.581862275111387e-05, + "loss": 0.0, + "step": 15785 + }, + { + "epoch": 0.89, + "learning_rate": 3.573402515368563e-05, + "loss": 0.0, + "step": 15790 + }, + { + "epoch": 0.89, + "learning_rate": 3.5649427556257404e-05, + "loss": 0.0, + "step": 15795 + }, + { + "epoch": 0.89, + "learning_rate": 3.556482995882917e-05, + "loss": 0.0, + "step": 15800 + }, + { + "epoch": 0.89, + "learning_rate": 3.548023236140093e-05, + "loss": 0.0, + "step": 15805 + }, + { + "epoch": 0.89, + "learning_rate": 3.53956347639727e-05, + "loss": 0.0, + "step": 15810 + }, + { + "epoch": 0.89, + "learning_rate": 3.531103716654447e-05, + "loss": 0.0, + "step": 15815 + }, + { + "epoch": 0.89, + "learning_rate": 3.522643956911624e-05, + "loss": 0.0, + "step": 15820 + }, + { + "epoch": 0.89, + "learning_rate": 3.5141841971688e-05, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 0.89, + "learning_rate": 3.505724437425977e-05, + "loss": 0.0, + "step": 15830 + }, + { + "epoch": 0.89, + "learning_rate": 3.497264677683154e-05, + "loss": 0.0, + "step": 15835 + }, + { + "epoch": 0.89, + "learning_rate": 3.48880491794033e-05, + "loss": 0.0, + "step": 15840 + }, + { + "epoch": 0.89, + "learning_rate": 3.4803451581975066e-05, + "loss": 0.0, + "step": 15845 + }, + { + "epoch": 0.89, + "learning_rate": 3.471885398454684e-05, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 0.89, + "learning_rate": 3.46342563871186e-05, + "loss": 0.0, + "step": 15855 + }, + { + "epoch": 0.89, + "learning_rate": 3.454965878969037e-05, + "loss": 0.0, + "step": 15860 + }, + { + "epoch": 0.89, + "learning_rate": 3.446506119226214e-05, + "loss": 0.0, + "step": 15865 + }, + { + "epoch": 0.89, + "learning_rate": 3.43804635948339e-05, + "loss": 0.0, + "step": 15870 + }, + { + "epoch": 0.9, + "learning_rate": 3.429586599740567e-05, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 0.9, + "learning_rate": 3.4211268399977436e-05, + "loss": 0.0, + "step": 15880 + }, + { + "epoch": 0.9, + "learning_rate": 3.412667080254921e-05, + "loss": 0.0, + "step": 15885 + }, + { + "epoch": 0.9, + "learning_rate": 3.404207320512097e-05, + "loss": 0.0, + "step": 15890 + }, + { + "epoch": 0.9, + "learning_rate": 3.3957475607692736e-05, + "loss": 0.0, + "step": 15895 + }, + { + "epoch": 0.9, + "learning_rate": 3.3872878010264507e-05, + "loss": 0.0, + "step": 15900 + }, + { + "epoch": 0.9, + "learning_rate": 3.378828041283627e-05, + "loss": 0.0, + "step": 15905 + }, + { + "epoch": 0.9, + "learning_rate": 3.3703682815408035e-05, + "loss": 0.0, + "step": 15910 + }, + { + "epoch": 0.9, + "learning_rate": 3.3619085217979806e-05, + "loss": 0.0, + "step": 15915 + }, + { + "epoch": 0.9, + "learning_rate": 3.353448762055157e-05, + "loss": 0.0, + "step": 15920 + }, + { + "epoch": 0.9, + "learning_rate": 3.344989002312334e-05, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 0.9, + "learning_rate": 3.3365292425695105e-05, + "loss": 0.0, + "step": 15930 + }, + { + "epoch": 0.9, + "learning_rate": 3.328069482826687e-05, + "loss": 0.0, + "step": 15935 + }, + { + "epoch": 0.9, + "learning_rate": 3.319609723083864e-05, + "loss": 0.0, + "step": 15940 + }, + { + "epoch": 0.9, + "learning_rate": 3.3111499633410405e-05, + "loss": 0.0, + "step": 15945 + }, + { + "epoch": 0.9, + "learning_rate": 3.3026902035982176e-05, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 0.9, + "learning_rate": 3.294230443855394e-05, + "loss": 0.0, + "step": 15955 + }, + { + "epoch": 0.9, + "learning_rate": 3.285770684112571e-05, + "loss": 0.0, + "step": 15960 + }, + { + "epoch": 0.9, + "learning_rate": 3.2773109243697475e-05, + "loss": 0.0, + "step": 15965 + }, + { + "epoch": 0.9, + "learning_rate": 3.268851164626924e-05, + "loss": 0.0, + "step": 15970 + }, + { + "epoch": 0.9, + "learning_rate": 3.260391404884101e-05, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 0.9, + "learning_rate": 3.2519316451412775e-05, + "loss": 0.0, + "step": 15980 + }, + { + "epoch": 0.9, + "learning_rate": 3.2434718853984546e-05, + "loss": 0.0, + "step": 15985 + }, + { + "epoch": 0.9, + "learning_rate": 3.235012125655631e-05, + "loss": 0.0, + "step": 15990 + }, + { + "epoch": 0.9, + "learning_rate": 3.226552365912808e-05, + "loss": 0.0, + "step": 15995 + }, + { + "epoch": 0.9, + "learning_rate": 3.2180926061699845e-05, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 0.9, + "learning_rate": 3.209632846427161e-05, + "loss": 0.0, + "step": 16005 + }, + { + "epoch": 0.9, + "learning_rate": 3.201173086684338e-05, + "loss": 0.0, + "step": 16010 + }, + { + "epoch": 0.9, + "learning_rate": 3.1927133269415144e-05, + "loss": 0.0, + "step": 16015 + }, + { + "epoch": 0.9, + "learning_rate": 3.1842535671986915e-05, + "loss": 0.0, + "step": 16020 + }, + { + "epoch": 0.9, + "learning_rate": 3.175793807455868e-05, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 0.9, + "learning_rate": 3.1673340477130444e-05, + "loss": 0.0, + "step": 16030 + }, + { + "epoch": 0.9, + "learning_rate": 3.1588742879702215e-05, + "loss": 0.0, + "step": 16035 + }, + { + "epoch": 0.9, + "learning_rate": 3.150414528227398e-05, + "loss": 0.0, + "step": 16040 + }, + { + "epoch": 0.9, + "learning_rate": 3.141954768484575e-05, + "loss": 0.0, + "step": 16045 + }, + { + "epoch": 0.9, + "learning_rate": 3.1334950087417514e-05, + "loss": 0.0, + "step": 16050 + }, + { + "epoch": 0.91, + "learning_rate": 3.1250352489989285e-05, + "loss": 0.0, + "step": 16055 + }, + { + "epoch": 0.91, + "learning_rate": 3.116575489256105e-05, + "loss": 0.0, + "step": 16060 + }, + { + "epoch": 0.91, + "learning_rate": 3.1081157295132814e-05, + "loss": 0.0, + "step": 16065 + }, + { + "epoch": 0.91, + "learning_rate": 3.0996559697704585e-05, + "loss": 0.0, + "step": 16070 + }, + { + "epoch": 0.91, + "learning_rate": 3.091196210027635e-05, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 0.91, + "learning_rate": 3.082736450284812e-05, + "loss": 0.0, + "step": 16080 + }, + { + "epoch": 0.91, + "learning_rate": 3.0742766905419884e-05, + "loss": 0.0, + "step": 16085 + }, + { + "epoch": 0.91, + "learning_rate": 3.0658169307991655e-05, + "loss": 0.0, + "step": 16090 + }, + { + "epoch": 0.91, + "learning_rate": 3.057357171056342e-05, + "loss": 0.0, + "step": 16095 + }, + { + "epoch": 0.91, + "learning_rate": 3.0488974113135183e-05, + "loss": 0.0, + "step": 16100 + }, + { + "epoch": 0.91, + "learning_rate": 3.040437651570695e-05, + "loss": 0.0, + "step": 16105 + }, + { + "epoch": 0.91, + "learning_rate": 3.031977891827872e-05, + "loss": 0.0, + "step": 16110 + }, + { + "epoch": 0.91, + "learning_rate": 3.0235181320850486e-05, + "loss": 0.0, + "step": 16115 + }, + { + "epoch": 0.91, + "learning_rate": 3.015058372342225e-05, + "loss": 0.0, + "step": 16120 + }, + { + "epoch": 0.91, + "learning_rate": 3.006598612599402e-05, + "loss": 0.0, + "step": 16125 + }, + { + "epoch": 0.91, + "learning_rate": 2.9981388528565785e-05, + "loss": 0.0, + "step": 16130 + }, + { + "epoch": 0.91, + "learning_rate": 2.9896790931137553e-05, + "loss": 0.0, + "step": 16135 + }, + { + "epoch": 0.91, + "learning_rate": 2.981219333370932e-05, + "loss": 0.0, + "step": 16140 + }, + { + "epoch": 0.91, + "learning_rate": 2.9727595736281085e-05, + "loss": 0.0, + "step": 16145 + }, + { + "epoch": 0.91, + "learning_rate": 2.9642998138852856e-05, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 0.91, + "learning_rate": 2.955840054142462e-05, + "loss": 0.0, + "step": 16155 + }, + { + "epoch": 0.91, + "learning_rate": 2.947380294399639e-05, + "loss": 0.0, + "step": 16160 + }, + { + "epoch": 0.91, + "learning_rate": 2.9389205346568155e-05, + "loss": 0.0, + "step": 16165 + }, + { + "epoch": 0.91, + "learning_rate": 2.930460774913992e-05, + "loss": 0.0, + "step": 16170 + }, + { + "epoch": 0.91, + "learning_rate": 2.922001015171169e-05, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 0.91, + "learning_rate": 2.9135412554283455e-05, + "loss": 0.0, + "step": 16180 + }, + { + "epoch": 0.91, + "learning_rate": 2.9050814956855226e-05, + "loss": 0.0, + "step": 16185 + }, + { + "epoch": 0.91, + "learning_rate": 2.896621735942699e-05, + "loss": 0.0, + "step": 16190 + }, + { + "epoch": 0.91, + "learning_rate": 2.8881619761998754e-05, + "loss": 0.0, + "step": 16195 + }, + { + "epoch": 0.91, + "learning_rate": 2.8797022164570525e-05, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 0.91, + "learning_rate": 2.871242456714229e-05, + "loss": 0.0, + "step": 16205 + }, + { + "epoch": 0.91, + "learning_rate": 2.862782696971406e-05, + "loss": 0.0, + "step": 16210 + }, + { + "epoch": 0.91, + "learning_rate": 2.8543229372285824e-05, + "loss": 0.0, + "step": 16215 + }, + { + "epoch": 0.91, + "learning_rate": 2.8458631774857592e-05, + "loss": 0.0, + "step": 16220 + }, + { + "epoch": 0.91, + "learning_rate": 2.837403417742936e-05, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 0.92, + "learning_rate": 2.8289436580001124e-05, + "loss": 0.0, + "step": 16230 + }, + { + "epoch": 0.92, + "learning_rate": 2.820483898257289e-05, + "loss": 0.0, + "step": 16235 + }, + { + "epoch": 0.92, + "learning_rate": 2.812024138514466e-05, + "loss": 0.0, + "step": 16240 + }, + { + "epoch": 0.92, + "learning_rate": 2.8035643787716427e-05, + "loss": 0.0, + "step": 16245 + }, + { + "epoch": 0.92, + "learning_rate": 2.7951046190288194e-05, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 0.92, + "learning_rate": 2.7866448592859962e-05, + "loss": 0.0, + "step": 16255 + }, + { + "epoch": 0.92, + "learning_rate": 2.7781850995431726e-05, + "loss": 0.0, + "step": 16260 + }, + { + "epoch": 0.92, + "learning_rate": 2.7697253398003494e-05, + "loss": 0.0, + "step": 16265 + }, + { + "epoch": 0.92, + "learning_rate": 2.761265580057526e-05, + "loss": 0.0, + "step": 16270 + }, + { + "epoch": 0.92, + "learning_rate": 2.7528058203147025e-05, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 0.92, + "learning_rate": 2.7443460605718796e-05, + "loss": 0.0, + "step": 16280 + }, + { + "epoch": 0.92, + "learning_rate": 2.735886300829056e-05, + "loss": 0.0, + "step": 16285 + }, + { + "epoch": 0.92, + "learning_rate": 2.727426541086233e-05, + "loss": 0.0, + "step": 16290 + }, + { + "epoch": 0.92, + "learning_rate": 2.7189667813434096e-05, + "loss": 0.0, + "step": 16295 + }, + { + "epoch": 0.92, + "learning_rate": 2.710507021600586e-05, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 0.92, + "learning_rate": 2.702047261857763e-05, + "loss": 0.0, + "step": 16305 + }, + { + "epoch": 0.92, + "learning_rate": 2.6935875021149395e-05, + "loss": 0.0, + "step": 16310 + }, + { + "epoch": 0.92, + "learning_rate": 2.6851277423721166e-05, + "loss": 0.0, + "step": 16315 + }, + { + "epoch": 0.92, + "learning_rate": 2.676667982629293e-05, + "loss": 0.0, + "step": 16320 + }, + { + "epoch": 0.92, + "learning_rate": 2.66820822288647e-05, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 0.92, + "learning_rate": 2.6597484631436466e-05, + "loss": 0.0, + "step": 16330 + }, + { + "epoch": 0.92, + "learning_rate": 2.651288703400823e-05, + "loss": 0.0, + "step": 16335 + }, + { + "epoch": 0.92, + "learning_rate": 2.642828943658e-05, + "loss": 0.0, + "step": 16340 + }, + { + "epoch": 0.92, + "learning_rate": 2.6343691839151765e-05, + "loss": 0.0, + "step": 16345 + }, + { + "epoch": 0.92, + "learning_rate": 2.6259094241723533e-05, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 0.92, + "learning_rate": 2.61744966442953e-05, + "loss": 0.0, + "step": 16355 + }, + { + "epoch": 0.92, + "learning_rate": 2.6089899046867064e-05, + "loss": 0.0, + "step": 16360 + }, + { + "epoch": 0.92, + "learning_rate": 2.6005301449438835e-05, + "loss": 0.0, + "step": 16365 + }, + { + "epoch": 0.92, + "learning_rate": 2.59207038520106e-05, + "loss": 0.0, + "step": 16370 + }, + { + "epoch": 0.92, + "learning_rate": 2.5836106254582367e-05, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 0.92, + "learning_rate": 2.5751508657154135e-05, + "loss": 0.0, + "step": 16380 + }, + { + "epoch": 0.92, + "learning_rate": 2.5666911059725902e-05, + "loss": 0.0, + "step": 16385 + }, + { + "epoch": 0.92, + "learning_rate": 2.558231346229767e-05, + "loss": 0.0, + "step": 16390 + }, + { + "epoch": 0.92, + "learning_rate": 2.5497715864869434e-05, + "loss": 0.0, + "step": 16395 + }, + { + "epoch": 0.92, + "learning_rate": 2.5413118267441202e-05, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 0.92, + "learning_rate": 2.532852067001297e-05, + "loss": 0.0, + "step": 16405 + }, + { + "epoch": 0.93, + "learning_rate": 2.5243923072584737e-05, + "loss": 0.0, + "step": 16410 + }, + { + "epoch": 0.93, + "learning_rate": 2.51593254751565e-05, + "loss": 0.0, + "step": 16415 + }, + { + "epoch": 0.93, + "learning_rate": 2.5074727877728272e-05, + "loss": 0.0, + "step": 16420 + }, + { + "epoch": 0.93, + "learning_rate": 2.4990130280300036e-05, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 0.93, + "learning_rate": 2.4905532682871804e-05, + "loss": 0.0, + "step": 16430 + }, + { + "epoch": 0.93, + "learning_rate": 2.482093508544357e-05, + "loss": 0.0, + "step": 16435 + }, + { + "epoch": 0.93, + "learning_rate": 2.4736337488015336e-05, + "loss": 0.0, + "step": 16440 + }, + { + "epoch": 0.93, + "learning_rate": 2.4651739890587107e-05, + "loss": 0.0, + "step": 16445 + }, + { + "epoch": 0.93, + "learning_rate": 2.456714229315887e-05, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 0.93, + "learning_rate": 2.4482544695730642e-05, + "loss": 0.0, + "step": 16455 + }, + { + "epoch": 0.93, + "learning_rate": 2.4397947098302406e-05, + "loss": 0.0, + "step": 16460 + }, + { + "epoch": 0.93, + "learning_rate": 2.431334950087417e-05, + "loss": 0.0, + "step": 16465 + }, + { + "epoch": 0.93, + "learning_rate": 2.422875190344594e-05, + "loss": 0.0, + "step": 16470 + }, + { + "epoch": 0.93, + "learning_rate": 2.4144154306017706e-05, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 0.93, + "learning_rate": 2.4059556708589477e-05, + "loss": 0.0, + "step": 16480 + }, + { + "epoch": 0.93, + "learning_rate": 2.397495911116124e-05, + "loss": 0.0, + "step": 16485 + }, + { + "epoch": 0.93, + "learning_rate": 2.389036151373301e-05, + "loss": 0.0, + "step": 16490 + }, + { + "epoch": 0.93, + "learning_rate": 2.3805763916304776e-05, + "loss": 0.0, + "step": 16495 + }, + { + "epoch": 0.93, + "learning_rate": 2.372116631887654e-05, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 0.93, + "learning_rate": 2.363656872144831e-05, + "loss": 0.0, + "step": 16505 + }, + { + "epoch": 0.93, + "learning_rate": 2.3551971124020075e-05, + "loss": 0.0, + "step": 16510 + }, + { + "epoch": 0.93, + "learning_rate": 2.3467373526591843e-05, + "loss": 0.0, + "step": 16515 + }, + { + "epoch": 0.93, + "learning_rate": 2.338277592916361e-05, + "loss": 0.0, + "step": 16520 + }, + { + "epoch": 0.93, + "learning_rate": 2.3298178331735378e-05, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 0.93, + "learning_rate": 2.3213580734307142e-05, + "loss": 0.0, + "step": 16530 + }, + { + "epoch": 0.93, + "learning_rate": 2.312898313687891e-05, + "loss": 0.0, + "step": 16535 + }, + { + "epoch": 0.93, + "learning_rate": 2.3044385539450678e-05, + "loss": 0.0, + "step": 16540 + }, + { + "epoch": 0.93, + "learning_rate": 2.2959787942022445e-05, + "loss": 0.0, + "step": 16545 + }, + { + "epoch": 0.93, + "learning_rate": 2.2875190344594213e-05, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 0.93, + "learning_rate": 2.2790592747165977e-05, + "loss": 0.0, + "step": 16555 + }, + { + "epoch": 0.93, + "learning_rate": 2.2705995149737745e-05, + "loss": 0.0, + "step": 16560 + }, + { + "epoch": 0.93, + "learning_rate": 2.2621397552309512e-05, + "loss": 0.0, + "step": 16565 + }, + { + "epoch": 0.93, + "learning_rate": 2.2536799954881276e-05, + "loss": 0.0, + "step": 16570 + }, + { + "epoch": 0.93, + "learning_rate": 2.2452202357453047e-05, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 0.93, + "learning_rate": 2.236760476002481e-05, + "loss": 0.0, + "step": 16580 + }, + { + "epoch": 0.94, + "learning_rate": 2.2283007162596583e-05, + "loss": 0.0, + "step": 16585 + }, + { + "epoch": 0.94, + "learning_rate": 2.2198409565168347e-05, + "loss": 0.0, + "step": 16590 + }, + { + "epoch": 0.94, + "learning_rate": 2.211381196774011e-05, + "loss": 0.0, + "step": 16595 + }, + { + "epoch": 0.94, + "learning_rate": 2.2029214370311882e-05, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 0.94, + "learning_rate": 2.1944616772883646e-05, + "loss": 0.0, + "step": 16605 + }, + { + "epoch": 0.94, + "learning_rate": 2.1860019175455417e-05, + "loss": 0.0, + "step": 16610 + }, + { + "epoch": 0.94, + "learning_rate": 2.177542157802718e-05, + "loss": 0.0, + "step": 16615 + }, + { + "epoch": 0.94, + "learning_rate": 2.1690823980598952e-05, + "loss": 0.0, + "step": 16620 + }, + { + "epoch": 0.94, + "learning_rate": 2.1606226383170717e-05, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 0.94, + "learning_rate": 2.152162878574248e-05, + "loss": 0.0, + "step": 16630 + }, + { + "epoch": 0.94, + "learning_rate": 2.1437031188314252e-05, + "loss": 0.0, + "step": 16635 + }, + { + "epoch": 0.94, + "learning_rate": 2.1352433590886016e-05, + "loss": 0.0, + "step": 16640 + }, + { + "epoch": 0.94, + "learning_rate": 2.1267835993457784e-05, + "loss": 0.0, + "step": 16645 + }, + { + "epoch": 0.94, + "learning_rate": 2.118323839602955e-05, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 0.94, + "learning_rate": 2.109864079860132e-05, + "loss": 0.0, + "step": 16655 + }, + { + "epoch": 0.94, + "learning_rate": 2.1014043201173086e-05, + "loss": 0.0, + "step": 16660 + }, + { + "epoch": 0.94, + "learning_rate": 2.092944560374485e-05, + "loss": 0.0, + "step": 16665 + }, + { + "epoch": 0.94, + "learning_rate": 2.0844848006316618e-05, + "loss": 0.0, + "step": 16670 + }, + { + "epoch": 0.94, + "learning_rate": 2.0760250408888386e-05, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 0.94, + "learning_rate": 2.0675652811460153e-05, + "loss": 0.0, + "step": 16680 + }, + { + "epoch": 0.94, + "learning_rate": 2.059105521403192e-05, + "loss": 0.0, + "step": 16685 + }, + { + "epoch": 0.94, + "learning_rate": 2.050645761660369e-05, + "loss": 0.0, + "step": 16690 + }, + { + "epoch": 0.94, + "learning_rate": 2.0421860019175453e-05, + "loss": 0.0, + "step": 16695 + }, + { + "epoch": 0.94, + "learning_rate": 2.033726242174722e-05, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 0.94, + "learning_rate": 2.0252664824318988e-05, + "loss": 0.0, + "step": 16705 + }, + { + "epoch": 0.94, + "learning_rate": 2.0168067226890752e-05, + "loss": 0.0, + "step": 16710 + }, + { + "epoch": 0.94, + "learning_rate": 2.0083469629462523e-05, + "loss": 0.0, + "step": 16715 + }, + { + "epoch": 0.94, + "learning_rate": 1.9998872032034287e-05, + "loss": 0.0, + "step": 16720 + }, + { + "epoch": 0.94, + "learning_rate": 1.9914274434606055e-05, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 0.94, + "learning_rate": 1.9829676837177823e-05, + "loss": 0.0, + "step": 16730 + }, + { + "epoch": 0.94, + "learning_rate": 1.9745079239749587e-05, + "loss": 0.0, + "step": 16735 + }, + { + "epoch": 0.94, + "learning_rate": 1.9660481642321358e-05, + "loss": 0.0, + "step": 16740 + }, + { + "epoch": 0.94, + "learning_rate": 1.9575884044893122e-05, + "loss": 0.0, + "step": 16745 + }, + { + "epoch": 0.94, + "learning_rate": 1.9491286447464893e-05, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 0.94, + "learning_rate": 1.9406688850036657e-05, + "loss": 0.0, + "step": 16755 + }, + { + "epoch": 0.94, + "learning_rate": 1.932209125260842e-05, + "loss": 0.0, + "step": 16760 + }, + { + "epoch": 0.95, + "learning_rate": 1.9237493655180192e-05, + "loss": 0.0, + "step": 16765 + }, + { + "epoch": 0.95, + "learning_rate": 1.9152896057751956e-05, + "loss": 0.0, + "step": 16770 + }, + { + "epoch": 0.95, + "learning_rate": 1.9068298460323727e-05, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 0.95, + "learning_rate": 1.898370086289549e-05, + "loss": 0.0, + "step": 16780 + }, + { + "epoch": 0.95, + "learning_rate": 1.889910326546726e-05, + "loss": 0.0, + "step": 16785 + }, + { + "epoch": 0.95, + "learning_rate": 1.8814505668039027e-05, + "loss": 0.0, + "step": 16790 + }, + { + "epoch": 0.95, + "learning_rate": 1.8729908070610794e-05, + "loss": 0.0, + "step": 16795 + }, + { + "epoch": 0.95, + "learning_rate": 1.8645310473182562e-05, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 0.95, + "learning_rate": 1.8560712875754326e-05, + "loss": 0.0, + "step": 16805 + }, + { + "epoch": 0.95, + "learning_rate": 1.8476115278326094e-05, + "loss": 0.0, + "step": 16810 + }, + { + "epoch": 0.95, + "learning_rate": 1.839151768089786e-05, + "loss": 0.0, + "step": 16815 + }, + { + "epoch": 0.95, + "learning_rate": 1.830692008346963e-05, + "loss": 0.0, + "step": 16820 + }, + { + "epoch": 0.95, + "learning_rate": 1.8222322486041393e-05, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 0.95, + "learning_rate": 1.813772488861316e-05, + "loss": 0.0, + "step": 16830 + }, + { + "epoch": 0.95, + "learning_rate": 1.805312729118493e-05, + "loss": 0.0, + "step": 16835 + }, + { + "epoch": 0.95, + "learning_rate": 1.7968529693756696e-05, + "loss": 0.0, + "step": 16840 + }, + { + "epoch": 0.95, + "learning_rate": 1.7883932096328464e-05, + "loss": 0.0, + "step": 16845 + }, + { + "epoch": 0.95, + "learning_rate": 1.7799334498900228e-05, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 0.95, + "learning_rate": 1.7714736901471995e-05, + "loss": 0.0, + "step": 16855 + }, + { + "epoch": 0.95, + "learning_rate": 1.7630139304043763e-05, + "loss": 0.0, + "step": 16860 + }, + { + "epoch": 0.95, + "learning_rate": 1.754554170661553e-05, + "loss": 0.0, + "step": 16865 + }, + { + "epoch": 0.95, + "learning_rate": 1.7460944109187298e-05, + "loss": 0.0, + "step": 16870 + }, + { + "epoch": 0.95, + "learning_rate": 1.7376346511759062e-05, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 0.95, + "learning_rate": 1.729174891433083e-05, + "loss": 0.0, + "step": 16880 + }, + { + "epoch": 0.95, + "learning_rate": 1.7207151316902598e-05, + "loss": 0.0, + "step": 16885 + }, + { + "epoch": 0.95, + "learning_rate": 1.7122553719474365e-05, + "loss": 0.0, + "step": 16890 + }, + { + "epoch": 0.95, + "learning_rate": 1.7037956122046133e-05, + "loss": 0.0, + "step": 16895 + }, + { + "epoch": 0.95, + "learning_rate": 1.69533585246179e-05, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 0.95, + "learning_rate": 1.6868760927189668e-05, + "loss": 0.0, + "step": 16905 + }, + { + "epoch": 0.95, + "learning_rate": 1.6784163329761432e-05, + "loss": 0.0, + "step": 16910 + }, + { + "epoch": 0.95, + "learning_rate": 1.66995657323332e-05, + "loss": 0.0, + "step": 16915 + }, + { + "epoch": 0.95, + "learning_rate": 1.6614968134904967e-05, + "loss": 0.0, + "step": 16920 + }, + { + "epoch": 0.95, + "learning_rate": 1.6530370537476735e-05, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 0.95, + "learning_rate": 1.6445772940048503e-05, + "loss": 0.0, + "step": 16930 + }, + { + "epoch": 0.95, + "learning_rate": 1.636117534262027e-05, + "loss": 0.0, + "step": 16935 + }, + { + "epoch": 0.96, + "learning_rate": 1.6276577745192034e-05, + "loss": 0.0, + "step": 16940 + }, + { + "epoch": 0.96, + "learning_rate": 1.6191980147763802e-05, + "loss": 0.0, + "step": 16945 + }, + { + "epoch": 0.96, + "learning_rate": 1.610738255033557e-05, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 0.96, + "learning_rate": 1.6022784952907337e-05, + "loss": 0.0, + "step": 16955 + }, + { + "epoch": 0.96, + "learning_rate": 1.5938187355479105e-05, + "loss": 0.0, + "step": 16960 + }, + { + "epoch": 0.96, + "learning_rate": 1.585358975805087e-05, + "loss": 0.0, + "step": 16965 + }, + { + "epoch": 0.96, + "learning_rate": 1.5768992160622637e-05, + "loss": 0.0, + "step": 16970 + }, + { + "epoch": 0.96, + "learning_rate": 1.5684394563194404e-05, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 0.96, + "learning_rate": 1.5599796965766172e-05, + "loss": 0.0, + "step": 16980 + }, + { + "epoch": 0.96, + "learning_rate": 1.5515199368337936e-05, + "loss": 0.0, + "step": 16985 + }, + { + "epoch": 0.96, + "learning_rate": 1.5430601770909704e-05, + "loss": 0.0, + "step": 16990 + }, + { + "epoch": 0.96, + "learning_rate": 1.534600417348147e-05, + "loss": 0.0, + "step": 16995 + }, + { + "epoch": 0.96, + "learning_rate": 1.526140657605324e-05, + "loss": 0.0, + "step": 17000 + }, + { + "epoch": 0.96, + "learning_rate": 1.5176808978625006e-05, + "loss": 0.0, + "step": 17005 + }, + { + "epoch": 0.96, + "learning_rate": 1.5092211381196772e-05, + "loss": 0.0, + "step": 17010 + }, + { + "epoch": 0.96, + "learning_rate": 1.500761378376854e-05, + "loss": 0.0, + "step": 17015 + }, + { + "epoch": 0.96, + "learning_rate": 1.4923016186340306e-05, + "loss": 0.0, + "step": 17020 + }, + { + "epoch": 0.96, + "learning_rate": 1.4838418588912073e-05, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 0.96, + "learning_rate": 1.4753820991483841e-05, + "loss": 0.0, + "step": 17030 + }, + { + "epoch": 0.96, + "learning_rate": 1.4669223394055609e-05, + "loss": 0.0, + "step": 17035 + }, + { + "epoch": 0.96, + "learning_rate": 1.4584625796627373e-05, + "loss": 0.0, + "step": 17040 + }, + { + "epoch": 0.96, + "learning_rate": 1.450002819919914e-05, + "loss": 0.0, + "step": 17045 + }, + { + "epoch": 0.96, + "learning_rate": 1.4415430601770908e-05, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 0.96, + "learning_rate": 1.4330833004342676e-05, + "loss": 0.0, + "step": 17055 + }, + { + "epoch": 0.96, + "learning_rate": 1.4246235406914443e-05, + "loss": 0.0, + "step": 17060 + }, + { + "epoch": 0.96, + "learning_rate": 1.416163780948621e-05, + "loss": 0.0, + "step": 17065 + }, + { + "epoch": 0.96, + "learning_rate": 1.4077040212057977e-05, + "loss": 0.0, + "step": 17070 + }, + { + "epoch": 0.96, + "learning_rate": 1.3992442614629743e-05, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 0.96, + "learning_rate": 1.390784501720151e-05, + "loss": 0.0, + "step": 17080 + }, + { + "epoch": 0.96, + "learning_rate": 1.3823247419773278e-05, + "loss": 0.0, + "step": 17085 + }, + { + "epoch": 0.96, + "learning_rate": 1.3738649822345044e-05, + "loss": 0.0, + "step": 17090 + }, + { + "epoch": 0.96, + "learning_rate": 1.3654052224916811e-05, + "loss": 0.0, + "step": 17095 + }, + { + "epoch": 0.96, + "learning_rate": 1.3569454627488579e-05, + "loss": 0.0, + "step": 17100 + }, + { + "epoch": 0.96, + "learning_rate": 1.3484857030060345e-05, + "loss": 0.0, + "step": 17105 + }, + { + "epoch": 0.96, + "learning_rate": 1.340025943263211e-05, + "loss": 0.0, + "step": 17110 + }, + { + "epoch": 0.96, + "learning_rate": 1.3315661835203878e-05, + "loss": 0.0, + "step": 17115 + }, + { + "epoch": 0.97, + "learning_rate": 1.3231064237775646e-05, + "loss": 0.0, + "step": 17120 + }, + { + "epoch": 0.97, + "learning_rate": 1.3146466640347413e-05, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 0.97, + "learning_rate": 1.3061869042919181e-05, + "loss": 0.0, + "step": 17130 + }, + { + "epoch": 0.97, + "learning_rate": 1.2977271445490947e-05, + "loss": 0.0, + "step": 17135 + }, + { + "epoch": 0.97, + "learning_rate": 1.2892673848062713e-05, + "loss": 0.0, + "step": 17140 + }, + { + "epoch": 0.97, + "learning_rate": 1.280807625063448e-05, + "loss": 0.0, + "step": 17145 + }, + { + "epoch": 0.97, + "learning_rate": 1.2723478653206248e-05, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 0.97, + "learning_rate": 1.2638881055778016e-05, + "loss": 0.0, + "step": 17155 + }, + { + "epoch": 0.97, + "learning_rate": 1.2554283458349782e-05, + "loss": 0.0, + "step": 17160 + }, + { + "epoch": 0.97, + "learning_rate": 1.246968586092155e-05, + "loss": 0.0, + "step": 17165 + }, + { + "epoch": 0.97, + "learning_rate": 1.2385088263493317e-05, + "loss": 0.0, + "step": 17170 + }, + { + "epoch": 0.97, + "learning_rate": 1.2300490666065083e-05, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 0.97, + "learning_rate": 1.2215893068636849e-05, + "loss": 0.0, + "step": 17180 + }, + { + "epoch": 0.97, + "learning_rate": 1.2131295471208616e-05, + "loss": 0.0, + "step": 17185 + }, + { + "epoch": 0.97, + "learning_rate": 1.2046697873780384e-05, + "loss": 0.0, + "step": 17190 + }, + { + "epoch": 0.97, + "learning_rate": 1.1962100276352151e-05, + "loss": 0.0, + "step": 17195 + }, + { + "epoch": 0.97, + "learning_rate": 1.1877502678923919e-05, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 0.97, + "learning_rate": 1.1792905081495683e-05, + "loss": 0.0, + "step": 17205 + }, + { + "epoch": 0.97, + "learning_rate": 1.170830748406745e-05, + "loss": 0.0, + "step": 17210 + }, + { + "epoch": 0.97, + "learning_rate": 1.1623709886639218e-05, + "loss": 0.0, + "step": 17215 + }, + { + "epoch": 0.97, + "learning_rate": 1.1539112289210986e-05, + "loss": 0.0, + "step": 17220 + }, + { + "epoch": 0.97, + "learning_rate": 1.1454514691782752e-05, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 0.97, + "learning_rate": 1.136991709435452e-05, + "loss": 0.0, + "step": 17230 + }, + { + "epoch": 0.97, + "learning_rate": 1.1285319496926287e-05, + "loss": 0.0, + "step": 17235 + }, + { + "epoch": 0.97, + "learning_rate": 1.1200721899498053e-05, + "loss": 0.0, + "step": 17240 + }, + { + "epoch": 0.97, + "learning_rate": 1.111612430206982e-05, + "loss": 0.0, + "step": 17245 + }, + { + "epoch": 0.97, + "learning_rate": 1.1031526704641586e-05, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 0.97, + "learning_rate": 1.0946929107213354e-05, + "loss": 0.0, + "step": 17255 + }, + { + "epoch": 0.97, + "learning_rate": 1.0862331509785122e-05, + "loss": 0.0, + "step": 17260 + }, + { + "epoch": 0.97, + "learning_rate": 1.077773391235689e-05, + "loss": 0.0, + "step": 17265 + }, + { + "epoch": 0.97, + "learning_rate": 1.0693136314928653e-05, + "loss": 0.0, + "step": 17270 + }, + { + "epoch": 0.97, + "learning_rate": 1.0608538717500421e-05, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 0.97, + "learning_rate": 1.0523941120072189e-05, + "loss": 0.0, + "step": 17280 + }, + { + "epoch": 0.97, + "learning_rate": 1.0439343522643956e-05, + "loss": 0.0, + "step": 17285 + }, + { + "epoch": 0.97, + "learning_rate": 1.0354745925215724e-05, + "loss": 0.0, + "step": 17290 + }, + { + "epoch": 0.98, + "learning_rate": 1.027014832778749e-05, + "loss": 0.0, + "step": 17295 + }, + { + "epoch": 0.98, + "learning_rate": 1.0185550730359257e-05, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 0.98, + "learning_rate": 1.0100953132931023e-05, + "loss": 0.0, + "step": 17305 + }, + { + "epoch": 0.98, + "learning_rate": 1.001635553550279e-05, + "loss": 0.0, + "step": 17310 + }, + { + "epoch": 0.98, + "learning_rate": 9.931757938074557e-06, + "loss": 0.0, + "step": 17315 + }, + { + "epoch": 0.98, + "learning_rate": 9.847160340646324e-06, + "loss": 0.0, + "step": 17320 + }, + { + "epoch": 0.98, + "learning_rate": 9.762562743218092e-06, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 0.98, + "learning_rate": 9.67796514578986e-06, + "loss": 0.0, + "step": 17330 + }, + { + "epoch": 0.98, + "learning_rate": 9.593367548361627e-06, + "loss": 0.0, + "step": 17335 + }, + { + "epoch": 0.98, + "learning_rate": 9.508769950933391e-06, + "loss": 0.0, + "step": 17340 + }, + { + "epoch": 0.98, + "learning_rate": 9.424172353505159e-06, + "loss": 0.0, + "step": 17345 + }, + { + "epoch": 0.98, + "learning_rate": 9.339574756076926e-06, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 0.98, + "learning_rate": 9.254977158648694e-06, + "loss": 0.0, + "step": 17355 + }, + { + "epoch": 0.98, + "learning_rate": 9.170379561220462e-06, + "loss": 0.0, + "step": 17360 + }, + { + "epoch": 0.98, + "learning_rate": 9.085781963792228e-06, + "loss": 0.0, + "step": 17365 + }, + { + "epoch": 0.98, + "learning_rate": 9.001184366363995e-06, + "loss": 0.0, + "step": 17370 + }, + { + "epoch": 0.98, + "learning_rate": 8.916586768935761e-06, + "loss": 0.0, + "step": 17375 + }, + { + "epoch": 0.98, + "learning_rate": 8.831989171507529e-06, + "loss": 0.0, + "step": 17380 + }, + { + "epoch": 0.98, + "learning_rate": 8.747391574079295e-06, + "loss": 0.0, + "step": 17385 + }, + { + "epoch": 0.98, + "learning_rate": 8.662793976651062e-06, + "loss": 0.0, + "step": 17390 + }, + { + "epoch": 0.98, + "learning_rate": 8.578196379222828e-06, + "loss": 0.0, + "step": 17395 + }, + { + "epoch": 0.98, + "learning_rate": 8.493598781794596e-06, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 0.98, + "learning_rate": 8.409001184366363e-06, + "loss": 0.0, + "step": 17405 + }, + { + "epoch": 0.98, + "learning_rate": 8.324403586938131e-06, + "loss": 0.0, + "step": 17410 + }, + { + "epoch": 0.98, + "learning_rate": 8.239805989509897e-06, + "loss": 0.0, + "step": 17415 + }, + { + "epoch": 0.98, + "learning_rate": 8.155208392081664e-06, + "loss": 0.0, + "step": 17420 + }, + { + "epoch": 0.98, + "learning_rate": 8.070610794653432e-06, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 0.98, + "learning_rate": 7.986013197225198e-06, + "loss": 0.0, + "step": 17430 + }, + { + "epoch": 0.98, + "learning_rate": 7.901415599796965e-06, + "loss": 0.0, + "step": 17435 + }, + { + "epoch": 0.98, + "learning_rate": 7.816818002368733e-06, + "loss": 0.0, + "step": 17440 + }, + { + "epoch": 0.98, + "learning_rate": 7.732220404940499e-06, + "loss": 0.0, + "step": 17445 + }, + { + "epoch": 0.98, + "learning_rate": 7.647622807512267e-06, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 0.98, + "learning_rate": 7.563025210084033e-06, + "loss": 0.0, + "step": 17455 + }, + { + "epoch": 0.98, + "learning_rate": 7.4784276126558e-06, + "loss": 0.0, + "step": 17460 + }, + { + "epoch": 0.98, + "learning_rate": 7.393830015227567e-06, + "loss": 0.0, + "step": 17465 + }, + { + "epoch": 0.99, + "learning_rate": 7.3092324177993336e-06, + "loss": 0.0, + "step": 17470 + }, + { + "epoch": 0.99, + "learning_rate": 7.224634820371101e-06, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 0.99, + "learning_rate": 7.140037222942867e-06, + "loss": 0.0, + "step": 17480 + }, + { + "epoch": 0.99, + "learning_rate": 7.055439625514635e-06, + "loss": 0.0, + "step": 17485 + }, + { + "epoch": 0.99, + "learning_rate": 6.970842028086402e-06, + "loss": 0.0, + "step": 17490 + }, + { + "epoch": 0.99, + "learning_rate": 6.886244430658168e-06, + "loss": 0.0, + "step": 17495 + }, + { + "epoch": 0.99, + "learning_rate": 6.801646833229936e-06, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 0.99, + "learning_rate": 6.7170492358017025e-06, + "loss": 0.0, + "step": 17505 + }, + { + "epoch": 0.99, + "learning_rate": 6.632451638373469e-06, + "loss": 0.0, + "step": 17510 + }, + { + "epoch": 0.99, + "learning_rate": 6.547854040945236e-06, + "loss": 0.0, + "step": 17515 + }, + { + "epoch": 0.99, + "learning_rate": 6.463256443517004e-06, + "loss": 0.0, + "step": 17520 + }, + { + "epoch": 0.99, + "learning_rate": 6.378658846088771e-06, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 0.99, + "learning_rate": 6.294061248660537e-06, + "loss": 0.0, + "step": 17530 + }, + { + "epoch": 0.99, + "learning_rate": 6.209463651232305e-06, + "loss": 0.0, + "step": 17535 + }, + { + "epoch": 0.99, + "learning_rate": 6.1248660538040714e-06, + "loss": 0.0, + "step": 17540 + }, + { + "epoch": 0.99, + "learning_rate": 6.040268456375838e-06, + "loss": 0.0, + "step": 17545 + }, + { + "epoch": 0.99, + "learning_rate": 5.955670858947605e-06, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 0.99, + "learning_rate": 5.8710732615193725e-06, + "loss": 0.0, + "step": 17555 + }, + { + "epoch": 0.99, + "learning_rate": 5.7864756640911384e-06, + "loss": 0.0, + "step": 17560 + }, + { + "epoch": 0.99, + "learning_rate": 5.701878066662906e-06, + "loss": 0.0, + "step": 17565 + }, + { + "epoch": 0.99, + "learning_rate": 5.617280469234674e-06, + "loss": 0.0, + "step": 17570 + }, + { + "epoch": 0.99, + "learning_rate": 5.53268287180644e-06, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 0.99, + "learning_rate": 5.448085274378207e-06, + "loss": 0.0, + "step": 17580 + }, + { + "epoch": 0.99, + "learning_rate": 5.363487676949974e-06, + "loss": 0.0, + "step": 17585 + }, + { + "epoch": 0.99, + "learning_rate": 5.2788900795217415e-06, + "loss": 0.0, + "step": 17590 + }, + { + "epoch": 0.99, + "learning_rate": 5.194292482093507e-06, + "loss": 0.0, + "step": 17595 + }, + { + "epoch": 0.99, + "learning_rate": 5.109694884665275e-06, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 0.99, + "learning_rate": 5.0250972872370426e-06, + "loss": 0.0, + "step": 17605 + }, + { + "epoch": 0.99, + "learning_rate": 4.9404996898088085e-06, + "loss": 0.0, + "step": 17610 + }, + { + "epoch": 0.99, + "learning_rate": 4.855902092380576e-06, + "loss": 0.0, + "step": 17615 + }, + { + "epoch": 0.99, + "learning_rate": 4.771304494952343e-06, + "loss": 0.0, + "step": 17620 + }, + { + "epoch": 0.99, + "learning_rate": 4.6867068975241096e-06, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 0.99, + "learning_rate": 4.602109300095876e-06, + "loss": 0.0, + "step": 17630 + }, + { + "epoch": 0.99, + "learning_rate": 4.517511702667644e-06, + "loss": 0.0, + "step": 17635 + }, + { + "epoch": 0.99, + "learning_rate": 4.432914105239411e-06, + "loss": 0.0, + "step": 17640 + }, + { + "epoch": 0.99, + "learning_rate": 4.348316507811178e-06, + "loss": 0.0, + "step": 17645 + }, + { + "epoch": 1.0, + "learning_rate": 4.263718910382945e-06, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 1.0, + "learning_rate": 4.179121312954712e-06, + "loss": 0.0, + "step": 17655 + }, + { + "epoch": 1.0, + "learning_rate": 4.0945237155264785e-06, + "loss": 0.0, + "step": 17660 + }, + { + "epoch": 1.0, + "learning_rate": 4.009926118098245e-06, + "loss": 0.0, + "step": 17665 + }, + { + "epoch": 1.0, + "learning_rate": 3.925328520670013e-06, + "loss": 0.0, + "step": 17670 + }, + { + "epoch": 1.0, + "learning_rate": 3.84073092324178e-06, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 1.0, + "learning_rate": 3.7561333258135463e-06, + "loss": 0.0, + "step": 17680 + }, + { + "epoch": 1.0, + "learning_rate": 3.6715357283853135e-06, + "loss": 0.0, + "step": 17685 + }, + { + "epoch": 1.0, + "learning_rate": 3.5869381309570803e-06, + "loss": 0.0, + "step": 17690 + }, + { + "epoch": 1.0, + "learning_rate": 3.502340533528848e-06, + "loss": 0.0, + "step": 17695 + }, + { + "epoch": 1.0, + "learning_rate": 3.4177429361006146e-06, + "loss": 0.0, + "step": 17700 + } + ], + "max_steps": 17736, + "num_train_epochs": 1, + "total_flos": 2.0527377490744934e+17, + "trial_name": null, + "trial_params": null +}