futurewei-test-1 / last-checkpoint /trainer_state.json
hemanth-kj's picture
Training in progress, step 3750
6715c53
raw
history blame contribute delete
No virus
88.9 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 1.440645409143296,
"global_step": 3750,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 0.00013979400086720374,
"loss": 0.0,
"step": 5
},
{
"epoch": 0.0,
"learning_rate": 0.00019999999999999998,
"loss": 0.0,
"step": 10
},
{
"epoch": 0.01,
"learning_rate": 0.0001998460354118553,
"loss": 0.0,
"step": 15
},
{
"epoch": 0.01,
"learning_rate": 0.00019965357967667439,
"loss": 0.0,
"step": 20
},
{
"epoch": 0.01,
"learning_rate": 0.00019946112394149347,
"loss": 0.0,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 0.00019926866820631255,
"loss": 0.0,
"step": 30
},
{
"epoch": 0.01,
"learning_rate": 0.00019907621247113163,
"loss": 0.0,
"step": 35
},
{
"epoch": 0.02,
"learning_rate": 0.00019888375673595074,
"loss": 0.0,
"step": 40
},
{
"epoch": 0.02,
"learning_rate": 0.00019869130100076983,
"loss": 0.0,
"step": 45
},
{
"epoch": 0.02,
"learning_rate": 0.00019849884526558894,
"loss": 0.0,
"step": 50
},
{
"epoch": 0.02,
"learning_rate": 0.00019830638953040802,
"loss": 0.0,
"step": 55
},
{
"epoch": 0.02,
"learning_rate": 0.0001981139337952271,
"loss": 0.0,
"step": 60
},
{
"epoch": 0.02,
"learning_rate": 0.00019792147806004618,
"loss": 0.0,
"step": 65
},
{
"epoch": 0.03,
"learning_rate": 0.0001977290223248653,
"loss": 0.0,
"step": 70
},
{
"epoch": 0.03,
"learning_rate": 0.00019753656658968438,
"loss": 0.0,
"step": 75
},
{
"epoch": 0.03,
"learning_rate": 0.0001973441108545035,
"loss": 0.0,
"step": 80
},
{
"epoch": 0.03,
"learning_rate": 0.00019715165511932257,
"loss": 0.0,
"step": 85
},
{
"epoch": 0.03,
"learning_rate": 0.00019695919938414165,
"loss": 0.0,
"step": 90
},
{
"epoch": 0.04,
"learning_rate": 0.00019676674364896076,
"loss": 0.0,
"step": 95
},
{
"epoch": 0.04,
"learning_rate": 0.00019657428791377982,
"loss": 0.0,
"step": 100
},
{
"epoch": 0.04,
"learning_rate": 0.00019638183217859893,
"loss": 0.0,
"step": 105
},
{
"epoch": 0.04,
"learning_rate": 0.000196189376443418,
"loss": 0.0,
"step": 110
},
{
"epoch": 0.04,
"learning_rate": 0.00019599692070823712,
"loss": 0.0,
"step": 115
},
{
"epoch": 0.05,
"learning_rate": 0.0001958044649730562,
"loss": 0.0,
"step": 120
},
{
"epoch": 0.05,
"learning_rate": 0.0001956120092378753,
"loss": 0.0,
"step": 125
},
{
"epoch": 0.05,
"learning_rate": 0.0001954195535026944,
"loss": 0.0,
"step": 130
},
{
"epoch": 0.05,
"learning_rate": 0.00019522709776751348,
"loss": 0.0,
"step": 135
},
{
"epoch": 0.05,
"learning_rate": 0.00019503464203233256,
"loss": 0.0,
"step": 140
},
{
"epoch": 0.06,
"learning_rate": 0.00019484218629715167,
"loss": 0.0,
"step": 145
},
{
"epoch": 0.06,
"learning_rate": 0.00019464973056197075,
"loss": 0.0,
"step": 150
},
{
"epoch": 0.06,
"learning_rate": 0.00019445727482678984,
"loss": 0.0,
"step": 155
},
{
"epoch": 0.06,
"learning_rate": 0.00019426481909160895,
"loss": 0.0,
"step": 160
},
{
"epoch": 0.06,
"learning_rate": 0.00019407236335642803,
"loss": 0.0,
"step": 165
},
{
"epoch": 0.07,
"learning_rate": 0.0001938799076212471,
"loss": 0.0,
"step": 170
},
{
"epoch": 0.07,
"learning_rate": 0.0001936874518860662,
"loss": 0.0,
"step": 175
},
{
"epoch": 0.07,
"learning_rate": 0.0001934949961508853,
"loss": 0.0,
"step": 180
},
{
"epoch": 0.07,
"learning_rate": 0.0001933025404157044,
"loss": 0.0,
"step": 185
},
{
"epoch": 0.07,
"learning_rate": 0.0001931100846805235,
"loss": 0.0,
"step": 190
},
{
"epoch": 0.07,
"learning_rate": 0.00019291762894534258,
"loss": 0.0,
"step": 195
},
{
"epoch": 0.08,
"learning_rate": 0.0001927251732101617,
"loss": 0.0,
"step": 200
},
{
"epoch": 0.08,
"learning_rate": 0.00019253271747498077,
"loss": 0.0,
"step": 205
},
{
"epoch": 0.08,
"learning_rate": 0.00019234026173979986,
"loss": 0.0,
"step": 210
},
{
"epoch": 0.08,
"learning_rate": 0.00019214780600461894,
"loss": 0.0,
"step": 215
},
{
"epoch": 0.08,
"learning_rate": 0.00019195535026943802,
"loss": 0.0,
"step": 220
},
{
"epoch": 0.09,
"learning_rate": 0.00019176289453425713,
"loss": 0.0,
"step": 225
},
{
"epoch": 0.09,
"learning_rate": 0.00019157043879907621,
"loss": 0.0,
"step": 230
},
{
"epoch": 0.09,
"learning_rate": 0.00019137798306389532,
"loss": 0.0,
"step": 235
},
{
"epoch": 0.09,
"learning_rate": 0.0001911855273287144,
"loss": 0.0,
"step": 240
},
{
"epoch": 0.09,
"learning_rate": 0.0001909930715935335,
"loss": 0.0,
"step": 245
},
{
"epoch": 0.1,
"learning_rate": 0.00019080061585835257,
"loss": 0.0,
"step": 250
},
{
"epoch": 0.1,
"learning_rate": 0.00019060816012317168,
"loss": 0.0,
"step": 255
},
{
"epoch": 0.1,
"learning_rate": 0.00019041570438799076,
"loss": 0.0,
"step": 260
},
{
"epoch": 0.1,
"learning_rate": 0.00019022324865280987,
"loss": 0.0,
"step": 265
},
{
"epoch": 0.1,
"learning_rate": 0.00019003079291762896,
"loss": 0.0,
"step": 270
},
{
"epoch": 0.11,
"learning_rate": 0.00018983833718244807,
"loss": 0.0,
"step": 275
},
{
"epoch": 0.11,
"learning_rate": 0.00018964588144726715,
"loss": 0.0,
"step": 280
},
{
"epoch": 0.11,
"learning_rate": 0.0001894534257120862,
"loss": 0.0,
"step": 285
},
{
"epoch": 0.11,
"learning_rate": 0.00018926096997690532,
"loss": 0.0,
"step": 290
},
{
"epoch": 0.11,
"learning_rate": 0.0001890685142417244,
"loss": 0.0,
"step": 295
},
{
"epoch": 0.12,
"learning_rate": 0.0001888760585065435,
"loss": 0.0,
"step": 300
},
{
"epoch": 0.12,
"learning_rate": 0.0001886836027713626,
"loss": 0.0,
"step": 305
},
{
"epoch": 0.12,
"learning_rate": 0.0001884911470361817,
"loss": 0.0,
"step": 310
},
{
"epoch": 0.12,
"learning_rate": 0.00018829869130100078,
"loss": 0.0,
"step": 315
},
{
"epoch": 0.12,
"learning_rate": 0.00018810623556581987,
"loss": 0.0,
"step": 320
},
{
"epoch": 0.12,
"learning_rate": 0.00018791377983063895,
"loss": 0.0,
"step": 325
},
{
"epoch": 0.13,
"learning_rate": 0.00018772132409545806,
"loss": 0.0,
"step": 330
},
{
"epoch": 0.13,
"learning_rate": 0.00018752886836027714,
"loss": 0.0,
"step": 335
},
{
"epoch": 0.13,
"learning_rate": 0.00018733641262509625,
"loss": 0.0,
"step": 340
},
{
"epoch": 0.13,
"learning_rate": 0.00018714395688991533,
"loss": 0.0,
"step": 345
},
{
"epoch": 0.13,
"learning_rate": 0.00018695150115473442,
"loss": 0.0,
"step": 350
},
{
"epoch": 0.14,
"learning_rate": 0.00018675904541955353,
"loss": 0.0,
"step": 355
},
{
"epoch": 0.14,
"learning_rate": 0.00018656658968437258,
"loss": 0.0,
"step": 360
},
{
"epoch": 0.14,
"learning_rate": 0.0001863741339491917,
"loss": 0.0,
"step": 365
},
{
"epoch": 0.14,
"learning_rate": 0.00018618167821401078,
"loss": 0.0,
"step": 370
},
{
"epoch": 0.14,
"learning_rate": 0.00018598922247882988,
"loss": 0.0,
"step": 375
},
{
"epoch": 0.15,
"learning_rate": 0.00018579676674364897,
"loss": 0.0,
"step": 380
},
{
"epoch": 0.15,
"learning_rate": 0.00018560431100846808,
"loss": 0.0,
"step": 385
},
{
"epoch": 0.15,
"learning_rate": 0.00018541185527328716,
"loss": 0.0,
"step": 390
},
{
"epoch": 0.15,
"learning_rate": 0.00018521939953810624,
"loss": 0.0,
"step": 395
},
{
"epoch": 0.15,
"learning_rate": 0.00018502694380292533,
"loss": 0.0,
"step": 400
},
{
"epoch": 0.16,
"learning_rate": 0.00018483448806774444,
"loss": 0.0,
"step": 405
},
{
"epoch": 0.16,
"learning_rate": 0.00018464203233256352,
"loss": 0.0,
"step": 410
},
{
"epoch": 0.16,
"learning_rate": 0.0001844495765973826,
"loss": 0.0,
"step": 415
},
{
"epoch": 0.16,
"learning_rate": 0.0001842571208622017,
"loss": 0.0,
"step": 420
},
{
"epoch": 0.16,
"learning_rate": 0.0001840646651270208,
"loss": 0.0,
"step": 425
},
{
"epoch": 0.17,
"learning_rate": 0.00018387220939183988,
"loss": 0.0,
"step": 430
},
{
"epoch": 0.17,
"learning_rate": 0.00018367975365665896,
"loss": 0.0,
"step": 435
},
{
"epoch": 0.17,
"learning_rate": 0.00018348729792147807,
"loss": 0.0,
"step": 440
},
{
"epoch": 0.17,
"learning_rate": 0.00018329484218629715,
"loss": 0.0,
"step": 445
},
{
"epoch": 0.17,
"learning_rate": 0.00018310238645111626,
"loss": 0.0,
"step": 450
},
{
"epoch": 0.17,
"learning_rate": 0.00018290993071593534,
"loss": 0.0,
"step": 455
},
{
"epoch": 0.18,
"learning_rate": 0.00018271747498075445,
"loss": 0.0,
"step": 460
},
{
"epoch": 0.18,
"learning_rate": 0.00018252501924557354,
"loss": 0.0,
"step": 465
},
{
"epoch": 0.18,
"learning_rate": 0.00018233256351039262,
"loss": 0.0,
"step": 470
},
{
"epoch": 0.18,
"learning_rate": 0.0001821401077752117,
"loss": 0.0,
"step": 475
},
{
"epoch": 0.18,
"learning_rate": 0.00018194765204003079,
"loss": 0.0,
"step": 480
},
{
"epoch": 0.19,
"learning_rate": 0.0001817551963048499,
"loss": 0.0,
"step": 485
},
{
"epoch": 0.19,
"learning_rate": 0.00018156274056966898,
"loss": 0.0,
"step": 490
},
{
"epoch": 0.19,
"learning_rate": 0.0001813702848344881,
"loss": 0.0,
"step": 495
},
{
"epoch": 0.19,
"learning_rate": 0.00018117782909930717,
"loss": 0.0,
"step": 500
},
{
"epoch": 0.19,
"learning_rate": 0.00018098537336412625,
"loss": 0.0,
"step": 505
},
{
"epoch": 0.2,
"learning_rate": 0.00018079291762894534,
"loss": 0.0,
"step": 510
},
{
"epoch": 0.2,
"learning_rate": 0.00018060046189376445,
"loss": 0.0,
"step": 515
},
{
"epoch": 0.2,
"learning_rate": 0.00018040800615858353,
"loss": 0.0,
"step": 520
},
{
"epoch": 0.2,
"learning_rate": 0.00018021555042340264,
"loss": 0.0,
"step": 525
},
{
"epoch": 0.2,
"learning_rate": 0.00018002309468822172,
"loss": 0.0,
"step": 530
},
{
"epoch": 0.21,
"learning_rate": 0.00017983063895304083,
"loss": 0.0,
"step": 535
},
{
"epoch": 0.21,
"learning_rate": 0.00017963818321785991,
"loss": 0.0,
"step": 540
},
{
"epoch": 0.21,
"learning_rate": 0.00017944572748267897,
"loss": 0.0,
"step": 545
},
{
"epoch": 0.21,
"learning_rate": 0.00017925327174749808,
"loss": 0.0,
"step": 550
},
{
"epoch": 0.21,
"learning_rate": 0.00017906081601231716,
"loss": 0.0,
"step": 555
},
{
"epoch": 0.22,
"learning_rate": 0.00017886836027713627,
"loss": 0.0,
"step": 560
},
{
"epoch": 0.22,
"learning_rate": 0.00017867590454195535,
"loss": 0.0,
"step": 565
},
{
"epoch": 0.22,
"learning_rate": 0.00017848344880677446,
"loss": 0.0,
"step": 570
},
{
"epoch": 0.22,
"learning_rate": 0.00017829099307159355,
"loss": 0.0,
"step": 575
},
{
"epoch": 0.22,
"learning_rate": 0.00017809853733641263,
"loss": 0.0,
"step": 580
},
{
"epoch": 0.22,
"learning_rate": 0.0001779060816012317,
"loss": 0.0,
"step": 585
},
{
"epoch": 0.23,
"learning_rate": 0.00017771362586605082,
"loss": 0.0,
"step": 590
},
{
"epoch": 0.23,
"learning_rate": 0.0001775211701308699,
"loss": 0.0,
"step": 595
},
{
"epoch": 0.23,
"learning_rate": 0.00017732871439568902,
"loss": 0.0,
"step": 600
},
{
"epoch": 0.23,
"learning_rate": 0.0001771362586605081,
"loss": 0.0,
"step": 605
},
{
"epoch": 0.23,
"learning_rate": 0.00017694380292532718,
"loss": 0.0,
"step": 610
},
{
"epoch": 0.24,
"learning_rate": 0.00017675134719014626,
"loss": 0.0,
"step": 615
},
{
"epoch": 0.24,
"learning_rate": 0.00017655889145496535,
"loss": 0.0,
"step": 620
},
{
"epoch": 0.24,
"learning_rate": 0.00017636643571978446,
"loss": 0.0,
"step": 625
},
{
"epoch": 0.24,
"learning_rate": 0.00017617397998460354,
"loss": 0.0,
"step": 630
},
{
"epoch": 0.24,
"learning_rate": 0.00017598152424942265,
"loss": 0.0,
"step": 635
},
{
"epoch": 0.25,
"learning_rate": 0.00017578906851424173,
"loss": 0.0,
"step": 640
},
{
"epoch": 0.25,
"learning_rate": 0.00017559661277906084,
"loss": 0.0,
"step": 645
},
{
"epoch": 0.25,
"learning_rate": 0.00017540415704387992,
"loss": 0.0,
"step": 650
},
{
"epoch": 0.25,
"learning_rate": 0.000175211701308699,
"loss": 0.0,
"step": 655
},
{
"epoch": 0.25,
"learning_rate": 0.0001750192455735181,
"loss": 0.0,
"step": 660
},
{
"epoch": 0.26,
"learning_rate": 0.0001748267898383372,
"loss": 0.0,
"step": 665
},
{
"epoch": 0.26,
"learning_rate": 0.00017463433410315628,
"loss": 0.0,
"step": 670
},
{
"epoch": 0.26,
"learning_rate": 0.00017444187836797537,
"loss": 0.0,
"step": 675
},
{
"epoch": 0.26,
"learning_rate": 0.00017424942263279448,
"loss": 0.0,
"step": 680
},
{
"epoch": 0.26,
"learning_rate": 0.00017405696689761356,
"loss": 0.0,
"step": 685
},
{
"epoch": 0.27,
"learning_rate": 0.00017386451116243264,
"loss": 0.0,
"step": 690
},
{
"epoch": 0.27,
"learning_rate": 0.00017367205542725172,
"loss": 0.0,
"step": 695
},
{
"epoch": 0.27,
"learning_rate": 0.00017347959969207083,
"loss": 0.0,
"step": 700
},
{
"epoch": 0.27,
"learning_rate": 0.00017328714395688992,
"loss": 0.0,
"step": 705
},
{
"epoch": 0.27,
"learning_rate": 0.00017309468822170903,
"loss": 0.0,
"step": 710
},
{
"epoch": 0.27,
"learning_rate": 0.0001729022324865281,
"loss": 0.0,
"step": 715
},
{
"epoch": 0.28,
"learning_rate": 0.00017270977675134722,
"loss": 0.0,
"step": 720
},
{
"epoch": 0.28,
"learning_rate": 0.0001725173210161663,
"loss": 0.0,
"step": 725
},
{
"epoch": 0.28,
"learning_rate": 0.00017232486528098538,
"loss": 0.0,
"step": 730
},
{
"epoch": 0.28,
"learning_rate": 0.00017213240954580447,
"loss": 0.0,
"step": 735
},
{
"epoch": 0.28,
"learning_rate": 0.00017193995381062355,
"loss": 0.0,
"step": 740
},
{
"epoch": 0.29,
"learning_rate": 0.00017174749807544266,
"loss": 0.0,
"step": 745
},
{
"epoch": 0.29,
"learning_rate": 0.00017155504234026174,
"loss": 0.0,
"step": 750
},
{
"epoch": 0.29,
"learning_rate": 0.00017136258660508085,
"loss": 0.0,
"step": 755
},
{
"epoch": 0.29,
"learning_rate": 0.00017117013086989993,
"loss": 0.0,
"step": 760
},
{
"epoch": 0.29,
"learning_rate": 0.00017097767513471902,
"loss": 0.0,
"step": 765
},
{
"epoch": 0.3,
"learning_rate": 0.0001707852193995381,
"loss": 0.0,
"step": 770
},
{
"epoch": 0.3,
"learning_rate": 0.0001705927636643572,
"loss": 0.0,
"step": 775
},
{
"epoch": 0.3,
"learning_rate": 0.0001704003079291763,
"loss": 0.0,
"step": 780
},
{
"epoch": 0.3,
"learning_rate": 0.0001702078521939954,
"loss": 0.0,
"step": 785
},
{
"epoch": 0.3,
"learning_rate": 0.00017001539645881449,
"loss": 0.0,
"step": 790
},
{
"epoch": 0.31,
"learning_rate": 0.0001698229407236336,
"loss": 0.0,
"step": 795
},
{
"epoch": 0.31,
"learning_rate": 0.00016963048498845268,
"loss": 0.0,
"step": 800
},
{
"epoch": 0.31,
"learning_rate": 0.00016943802925327173,
"loss": 0.0,
"step": 805
},
{
"epoch": 0.31,
"learning_rate": 0.00016924557351809084,
"loss": 0.0,
"step": 810
},
{
"epoch": 0.31,
"learning_rate": 0.00016905311778290993,
"loss": 0.0,
"step": 815
},
{
"epoch": 0.32,
"learning_rate": 0.00016886066204772904,
"loss": 0.0,
"step": 820
},
{
"epoch": 0.32,
"learning_rate": 0.00016866820631254812,
"loss": 0.0,
"step": 825
},
{
"epoch": 0.32,
"learning_rate": 0.00016847575057736723,
"loss": 0.0,
"step": 830
},
{
"epoch": 0.32,
"learning_rate": 0.0001682832948421863,
"loss": 0.0,
"step": 835
},
{
"epoch": 0.32,
"learning_rate": 0.0001680908391070054,
"loss": 0.0,
"step": 840
},
{
"epoch": 0.32,
"learning_rate": 0.00016789838337182448,
"loss": 0.0,
"step": 845
},
{
"epoch": 0.33,
"learning_rate": 0.0001677059276366436,
"loss": 0.0,
"step": 850
},
{
"epoch": 0.33,
"learning_rate": 0.00016751347190146267,
"loss": 0.0,
"step": 855
},
{
"epoch": 0.33,
"learning_rate": 0.00016732101616628175,
"loss": 0.0,
"step": 860
},
{
"epoch": 0.33,
"learning_rate": 0.00016712856043110086,
"loss": 0.0,
"step": 865
},
{
"epoch": 0.33,
"learning_rate": 0.00016693610469591995,
"loss": 0.0,
"step": 870
},
{
"epoch": 0.34,
"learning_rate": 0.00016674364896073903,
"loss": 0.0,
"step": 875
},
{
"epoch": 0.34,
"learning_rate": 0.0001665511932255581,
"loss": 0.0,
"step": 880
},
{
"epoch": 0.34,
"learning_rate": 0.00016635873749037722,
"loss": 0.0,
"step": 885
},
{
"epoch": 0.34,
"learning_rate": 0.0001661662817551963,
"loss": 0.0,
"step": 890
},
{
"epoch": 0.34,
"learning_rate": 0.0001659738260200154,
"loss": 0.0,
"step": 895
},
{
"epoch": 0.35,
"learning_rate": 0.0001657813702848345,
"loss": 0.0,
"step": 900
},
{
"epoch": 0.35,
"learning_rate": 0.0001655889145496536,
"loss": 0.0,
"step": 905
},
{
"epoch": 0.35,
"learning_rate": 0.0001653964588144727,
"loss": 0.0,
"step": 910
},
{
"epoch": 0.35,
"learning_rate": 0.00016520400307929177,
"loss": 0.0,
"step": 915
},
{
"epoch": 0.35,
"learning_rate": 0.00016501154734411085,
"loss": 0.0,
"step": 920
},
{
"epoch": 0.36,
"learning_rate": 0.00016481909160892994,
"loss": 0.0,
"step": 925
},
{
"epoch": 0.36,
"learning_rate": 0.00016462663587374905,
"loss": 0.0,
"step": 930
},
{
"epoch": 0.36,
"learning_rate": 0.00016443418013856813,
"loss": 0.0,
"step": 935
},
{
"epoch": 0.36,
"learning_rate": 0.00016424172440338724,
"loss": 0.0,
"step": 940
},
{
"epoch": 0.36,
"learning_rate": 0.00016404926866820632,
"loss": 0.0,
"step": 945
},
{
"epoch": 0.36,
"learning_rate": 0.0001638568129330254,
"loss": 0.0,
"step": 950
},
{
"epoch": 0.37,
"learning_rate": 0.0001636643571978445,
"loss": 0.0,
"step": 955
},
{
"epoch": 0.37,
"learning_rate": 0.0001634719014626636,
"loss": 0.0,
"step": 960
},
{
"epoch": 0.37,
"learning_rate": 0.00016327944572748268,
"loss": 0.0,
"step": 965
},
{
"epoch": 0.37,
"learning_rate": 0.0001630869899923018,
"loss": 0.0,
"step": 970
},
{
"epoch": 0.37,
"learning_rate": 0.00016289453425712087,
"loss": 0.0,
"step": 975
},
{
"epoch": 0.38,
"learning_rate": 0.00016270207852193998,
"loss": 0.0,
"step": 980
},
{
"epoch": 0.38,
"learning_rate": 0.00016250962278675907,
"loss": 0.0,
"step": 985
},
{
"epoch": 0.38,
"learning_rate": 0.00016231716705157812,
"loss": 0.0,
"step": 990
},
{
"epoch": 0.38,
"learning_rate": 0.00016212471131639723,
"loss": 0.0,
"step": 995
},
{
"epoch": 0.38,
"learning_rate": 0.00016193225558121631,
"loss": 0.0,
"step": 1000
},
{
"epoch": 0.39,
"learning_rate": 0.00016173979984603542,
"loss": 0.0,
"step": 1005
},
{
"epoch": 0.39,
"learning_rate": 0.0001615473441108545,
"loss": 0.0,
"step": 1010
},
{
"epoch": 0.39,
"learning_rate": 0.00016135488837567362,
"loss": 0.0,
"step": 1015
},
{
"epoch": 0.39,
"learning_rate": 0.0001611624326404927,
"loss": 0.0,
"step": 1020
},
{
"epoch": 0.39,
"learning_rate": 0.00016096997690531178,
"loss": 0.0,
"step": 1025
},
{
"epoch": 0.4,
"learning_rate": 0.00016077752117013086,
"loss": 0.0,
"step": 1030
},
{
"epoch": 0.4,
"learning_rate": 0.00016058506543494997,
"loss": 0.0,
"step": 1035
},
{
"epoch": 0.4,
"learning_rate": 0.00016039260969976906,
"loss": 0.0,
"step": 1040
},
{
"epoch": 0.4,
"learning_rate": 0.00016020015396458817,
"loss": 0.0,
"step": 1045
},
{
"epoch": 0.4,
"learning_rate": 0.00016000769822940725,
"loss": 0.0,
"step": 1050
},
{
"epoch": 0.41,
"learning_rate": 0.00015981524249422633,
"loss": 0.0,
"step": 1055
},
{
"epoch": 0.41,
"learning_rate": 0.00015962278675904542,
"loss": 0.0,
"step": 1060
},
{
"epoch": 0.41,
"learning_rate": 0.0001594303310238645,
"loss": 0.0,
"step": 1065
},
{
"epoch": 0.41,
"learning_rate": 0.0001592378752886836,
"loss": 0.0,
"step": 1070
},
{
"epoch": 0.41,
"learning_rate": 0.0001590454195535027,
"loss": 0.0,
"step": 1075
},
{
"epoch": 0.41,
"learning_rate": 0.0001588529638183218,
"loss": 0.0,
"step": 1080
},
{
"epoch": 0.42,
"learning_rate": 0.00015866050808314088,
"loss": 0.0,
"step": 1085
},
{
"epoch": 0.42,
"learning_rate": 0.00015846805234796,
"loss": 0.0,
"step": 1090
},
{
"epoch": 0.42,
"learning_rate": 0.00015827559661277908,
"loss": 0.0,
"step": 1095
},
{
"epoch": 0.42,
"learning_rate": 0.00015808314087759816,
"loss": 0.0,
"step": 1100
},
{
"epoch": 0.42,
"learning_rate": 0.00015789068514241724,
"loss": 0.0,
"step": 1105
},
{
"epoch": 0.43,
"learning_rate": 0.00015769822940723635,
"loss": 0.0,
"step": 1110
},
{
"epoch": 0.43,
"learning_rate": 0.00015750577367205543,
"loss": 0.0,
"step": 1115
},
{
"epoch": 0.43,
"learning_rate": 0.00015731331793687452,
"loss": 0.0,
"step": 1120
},
{
"epoch": 0.43,
"learning_rate": 0.00015712086220169363,
"loss": 0.0,
"step": 1125
},
{
"epoch": 0.43,
"learning_rate": 0.0001569284064665127,
"loss": 0.0,
"step": 1130
},
{
"epoch": 0.44,
"learning_rate": 0.0001567359507313318,
"loss": 0.0,
"step": 1135
},
{
"epoch": 0.44,
"learning_rate": 0.00015654349499615087,
"loss": 0.0,
"step": 1140
},
{
"epoch": 0.44,
"learning_rate": 0.00015635103926096998,
"loss": 0.0,
"step": 1145
},
{
"epoch": 0.44,
"learning_rate": 0.00015615858352578907,
"loss": 0.0,
"step": 1150
},
{
"epoch": 0.44,
"learning_rate": 0.00015596612779060818,
"loss": 0.0,
"step": 1155
},
{
"epoch": 0.45,
"learning_rate": 0.00015577367205542726,
"loss": 0.0,
"step": 1160
},
{
"epoch": 0.45,
"learning_rate": 0.00015558121632024637,
"loss": 0.0,
"step": 1165
},
{
"epoch": 0.45,
"learning_rate": 0.00015538876058506545,
"loss": 0.0,
"step": 1170
},
{
"epoch": 0.45,
"learning_rate": 0.00015519630484988454,
"loss": 0.0,
"step": 1175
},
{
"epoch": 0.45,
"learning_rate": 0.00015500384911470362,
"loss": 0.0,
"step": 1180
},
{
"epoch": 0.46,
"learning_rate": 0.0001548113933795227,
"loss": 0.0,
"step": 1185
},
{
"epoch": 0.46,
"learning_rate": 0.0001546189376443418,
"loss": 0.0,
"step": 1190
},
{
"epoch": 0.46,
"learning_rate": 0.0001544264819091609,
"loss": 0.0,
"step": 1195
},
{
"epoch": 0.46,
"learning_rate": 0.00015423402617398,
"loss": 0.0,
"step": 1200
},
{
"epoch": 0.46,
"learning_rate": 0.00015404157043879909,
"loss": 0.0,
"step": 1205
},
{
"epoch": 0.46,
"learning_rate": 0.00015384911470361817,
"loss": 0.0,
"step": 1210
},
{
"epoch": 0.47,
"learning_rate": 0.00015365665896843725,
"loss": 0.0,
"step": 1215
},
{
"epoch": 0.47,
"learning_rate": 0.00015346420323325636,
"loss": 0.0,
"step": 1220
},
{
"epoch": 0.47,
"learning_rate": 0.00015327174749807544,
"loss": 0.0,
"step": 1225
},
{
"epoch": 0.47,
"learning_rate": 0.00015307929176289455,
"loss": 0.0,
"step": 1230
},
{
"epoch": 0.47,
"learning_rate": 0.00015288683602771364,
"loss": 0.0,
"step": 1235
},
{
"epoch": 0.48,
"learning_rate": 0.00015269438029253275,
"loss": 0.0,
"step": 1240
},
{
"epoch": 0.48,
"learning_rate": 0.00015250192455735183,
"loss": 0.0,
"step": 1245
},
{
"epoch": 0.48,
"learning_rate": 0.00015230946882217089,
"loss": 0.0,
"step": 1250
},
{
"epoch": 0.48,
"learning_rate": 0.00015211701308699,
"loss": 0.0,
"step": 1255
},
{
"epoch": 0.48,
"learning_rate": 0.00015192455735180908,
"loss": 0.0,
"step": 1260
},
{
"epoch": 0.49,
"learning_rate": 0.0001517321016166282,
"loss": 0.0,
"step": 1265
},
{
"epoch": 0.49,
"learning_rate": 0.00015153964588144727,
"loss": 0.0,
"step": 1270
},
{
"epoch": 0.49,
"learning_rate": 0.00015134719014626638,
"loss": 0.0,
"step": 1275
},
{
"epoch": 0.49,
"learning_rate": 0.00015115473441108546,
"loss": 0.0,
"step": 1280
},
{
"epoch": 0.49,
"learning_rate": 0.00015096227867590455,
"loss": 0.0,
"step": 1285
},
{
"epoch": 0.5,
"learning_rate": 0.00015076982294072363,
"loss": 0.0,
"step": 1290
},
{
"epoch": 0.5,
"learning_rate": 0.00015057736720554274,
"loss": 0.0,
"step": 1295
},
{
"epoch": 0.5,
"learning_rate": 0.00015038491147036182,
"loss": 0.0,
"step": 1300
},
{
"epoch": 0.5,
"learning_rate": 0.00015019245573518093,
"loss": 0.0,
"step": 1305
},
{
"epoch": 0.5,
"learning_rate": 0.00015000000000000001,
"loss": 0.0,
"step": 1310
},
{
"epoch": 0.51,
"learning_rate": 0.0001498075442648191,
"loss": 0.0,
"step": 1315
},
{
"epoch": 0.51,
"learning_rate": 0.00014961508852963818,
"loss": 0.0,
"step": 1320
},
{
"epoch": 0.51,
"learning_rate": 0.00014942263279445726,
"loss": 0.0,
"step": 1325
},
{
"epoch": 0.51,
"learning_rate": 0.00014923017705927637,
"loss": 0.0,
"step": 1330
},
{
"epoch": 0.51,
"learning_rate": 0.00014903772132409545,
"loss": 0.0,
"step": 1335
},
{
"epoch": 0.51,
"learning_rate": 0.00014884526558891456,
"loss": 0.0,
"step": 1340
},
{
"epoch": 0.52,
"learning_rate": 0.00014865280985373365,
"loss": 0.0,
"step": 1345
},
{
"epoch": 0.52,
"learning_rate": 0.00014846035411855276,
"loss": 0.0,
"step": 1350
},
{
"epoch": 0.52,
"learning_rate": 0.00014826789838337184,
"loss": 0.0,
"step": 1355
},
{
"epoch": 0.52,
"learning_rate": 0.00014807544264819092,
"loss": 0.0,
"step": 1360
},
{
"epoch": 0.52,
"learning_rate": 0.00014788298691301,
"loss": 0.0,
"step": 1365
},
{
"epoch": 0.53,
"learning_rate": 0.00014769053117782912,
"loss": 0.0,
"step": 1370
},
{
"epoch": 0.53,
"learning_rate": 0.0001474980754426482,
"loss": 0.0,
"step": 1375
},
{
"epoch": 0.53,
"learning_rate": 0.00014730561970746728,
"loss": 0.0,
"step": 1380
},
{
"epoch": 0.53,
"learning_rate": 0.0001471131639722864,
"loss": 0.0,
"step": 1385
},
{
"epoch": 0.53,
"learning_rate": 0.00014692070823710547,
"loss": 0.0,
"step": 1390
},
{
"epoch": 0.54,
"learning_rate": 0.00014672825250192456,
"loss": 0.0,
"step": 1395
},
{
"epoch": 0.54,
"learning_rate": 0.00014653579676674364,
"loss": 0.0,
"step": 1400
},
{
"epoch": 0.54,
"learning_rate": 0.00014634334103156275,
"loss": 0.0,
"step": 1405
},
{
"epoch": 0.54,
"learning_rate": 0.00014615088529638183,
"loss": 0.0,
"step": 1410
},
{
"epoch": 0.54,
"learning_rate": 0.00014595842956120094,
"loss": 0.0,
"step": 1415
},
{
"epoch": 0.55,
"learning_rate": 0.00014576597382602002,
"loss": 0.0,
"step": 1420
},
{
"epoch": 0.55,
"learning_rate": 0.00014557351809083913,
"loss": 0.0,
"step": 1425
},
{
"epoch": 0.55,
"learning_rate": 0.00014538106235565822,
"loss": 0.0,
"step": 1430
},
{
"epoch": 0.55,
"learning_rate": 0.0001451886066204773,
"loss": 0.0,
"step": 1435
},
{
"epoch": 0.55,
"learning_rate": 0.00014499615088529638,
"loss": 0.0,
"step": 1440
},
{
"epoch": 0.56,
"learning_rate": 0.00014480369515011547,
"loss": 0.0,
"step": 1445
},
{
"epoch": 0.56,
"learning_rate": 0.00014461123941493458,
"loss": 0.0,
"step": 1450
},
{
"epoch": 0.56,
"learning_rate": 0.00014441878367975366,
"loss": 0.0,
"step": 1455
},
{
"epoch": 0.56,
"learning_rate": 0.00014422632794457277,
"loss": 0.0,
"step": 1460
},
{
"epoch": 0.56,
"learning_rate": 0.00014403387220939185,
"loss": 0.0,
"step": 1465
},
{
"epoch": 0.56,
"learning_rate": 0.00014384141647421093,
"loss": 0.0,
"step": 1470
},
{
"epoch": 0.57,
"learning_rate": 0.00014364896073903002,
"loss": 0.0,
"step": 1475
},
{
"epoch": 0.57,
"learning_rate": 0.00014345650500384913,
"loss": 0.0,
"step": 1480
},
{
"epoch": 0.57,
"learning_rate": 0.0001432640492686682,
"loss": 0.0,
"step": 1485
},
{
"epoch": 0.57,
"learning_rate": 0.00014307159353348732,
"loss": 0.0,
"step": 1490
},
{
"epoch": 0.57,
"learning_rate": 0.0001428791377983064,
"loss": 0.0,
"step": 1495
},
{
"epoch": 0.58,
"learning_rate": 0.0001426866820631255,
"loss": 0.0,
"step": 1500
},
{
"epoch": 0.58,
"learning_rate": 0.00014249422632794457,
"loss": 0.0,
"step": 1505
},
{
"epoch": 0.58,
"learning_rate": 0.00014230177059276365,
"loss": 0.0,
"step": 1510
},
{
"epoch": 0.58,
"learning_rate": 0.00014210931485758276,
"loss": 0.0,
"step": 1515
},
{
"epoch": 0.58,
"learning_rate": 0.00014191685912240184,
"loss": 0.0,
"step": 1520
},
{
"epoch": 0.59,
"learning_rate": 0.00014172440338722095,
"loss": 0.0,
"step": 1525
},
{
"epoch": 0.59,
"learning_rate": 0.00014153194765204003,
"loss": 0.0,
"step": 1530
},
{
"epoch": 0.59,
"learning_rate": 0.00014133949191685914,
"loss": 0.0,
"step": 1535
},
{
"epoch": 0.59,
"learning_rate": 0.00014114703618167823,
"loss": 0.0,
"step": 1540
},
{
"epoch": 0.59,
"learning_rate": 0.0001409545804464973,
"loss": 0.0,
"step": 1545
},
{
"epoch": 0.6,
"learning_rate": 0.0001407621247113164,
"loss": 0.0,
"step": 1550
},
{
"epoch": 0.6,
"learning_rate": 0.0001405696689761355,
"loss": 0.0,
"step": 1555
},
{
"epoch": 0.6,
"learning_rate": 0.00014037721324095459,
"loss": 0.0,
"step": 1560
},
{
"epoch": 0.6,
"learning_rate": 0.0001401847575057737,
"loss": 0.0,
"step": 1565
},
{
"epoch": 0.6,
"learning_rate": 0.00013999230177059278,
"loss": 0.0,
"step": 1570
},
{
"epoch": 0.61,
"learning_rate": 0.00013979984603541186,
"loss": 0.0,
"step": 1575
},
{
"epoch": 0.61,
"learning_rate": 0.00013960739030023094,
"loss": 0.0,
"step": 1580
},
{
"epoch": 0.61,
"learning_rate": 0.00013941493456505003,
"loss": 0.0,
"step": 1585
},
{
"epoch": 0.61,
"learning_rate": 0.00013922247882986914,
"loss": 0.0,
"step": 1590
},
{
"epoch": 0.61,
"learning_rate": 0.00013903002309468822,
"loss": 0.0,
"step": 1595
},
{
"epoch": 0.61,
"learning_rate": 0.00013883756735950733,
"loss": 0.0,
"step": 1600
},
{
"epoch": 0.62,
"learning_rate": 0.0001386451116243264,
"loss": 0.0,
"step": 1605
},
{
"epoch": 0.62,
"learning_rate": 0.00013845265588914552,
"loss": 0.0,
"step": 1610
},
{
"epoch": 0.62,
"learning_rate": 0.0001382602001539646,
"loss": 0.0,
"step": 1615
},
{
"epoch": 0.62,
"learning_rate": 0.0001380677444187837,
"loss": 0.0,
"step": 1620
},
{
"epoch": 0.62,
"learning_rate": 0.00013787528868360277,
"loss": 0.0,
"step": 1625
},
{
"epoch": 0.63,
"learning_rate": 0.00013768283294842188,
"loss": 0.0,
"step": 1630
},
{
"epoch": 0.63,
"learning_rate": 0.00013749037721324096,
"loss": 0.0,
"step": 1635
},
{
"epoch": 0.63,
"learning_rate": 0.00013729792147806005,
"loss": 0.0,
"step": 1640
},
{
"epoch": 0.63,
"learning_rate": 0.00013710546574287916,
"loss": 0.0,
"step": 1645
},
{
"epoch": 0.63,
"learning_rate": 0.00013691301000769824,
"loss": 0.0,
"step": 1650
},
{
"epoch": 0.64,
"learning_rate": 0.00013672055427251732,
"loss": 0.0,
"step": 1655
},
{
"epoch": 0.64,
"learning_rate": 0.0001365280985373364,
"loss": 0.0,
"step": 1660
},
{
"epoch": 0.64,
"learning_rate": 0.0001363356428021555,
"loss": 0.0,
"step": 1665
},
{
"epoch": 0.64,
"learning_rate": 0.0001361431870669746,
"loss": 0.0,
"step": 1670
},
{
"epoch": 0.64,
"learning_rate": 0.0001359507313317937,
"loss": 0.0,
"step": 1675
},
{
"epoch": 0.65,
"learning_rate": 0.0001357582755966128,
"loss": 0.0,
"step": 1680
},
{
"epoch": 0.65,
"learning_rate": 0.0001355658198614319,
"loss": 0.0,
"step": 1685
},
{
"epoch": 0.65,
"learning_rate": 0.00013537336412625098,
"loss": 0.0,
"step": 1690
},
{
"epoch": 0.65,
"learning_rate": 0.00013518090839107006,
"loss": 0.0,
"step": 1695
},
{
"epoch": 0.65,
"learning_rate": 0.00013498845265588915,
"loss": 0.0,
"step": 1700
},
{
"epoch": 0.66,
"learning_rate": 0.00013479599692070823,
"loss": 0.0,
"step": 1705
},
{
"epoch": 0.66,
"learning_rate": 0.00013460354118552734,
"loss": 0.0,
"step": 1710
},
{
"epoch": 0.66,
"learning_rate": 0.00013441108545034642,
"loss": 0.0,
"step": 1715
},
{
"epoch": 0.66,
"learning_rate": 0.00013421862971516553,
"loss": 0.0,
"step": 1720
},
{
"epoch": 0.66,
"learning_rate": 0.00013402617397998461,
"loss": 0.0,
"step": 1725
},
{
"epoch": 0.66,
"learning_rate": 0.0001338337182448037,
"loss": 0.0,
"step": 1730
},
{
"epoch": 0.67,
"learning_rate": 0.00013364126250962278,
"loss": 0.0,
"step": 1735
},
{
"epoch": 0.67,
"learning_rate": 0.0001334488067744419,
"loss": 0.0,
"step": 1740
},
{
"epoch": 0.67,
"learning_rate": 0.00013325635103926097,
"loss": 0.0,
"step": 1745
},
{
"epoch": 0.67,
"learning_rate": 0.00013306389530408008,
"loss": 0.0,
"step": 1750
},
{
"epoch": 0.67,
"learning_rate": 0.00013287143956889917,
"loss": 0.0,
"step": 1755
},
{
"epoch": 0.68,
"learning_rate": 0.00013267898383371828,
"loss": 0.0,
"step": 1760
},
{
"epoch": 0.68,
"learning_rate": 0.00013248652809853733,
"loss": 0.0,
"step": 1765
},
{
"epoch": 0.68,
"learning_rate": 0.00013229407236335641,
"loss": 0.0,
"step": 1770
},
{
"epoch": 0.68,
"learning_rate": 0.00013210161662817552,
"loss": 0.0,
"step": 1775
},
{
"epoch": 0.68,
"learning_rate": 0.0001319091608929946,
"loss": 0.0,
"step": 1780
},
{
"epoch": 0.69,
"learning_rate": 0.00013171670515781372,
"loss": 0.0,
"step": 1785
},
{
"epoch": 0.69,
"learning_rate": 0.0001315242494226328,
"loss": 0.0,
"step": 1790
},
{
"epoch": 0.69,
"learning_rate": 0.0001313317936874519,
"loss": 0.0,
"step": 1795
},
{
"epoch": 0.69,
"learning_rate": 0.000131139337952271,
"loss": 0.0,
"step": 1800
},
{
"epoch": 0.69,
"learning_rate": 0.00013094688221709007,
"loss": 0.0,
"step": 1805
},
{
"epoch": 0.7,
"learning_rate": 0.00013075442648190916,
"loss": 0.0,
"step": 1810
},
{
"epoch": 0.7,
"learning_rate": 0.00013056197074672827,
"loss": 0.0,
"step": 1815
},
{
"epoch": 0.7,
"learning_rate": 0.00013036951501154735,
"loss": 0.0,
"step": 1820
},
{
"epoch": 0.7,
"learning_rate": 0.00013017705927636643,
"loss": 0.0,
"step": 1825
},
{
"epoch": 0.7,
"learning_rate": 0.00012998460354118554,
"loss": 0.0,
"step": 1830
},
{
"epoch": 0.7,
"learning_rate": 0.00012979214780600463,
"loss": 0.0,
"step": 1835
},
{
"epoch": 0.71,
"learning_rate": 0.0001295996920708237,
"loss": 0.0,
"step": 1840
},
{
"epoch": 0.71,
"learning_rate": 0.0001294072363356428,
"loss": 0.0,
"step": 1845
},
{
"epoch": 0.71,
"learning_rate": 0.0001292147806004619,
"loss": 0.0,
"step": 1850
},
{
"epoch": 0.71,
"learning_rate": 0.00012902232486528098,
"loss": 0.0,
"step": 1855
},
{
"epoch": 0.71,
"learning_rate": 0.0001288298691301001,
"loss": 0.0,
"step": 1860
},
{
"epoch": 0.72,
"learning_rate": 0.00012863741339491918,
"loss": 0.0,
"step": 1865
},
{
"epoch": 0.72,
"learning_rate": 0.00012844495765973829,
"loss": 0.0,
"step": 1870
},
{
"epoch": 0.72,
"learning_rate": 0.00012825250192455737,
"loss": 0.0,
"step": 1875
},
{
"epoch": 0.72,
"learning_rate": 0.00012806004618937645,
"loss": 0.0,
"step": 1880
},
{
"epoch": 0.72,
"learning_rate": 0.00012786759045419553,
"loss": 0.0,
"step": 1885
},
{
"epoch": 0.73,
"learning_rate": 0.00012767513471901462,
"loss": 0.0,
"step": 1890
},
{
"epoch": 0.73,
"learning_rate": 0.00012748267898383373,
"loss": 0.0,
"step": 1895
},
{
"epoch": 0.73,
"learning_rate": 0.0001272902232486528,
"loss": 0.0,
"step": 1900
},
{
"epoch": 0.73,
"learning_rate": 0.00012709776751347192,
"loss": 0.0,
"step": 1905
},
{
"epoch": 0.73,
"learning_rate": 0.000126905311778291,
"loss": 0.0,
"step": 1910
},
{
"epoch": 0.74,
"learning_rate": 0.00012671285604311008,
"loss": 0.0,
"step": 1915
},
{
"epoch": 0.74,
"learning_rate": 0.00012652040030792917,
"loss": 0.0,
"step": 1920
},
{
"epoch": 0.74,
"learning_rate": 0.00012632794457274828,
"loss": 0.0,
"step": 1925
},
{
"epoch": 0.74,
"learning_rate": 0.00012613548883756736,
"loss": 0.0,
"step": 1930
},
{
"epoch": 0.74,
"learning_rate": 0.00012594303310238647,
"loss": 0.0,
"step": 1935
},
{
"epoch": 0.75,
"learning_rate": 0.00012575057736720555,
"loss": 0.0,
"step": 1940
},
{
"epoch": 0.75,
"learning_rate": 0.00012555812163202466,
"loss": 0.0,
"step": 1945
},
{
"epoch": 0.75,
"learning_rate": 0.00012536566589684372,
"loss": 0.0,
"step": 1950
},
{
"epoch": 0.75,
"learning_rate": 0.0001251732101616628,
"loss": 0.0,
"step": 1955
},
{
"epoch": 0.75,
"learning_rate": 0.0001249807544264819,
"loss": 0.0,
"step": 1960
},
{
"epoch": 0.75,
"learning_rate": 0.000124788298691301,
"loss": 0.0,
"step": 1965
},
{
"epoch": 0.76,
"learning_rate": 0.0001245958429561201,
"loss": 0.0,
"step": 1970
},
{
"epoch": 0.76,
"learning_rate": 0.00012440338722093919,
"loss": 0.0,
"step": 1975
},
{
"epoch": 0.76,
"learning_rate": 0.0001242109314857583,
"loss": 0.0,
"step": 1980
},
{
"epoch": 0.76,
"learning_rate": 0.00012401847575057738,
"loss": 0.0,
"step": 1985
},
{
"epoch": 0.76,
"learning_rate": 0.00012382602001539646,
"loss": 0.0,
"step": 1990
},
{
"epoch": 0.77,
"learning_rate": 0.00012363356428021554,
"loss": 0.0,
"step": 1995
},
{
"epoch": 0.77,
"learning_rate": 0.00012344110854503465,
"loss": 0.0,
"step": 2000
},
{
"epoch": 0.77,
"learning_rate": 0.00012324865280985374,
"loss": 0.0,
"step": 2005
},
{
"epoch": 0.77,
"learning_rate": 0.00012305619707467285,
"loss": 0.0,
"step": 2010
},
{
"epoch": 0.77,
"learning_rate": 0.00012286374133949193,
"loss": 0.0,
"step": 2015
},
{
"epoch": 0.78,
"learning_rate": 0.000122671285604311,
"loss": 0.0,
"step": 2020
},
{
"epoch": 0.78,
"learning_rate": 0.0001224788298691301,
"loss": 0.0,
"step": 2025
},
{
"epoch": 0.78,
"learning_rate": 0.00012228637413394918,
"loss": 0.0,
"step": 2030
},
{
"epoch": 0.78,
"learning_rate": 0.0001220939183987683,
"loss": 0.0,
"step": 2035
},
{
"epoch": 0.78,
"learning_rate": 0.00012190146266358737,
"loss": 0.0,
"step": 2040
},
{
"epoch": 0.79,
"learning_rate": 0.00012170900692840648,
"loss": 0.0,
"step": 2045
},
{
"epoch": 0.79,
"learning_rate": 0.00012151655119322556,
"loss": 0.0,
"step": 2050
},
{
"epoch": 0.79,
"learning_rate": 0.00012132409545804466,
"loss": 0.0,
"step": 2055
},
{
"epoch": 0.79,
"learning_rate": 0.00012113163972286374,
"loss": 0.0,
"step": 2060
},
{
"epoch": 0.79,
"learning_rate": 0.00012093918398768285,
"loss": 0.0,
"step": 2065
},
{
"epoch": 0.8,
"learning_rate": 0.00012074672825250193,
"loss": 0.0,
"step": 2070
},
{
"epoch": 0.8,
"learning_rate": 0.00012055427251732103,
"loss": 0.0,
"step": 2075
},
{
"epoch": 0.8,
"learning_rate": 0.00012036181678214011,
"loss": 0.0,
"step": 2080
},
{
"epoch": 0.8,
"learning_rate": 0.0001201693610469592,
"loss": 0.0,
"step": 2085
},
{
"epoch": 0.8,
"learning_rate": 0.00011997690531177829,
"loss": 0.0,
"step": 2090
},
{
"epoch": 0.8,
"learning_rate": 0.00011978444957659738,
"loss": 0.0,
"step": 2095
},
{
"epoch": 0.81,
"learning_rate": 0.00011959199384141649,
"loss": 0.0,
"step": 2100
},
{
"epoch": 0.81,
"learning_rate": 0.00011939953810623557,
"loss": 0.0,
"step": 2105
},
{
"epoch": 0.81,
"learning_rate": 0.00011920708237105466,
"loss": 0.0,
"step": 2110
},
{
"epoch": 0.81,
"learning_rate": 0.00011901462663587375,
"loss": 0.0,
"step": 2115
},
{
"epoch": 0.81,
"learning_rate": 0.00011882217090069286,
"loss": 0.0,
"step": 2120
},
{
"epoch": 0.82,
"learning_rate": 0.00011862971516551194,
"loss": 0.0,
"step": 2125
},
{
"epoch": 0.82,
"learning_rate": 0.00011843725943033104,
"loss": 0.0,
"step": 2130
},
{
"epoch": 0.82,
"learning_rate": 0.00011824480369515012,
"loss": 0.0,
"step": 2135
},
{
"epoch": 0.82,
"learning_rate": 0.00011805234795996923,
"loss": 0.0,
"step": 2140
},
{
"epoch": 0.82,
"learning_rate": 0.0001178598922247883,
"loss": 0.0,
"step": 2145
},
{
"epoch": 0.83,
"learning_rate": 0.00011766743648960738,
"loss": 0.0,
"step": 2150
},
{
"epoch": 0.83,
"learning_rate": 0.00011747498075442649,
"loss": 0.0,
"step": 2155
},
{
"epoch": 0.83,
"learning_rate": 0.00011728252501924557,
"loss": 0.0,
"step": 2160
},
{
"epoch": 0.83,
"learning_rate": 0.00011709006928406467,
"loss": 0.0,
"step": 2165
},
{
"epoch": 0.83,
"learning_rate": 0.00011689761354888375,
"loss": 0.0,
"step": 2170
},
{
"epoch": 0.84,
"learning_rate": 0.00011670515781370286,
"loss": 0.0,
"step": 2175
},
{
"epoch": 0.84,
"learning_rate": 0.00011651270207852195,
"loss": 0.0,
"step": 2180
},
{
"epoch": 0.84,
"learning_rate": 0.00011632024634334104,
"loss": 0.0,
"step": 2185
},
{
"epoch": 0.84,
"learning_rate": 0.00011612779060816012,
"loss": 0.0,
"step": 2190
},
{
"epoch": 0.84,
"learning_rate": 0.00011593533487297923,
"loss": 0.0,
"step": 2195
},
{
"epoch": 0.85,
"learning_rate": 0.00011574287913779832,
"loss": 0.0,
"step": 2200
},
{
"epoch": 0.85,
"learning_rate": 0.00011555042340261741,
"loss": 0.0,
"step": 2205
},
{
"epoch": 0.85,
"learning_rate": 0.0001153579676674365,
"loss": 0.0,
"step": 2210
},
{
"epoch": 0.85,
"learning_rate": 0.00011516551193225558,
"loss": 0.0,
"step": 2215
},
{
"epoch": 0.85,
"learning_rate": 0.00011497305619707468,
"loss": 0.0,
"step": 2220
},
{
"epoch": 0.85,
"learning_rate": 0.00011478060046189376,
"loss": 0.0,
"step": 2225
},
{
"epoch": 0.86,
"learning_rate": 0.00011458814472671287,
"loss": 0.0,
"step": 2230
},
{
"epoch": 0.86,
"learning_rate": 0.00011439568899153195,
"loss": 0.0,
"step": 2235
},
{
"epoch": 0.86,
"learning_rate": 0.00011420323325635105,
"loss": 0.0,
"step": 2240
},
{
"epoch": 0.86,
"learning_rate": 0.00011401077752117013,
"loss": 0.0,
"step": 2245
},
{
"epoch": 0.86,
"learning_rate": 0.00011381832178598924,
"loss": 0.0,
"step": 2250
},
{
"epoch": 0.87,
"learning_rate": 0.00011362586605080832,
"loss": 0.0,
"step": 2255
},
{
"epoch": 0.87,
"learning_rate": 0.00011343341031562742,
"loss": 0.0,
"step": 2260
},
{
"epoch": 0.87,
"learning_rate": 0.0001132409545804465,
"loss": 0.0,
"step": 2265
},
{
"epoch": 0.87,
"learning_rate": 0.00011304849884526561,
"loss": 0.0,
"step": 2270
},
{
"epoch": 0.87,
"learning_rate": 0.00011285604311008468,
"loss": 0.0,
"step": 2275
},
{
"epoch": 0.88,
"learning_rate": 0.00011266358737490376,
"loss": 0.0,
"step": 2280
},
{
"epoch": 0.88,
"learning_rate": 0.00011247113163972287,
"loss": 0.0,
"step": 2285
},
{
"epoch": 0.88,
"learning_rate": 0.00011227867590454196,
"loss": 0.0,
"step": 2290
},
{
"epoch": 0.88,
"learning_rate": 0.00011208622016936105,
"loss": 0.0,
"step": 2295
},
{
"epoch": 0.88,
"learning_rate": 0.00011189376443418013,
"loss": 0.0,
"step": 2300
},
{
"epoch": 0.89,
"learning_rate": 0.00011170130869899924,
"loss": 0.0,
"step": 2305
},
{
"epoch": 0.89,
"learning_rate": 0.00011150885296381833,
"loss": 0.0,
"step": 2310
},
{
"epoch": 0.89,
"learning_rate": 0.00011131639722863742,
"loss": 0.0,
"step": 2315
},
{
"epoch": 0.89,
"learning_rate": 0.0001111239414934565,
"loss": 0.0,
"step": 2320
},
{
"epoch": 0.89,
"learning_rate": 0.00011093148575827562,
"loss": 0.0,
"step": 2325
},
{
"epoch": 0.9,
"learning_rate": 0.00011073903002309469,
"loss": 0.0,
"step": 2330
},
{
"epoch": 0.9,
"learning_rate": 0.0001105465742879138,
"loss": 0.0,
"step": 2335
},
{
"epoch": 0.9,
"learning_rate": 0.00011035411855273288,
"loss": 0.0,
"step": 2340
},
{
"epoch": 0.9,
"learning_rate": 0.00011016166281755196,
"loss": 0.0,
"step": 2345
},
{
"epoch": 0.9,
"learning_rate": 0.00010996920708237106,
"loss": 0.0,
"step": 2350
},
{
"epoch": 0.9,
"learning_rate": 0.00010977675134719014,
"loss": 0.0,
"step": 2355
},
{
"epoch": 0.91,
"learning_rate": 0.00010958429561200925,
"loss": 0.0,
"step": 2360
},
{
"epoch": 0.91,
"learning_rate": 0.00010939183987682833,
"loss": 0.0,
"step": 2365
},
{
"epoch": 0.91,
"learning_rate": 0.00010919938414164743,
"loss": 0.0,
"step": 2370
},
{
"epoch": 0.91,
"learning_rate": 0.00010900692840646651,
"loss": 0.0,
"step": 2375
},
{
"epoch": 0.91,
"learning_rate": 0.00010881447267128562,
"loss": 0.0,
"step": 2380
},
{
"epoch": 0.92,
"learning_rate": 0.0001086220169361047,
"loss": 0.0,
"step": 2385
},
{
"epoch": 0.92,
"learning_rate": 0.0001084295612009238,
"loss": 0.0,
"step": 2390
},
{
"epoch": 0.92,
"learning_rate": 0.00010823710546574288,
"loss": 0.0,
"step": 2395
},
{
"epoch": 0.92,
"learning_rate": 0.00010804464973056199,
"loss": 0.0,
"step": 2400
},
{
"epoch": 0.92,
"learning_rate": 0.00010785219399538106,
"loss": 0.0,
"step": 2405
},
{
"epoch": 0.93,
"learning_rate": 0.00010765973826020015,
"loss": 0.0,
"step": 2410
},
{
"epoch": 0.93,
"learning_rate": 0.00010746728252501925,
"loss": 0.0,
"step": 2415
},
{
"epoch": 0.93,
"learning_rate": 0.00010727482678983834,
"loss": 0.0,
"step": 2420
},
{
"epoch": 0.93,
"learning_rate": 0.00010708237105465743,
"loss": 0.0,
"step": 2425
},
{
"epoch": 0.93,
"learning_rate": 0.00010688991531947652,
"loss": 0.0,
"step": 2430
},
{
"epoch": 0.94,
"learning_rate": 0.00010669745958429563,
"loss": 0.0,
"step": 2435
},
{
"epoch": 0.94,
"learning_rate": 0.00010650500384911471,
"loss": 0.0,
"step": 2440
},
{
"epoch": 0.94,
"learning_rate": 0.0001063125481139338,
"loss": 0.0,
"step": 2445
},
{
"epoch": 0.94,
"learning_rate": 0.00010612009237875289,
"loss": 0.0,
"step": 2450
},
{
"epoch": 0.94,
"learning_rate": 0.000105927636643572,
"loss": 0.0,
"step": 2455
},
{
"epoch": 0.95,
"learning_rate": 0.00010573518090839107,
"loss": 0.0,
"step": 2460
},
{
"epoch": 0.95,
"learning_rate": 0.00010554272517321018,
"loss": 0.0,
"step": 2465
},
{
"epoch": 0.95,
"learning_rate": 0.00010535026943802926,
"loss": 0.0,
"step": 2470
},
{
"epoch": 0.95,
"learning_rate": 0.00010515781370284834,
"loss": 0.0,
"step": 2475
},
{
"epoch": 0.95,
"learning_rate": 0.00010496535796766744,
"loss": 0.0,
"step": 2480
},
{
"epoch": 0.95,
"learning_rate": 0.00010477290223248652,
"loss": 0.0,
"step": 2485
},
{
"epoch": 0.96,
"learning_rate": 0.00010458044649730563,
"loss": 0.0,
"step": 2490
},
{
"epoch": 0.96,
"learning_rate": 0.00010438799076212471,
"loss": 0.0,
"step": 2495
},
{
"epoch": 0.96,
"learning_rate": 0.00010419553502694381,
"loss": 0.0,
"step": 2500
},
{
"epoch": 0.96,
"learning_rate": 0.0001040030792917629,
"loss": 0.0,
"step": 2505
},
{
"epoch": 0.96,
"learning_rate": 0.000103810623556582,
"loss": 0.0,
"step": 2510
},
{
"epoch": 0.97,
"learning_rate": 0.00010361816782140109,
"loss": 0.0,
"step": 2515
},
{
"epoch": 0.97,
"learning_rate": 0.00010342571208622018,
"loss": 0.0,
"step": 2520
},
{
"epoch": 0.97,
"learning_rate": 0.00010323325635103927,
"loss": 0.0,
"step": 2525
},
{
"epoch": 0.97,
"learning_rate": 0.00010304080061585836,
"loss": 0.0,
"step": 2530
},
{
"epoch": 0.97,
"learning_rate": 0.00010284834488067744,
"loss": 0.0,
"step": 2535
},
{
"epoch": 0.98,
"learning_rate": 0.00010265588914549653,
"loss": 0.0,
"step": 2540
},
{
"epoch": 0.98,
"learning_rate": 0.00010246343341031564,
"loss": 0.0,
"step": 2545
},
{
"epoch": 0.98,
"learning_rate": 0.00010227097767513472,
"loss": 0.0,
"step": 2550
},
{
"epoch": 0.98,
"learning_rate": 0.00010207852193995382,
"loss": 0.0,
"step": 2555
},
{
"epoch": 0.98,
"learning_rate": 0.0001018860662047729,
"loss": 0.0,
"step": 2560
},
{
"epoch": 0.99,
"learning_rate": 0.00010169361046959201,
"loss": 0.0,
"step": 2565
},
{
"epoch": 0.99,
"learning_rate": 0.00010150115473441109,
"loss": 0.0,
"step": 2570
},
{
"epoch": 0.99,
"learning_rate": 0.00010130869899923019,
"loss": 0.0,
"step": 2575
},
{
"epoch": 0.99,
"learning_rate": 0.00010111624326404927,
"loss": 0.0,
"step": 2580
},
{
"epoch": 0.99,
"learning_rate": 0.00010092378752886838,
"loss": 0.0,
"step": 2585
},
{
"epoch": 1.0,
"learning_rate": 0.00010073133179368745,
"loss": 0.0,
"step": 2590
},
{
"epoch": 1.0,
"learning_rate": 0.00010053887605850656,
"loss": 0.0,
"step": 2595
},
{
"epoch": 1.0,
"learning_rate": 0.00010034642032332564,
"loss": 0.0,
"step": 2600
},
{
"epoch": 1.0,
"learning_rate": 0.00010015396458814473,
"loss": 0.0,
"step": 2605
},
{
"epoch": 1.0,
"learning_rate": 9.996150885296382e-05,
"loss": 0.0,
"step": 2610
},
{
"epoch": 1.0,
"learning_rate": 9.976905311778292e-05,
"loss": 0.0,
"step": 2615
},
{
"epoch": 1.01,
"learning_rate": 9.957659738260201e-05,
"loss": 0.0,
"step": 2620
},
{
"epoch": 1.01,
"learning_rate": 9.93841416474211e-05,
"loss": 0.0,
"step": 2625
},
{
"epoch": 1.01,
"learning_rate": 9.919168591224018e-05,
"loss": 0.0,
"step": 2630
},
{
"epoch": 1.01,
"learning_rate": 9.899923017705928e-05,
"loss": 0.0,
"step": 2635
},
{
"epoch": 1.01,
"learning_rate": 9.880677444187837e-05,
"loss": 0.0,
"step": 2640
},
{
"epoch": 1.02,
"learning_rate": 9.861431870669747e-05,
"loss": 0.0,
"step": 2645
},
{
"epoch": 1.02,
"learning_rate": 9.842186297151655e-05,
"loss": 0.0,
"step": 2650
},
{
"epoch": 1.02,
"learning_rate": 9.822940723633565e-05,
"loss": 0.0,
"step": 2655
},
{
"epoch": 1.02,
"learning_rate": 9.803695150115474e-05,
"loss": 0.0,
"step": 2660
},
{
"epoch": 1.02,
"learning_rate": 9.784449576597383e-05,
"loss": 0.0,
"step": 2665
},
{
"epoch": 1.03,
"learning_rate": 9.765204003079292e-05,
"loss": 0.0,
"step": 2670
},
{
"epoch": 1.03,
"learning_rate": 9.745958429561202e-05,
"loss": 0.0,
"step": 2675
},
{
"epoch": 1.03,
"learning_rate": 9.726712856043112e-05,
"loss": 0.0,
"step": 2680
},
{
"epoch": 1.03,
"learning_rate": 9.70746728252502e-05,
"loss": 0.0,
"step": 2685
},
{
"epoch": 1.03,
"learning_rate": 9.68822170900693e-05,
"loss": 0.0,
"step": 2690
},
{
"epoch": 1.04,
"learning_rate": 9.668976135488838e-05,
"loss": 0.0,
"step": 2695
},
{
"epoch": 1.04,
"learning_rate": 9.649730561970747e-05,
"loss": 0.0,
"step": 2700
},
{
"epoch": 1.04,
"learning_rate": 9.630484988452656e-05,
"loss": 0.0,
"step": 2705
},
{
"epoch": 1.04,
"learning_rate": 9.611239414934565e-05,
"loss": 0.0,
"step": 2710
},
{
"epoch": 1.04,
"learning_rate": 9.591993841416475e-05,
"loss": 0.0,
"step": 2715
},
{
"epoch": 1.04,
"learning_rate": 9.572748267898383e-05,
"loss": 0.0,
"step": 2720
},
{
"epoch": 1.05,
"learning_rate": 9.553502694380293e-05,
"loss": 0.0,
"step": 2725
},
{
"epoch": 1.05,
"learning_rate": 9.534257120862202e-05,
"loss": 0.0,
"step": 2730
},
{
"epoch": 1.05,
"learning_rate": 9.515011547344112e-05,
"loss": 0.0,
"step": 2735
},
{
"epoch": 1.05,
"learning_rate": 9.49576597382602e-05,
"loss": 0.0,
"step": 2740
},
{
"epoch": 1.05,
"learning_rate": 9.47652040030793e-05,
"loss": 0.0,
"step": 2745
},
{
"epoch": 1.06,
"learning_rate": 9.45727482678984e-05,
"loss": 0.0,
"step": 2750
},
{
"epoch": 1.06,
"learning_rate": 9.438029253271748e-05,
"loss": 0.0,
"step": 2755
},
{
"epoch": 1.06,
"learning_rate": 9.418783679753656e-05,
"loss": 0.0,
"step": 2760
},
{
"epoch": 1.06,
"learning_rate": 9.399538106235566e-05,
"loss": 0.0,
"step": 2765
},
{
"epoch": 1.06,
"learning_rate": 9.380292532717475e-05,
"loss": 0.0,
"step": 2770
},
{
"epoch": 1.07,
"learning_rate": 9.361046959199384e-05,
"loss": 0.0,
"step": 2775
},
{
"epoch": 1.07,
"learning_rate": 9.341801385681293e-05,
"loss": 0.0,
"step": 2780
},
{
"epoch": 1.07,
"learning_rate": 9.322555812163203e-05,
"loss": 0.0,
"step": 2785
},
{
"epoch": 1.07,
"learning_rate": 9.303310238645113e-05,
"loss": 0.0,
"step": 2790
},
{
"epoch": 1.07,
"learning_rate": 9.284064665127021e-05,
"loss": 0.0,
"step": 2795
},
{
"epoch": 1.08,
"learning_rate": 9.26481909160893e-05,
"loss": 0.0,
"step": 2800
},
{
"epoch": 1.08,
"learning_rate": 9.24557351809084e-05,
"loss": 0.0,
"step": 2805
},
{
"epoch": 1.08,
"learning_rate": 9.226327944572748e-05,
"loss": 0.0,
"step": 2810
},
{
"epoch": 1.08,
"learning_rate": 9.207082371054658e-05,
"loss": 0.0,
"step": 2815
},
{
"epoch": 1.08,
"learning_rate": 9.187836797536568e-05,
"loss": 0.0,
"step": 2820
},
{
"epoch": 1.09,
"learning_rate": 9.168591224018476e-05,
"loss": 0.0,
"step": 2825
},
{
"epoch": 1.09,
"learning_rate": 9.149345650500386e-05,
"loss": 0.0,
"step": 2830
},
{
"epoch": 1.09,
"learning_rate": 9.130100076982294e-05,
"loss": 0.0,
"step": 2835
},
{
"epoch": 1.09,
"learning_rate": 9.110854503464203e-05,
"loss": 0.0,
"step": 2840
},
{
"epoch": 1.09,
"learning_rate": 9.091608929946113e-05,
"loss": 0.0,
"step": 2845
},
{
"epoch": 1.09,
"learning_rate": 9.072363356428021e-05,
"loss": 0.0,
"step": 2850
},
{
"epoch": 1.1,
"learning_rate": 9.053117782909931e-05,
"loss": 0.0,
"step": 2855
},
{
"epoch": 1.1,
"learning_rate": 9.03387220939184e-05,
"loss": 0.0,
"step": 2860
},
{
"epoch": 1.1,
"learning_rate": 9.01462663587375e-05,
"loss": 0.0,
"step": 2865
},
{
"epoch": 1.1,
"learning_rate": 8.995381062355659e-05,
"loss": 0.0,
"step": 2870
},
{
"epoch": 1.1,
"learning_rate": 8.976135488837568e-05,
"loss": 0.0,
"step": 2875
},
{
"epoch": 1.11,
"learning_rate": 8.956889915319478e-05,
"loss": 0.0,
"step": 2880
},
{
"epoch": 1.11,
"learning_rate": 8.937644341801386e-05,
"loss": 0.0,
"step": 2885
},
{
"epoch": 1.11,
"learning_rate": 8.918398768283294e-05,
"loss": 0.0,
"step": 2890
},
{
"epoch": 1.11,
"learning_rate": 8.899153194765204e-05,
"loss": 0.0,
"step": 2895
},
{
"epoch": 1.11,
"learning_rate": 8.879907621247114e-05,
"loss": 0.0,
"step": 2900
},
{
"epoch": 1.12,
"learning_rate": 8.860662047729022e-05,
"loss": 0.0,
"step": 2905
},
{
"epoch": 1.12,
"learning_rate": 8.841416474210932e-05,
"loss": 0.0,
"step": 2910
},
{
"epoch": 1.12,
"learning_rate": 8.822170900692841e-05,
"loss": 0.0,
"step": 2915
},
{
"epoch": 1.12,
"learning_rate": 8.802925327174751e-05,
"loss": 0.0,
"step": 2920
},
{
"epoch": 1.12,
"learning_rate": 8.783679753656659e-05,
"loss": 0.0,
"step": 2925
},
{
"epoch": 1.13,
"learning_rate": 8.764434180138569e-05,
"loss": 0.0,
"step": 2930
},
{
"epoch": 1.13,
"learning_rate": 8.745188606620478e-05,
"loss": 0.0,
"step": 2935
},
{
"epoch": 1.13,
"learning_rate": 8.725943033102387e-05,
"loss": 0.0,
"step": 2940
},
{
"epoch": 1.13,
"learning_rate": 8.706697459584296e-05,
"loss": 0.0,
"step": 2945
},
{
"epoch": 1.13,
"learning_rate": 8.687451886066205e-05,
"loss": 0.0,
"step": 2950
},
{
"epoch": 1.14,
"learning_rate": 8.668206312548114e-05,
"loss": 0.0,
"step": 2955
},
{
"epoch": 1.14,
"learning_rate": 8.648960739030024e-05,
"loss": 0.0,
"step": 2960
},
{
"epoch": 1.14,
"learning_rate": 8.629715165511932e-05,
"loss": 0.0,
"step": 2965
},
{
"epoch": 1.14,
"learning_rate": 8.610469591993842e-05,
"loss": 0.0,
"step": 2970
},
{
"epoch": 1.14,
"learning_rate": 8.591224018475751e-05,
"loss": 0.0,
"step": 2975
},
{
"epoch": 1.14,
"learning_rate": 8.57197844495766e-05,
"loss": 0.0,
"step": 2980
},
{
"epoch": 1.15,
"learning_rate": 8.552732871439569e-05,
"loss": 0.0,
"step": 2985
},
{
"epoch": 1.15,
"learning_rate": 8.533487297921479e-05,
"loss": 0.0,
"step": 2990
},
{
"epoch": 1.15,
"learning_rate": 8.514241724403388e-05,
"loss": 0.0,
"step": 2995
},
{
"epoch": 1.15,
"learning_rate": 8.494996150885297e-05,
"loss": 0.0,
"step": 3000
},
{
"epoch": 1.15,
"learning_rate": 8.475750577367206e-05,
"loss": 0.0,
"step": 3005
},
{
"epoch": 1.16,
"learning_rate": 8.456505003849116e-05,
"loss": 0.0,
"step": 3010
},
{
"epoch": 1.16,
"learning_rate": 8.437259430331024e-05,
"loss": 0.0,
"step": 3015
},
{
"epoch": 1.16,
"learning_rate": 8.418013856812933e-05,
"loss": 0.0,
"step": 3020
},
{
"epoch": 1.16,
"learning_rate": 8.398768283294842e-05,
"loss": 0.0,
"step": 3025
},
{
"epoch": 1.16,
"learning_rate": 8.379522709776752e-05,
"loss": 0.0,
"step": 3030
},
{
"epoch": 1.17,
"learning_rate": 8.36027713625866e-05,
"loss": 0.0,
"step": 3035
},
{
"epoch": 1.17,
"learning_rate": 8.34103156274057e-05,
"loss": 0.0,
"step": 3040
},
{
"epoch": 1.17,
"learning_rate": 8.32178598922248e-05,
"loss": 0.0,
"step": 3045
},
{
"epoch": 1.17,
"learning_rate": 8.302540415704389e-05,
"loss": 0.0,
"step": 3050
},
{
"epoch": 1.17,
"learning_rate": 8.283294842186297e-05,
"loss": 0.0,
"step": 3055
},
{
"epoch": 1.18,
"learning_rate": 8.264049268668207e-05,
"loss": 0.0,
"step": 3060
},
{
"epoch": 1.18,
"learning_rate": 8.244803695150117e-05,
"loss": 0.0,
"step": 3065
},
{
"epoch": 1.18,
"learning_rate": 8.225558121632025e-05,
"loss": 0.0,
"step": 3070
},
{
"epoch": 1.18,
"learning_rate": 8.206312548113934e-05,
"loss": 0.0,
"step": 3075
},
{
"epoch": 1.18,
"learning_rate": 8.187066974595843e-05,
"loss": 0.0,
"step": 3080
},
{
"epoch": 1.19,
"learning_rate": 8.167821401077752e-05,
"loss": 0.0,
"step": 3085
},
{
"epoch": 1.19,
"learning_rate": 8.148575827559662e-05,
"loss": 0.0,
"step": 3090
},
{
"epoch": 1.19,
"learning_rate": 8.12933025404157e-05,
"loss": 0.0,
"step": 3095
},
{
"epoch": 1.19,
"learning_rate": 8.11008468052348e-05,
"loss": 0.0,
"step": 3100
},
{
"epoch": 1.19,
"learning_rate": 8.09083910700539e-05,
"loss": 0.0,
"step": 3105
},
{
"epoch": 1.19,
"learning_rate": 8.071593533487298e-05,
"loss": 0.0,
"step": 3110
},
{
"epoch": 1.2,
"learning_rate": 8.052347959969207e-05,
"loss": 0.0,
"step": 3115
},
{
"epoch": 1.2,
"learning_rate": 8.033102386451117e-05,
"loss": 0.0,
"step": 3120
},
{
"epoch": 1.2,
"learning_rate": 8.013856812933027e-05,
"loss": 0.0,
"step": 3125
},
{
"epoch": 1.2,
"learning_rate": 7.994611239414935e-05,
"loss": 0.0,
"step": 3130
},
{
"epoch": 1.2,
"learning_rate": 7.975365665896845e-05,
"loss": 0.0,
"step": 3135
},
{
"epoch": 1.21,
"learning_rate": 7.956120092378754e-05,
"loss": 0.0,
"step": 3140
},
{
"epoch": 1.21,
"learning_rate": 7.936874518860663e-05,
"loss": 0.0,
"step": 3145
},
{
"epoch": 1.21,
"learning_rate": 7.917628945342571e-05,
"loss": 0.0,
"step": 3150
},
{
"epoch": 1.21,
"learning_rate": 7.89838337182448e-05,
"loss": 0.0,
"step": 3155
},
{
"epoch": 1.21,
"learning_rate": 7.87913779830639e-05,
"loss": 0.0,
"step": 3160
},
{
"epoch": 1.22,
"learning_rate": 7.859892224788298e-05,
"loss": 0.0,
"step": 3165
},
{
"epoch": 1.22,
"learning_rate": 7.840646651270208e-05,
"loss": 0.0,
"step": 3170
},
{
"epoch": 1.22,
"learning_rate": 7.821401077752118e-05,
"loss": 0.0,
"step": 3175
},
{
"epoch": 1.22,
"learning_rate": 7.802155504234027e-05,
"loss": 0.0,
"step": 3180
},
{
"epoch": 1.22,
"learning_rate": 7.782909930715935e-05,
"loss": 0.0,
"step": 3185
},
{
"epoch": 1.23,
"learning_rate": 7.763664357197845e-05,
"loss": 0.0,
"step": 3190
},
{
"epoch": 1.23,
"learning_rate": 7.744418783679755e-05,
"loss": 0.0,
"step": 3195
},
{
"epoch": 1.23,
"learning_rate": 7.725173210161663e-05,
"loss": 0.0,
"step": 3200
},
{
"epoch": 1.23,
"learning_rate": 7.705927636643573e-05,
"loss": 0.0,
"step": 3205
},
{
"epoch": 1.23,
"learning_rate": 7.686682063125481e-05,
"loss": 0.0,
"step": 3210
},
{
"epoch": 1.24,
"learning_rate": 7.66743648960739e-05,
"loss": 0.0,
"step": 3215
},
{
"epoch": 1.24,
"learning_rate": 7.648190916089299e-05,
"loss": 0.0,
"step": 3220
},
{
"epoch": 1.24,
"learning_rate": 7.628945342571208e-05,
"loss": 0.0,
"step": 3225
},
{
"epoch": 1.24,
"learning_rate": 7.609699769053118e-05,
"loss": 0.0,
"step": 3230
},
{
"epoch": 1.24,
"learning_rate": 7.590454195535028e-05,
"loss": 0.0,
"step": 3235
},
{
"epoch": 1.24,
"learning_rate": 7.571208622016936e-05,
"loss": 0.0,
"step": 3240
},
{
"epoch": 1.25,
"learning_rate": 7.551963048498846e-05,
"loss": 0.0,
"step": 3245
},
{
"epoch": 1.25,
"learning_rate": 7.532717474980755e-05,
"loss": 0.0,
"step": 3250
},
{
"epoch": 1.25,
"learning_rate": 7.513471901462664e-05,
"loss": 0.0,
"step": 3255
},
{
"epoch": 1.25,
"learning_rate": 7.494226327944573e-05,
"loss": 0.0,
"step": 3260
},
{
"epoch": 1.25,
"learning_rate": 7.474980754426483e-05,
"loss": 0.0,
"step": 3265
},
{
"epoch": 1.26,
"learning_rate": 7.455735180908392e-05,
"loss": 0.0,
"step": 3270
},
{
"epoch": 1.26,
"learning_rate": 7.436489607390301e-05,
"loss": 0.0,
"step": 3275
},
{
"epoch": 1.26,
"learning_rate": 7.417244033872209e-05,
"loss": 0.0,
"step": 3280
},
{
"epoch": 1.26,
"learning_rate": 7.397998460354119e-05,
"loss": 0.0,
"step": 3285
},
{
"epoch": 1.26,
"learning_rate": 7.378752886836028e-05,
"loss": 0.0,
"step": 3290
},
{
"epoch": 1.27,
"learning_rate": 7.359507313317937e-05,
"loss": 0.0,
"step": 3295
},
{
"epoch": 1.27,
"learning_rate": 7.340261739799846e-05,
"loss": 0.0,
"step": 3300
},
{
"epoch": 1.27,
"learning_rate": 7.321016166281756e-05,
"loss": 0.0,
"step": 3305
},
{
"epoch": 1.27,
"learning_rate": 7.301770592763665e-05,
"loss": 0.0,
"step": 3310
},
{
"epoch": 1.27,
"learning_rate": 7.282525019245574e-05,
"loss": 0.0,
"step": 3315
},
{
"epoch": 1.28,
"learning_rate": 7.263279445727483e-05,
"loss": 0.0,
"step": 3320
},
{
"epoch": 1.28,
"learning_rate": 7.244033872209393e-05,
"loss": 0.0,
"step": 3325
},
{
"epoch": 1.28,
"learning_rate": 7.224788298691301e-05,
"loss": 0.0,
"step": 3330
},
{
"epoch": 1.28,
"learning_rate": 7.205542725173211e-05,
"loss": 0.0,
"step": 3335
},
{
"epoch": 1.28,
"learning_rate": 7.186297151655119e-05,
"loss": 0.0,
"step": 3340
},
{
"epoch": 1.29,
"learning_rate": 7.167051578137029e-05,
"loss": 0.0,
"step": 3345
},
{
"epoch": 1.29,
"learning_rate": 7.147806004618937e-05,
"loss": 0.0,
"step": 3350
},
{
"epoch": 1.29,
"learning_rate": 7.128560431100847e-05,
"loss": 0.0,
"step": 3355
},
{
"epoch": 1.29,
"learning_rate": 7.109314857582756e-05,
"loss": 0.0,
"step": 3360
},
{
"epoch": 1.29,
"learning_rate": 7.090069284064666e-05,
"loss": 0.0,
"step": 3365
},
{
"epoch": 1.29,
"learning_rate": 7.070823710546574e-05,
"loss": 0.0,
"step": 3370
},
{
"epoch": 1.3,
"learning_rate": 7.051578137028484e-05,
"loss": 0.0,
"step": 3375
},
{
"epoch": 1.3,
"learning_rate": 7.032332563510393e-05,
"loss": 0.0,
"step": 3380
},
{
"epoch": 1.3,
"learning_rate": 7.013086989992302e-05,
"loss": 0.0,
"step": 3385
},
{
"epoch": 1.3,
"learning_rate": 6.993841416474211e-05,
"loss": 0.0,
"step": 3390
},
{
"epoch": 1.3,
"learning_rate": 6.974595842956121e-05,
"loss": 0.0,
"step": 3395
},
{
"epoch": 1.31,
"learning_rate": 6.955350269438029e-05,
"loss": 0.0,
"step": 3400
},
{
"epoch": 1.31,
"learning_rate": 6.936104695919939e-05,
"loss": 0.0,
"step": 3405
},
{
"epoch": 1.31,
"learning_rate": 6.916859122401847e-05,
"loss": 0.0,
"step": 3410
},
{
"epoch": 1.31,
"learning_rate": 6.897613548883757e-05,
"loss": 0.0,
"step": 3415
},
{
"epoch": 1.31,
"learning_rate": 6.878367975365666e-05,
"loss": 0.0,
"step": 3420
},
{
"epoch": 1.32,
"learning_rate": 6.859122401847575e-05,
"loss": 0.0,
"step": 3425
},
{
"epoch": 1.32,
"learning_rate": 6.839876828329484e-05,
"loss": 0.0,
"step": 3430
},
{
"epoch": 1.32,
"learning_rate": 6.820631254811394e-05,
"loss": 0.0,
"step": 3435
},
{
"epoch": 1.32,
"learning_rate": 6.801385681293304e-05,
"loss": 0.0,
"step": 3440
},
{
"epoch": 1.32,
"learning_rate": 6.782140107775212e-05,
"loss": 0.0,
"step": 3445
},
{
"epoch": 1.33,
"learning_rate": 6.762894534257122e-05,
"loss": 0.0,
"step": 3450
},
{
"epoch": 1.33,
"learning_rate": 6.743648960739031e-05,
"loss": 0.0,
"step": 3455
},
{
"epoch": 1.33,
"learning_rate": 6.72440338722094e-05,
"loss": 0.0,
"step": 3460
},
{
"epoch": 1.33,
"learning_rate": 6.705157813702848e-05,
"loss": 0.0,
"step": 3465
},
{
"epoch": 1.33,
"learning_rate": 6.685912240184757e-05,
"loss": 0.0,
"step": 3470
},
{
"epoch": 1.33,
"learning_rate": 6.666666666666667e-05,
"loss": 0.0,
"step": 3475
},
{
"epoch": 1.34,
"learning_rate": 6.647421093148575e-05,
"loss": 0.0,
"step": 3480
},
{
"epoch": 1.34,
"learning_rate": 6.628175519630485e-05,
"loss": 0.0,
"step": 3485
},
{
"epoch": 1.34,
"learning_rate": 6.608929946112395e-05,
"loss": 0.0,
"step": 3490
},
{
"epoch": 1.34,
"learning_rate": 6.589684372594304e-05,
"loss": 0.0,
"step": 3495
},
{
"epoch": 1.34,
"learning_rate": 6.570438799076212e-05,
"loss": 0.0,
"step": 3500
},
{
"epoch": 1.35,
"learning_rate": 6.551193225558122e-05,
"loss": 0.0,
"step": 3505
},
{
"epoch": 1.35,
"learning_rate": 6.531947652040032e-05,
"loss": 0.0,
"step": 3510
},
{
"epoch": 1.35,
"learning_rate": 6.51270207852194e-05,
"loss": 0.0,
"step": 3515
},
{
"epoch": 1.35,
"learning_rate": 6.49345650500385e-05,
"loss": 0.0,
"step": 3520
},
{
"epoch": 1.35,
"learning_rate": 6.474210931485759e-05,
"loss": 0.0,
"step": 3525
},
{
"epoch": 1.36,
"learning_rate": 6.454965357967668e-05,
"loss": 0.0,
"step": 3530
},
{
"epoch": 1.36,
"learning_rate": 6.435719784449577e-05,
"loss": 0.0,
"step": 3535
},
{
"epoch": 1.36,
"learning_rate": 6.416474210931485e-05,
"loss": 0.0,
"step": 3540
},
{
"epoch": 1.36,
"learning_rate": 6.397228637413395e-05,
"loss": 0.0,
"step": 3545
},
{
"epoch": 1.36,
"learning_rate": 6.377983063895305e-05,
"loss": 0.0,
"step": 3550
},
{
"epoch": 1.37,
"learning_rate": 6.358737490377213e-05,
"loss": 0.0,
"step": 3555
},
{
"epoch": 1.37,
"learning_rate": 6.339491916859123e-05,
"loss": 0.0,
"step": 3560
},
{
"epoch": 1.37,
"learning_rate": 6.320246343341032e-05,
"loss": 0.0,
"step": 3565
},
{
"epoch": 1.37,
"learning_rate": 6.301000769822942e-05,
"loss": 0.0,
"step": 3570
},
{
"epoch": 1.37,
"learning_rate": 6.28175519630485e-05,
"loss": 0.0,
"step": 3575
},
{
"epoch": 1.38,
"learning_rate": 6.26250962278676e-05,
"loss": 0.0,
"step": 3580
},
{
"epoch": 1.38,
"learning_rate": 6.24326404926867e-05,
"loss": 0.0,
"step": 3585
},
{
"epoch": 1.38,
"learning_rate": 6.224018475750578e-05,
"loss": 0.0,
"step": 3590
},
{
"epoch": 1.38,
"learning_rate": 6.204772902232486e-05,
"loss": 0.0,
"step": 3595
},
{
"epoch": 1.38,
"learning_rate": 6.185527328714396e-05,
"loss": 0.0,
"step": 3600
},
{
"epoch": 1.38,
"learning_rate": 6.166281755196305e-05,
"loss": 0.0,
"step": 3605
},
{
"epoch": 1.39,
"learning_rate": 6.147036181678213e-05,
"loss": 0.0,
"step": 3610
},
{
"epoch": 1.39,
"learning_rate": 6.127790608160123e-05,
"loss": 0.0,
"step": 3615
},
{
"epoch": 1.39,
"learning_rate": 6.108545034642033e-05,
"loss": 0.0,
"step": 3620
},
{
"epoch": 1.39,
"learning_rate": 6.089299461123942e-05,
"loss": 0.0,
"step": 3625
},
{
"epoch": 1.39,
"learning_rate": 6.070053887605851e-05,
"loss": 0.0,
"step": 3630
},
{
"epoch": 1.4,
"learning_rate": 6.05080831408776e-05,
"loss": 0.0,
"step": 3635
},
{
"epoch": 1.4,
"learning_rate": 6.031562740569669e-05,
"loss": 0.0,
"step": 3640
},
{
"epoch": 1.4,
"learning_rate": 6.012317167051579e-05,
"loss": 0.0,
"step": 3645
},
{
"epoch": 1.4,
"learning_rate": 5.993071593533488e-05,
"loss": 0.0,
"step": 3650
},
{
"epoch": 1.4,
"learning_rate": 5.9738260200153974e-05,
"loss": 0.0,
"step": 3655
},
{
"epoch": 1.41,
"learning_rate": 5.954580446497306e-05,
"loss": 0.0,
"step": 3660
},
{
"epoch": 1.41,
"learning_rate": 5.935334872979215e-05,
"loss": 0.0,
"step": 3665
},
{
"epoch": 1.41,
"learning_rate": 5.9160892994611236e-05,
"loss": 0.0,
"step": 3670
},
{
"epoch": 1.41,
"learning_rate": 5.896843725943033e-05,
"loss": 0.0,
"step": 3675
},
{
"epoch": 1.41,
"learning_rate": 5.877598152424942e-05,
"loss": 0.0,
"step": 3680
},
{
"epoch": 1.42,
"learning_rate": 5.858352578906852e-05,
"loss": 0.0,
"step": 3685
},
{
"epoch": 1.42,
"learning_rate": 5.839107005388761e-05,
"loss": 0.0,
"step": 3690
},
{
"epoch": 1.42,
"learning_rate": 5.8198614318706704e-05,
"loss": 0.0,
"step": 3695
},
{
"epoch": 1.42,
"learning_rate": 5.8006158583525794e-05,
"loss": 0.0,
"step": 3700
},
{
"epoch": 1.42,
"learning_rate": 5.781370284834488e-05,
"loss": 0.0,
"step": 3705
},
{
"epoch": 1.43,
"learning_rate": 5.762124711316398e-05,
"loss": 0.0,
"step": 3710
},
{
"epoch": 1.43,
"learning_rate": 5.742879137798307e-05,
"loss": 0.0,
"step": 3715
},
{
"epoch": 1.43,
"learning_rate": 5.7236335642802165e-05,
"loss": 0.0,
"step": 3720
},
{
"epoch": 1.43,
"learning_rate": 5.704387990762125e-05,
"loss": 0.0,
"step": 3725
},
{
"epoch": 1.43,
"learning_rate": 5.685142417244034e-05,
"loss": 0.0,
"step": 3730
},
{
"epoch": 1.43,
"learning_rate": 5.665896843725943e-05,
"loss": 0.0,
"step": 3735
},
{
"epoch": 1.44,
"learning_rate": 5.6466512702078524e-05,
"loss": 0.0,
"step": 3740
},
{
"epoch": 1.44,
"learning_rate": 5.627405696689761e-05,
"loss": 0.0,
"step": 3745
},
{
"epoch": 1.44,
"learning_rate": 5.608160123171671e-05,
"loss": 0.0,
"step": 3750
}
],
"max_steps": 5206,
"num_train_epochs": 2,
"total_flos": 57215852847104.0,
"trial_name": null,
"trial_params": null
}