diff --git "a/checkpoint-73250/trainer_state.json" "b/checkpoint-73250/trainer_state.json"
deleted file mode 100644--- "a/checkpoint-73250/trainer_state.json"
+++ /dev/null
@@ -1,43966 +0,0 @@
-{
-  "best_metric": null,
-  "best_model_checkpoint": null,
-  "epoch": 2.994746417547374,
-  "global_step": 73250,
-  "is_hyper_param_search": false,
-  "is_local_process_zero": true,
-  "is_world_process_zero": true,
-  "log_history": [
-    {
-      "epoch": 0.0,
-      "learning_rate": 3.2000000000000005e-05,
-      "loss": 2.1894,
-      "step": 10
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 7.2e-05,
-      "loss": 1.9199,
-      "step": 20
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00011200000000000001,
-      "loss": 1.7689,
-      "step": 30
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.000152,
-      "loss": 1.178,
-      "step": 40
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.000192,
-      "loss": 0.9155,
-      "step": 50
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00019997817993372156,
-      "loss": 0.9795,
-      "step": 60
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.0001999509048508735,
-      "loss": 0.7469,
-      "step": 70
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00019992362976802542,
-      "loss": 0.9303,
-      "step": 80
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00019989635468517736,
-      "loss": 0.836,
-      "step": 90
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.0001998690796023293,
-      "loss": 0.9109,
-      "step": 100
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00019984180451948125,
-      "loss": 0.8246,
-      "step": 110
-    },
-    {
-      "epoch": 0.0,
-      "learning_rate": 0.00019981452943663316,
-      "loss": 0.869,
-      "step": 120
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019978725435378512,
-      "loss": 0.7522,
-      "step": 130
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019975997927093703,
-      "loss": 0.8445,
-      "step": 140
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.000199732704188089,
-      "loss": 0.7912,
-      "step": 150
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019970542910524092,
-      "loss": 0.8906,
-      "step": 160
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019967815402239286,
-      "loss": 0.8914,
-      "step": 170
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001996508789395448,
-      "loss": 0.8485,
-      "step": 180
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019962360385669672,
-      "loss": 0.8542,
-      "step": 190
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019959632877384866,
-      "loss": 0.7495,
-      "step": 200
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019956905369100062,
-      "loss": 0.8024,
-      "step": 210
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019954177860815253,
-      "loss": 0.8211,
-      "step": 220
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001995145035253045,
-      "loss": 0.9335,
-      "step": 230
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001994872284424564,
-      "loss": 0.8824,
-      "step": 240
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019945995335960836,
-      "loss": 0.8261,
-      "step": 250
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019943267827676026,
-      "loss": 0.8094,
-      "step": 260
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019940540319391222,
-      "loss": 0.8212,
-      "step": 270
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019937812811106416,
-      "loss": 0.8513,
-      "step": 280
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001993508530282161,
-      "loss": 0.8105,
-      "step": 290
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019932357794536803,
-      "loss": 0.795,
-      "step": 300
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019929630286251996,
-      "loss": 0.7169,
-      "step": 310
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001992690277796719,
-      "loss": 0.8297,
-      "step": 320
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019924175269682383,
-      "loss": 0.7912,
-      "step": 330
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019921447761397576,
-      "loss": 0.7396,
-      "step": 340
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.0001991872025311277,
-      "loss": 0.8041,
-      "step": 350
-    },
-    {
-      "epoch": 0.01,
-      "learning_rate": 0.00019915992744827963,
-      "loss": 0.7535,
-      "step": 360
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019913265236543156,
-      "loss": 0.7997,
-      "step": 370
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019910537728258352,
-      "loss": 0.8178,
-      "step": 380
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019907810219973543,
-      "loss": 0.8094,
-      "step": 390
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001990508271168874,
-      "loss": 0.8184,
-      "step": 400
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001990235520340393,
-      "loss": 0.7596,
-      "step": 410
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019899627695119126,
-      "loss": 0.8417,
-      "step": 420
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019896900186834317,
-      "loss": 0.7979,
-      "step": 430
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019894172678549513,
-      "loss": 0.8679,
-      "step": 440
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019891445170264706,
-      "loss": 0.9026,
-      "step": 450
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.000198887176619799,
-      "loss": 0.7829,
-      "step": 460
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019885990153695093,
-      "loss": 0.8704,
-      "step": 470
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019883262645410286,
-      "loss": 0.9009,
-      "step": 480
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001988053513712548,
-      "loss": 0.8974,
-      "step": 490
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019877807628840676,
-      "loss": 0.7383,
-      "step": 500
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019875080120555866,
-      "loss": 0.8161,
-      "step": 510
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019872352612271063,
-      "loss": 0.7545,
-      "step": 520
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019869625103986253,
-      "loss": 0.7352,
-      "step": 530
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001986689759570145,
-      "loss": 0.7991,
-      "step": 540
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019864170087416643,
-      "loss": 0.8307,
-      "step": 550
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019861442579131836,
-      "loss": 0.7038,
-      "step": 560
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001985871507084703,
-      "loss": 0.8171,
-      "step": 570
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019855987562562223,
-      "loss": 0.8189,
-      "step": 580
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019853260054277416,
-      "loss": 0.7346,
-      "step": 590
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.0001985053254599261,
-      "loss": 0.7593,
-      "step": 600
-    },
-    {
-      "epoch": 0.02,
-      "learning_rate": 0.00019847805037707803,
-      "loss": 0.7129,
-      "step": 610
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019845077529422996,
-      "loss": 0.7354,
-      "step": 620
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001984235002113819,
-      "loss": 0.8172,
-      "step": 630
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019839622512853383,
-      "loss": 0.8608,
-      "step": 640
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019836895004568577,
-      "loss": 0.718,
-      "step": 650
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001983416749628377,
-      "loss": 0.7164,
-      "step": 660
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019831439987998966,
-      "loss": 0.8195,
-      "step": 670
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019828712479714157,
-      "loss": 0.7818,
-      "step": 680
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019825984971429353,
-      "loss": 0.7303,
-      "step": 690
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019823257463144544,
-      "loss": 0.8328,
-      "step": 700
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001982052995485974,
-      "loss": 0.7938,
-      "step": 710
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019817802446574933,
-      "loss": 0.7457,
-      "step": 720
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019815074938290126,
-      "loss": 0.74,
-      "step": 730
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001981234743000532,
-      "loss": 0.7853,
-      "step": 740
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019809619921720513,
-      "loss": 0.741,
-      "step": 750
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019806892413435707,
-      "loss": 0.7168,
-      "step": 760
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.000198041649051509,
-      "loss": 0.7977,
-      "step": 770
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019801437396866093,
-      "loss": 0.6875,
-      "step": 780
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001979870988858129,
-      "loss": 0.6648,
-      "step": 790
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001979598238029648,
-      "loss": 0.6917,
-      "step": 800
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019793254872011676,
-      "loss": 0.7612,
-      "step": 810
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019790527363726867,
-      "loss": 0.7678,
-      "step": 820
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019787799855442063,
-      "loss": 0.7486,
-      "step": 830
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.00019785072347157257,
-      "loss": 0.858,
-      "step": 840
-    },
-    {
-      "epoch": 0.03,
-      "learning_rate": 0.0001978234483887245,
-      "loss": 0.7616,
-      "step": 850
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019779617330587643,
-      "loss": 0.9279,
-      "step": 860
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019776889822302837,
-      "loss": 0.8151,
-      "step": 870
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001977416231401803,
-      "loss": 0.8149,
-      "step": 880
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019771434805733223,
-      "loss": 0.8596,
-      "step": 890
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019768707297448417,
-      "loss": 0.7991,
-      "step": 900
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001976597978916361,
-      "loss": 0.8686,
-      "step": 910
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019763252280878804,
-      "loss": 0.8083,
-      "step": 920
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019760524772593997,
-      "loss": 0.7284,
-      "step": 930
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001975779726430919,
-      "loss": 0.7197,
-      "step": 940
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019755069756024384,
-      "loss": 0.7928,
-      "step": 950
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001975234224773958,
-      "loss": 0.7974,
-      "step": 960
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001974961473945477,
-      "loss": 0.7855,
-      "step": 970
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019746887231169967,
-      "loss": 0.7653,
-      "step": 980
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019744159722885157,
-      "loss": 0.7665,
-      "step": 990
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019741432214600353,
-      "loss": 0.8066,
-      "step": 1000
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019738704706315547,
-      "loss": 0.7081,
-      "step": 1010
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001973597719803074,
-      "loss": 0.7979,
-      "step": 1020
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019733249689745934,
-      "loss": 0.8332,
-      "step": 1030
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019730522181461127,
-      "loss": 0.8567,
-      "step": 1040
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001972779467317632,
-      "loss": 0.7407,
-      "step": 1050
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019725067164891517,
-      "loss": 0.8158,
-      "step": 1060
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019722339656606707,
-      "loss": 0.8064,
-      "step": 1070
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019719612148321903,
-      "loss": 0.7986,
-      "step": 1080
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.00019716884640037094,
-      "loss": 0.7648,
-      "step": 1090
-    },
-    {
-      "epoch": 0.04,
-      "learning_rate": 0.0001971415713175229,
-      "loss": 0.7436,
-      "step": 1100
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001971142962346748,
-      "loss": 0.8076,
-      "step": 1110
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019708702115182677,
-      "loss": 0.7235,
-      "step": 1120
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001970597460689787,
-      "loss": 0.7786,
-      "step": 1130
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019703247098613064,
-      "loss": 0.7485,
-      "step": 1140
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019700519590328257,
-      "loss": 0.7777,
-      "step": 1150
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001969779208204345,
-      "loss": 0.8417,
-      "step": 1160
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019695064573758644,
-      "loss": 0.7488,
-      "step": 1170
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019692337065473837,
-      "loss": 0.8449,
-      "step": 1180
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001968960955718903,
-      "loss": 0.7792,
-      "step": 1190
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019686882048904224,
-      "loss": 0.7278,
-      "step": 1200
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019684154540619417,
-      "loss": 0.739,
-      "step": 1210
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001968142703233461,
-      "loss": 0.771,
-      "step": 1220
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019678699524049807,
-      "loss": 0.7159,
-      "step": 1230
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019675972015764998,
-      "loss": 0.8045,
-      "step": 1240
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019673244507480194,
-      "loss": 0.7561,
-      "step": 1250
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019670516999195384,
-      "loss": 0.7819,
-      "step": 1260
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001966778949091058,
-      "loss": 0.7793,
-      "step": 1270
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019665061982625774,
-      "loss": 0.7774,
-      "step": 1280
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019662334474340967,
-      "loss": 0.7182,
-      "step": 1290
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001965960696605616,
-      "loss": 0.7454,
-      "step": 1300
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019656879457771354,
-      "loss": 0.7342,
-      "step": 1310
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019654151949486547,
-      "loss": 0.7678,
-      "step": 1320
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.0001965142444120174,
-      "loss": 0.6835,
-      "step": 1330
-    },
-    {
-      "epoch": 0.05,
-      "learning_rate": 0.00019648696932916934,
-      "loss": 0.7409,
-      "step": 1340
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001964596942463213,
-      "loss": 0.7728,
-      "step": 1350
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001964324191634732,
-      "loss": 0.7942,
-      "step": 1360
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019640514408062517,
-      "loss": 0.6703,
-      "step": 1370
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019637786899777708,
-      "loss": 0.8455,
-      "step": 1380
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019635059391492904,
-      "loss": 0.8712,
-      "step": 1390
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019632331883208097,
-      "loss": 0.7269,
-      "step": 1400
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001962960437492329,
-      "loss": 0.6555,
-      "step": 1410
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019626876866638484,
-      "loss": 0.7294,
-      "step": 1420
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019624149358353677,
-      "loss": 0.7668,
-      "step": 1430
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001962142185006887,
-      "loss": 0.6238,
-      "step": 1440
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019618694341784064,
-      "loss": 0.7063,
-      "step": 1450
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019615966833499258,
-      "loss": 0.8358,
-      "step": 1460
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001961323932521445,
-      "loss": 0.8186,
-      "step": 1470
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019610511816929644,
-      "loss": 0.7796,
-      "step": 1480
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019607784308644838,
-      "loss": 0.7581,
-      "step": 1490
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001960505680036003,
-      "loss": 0.8077,
-      "step": 1500
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019602329292075225,
-      "loss": 0.7916,
-      "step": 1510
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001959960178379042,
-      "loss": 0.69,
-      "step": 1520
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019596874275505611,
-      "loss": 0.7494,
-      "step": 1530
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019594146767220808,
-      "loss": 0.7366,
-      "step": 1540
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019591419258935998,
-      "loss": 0.7615,
-      "step": 1550
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019588691750651194,
-      "loss": 0.8025,
-      "step": 1560
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.00019585964242366388,
-      "loss": 0.7122,
-      "step": 1570
-    },
-    {
-      "epoch": 0.06,
-      "learning_rate": 0.0001958323673408158,
-      "loss": 0.8769,
-      "step": 1580
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019580509225796774,
-      "loss": 0.7903,
-      "step": 1590
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019577781717511968,
-      "loss": 0.7516,
-      "step": 1600
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.0001957505420922716,
-      "loss": 0.8179,
-      "step": 1610
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019572326700942357,
-      "loss": 0.7629,
-      "step": 1620
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019569599192657548,
-      "loss": 0.8284,
-      "step": 1630
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019566871684372744,
-      "loss": 0.6984,
-      "step": 1640
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019564144176087935,
-      "loss": 0.8675,
-      "step": 1650
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.0001956141666780313,
-      "loss": 0.8421,
-      "step": 1660
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019558689159518322,
-      "loss": 0.7852,
-      "step": 1670
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019555961651233518,
-      "loss": 0.782,
-      "step": 1680
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.0001955323414294871,
-      "loss": 0.8095,
-      "step": 1690
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019550506634663904,
-      "loss": 0.7419,
-      "step": 1700
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019547779126379098,
-      "loss": 0.7562,
-      "step": 1710
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.0001954505161809429,
-      "loss": 0.899,
-      "step": 1720
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019542324109809485,
-      "loss": 0.7506,
-      "step": 1730
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019539596601524678,
-      "loss": 0.7715,
-      "step": 1740
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019536869093239871,
-      "loss": 0.629,
-      "step": 1750
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019534141584955065,
-      "loss": 0.7627,
-      "step": 1760
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019531414076670258,
-      "loss": 0.7372,
-      "step": 1770
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019528686568385452,
-      "loss": 0.8072,
-      "step": 1780
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019525959060100648,
-      "loss": 0.8172,
-      "step": 1790
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019523231551815838,
-      "loss": 0.8318,
-      "step": 1800
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019520504043531035,
-      "loss": 0.7845,
-      "step": 1810
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.00019517776535246225,
-      "loss": 0.6585,
-      "step": 1820
-    },
-    {
-      "epoch": 0.07,
-      "learning_rate": 0.0001951504902696142,
-      "loss": 0.7144,
-      "step": 1830
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019512321518676612,
-      "loss": 0.7664,
-      "step": 1840
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019509594010391808,
-      "loss": 0.7786,
-      "step": 1850
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019506866502107001,
-      "loss": 0.8056,
-      "step": 1860
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019504138993822195,
-      "loss": 0.7825,
-      "step": 1870
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019501411485537388,
-      "loss": 0.7129,
-      "step": 1880
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019498683977252582,
-      "loss": 0.7286,
-      "step": 1890
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019495956468967775,
-      "loss": 0.7881,
-      "step": 1900
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.0001949322896068297,
-      "loss": 0.7765,
-      "step": 1910
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019490501452398162,
-      "loss": 0.8011,
-      "step": 1920
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019487773944113358,
-      "loss": 0.8162,
-      "step": 1930
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019485046435828549,
-      "loss": 0.8357,
-      "step": 1940
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019482318927543745,
-      "loss": 0.7396,
-      "step": 1950
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019479591419258938,
-      "loss": 0.7631,
-      "step": 1960
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019476863910974131,
-      "loss": 0.7968,
-      "step": 1970
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019474136402689325,
-      "loss": 0.7974,
-      "step": 1980
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019471408894404518,
-      "loss": 0.8044,
-      "step": 1990
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019468681386119712,
-      "loss": 0.784,
-      "step": 2000
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019465953877834905,
-      "loss": 0.8214,
-      "step": 2010
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019463226369550098,
-      "loss": 0.7617,
-      "step": 2020
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019460498861265292,
-      "loss": 0.7464,
-      "step": 2030
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019457771352980485,
-      "loss": 0.8577,
-      "step": 2040
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019455043844695679,
-      "loss": 0.7433,
-      "step": 2050
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019452316336410872,
-      "loss": 0.7728,
-      "step": 2060
-    },
-    {
-      "epoch": 0.08,
-      "learning_rate": 0.00019449588828126065,
-      "loss": 0.705,
-      "step": 2070
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019446861319841262,
-      "loss": 0.7368,
-      "step": 2080
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019444133811556452,
-      "loss": 0.8164,
-      "step": 2090
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019441406303271648,
-      "loss": 0.7692,
-      "step": 2100
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.0001943867879498684,
-      "loss": 0.7054,
-      "step": 2110
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019435951286702035,
-      "loss": 0.6927,
-      "step": 2120
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019433223778417228,
-      "loss": 0.7787,
-      "step": 2130
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019430496270132422,
-      "loss": 0.7931,
-      "step": 2140
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019427768761847615,
-      "loss": 0.7785,
-      "step": 2150
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.0001942504125356281,
-      "loss": 0.7494,
-      "step": 2160
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019422313745278002,
-      "loss": 0.728,
-      "step": 2170
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019419586236993198,
-      "loss": 0.7253,
-      "step": 2180
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.0001941685872870839,
-      "loss": 0.8201,
-      "step": 2190
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019414131220423585,
-      "loss": 0.784,
-      "step": 2200
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019411403712138776,
-      "loss": 0.7829,
-      "step": 2210
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019408676203853972,
-      "loss": 0.5782,
-      "step": 2220
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019405948695569162,
-      "loss": 0.809,
-      "step": 2230
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019403221187284358,
-      "loss": 0.8004,
-      "step": 2240
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019400493678999552,
-      "loss": 0.7749,
-      "step": 2250
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019397766170714745,
-      "loss": 0.8058,
-      "step": 2260
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.0001939503866242994,
-      "loss": 0.8014,
-      "step": 2270
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019392311154145132,
-      "loss": 0.7265,
-      "step": 2280
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019389583645860325,
-      "loss": 0.698,
-      "step": 2290
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.0001938685613757552,
-      "loss": 0.8185,
-      "step": 2300
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019384128629290712,
-      "loss": 0.7699,
-      "step": 2310
-    },
-    {
-      "epoch": 0.09,
-      "learning_rate": 0.00019381401121005906,
-      "loss": 0.7311,
-      "step": 2320
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.000193786736127211,
-      "loss": 0.7927,
-      "step": 2330
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019375946104436292,
-      "loss": 0.754,
-      "step": 2340
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019373218596151489,
-      "loss": 0.8419,
-      "step": 2350
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001937049108786668,
-      "loss": 0.7805,
-      "step": 2360
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019367763579581875,
-      "loss": 0.833,
-      "step": 2370
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019365036071297066,
-      "loss": 0.6548,
-      "step": 2380
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019362308563012262,
-      "loss": 0.7651,
-      "step": 2390
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019359581054727453,
-      "loss": 0.7923,
-      "step": 2400
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001935685354644265,
-      "loss": 0.7216,
-      "step": 2410
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019354126038157842,
-      "loss": 0.7023,
-      "step": 2420
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019351398529873036,
-      "loss": 0.7256,
-      "step": 2430
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001934867102158823,
-      "loss": 0.8266,
-      "step": 2440
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019345943513303422,
-      "loss": 0.7389,
-      "step": 2450
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019343216005018616,
-      "loss": 0.6974,
-      "step": 2460
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019340488496733812,
-      "loss": 0.7563,
-      "step": 2470
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019337760988449003,
-      "loss": 0.7301,
-      "step": 2480
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.000193350334801642,
-      "loss": 0.7601,
-      "step": 2490
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001933230597187939,
-      "loss": 0.7333,
-      "step": 2500
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019329578463594586,
-      "loss": 0.8393,
-      "step": 2510
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001932685095530978,
-      "loss": 0.707,
-      "step": 2520
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019324123447024972,
-      "loss": 0.706,
-      "step": 2530
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019321395938740166,
-      "loss": 0.7334,
-      "step": 2540
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.0001931866843045536,
-      "loss": 0.7116,
-      "step": 2550
-    },
-    {
-      "epoch": 0.1,
-      "learning_rate": 0.00019315940922170552,
-      "loss": 0.779,
-      "step": 2560
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019313213413885746,
-      "loss": 0.7977,
-      "step": 2570
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001931048590560094,
-      "loss": 0.7299,
-      "step": 2580
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019307758397316133,
-      "loss": 0.7139,
-      "step": 2590
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019305030889031326,
-      "loss": 0.7712,
-      "step": 2600
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001930230338074652,
-      "loss": 0.7013,
-      "step": 2610
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019299575872461713,
-      "loss": 0.9072,
-      "step": 2620
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019296848364176906,
-      "loss": 0.7471,
-      "step": 2630
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019294120855892102,
-      "loss": 0.8672,
-      "step": 2640
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019291393347607293,
-      "loss": 0.7076,
-      "step": 2650
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001928866583932249,
-      "loss": 0.7115,
-      "step": 2660
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001928593833103768,
-      "loss": 0.7304,
-      "step": 2670
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019283210822752876,
-      "loss": 0.7718,
-      "step": 2680
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001928048331446807,
-      "loss": 0.7878,
-      "step": 2690
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019277755806183263,
-      "loss": 0.7473,
-      "step": 2700
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019275028297898456,
-      "loss": 0.806,
-      "step": 2710
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001927230078961365,
-      "loss": 0.76,
-      "step": 2720
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019269573281328843,
-      "loss": 0.7109,
-      "step": 2730
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019266845773044036,
-      "loss": 0.7984,
-      "step": 2740
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.0001926411826475923,
-      "loss": 0.6713,
-      "step": 2750
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019261390756474426,
-      "loss": 0.6533,
-      "step": 2760
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019258663248189616,
-      "loss": 0.7444,
-      "step": 2770
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019255935739904813,
-      "loss": 0.8606,
-      "step": 2780
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019253208231620003,
-      "loss": 0.7134,
-      "step": 2790
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.000192504807233352,
-      "loss": 0.7865,
-      "step": 2800
-    },
-    {
-      "epoch": 0.11,
-      "learning_rate": 0.00019247753215050393,
-      "loss": 0.6729,
-      "step": 2810
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019245025706765586,
-      "loss": 0.8521,
-      "step": 2820
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001924229819848078,
-      "loss": 0.7446,
-      "step": 2830
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019239570690195973,
-      "loss": 0.8111,
-      "step": 2840
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019236843181911166,
-      "loss": 0.6949,
-      "step": 2850
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001923411567362636,
-      "loss": 0.7304,
-      "step": 2860
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019231388165341553,
-      "loss": 0.7117,
-      "step": 2870
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019228660657056746,
-      "loss": 0.7113,
-      "step": 2880
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001922593314877194,
-      "loss": 0.8574,
-      "step": 2890
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019223205640487133,
-      "loss": 0.8463,
-      "step": 2900
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019220478132202327,
-      "loss": 0.7655,
-      "step": 2910
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001921775062391752,
-      "loss": 0.751,
-      "step": 2920
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019215023115632716,
-      "loss": 0.7654,
-      "step": 2930
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019212295607347907,
-      "loss": 0.8262,
-      "step": 2940
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019209568099063103,
-      "loss": 0.7738,
-      "step": 2950
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019206840590778294,
-      "loss": 0.7648,
-      "step": 2960
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001920411308249349,
-      "loss": 0.672,
-      "step": 2970
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019201385574208683,
-      "loss": 0.7491,
-      "step": 2980
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019198658065923876,
-      "loss": 0.7305,
-      "step": 2990
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001919593055763907,
-      "loss": 0.7013,
-      "step": 3000
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019193203049354263,
-      "loss": 0.7445,
-      "step": 3010
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019190475541069457,
-      "loss": 0.6721,
-      "step": 3020
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019187748032784653,
-      "loss": 0.8046,
-      "step": 3030
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.00019185020524499843,
-      "loss": 0.7735,
-      "step": 3040
-    },
-    {
-      "epoch": 0.12,
-      "learning_rate": 0.0001918229301621504,
-      "loss": 0.8522,
-      "step": 3050
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001917956550793023,
-      "loss": 0.7202,
-      "step": 3060
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019176837999645426,
-      "loss": 0.7957,
-      "step": 3070
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019174110491360617,
-      "loss": 0.8553,
-      "step": 3080
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019171382983075813,
-      "loss": 0.7605,
-      "step": 3090
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019168655474791006,
-      "loss": 0.7073,
-      "step": 3100
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.000191659279665062,
-      "loss": 0.7208,
-      "step": 3110
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019163200458221393,
-      "loss": 0.791,
-      "step": 3120
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019160472949936587,
-      "loss": 0.7928,
-      "step": 3130
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001915774544165178,
-      "loss": 0.8049,
-      "step": 3140
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019155017933366973,
-      "loss": 0.8542,
-      "step": 3150
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019152290425082167,
-      "loss": 0.7502,
-      "step": 3160
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001914956291679736,
-      "loss": 0.7666,
-      "step": 3170
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019146835408512554,
-      "loss": 0.7069,
-      "step": 3180
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019144107900227747,
-      "loss": 0.7722,
-      "step": 3190
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019141380391942943,
-      "loss": 0.8119,
-      "step": 3200
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019138652883658134,
-      "loss": 0.7022,
-      "step": 3210
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001913592537537333,
-      "loss": 0.7936,
-      "step": 3220
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001913319786708852,
-      "loss": 0.7377,
-      "step": 3230
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019130470358803717,
-      "loss": 0.7365,
-      "step": 3240
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001912774285051891,
-      "loss": 0.7402,
-      "step": 3250
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019125015342234103,
-      "loss": 0.7518,
-      "step": 3260
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019122287833949297,
-      "loss": 0.6915,
-      "step": 3270
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.0001911956032566449,
-      "loss": 0.776,
-      "step": 3280
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019116832817379684,
-      "loss": 0.7617,
-      "step": 3290
-    },
-    {
-      "epoch": 0.13,
-      "learning_rate": 0.00019114105309094877,
-      "loss": 0.7469,
-      "step": 3300
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001911137780081007,
-      "loss": 0.7252,
-      "step": 3310
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019108650292525267,
-      "loss": 0.6838,
-      "step": 3320
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019105922784240457,
-      "loss": 0.7107,
-      "step": 3330
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019103195275955653,
-      "loss": 0.6875,
-      "step": 3340
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019100467767670844,
-      "loss": 0.7599,
-      "step": 3350
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001909774025938604,
-      "loss": 0.7564,
-      "step": 3360
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019095012751101233,
-      "loss": 0.7545,
-      "step": 3370
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019092285242816427,
-      "loss": 0.7294,
-      "step": 3380
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001908955773453162,
-      "loss": 0.8103,
-      "step": 3390
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019086830226246814,
-      "loss": 0.8223,
-      "step": 3400
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019084102717962007,
-      "loss": 0.6985,
-      "step": 3410
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.000190813752096772,
-      "loss": 0.7903,
-      "step": 3420
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019078647701392394,
-      "loss": 0.8292,
-      "step": 3430
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019075920193107587,
-      "loss": 0.7502,
-      "step": 3440
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001907319268482278,
-      "loss": 0.7429,
-      "step": 3450
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019070465176537974,
-      "loss": 0.8731,
-      "step": 3460
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019067737668253167,
-      "loss": 0.7065,
-      "step": 3470
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001906501015996836,
-      "loss": 0.8267,
-      "step": 3480
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019062282651683557,
-      "loss": 0.7469,
-      "step": 3490
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019059555143398748,
-      "loss": 0.6903,
-      "step": 3500
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019056827635113944,
-      "loss": 0.7084,
-      "step": 3510
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019054100126829134,
-      "loss": 0.7126,
-      "step": 3520
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.0001905137261854433,
-      "loss": 0.7514,
-      "step": 3530
-    },
-    {
-      "epoch": 0.14,
-      "learning_rate": 0.00019048645110259524,
-      "loss": 0.7133,
-      "step": 3540
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019045917601974717,
-      "loss": 0.7002,
-      "step": 3550
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001904319009368991,
-      "loss": 0.7759,
-      "step": 3560
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019040462585405104,
-      "loss": 0.662,
-      "step": 3570
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019037735077120297,
-      "loss": 0.8033,
-      "step": 3580
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019035007568835494,
-      "loss": 0.7457,
-      "step": 3590
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019032280060550684,
-      "loss": 0.8514,
-      "step": 3600
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001902955255226588,
-      "loss": 0.6947,
-      "step": 3610
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001902682504398107,
-      "loss": 0.6892,
-      "step": 3620
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019024097535696267,
-      "loss": 0.7859,
-      "step": 3630
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019021370027411458,
-      "loss": 0.7551,
-      "step": 3640
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019018642519126654,
-      "loss": 0.6992,
-      "step": 3650
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019015915010841847,
-      "loss": 0.7564,
-      "step": 3660
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001901318750255704,
-      "loss": 0.8138,
-      "step": 3670
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019010459994272234,
-      "loss": 0.7375,
-      "step": 3680
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019007732485987427,
-      "loss": 0.7743,
-      "step": 3690
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001900500497770262,
-      "loss": 0.7458,
-      "step": 3700
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00019002277469417814,
-      "loss": 0.8347,
-      "step": 3710
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018999549961133008,
-      "loss": 0.887,
-      "step": 3720
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.000189968224528482,
-      "loss": 0.796,
-      "step": 3730
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018994094944563394,
-      "loss": 0.7185,
-      "step": 3740
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018991367436278588,
-      "loss": 0.7902,
-      "step": 3750
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018988639927993784,
-      "loss": 0.7787,
-      "step": 3760
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018985912419708975,
-      "loss": 0.7981,
-      "step": 3770
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.0001898318491142417,
-      "loss": 0.7416,
-      "step": 3780
-    },
-    {
-      "epoch": 0.15,
-      "learning_rate": 0.00018980457403139361,
-      "loss": 0.7489,
-      "step": 3790
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018977729894854557,
-      "loss": 0.838,
-      "step": 3800
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018975002386569748,
-      "loss": 0.7013,
-      "step": 3810
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018972274878284944,
-      "loss": 0.7431,
-      "step": 3820
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018969547370000138,
-      "loss": 0.7944,
-      "step": 3830
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.0001896681986171533,
-      "loss": 0.7773,
-      "step": 3840
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018964092353430524,
-      "loss": 0.792,
-      "step": 3850
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018961364845145718,
-      "loss": 0.7438,
-      "step": 3860
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.0001895863733686091,
-      "loss": 0.7137,
-      "step": 3870
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018955909828576107,
-      "loss": 0.7392,
-      "step": 3880
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018953182320291298,
-      "loss": 0.8245,
-      "step": 3890
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018950454812006494,
-      "loss": 0.8365,
-      "step": 3900
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018947727303721685,
-      "loss": 0.8265,
-      "step": 3910
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.0001894499979543688,
-      "loss": 0.8601,
-      "step": 3920
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018942272287152074,
-      "loss": 0.8049,
-      "step": 3930
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018939544778867268,
-      "loss": 0.705,
-      "step": 3940
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.0001893681727058246,
-      "loss": 0.6951,
-      "step": 3950
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018934089762297654,
-      "loss": 0.8036,
-      "step": 3960
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018931362254012848,
-      "loss": 0.7563,
-      "step": 3970
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.0001892863474572804,
-      "loss": 0.7665,
-      "step": 3980
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018925907237443235,
-      "loss": 0.8214,
-      "step": 3990
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018923179729158428,
-      "loss": 0.8143,
-      "step": 4000
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018920452220873621,
-      "loss": 0.7851,
-      "step": 4010
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018917724712588815,
-      "loss": 0.7815,
-      "step": 4020
-    },
-    {
-      "epoch": 0.16,
-      "learning_rate": 0.00018914997204304008,
-      "loss": 0.703,
-      "step": 4030
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018912269696019202,
-      "loss": 0.726,
-      "step": 4040
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018909542187734398,
-      "loss": 0.8151,
-      "step": 4050
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018906814679449588,
-      "loss": 0.8327,
-      "step": 4060
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018904087171164784,
-      "loss": 0.7546,
-      "step": 4070
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018901359662879975,
-      "loss": 0.7908,
-      "step": 4080
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.0001889863215459517,
-      "loss": 0.634,
-      "step": 4090
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018895904646310365,
-      "loss": 0.7894,
-      "step": 4100
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018893177138025558,
-      "loss": 0.6834,
-      "step": 4110
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018890449629740751,
-      "loss": 0.7445,
-      "step": 4120
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018887722121455945,
-      "loss": 0.7518,
-      "step": 4130
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018884994613171138,
-      "loss": 0.8995,
-      "step": 4140
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018882267104886334,
-      "loss": 0.7294,
-      "step": 4150
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018879539596601525,
-      "loss": 0.6974,
-      "step": 4160
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.0001887681208831672,
-      "loss": 0.8339,
-      "step": 4170
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018874084580031912,
-      "loss": 0.7512,
-      "step": 4180
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018871357071747108,
-      "loss": 0.7821,
-      "step": 4190
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018868629563462299,
-      "loss": 0.7095,
-      "step": 4200
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018865902055177495,
-      "loss": 0.7181,
-      "step": 4210
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018863174546892688,
-      "loss": 0.7589,
-      "step": 4220
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018860447038607881,
-      "loss": 0.6855,
-      "step": 4230
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018857719530323075,
-      "loss": 0.7645,
-      "step": 4240
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018854992022038268,
-      "loss": 0.8303,
-      "step": 4250
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018852264513753462,
-      "loss": 0.7654,
-      "step": 4260
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018849537005468655,
-      "loss": 0.8759,
-      "step": 4270
-    },
-    {
-      "epoch": 0.17,
-      "learning_rate": 0.00018846809497183848,
-      "loss": 0.7827,
-      "step": 4280
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018844081988899042,
-      "loss": 0.7613,
-      "step": 4290
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018841354480614235,
-      "loss": 0.7079,
-      "step": 4300
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018838626972329429,
-      "loss": 0.8423,
-      "step": 4310
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018835899464044625,
-      "loss": 0.8603,
-      "step": 4320
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018833171955759815,
-      "loss": 0.7536,
-      "step": 4330
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018830444447475011,
-      "loss": 0.7622,
-      "step": 4340
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018827716939190202,
-      "loss": 0.7417,
-      "step": 4350
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018824989430905398,
-      "loss": 0.6914,
-      "step": 4360
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.0001882226192262059,
-      "loss": 0.779,
-      "step": 4370
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018819534414335785,
-      "loss": 0.7574,
-      "step": 4380
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018816806906050978,
-      "loss": 0.8239,
-      "step": 4390
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018814079397766172,
-      "loss": 0.7351,
-      "step": 4400
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018811351889481365,
-      "loss": 0.8136,
-      "step": 4410
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018808624381196559,
-      "loss": 0.7861,
-      "step": 4420
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018805896872911752,
-      "loss": 0.6225,
-      "step": 4430
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018803169364626948,
-      "loss": 0.7536,
-      "step": 4440
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.0001880044185634214,
-      "loss": 0.7312,
-      "step": 4450
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018797714348057335,
-      "loss": 0.6426,
-      "step": 4460
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018794986839772526,
-      "loss": 0.698,
-      "step": 4470
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018792259331487722,
-      "loss": 0.7697,
-      "step": 4480
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018789531823202915,
-      "loss": 0.7784,
-      "step": 4490
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018786804314918108,
-      "loss": 0.7555,
-      "step": 4500
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018784076806633302,
-      "loss": 0.7551,
-      "step": 4510
-    },
-    {
-      "epoch": 0.18,
-      "learning_rate": 0.00018781349298348495,
-      "loss": 0.6934,
-      "step": 4520
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018778621790063689,
-      "loss": 0.763,
-      "step": 4530
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018775894281778882,
-      "loss": 0.7143,
-      "step": 4540
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018773166773494075,
-      "loss": 0.7097,
-      "step": 4550
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001877043926520927,
-      "loss": 0.7461,
-      "step": 4560
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018767711756924462,
-      "loss": 0.8242,
-      "step": 4570
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018764984248639656,
-      "loss": 0.7361,
-      "step": 4580
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001876225674035485,
-      "loss": 0.7416,
-      "step": 4590
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018759529232070042,
-      "loss": 0.8099,
-      "step": 4600
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018756801723785238,
-      "loss": 0.6956,
-      "step": 4610
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001875407421550043,
-      "loss": 0.6938,
-      "step": 4620
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018751346707215625,
-      "loss": 0.7574,
-      "step": 4630
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018748619198930816,
-      "loss": 0.7287,
-      "step": 4640
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018745891690646012,
-      "loss": 0.7341,
-      "step": 4650
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018743164182361205,
-      "loss": 0.7871,
-      "step": 4660
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.000187404366740764,
-      "loss": 0.7441,
-      "step": 4670
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018737709165791592,
-      "loss": 0.8396,
-      "step": 4680
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018734981657506786,
-      "loss": 0.7488,
-      "step": 4690
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001873225414922198,
-      "loss": 0.7441,
-      "step": 4700
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018729526640937172,
-      "loss": 0.6785,
-      "step": 4710
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018726799132652366,
-      "loss": 0.7367,
-      "step": 4720
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018724071624367562,
-      "loss": 0.7063,
-      "step": 4730
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.00018721344116082753,
-      "loss": 0.7805,
-      "step": 4740
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001871861660779795,
-      "loss": 0.7065,
-      "step": 4750
-    },
-    {
-      "epoch": 0.19,
-      "learning_rate": 0.0001871588909951314,
-      "loss": 0.7059,
-      "step": 4760
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018713161591228335,
-      "loss": 0.7541,
-      "step": 4770
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001871043408294353,
-      "loss": 0.7503,
-      "step": 4780
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018707706574658722,
-      "loss": 0.8822,
-      "step": 4790
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018704979066373916,
-      "loss": 0.6715,
-      "step": 4800
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001870225155808911,
-      "loss": 0.752,
-      "step": 4810
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018699524049804302,
-      "loss": 0.8345,
-      "step": 4820
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018696796541519496,
-      "loss": 0.7023,
-      "step": 4830
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001869406903323469,
-      "loss": 0.7279,
-      "step": 4840
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018691341524949883,
-      "loss": 0.8689,
-      "step": 4850
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018688614016665076,
-      "loss": 0.8071,
-      "step": 4860
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001868588650838027,
-      "loss": 0.7425,
-      "step": 4870
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018683159000095463,
-      "loss": 0.7627,
-      "step": 4880
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018680431491810656,
-      "loss": 0.7401,
-      "step": 4890
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018677703983525852,
-      "loss": 0.6536,
-      "step": 4900
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018674976475241043,
-      "loss": 0.7052,
-      "step": 4910
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001867224896695624,
-      "loss": 0.7439,
-      "step": 4920
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001866952145867143,
-      "loss": 0.7994,
-      "step": 4930
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018666793950386626,
-      "loss": 0.7859,
-      "step": 4940
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001866406644210182,
-      "loss": 0.7379,
-      "step": 4950
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018661338933817013,
-      "loss": 0.7118,
-      "step": 4960
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018658611425532206,
-      "loss": 0.8095,
-      "step": 4970
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.000186558839172474,
-      "loss": 0.7642,
-      "step": 4980
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.00018653156408962593,
-      "loss": 0.7035,
-      "step": 4990
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001865042890067779,
-      "loss": 0.6789,
-      "step": 5000
-    },
-    {
-      "epoch": 0.2,
-      "learning_rate": 0.0001864770139239298,
-      "loss": 0.8717,
-      "step": 5010
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018644973884108176,
-      "loss": 0.7718,
-      "step": 5020
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018642246375823366,
-      "loss": 0.667,
-      "step": 5030
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018639518867538562,
-      "loss": 0.7277,
-      "step": 5040
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018636791359253753,
-      "loss": 0.7547,
-      "step": 5050
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001863406385096895,
-      "loss": 0.7676,
-      "step": 5060
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018631336342684143,
-      "loss": 0.719,
-      "step": 5070
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018628608834399336,
-      "loss": 0.6376,
-      "step": 5080
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001862588132611453,
-      "loss": 0.7389,
-      "step": 5090
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018623153817829723,
-      "loss": 0.7773,
-      "step": 5100
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018620426309544916,
-      "loss": 0.6306,
-      "step": 5110
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001861769880126011,
-      "loss": 0.7627,
-      "step": 5120
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018614971292975303,
-      "loss": 0.7685,
-      "step": 5130
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018612243784690496,
-      "loss": 0.8956,
-      "step": 5140
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001860951627640569,
-      "loss": 0.8233,
-      "step": 5150
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018606788768120883,
-      "loss": 0.7577,
-      "step": 5160
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001860406125983608,
-      "loss": 0.84,
-      "step": 5170
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001860133375155127,
-      "loss": 0.6736,
-      "step": 5180
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018598606243266466,
-      "loss": 0.7439,
-      "step": 5190
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018595878734981657,
-      "loss": 0.7088,
-      "step": 5200
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018593151226696853,
-      "loss": 0.7763,
-      "step": 5210
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018590423718412046,
-      "loss": 0.7303,
-      "step": 5220
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.0001858769621012724,
-      "loss": 0.7751,
-      "step": 5230
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018584968701842433,
-      "loss": 0.7251,
-      "step": 5240
-    },
-    {
-      "epoch": 0.21,
-      "learning_rate": 0.00018582241193557626,
-      "loss": 0.7476,
-      "step": 5250
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001857951368527282,
-      "loss": 0.7431,
-      "step": 5260
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018576786176988013,
-      "loss": 0.6888,
-      "step": 5270
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018574058668703207,
-      "loss": 0.8209,
-      "step": 5280
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018571331160418403,
-      "loss": 0.7517,
-      "step": 5290
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018568603652133593,
-      "loss": 0.697,
-      "step": 5300
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001856587614384879,
-      "loss": 0.8074,
-      "step": 5310
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001856314863556398,
-      "loss": 0.8582,
-      "step": 5320
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018560421127279176,
-      "loss": 0.6586,
-      "step": 5330
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001855769361899437,
-      "loss": 0.6915,
-      "step": 5340
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018554966110709563,
-      "loss": 0.7173,
-      "step": 5350
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018552238602424756,
-      "loss": 0.8273,
-      "step": 5360
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001854951109413995,
-      "loss": 0.6951,
-      "step": 5370
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018546783585855143,
-      "loss": 0.8317,
-      "step": 5380
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018544056077570337,
-      "loss": 0.7678,
-      "step": 5390
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001854132856928553,
-      "loss": 0.7704,
-      "step": 5400
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018538601061000723,
-      "loss": 0.6333,
-      "step": 5410
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018535873552715917,
-      "loss": 0.7398,
-      "step": 5420
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001853314604443111,
-      "loss": 0.7808,
-      "step": 5430
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018530418536146304,
-      "loss": 0.7535,
-      "step": 5440
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018527691027861497,
-      "loss": 0.8366,
-      "step": 5450
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018524963519576693,
-      "loss": 0.6556,
-      "step": 5460
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018522236011291884,
-      "loss": 0.8642,
-      "step": 5470
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001851950850300708,
-      "loss": 0.7346,
-      "step": 5480
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.0001851678099472227,
-      "loss": 0.7701,
-      "step": 5490
-    },
-    {
-      "epoch": 0.22,
-      "learning_rate": 0.00018514053486437467,
-      "loss": 0.7261,
-      "step": 5500
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001851132597815266,
-      "loss": 0.8296,
-      "step": 5510
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018508598469867853,
-      "loss": 0.6797,
-      "step": 5520
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018505870961583047,
-      "loss": 0.7323,
-      "step": 5530
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001850314345329824,
-      "loss": 0.7301,
-      "step": 5540
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018500415945013434,
-      "loss": 0.7499,
-      "step": 5550
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001849768843672863,
-      "loss": 0.8161,
-      "step": 5560
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001849496092844382,
-      "loss": 0.7545,
-      "step": 5570
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018492233420159016,
-      "loss": 0.6626,
-      "step": 5580
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018489505911874207,
-      "loss": 0.8168,
-      "step": 5590
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018486778403589403,
-      "loss": 0.8167,
-      "step": 5600
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018484050895304594,
-      "loss": 0.7413,
-      "step": 5610
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001848132338701979,
-      "loss": 0.8482,
-      "step": 5620
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018478595878734983,
-      "loss": 0.7672,
-      "step": 5630
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018475868370450177,
-      "loss": 0.5644,
-      "step": 5640
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001847314086216537,
-      "loss": 0.7944,
-      "step": 5650
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018470413353880564,
-      "loss": 0.7067,
-      "step": 5660
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018467685845595757,
-      "loss": 0.6828,
-      "step": 5670
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001846495833731095,
-      "loss": 0.8442,
-      "step": 5680
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018462230829026144,
-      "loss": 0.661,
-      "step": 5690
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018459503320741337,
-      "loss": 0.758,
-      "step": 5700
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001845677581245653,
-      "loss": 0.73,
-      "step": 5710
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.00018454048304171724,
-      "loss": 0.6359,
-      "step": 5720
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001845132079588692,
-      "loss": 0.7823,
-      "step": 5730
-    },
-    {
-      "epoch": 0.23,
-      "learning_rate": 0.0001844859328760211,
-      "loss": 0.7235,
-      "step": 5740
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018445865779317307,
-      "loss": 0.8074,
-      "step": 5750
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018443138271032498,
-      "loss": 0.6807,
-      "step": 5760
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018440410762747694,
-      "loss": 0.6972,
-      "step": 5770
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018437683254462884,
-      "loss": 0.7789,
-      "step": 5780
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001843495574617808,
-      "loss": 0.7167,
-      "step": 5790
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018432228237893274,
-      "loss": 0.816,
-      "step": 5800
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018429500729608467,
-      "loss": 0.7259,
-      "step": 5810
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001842677322132366,
-      "loss": 0.7342,
-      "step": 5820
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018424045713038854,
-      "loss": 0.6981,
-      "step": 5830
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018421318204754047,
-      "loss": 0.7484,
-      "step": 5840
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018418590696469243,
-      "loss": 0.8294,
-      "step": 5850
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018415863188184434,
-      "loss": 0.7915,
-      "step": 5860
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001841313567989963,
-      "loss": 0.7648,
-      "step": 5870
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001841040817161482,
-      "loss": 0.7677,
-      "step": 5880
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018407680663330017,
-      "loss": 0.7829,
-      "step": 5890
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001840495315504521,
-      "loss": 0.7239,
-      "step": 5900
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018402225646760404,
-      "loss": 0.6357,
-      "step": 5910
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018399498138475597,
-      "loss": 0.7132,
-      "step": 5920
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001839677063019079,
-      "loss": 0.7147,
-      "step": 5930
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018394043121905984,
-      "loss": 0.7284,
-      "step": 5940
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018391315613621177,
-      "loss": 0.8156,
-      "step": 5950
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001838858810533637,
-      "loss": 0.7974,
-      "step": 5960
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018385860597051564,
-      "loss": 0.7757,
-      "step": 5970
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.00018383133088766758,
-      "loss": 0.7051,
-      "step": 5980
-    },
-    {
-      "epoch": 0.24,
-      "learning_rate": 0.0001838040558048195,
-      "loss": 0.848,
-      "step": 5990
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018377678072197144,
-      "loss": 0.7828,
-      "step": 6000
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018374950563912338,
-      "loss": 0.6865,
-      "step": 6010
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018372223055627534,
-      "loss": 0.745,
-      "step": 6020
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018369495547342725,
-      "loss": 0.7393,
-      "step": 6030
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001836676803905792,
-      "loss": 0.6995,
-      "step": 6040
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001836404053077311,
-      "loss": 0.7658,
-      "step": 6050
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018361313022488307,
-      "loss": 0.7819,
-      "step": 6060
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.000183585855142035,
-      "loss": 0.8179,
-      "step": 6070
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018355858005918694,
-      "loss": 0.8054,
-      "step": 6080
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018353130497633888,
-      "loss": 0.6798,
-      "step": 6090
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001835040298934908,
-      "loss": 0.7034,
-      "step": 6100
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018347675481064274,
-      "loss": 0.7613,
-      "step": 6110
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001834494797277947,
-      "loss": 0.789,
-      "step": 6120
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001834222046449466,
-      "loss": 0.6595,
-      "step": 6130
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018339492956209857,
-      "loss": 0.8563,
-      "step": 6140
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018336765447925048,
-      "loss": 0.7226,
-      "step": 6150
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018334037939640244,
-      "loss": 0.8077,
-      "step": 6160
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018331310431355435,
-      "loss": 0.7192,
-      "step": 6170
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001832858292307063,
-      "loss": 0.7691,
-      "step": 6180
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018325855414785824,
-      "loss": 0.7978,
-      "step": 6190
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018323127906501018,
-      "loss": 0.7347,
-      "step": 6200
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.0001832040039821621,
-      "loss": 0.7104,
-      "step": 6210
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018317672889931404,
-      "loss": 0.7496,
-      "step": 6220
-    },
-    {
-      "epoch": 0.25,
-      "learning_rate": 0.00018314945381646598,
-      "loss": 0.7735,
-      "step": 6230
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.0001831221787336179,
-      "loss": 0.7028,
-      "step": 6240
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018309490365076985,
-      "loss": 0.8684,
-      "step": 6250
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018306762856792178,
-      "loss": 0.7671,
-      "step": 6260
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018304035348507371,
-      "loss": 0.7674,
-      "step": 6270
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018301307840222565,
-      "loss": 0.7501,
-      "step": 6280
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.0001829858033193776,
-      "loss": 0.7864,
-      "step": 6290
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018295852823652952,
-      "loss": 0.7411,
-      "step": 6300
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018293125315368148,
-      "loss": 0.8599,
-      "step": 6310
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018290397807083338,
-      "loss": 0.7887,
-      "step": 6320
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018287670298798534,
-      "loss": 0.7039,
-      "step": 6330
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018284942790513725,
-      "loss": 0.6865,
-      "step": 6340
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.0001828221528222892,
-      "loss": 0.794,
-      "step": 6350
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018279487773944115,
-      "loss": 0.7778,
-      "step": 6360
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018276760265659308,
-      "loss": 0.7182,
-      "step": 6370
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018274032757374501,
-      "loss": 0.8227,
-      "step": 6380
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018271305249089695,
-      "loss": 0.7458,
-      "step": 6390
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018268577740804888,
-      "loss": 0.698,
-      "step": 6400
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018265850232520084,
-      "loss": 0.8306,
-      "step": 6410
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018263122724235275,
-      "loss": 0.7473,
-      "step": 6420
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.0001826039521595047,
-      "loss": 0.703,
-      "step": 6430
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018257667707665662,
-      "loss": 0.8175,
-      "step": 6440
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018254940199380858,
-      "loss": 0.7634,
-      "step": 6450
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.0001825221269109605,
-      "loss": 0.6979,
-      "step": 6460
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018249485182811245,
-      "loss": 0.747,
-      "step": 6470
-    },
-    {
-      "epoch": 0.26,
-      "learning_rate": 0.00018246757674526438,
-      "loss": 0.633,
-      "step": 6480
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018244030166241631,
-      "loss": 0.7944,
-      "step": 6490
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018241302657956825,
-      "loss": 0.7999,
-      "step": 6500
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018238575149672018,
-      "loss": 0.743,
-      "step": 6510
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018235847641387212,
-      "loss": 0.7804,
-      "step": 6520
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018233120133102405,
-      "loss": 0.7141,
-      "step": 6530
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018230392624817598,
-      "loss": 0.7078,
-      "step": 6540
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018227665116532792,
-      "loss": 0.7235,
-      "step": 6550
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018224937608247985,
-      "loss": 0.8168,
-      "step": 6560
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018222210099963179,
-      "loss": 0.7496,
-      "step": 6570
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018219482591678375,
-      "loss": 0.8219,
-      "step": 6580
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018216755083393565,
-      "loss": 0.7182,
-      "step": 6590
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018214027575108761,
-      "loss": 0.7381,
-      "step": 6600
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018211300066823952,
-      "loss": 0.7947,
-      "step": 6610
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018208572558539148,
-      "loss": 0.725,
-      "step": 6620
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018205845050254342,
-      "loss": 0.6842,
-      "step": 6630
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018203117541969535,
-      "loss": 0.7382,
-      "step": 6640
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018200390033684728,
-      "loss": 0.7722,
-      "step": 6650
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018197662525399922,
-      "loss": 0.7996,
-      "step": 6660
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018194935017115115,
-      "loss": 0.7247,
-      "step": 6670
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018192207508830309,
-      "loss": 0.7524,
-      "step": 6680
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018189480000545502,
-      "loss": 0.7188,
-      "step": 6690
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018186752492260698,
-      "loss": 0.7457,
-      "step": 6700
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.0001818402498397589,
-      "loss": 0.7373,
-      "step": 6710
-    },
-    {
-      "epoch": 0.27,
-      "learning_rate": 0.00018181297475691085,
-      "loss": 0.7665,
-      "step": 6720
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018178569967406276,
-      "loss": 0.756,
-      "step": 6730
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018175842459121472,
-      "loss": 0.7958,
-      "step": 6740
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018173114950836665,
-      "loss": 0.8086,
-      "step": 6750
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018170387442551858,
-      "loss": 0.7308,
-      "step": 6760
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018167659934267052,
-      "loss": 0.6422,
-      "step": 6770
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018164932425982245,
-      "loss": 0.8206,
-      "step": 6780
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018162204917697439,
-      "loss": 0.6723,
-      "step": 6790
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018159477409412632,
-      "loss": 0.7986,
-      "step": 6800
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018156749901127825,
-      "loss": 0.6931,
-      "step": 6810
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.0001815402239284302,
-      "loss": 0.7267,
-      "step": 6820
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018151840386215173,
-      "loss": 0.7846,
-      "step": 6830
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018149658379587328,
-      "loss": 0.7305,
-      "step": 6840
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018146930871302524,
-      "loss": 0.6448,
-      "step": 6850
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018144203363017715,
-      "loss": 0.8702,
-      "step": 6860
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.0001814147585473291,
-      "loss": 0.7614,
-      "step": 6870
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018138748346448102,
-      "loss": 0.6904,
-      "step": 6880
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018136020838163298,
-      "loss": 0.6908,
-      "step": 6890
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.0001813329332987849,
-      "loss": 0.8213,
-      "step": 6900
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018130565821593685,
-      "loss": 0.6664,
-      "step": 6910
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018127838313308878,
-      "loss": 0.6944,
-      "step": 6920
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018125110805024071,
-      "loss": 0.8115,
-      "step": 6930
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018122383296739265,
-      "loss": 0.7094,
-      "step": 6940
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018119655788454458,
-      "loss": 0.7959,
-      "step": 6950
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018116928280169652,
-      "loss": 0.7496,
-      "step": 6960
-    },
-    {
-      "epoch": 0.28,
-      "learning_rate": 0.00018114200771884845,
-      "loss": 0.7125,
-      "step": 6970
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018111473263600038,
-      "loss": 0.7042,
-      "step": 6980
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018108745755315232,
-      "loss": 0.7315,
-      "step": 6990
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018106018247030425,
-      "loss": 0.6728,
-      "step": 7000
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018103290738745619,
-      "loss": 0.7577,
-      "step": 7010
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018100563230460815,
-      "loss": 0.6827,
-      "step": 7020
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018097835722176005,
-      "loss": 0.6082,
-      "step": 7030
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018095108213891201,
-      "loss": 0.8726,
-      "step": 7040
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018092380705606392,
-      "loss": 0.7027,
-      "step": 7050
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018089653197321588,
-      "loss": 0.7718,
-      "step": 7060
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018086925689036782,
-      "loss": 0.7776,
-      "step": 7070
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018084198180751975,
-      "loss": 0.735,
-      "step": 7080
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018081470672467168,
-      "loss": 0.7009,
-      "step": 7090
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018078743164182362,
-      "loss": 0.7317,
-      "step": 7100
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018076015655897555,
-      "loss": 0.7386,
-      "step": 7110
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.0001807328814761275,
-      "loss": 0.7576,
-      "step": 7120
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018070560639327942,
-      "loss": 0.818,
-      "step": 7130
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018067833131043138,
-      "loss": 0.7614,
-      "step": 7140
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.0001806510562275833,
-      "loss": 0.7396,
-      "step": 7150
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018062378114473525,
-      "loss": 0.6351,
-      "step": 7160
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018059650606188716,
-      "loss": 0.8007,
-      "step": 7170
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018056923097903912,
-      "loss": 0.8042,
-      "step": 7180
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018054195589619105,
-      "loss": 0.7506,
-      "step": 7190
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018051468081334298,
-      "loss": 0.6813,
-      "step": 7200
-    },
-    {
-      "epoch": 0.29,
-      "learning_rate": 0.00018048740573049492,
-      "loss": 0.7758,
-      "step": 7210
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018046013064764685,
-      "loss": 0.6604,
-      "step": 7220
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018043285556479879,
-      "loss": 0.7485,
-      "step": 7230
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018040558048195072,
-      "loss": 0.8363,
-      "step": 7240
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018037830539910265,
-      "loss": 0.7235,
-      "step": 7250
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.0001803510303162546,
-      "loss": 0.8581,
-      "step": 7260
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018032375523340652,
-      "loss": 0.7707,
-      "step": 7270
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018029648015055846,
-      "loss": 0.7631,
-      "step": 7280
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018026920506771042,
-      "loss": 0.745,
-      "step": 7290
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018024192998486232,
-      "loss": 0.7907,
-      "step": 7300
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018021465490201428,
-      "loss": 0.7483,
-      "step": 7310
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.0001801873798191662,
-      "loss": 0.7584,
-      "step": 7320
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018016010473631815,
-      "loss": 0.6741,
-      "step": 7330
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018013282965347009,
-      "loss": 0.794,
-      "step": 7340
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018010555457062202,
-      "loss": 0.7046,
-      "step": 7350
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018007827948777395,
-      "loss": 0.7418,
-      "step": 7360
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.0001800510044049259,
-      "loss": 0.7578,
-      "step": 7370
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00018002372932207782,
-      "loss": 0.7386,
-      "step": 7380
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017999645423922976,
-      "loss": 0.6995,
-      "step": 7390
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.0001799691791563817,
-      "loss": 0.793,
-      "step": 7400
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017994190407353365,
-      "loss": 0.6799,
-      "step": 7410
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017991462899068556,
-      "loss": 0.7616,
-      "step": 7420
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017988735390783752,
-      "loss": 0.8177,
-      "step": 7430
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017986007882498943,
-      "loss": 0.7199,
-      "step": 7440
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.0001798328037421414,
-      "loss": 0.7396,
-      "step": 7450
-    },
-    {
-      "epoch": 0.3,
-      "learning_rate": 0.00017980552865929332,
-      "loss": 0.7537,
-      "step": 7460
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017977825357644525,
-      "loss": 0.7451,
-      "step": 7470
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001797509784935972,
-      "loss": 0.7053,
-      "step": 7480
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017972370341074912,
-      "loss": 0.6983,
-      "step": 7490
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017969642832790106,
-      "loss": 0.7435,
-      "step": 7500
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.000179669153245053,
-      "loss": 0.7677,
-      "step": 7510
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017964187816220492,
-      "loss": 0.7974,
-      "step": 7520
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017961460307935686,
-      "loss": 0.7606,
-      "step": 7530
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001795873279965088,
-      "loss": 0.745,
-      "step": 7540
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017956005291366073,
-      "loss": 0.7927,
-      "step": 7550
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017953277783081266,
-      "loss": 0.719,
-      "step": 7560
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001795055027479646,
-      "loss": 0.8148,
-      "step": 7570
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017947822766511655,
-      "loss": 0.7412,
-      "step": 7580
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017945095258226846,
-      "loss": 0.7157,
-      "step": 7590
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017942367749942042,
-      "loss": 0.7385,
-      "step": 7600
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017939640241657233,
-      "loss": 0.6597,
-      "step": 7610
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001793691273337243,
-      "loss": 0.7367,
-      "step": 7620
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017934185225087622,
-      "loss": 0.874,
-      "step": 7630
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017931457716802816,
-      "loss": 0.8131,
-      "step": 7640
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001792873020851801,
-      "loss": 0.7316,
-      "step": 7650
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017926002700233203,
-      "loss": 0.737,
-      "step": 7660
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017923275191948396,
-      "loss": 0.7385,
-      "step": 7670
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017920547683663592,
-      "loss": 0.8374,
-      "step": 7680
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.00017917820175378783,
-      "loss": 0.7385,
-      "step": 7690
-    },
-    {
-      "epoch": 0.31,
-      "learning_rate": 0.0001791509266709398,
-      "loss": 0.7871,
-      "step": 7700
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001791236515880917,
-      "loss": 0.7529,
-      "step": 7710
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017909637650524366,
-      "loss": 0.7731,
-      "step": 7720
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017906910142239556,
-      "loss": 0.7531,
-      "step": 7730
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017904182633954752,
-      "loss": 0.8181,
-      "step": 7740
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017901455125669946,
-      "loss": 0.688,
-      "step": 7750
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001789872761738514,
-      "loss": 0.6154,
-      "step": 7760
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017896000109100333,
-      "loss": 0.8266,
-      "step": 7770
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017893272600815526,
-      "loss": 0.7339,
-      "step": 7780
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001789054509253072,
-      "loss": 0.815,
-      "step": 7790
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017887817584245913,
-      "loss": 0.7168,
-      "step": 7800
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017885090075961106,
-      "loss": 0.7216,
-      "step": 7810
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.000178823625676763,
-      "loss": 0.6906,
-      "step": 7820
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017879635059391493,
-      "loss": 0.7682,
-      "step": 7830
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017876907551106686,
-      "loss": 0.7207,
-      "step": 7840
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017874180042821882,
-      "loss": 0.7782,
-      "step": 7850
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017871452534537073,
-      "loss": 0.7035,
-      "step": 7860
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001786872502625227,
-      "loss": 0.8311,
-      "step": 7870
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001786599751796746,
-      "loss": 0.7765,
-      "step": 7880
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017863270009682656,
-      "loss": 0.643,
-      "step": 7890
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017860542501397847,
-      "loss": 0.7541,
-      "step": 7900
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017857814993113043,
-      "loss": 0.7931,
-      "step": 7910
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017855087484828236,
-      "loss": 0.7463,
-      "step": 7920
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.0001785235997654343,
-      "loss": 0.7416,
-      "step": 7930
-    },
-    {
-      "epoch": 0.32,
-      "learning_rate": 0.00017849632468258623,
-      "loss": 0.9026,
-      "step": 7940
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017846904959973816,
-      "loss": 0.8,
-      "step": 7950
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001784417745168901,
-      "loss": 0.7194,
-      "step": 7960
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017841449943404206,
-      "loss": 0.7303,
-      "step": 7970
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017838722435119397,
-      "loss": 0.7573,
-      "step": 7980
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017835994926834593,
-      "loss": 0.8316,
-      "step": 7990
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017833267418549783,
-      "loss": 0.6897,
-      "step": 8000
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001783053991026498,
-      "loss": 0.6358,
-      "step": 8010
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017827812401980173,
-      "loss": 0.7409,
-      "step": 8020
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017825084893695366,
-      "loss": 0.6538,
-      "step": 8030
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001782235738541056,
-      "loss": 0.7739,
-      "step": 8040
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017819629877125753,
-      "loss": 0.8345,
-      "step": 8050
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017816902368840946,
-      "loss": 0.6941,
-      "step": 8060
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001781417486055614,
-      "loss": 0.7498,
-      "step": 8070
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017811447352271333,
-      "loss": 0.7402,
-      "step": 8080
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017808719843986527,
-      "loss": 0.7412,
-      "step": 8090
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001780599233570172,
-      "loss": 0.7364,
-      "step": 8100
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017803264827416913,
-      "loss": 0.7498,
-      "step": 8110
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017800537319132107,
-      "loss": 0.6954,
-      "step": 8120
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.000177978098108473,
-      "loss": 0.6692,
-      "step": 8130
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017795082302562496,
-      "loss": 0.7182,
-      "step": 8140
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017792354794277687,
-      "loss": 0.8711,
-      "step": 8150
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017789627285992883,
-      "loss": 0.8027,
-      "step": 8160
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017786899777708074,
-      "loss": 0.8394,
-      "step": 8170
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.0001778417226942327,
-      "loss": 0.7273,
-      "step": 8180
-    },
-    {
-      "epoch": 0.33,
-      "learning_rate": 0.00017781444761138463,
-      "loss": 0.7211,
-      "step": 8190
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017778717252853657,
-      "loss": 0.7571,
-      "step": 8200
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001777598974456885,
-      "loss": 0.7787,
-      "step": 8210
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017773262236284043,
-      "loss": 0.6375,
-      "step": 8220
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017770534727999237,
-      "loss": 0.7746,
-      "step": 8230
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001776780721971443,
-      "loss": 0.6684,
-      "step": 8240
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017765079711429624,
-      "loss": 0.7746,
-      "step": 8250
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001776235220314482,
-      "loss": 0.6816,
-      "step": 8260
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001775962469486001,
-      "loss": 0.7877,
-      "step": 8270
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017756897186575206,
-      "loss": 0.7144,
-      "step": 8280
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017754169678290397,
-      "loss": 0.6435,
-      "step": 8290
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017751442170005593,
-      "loss": 0.8246,
-      "step": 8300
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017748714661720787,
-      "loss": 0.8477,
-      "step": 8310
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001774598715343598,
-      "loss": 0.7874,
-      "step": 8320
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017743259645151173,
-      "loss": 0.7087,
-      "step": 8330
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017740532136866367,
-      "loss": 0.6565,
-      "step": 8340
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001773780462858156,
-      "loss": 0.6717,
-      "step": 8350
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017735077120296754,
-      "loss": 0.6537,
-      "step": 8360
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017732349612011947,
-      "loss": 0.7261,
-      "step": 8370
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001772962210372714,
-      "loss": 0.6638,
-      "step": 8380
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017726894595442334,
-      "loss": 0.8403,
-      "step": 8390
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017724167087157527,
-      "loss": 0.8468,
-      "step": 8400
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017721439578872723,
-      "loss": 0.7903,
-      "step": 8410
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.00017718712070587914,
-      "loss": 0.6946,
-      "step": 8420
-    },
-    {
-      "epoch": 0.34,
-      "learning_rate": 0.0001771598456230311,
-      "loss": 0.6842,
-      "step": 8430
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.000177132570540183,
-      "loss": 0.686,
-      "step": 8440
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017710529545733497,
-      "loss": 0.7415,
-      "step": 8450
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017707802037448688,
-      "loss": 0.6995,
-      "step": 8460
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017705074529163884,
-      "loss": 0.7051,
-      "step": 8470
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017702347020879077,
-      "loss": 0.7786,
-      "step": 8480
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.0001769961951259427,
-      "loss": 0.7833,
-      "step": 8490
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017696892004309464,
-      "loss": 0.833,
-      "step": 8500
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017694164496024657,
-      "loss": 0.8662,
-      "step": 8510
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.0001769143698773985,
-      "loss": 0.7309,
-      "step": 8520
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017688709479455047,
-      "loss": 0.8636,
-      "step": 8530
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017685981971170237,
-      "loss": 0.7192,
-      "step": 8540
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017683254462885433,
-      "loss": 0.7332,
-      "step": 8550
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017680526954600624,
-      "loss": 0.6712,
-      "step": 8560
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.0001767779944631582,
-      "loss": 0.7437,
-      "step": 8570
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017675071938031014,
-      "loss": 0.7713,
-      "step": 8580
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017672344429746207,
-      "loss": 0.7788,
-      "step": 8590
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.000176696169214614,
-      "loss": 0.7272,
-      "step": 8600
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017666889413176594,
-      "loss": 0.8131,
-      "step": 8610
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017664161904891787,
-      "loss": 0.7418,
-      "step": 8620
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.0001766143439660698,
-      "loss": 0.7167,
-      "step": 8630
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017658706888322174,
-      "loss": 0.9113,
-      "step": 8640
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017655979380037367,
-      "loss": 0.7082,
-      "step": 8650
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.0001765325187175256,
-      "loss": 0.7468,
-      "step": 8660
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017650524363467754,
-      "loss": 0.6927,
-      "step": 8670
-    },
-    {
-      "epoch": 0.35,
-      "learning_rate": 0.00017647796855182948,
-      "loss": 0.8069,
-      "step": 8680
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001764506934689814,
-      "loss": 0.6716,
-      "step": 8690
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017642341838613337,
-      "loss": 0.8006,
-      "step": 8700
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017639614330328528,
-      "loss": 0.7479,
-      "step": 8710
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017636886822043724,
-      "loss": 0.743,
-      "step": 8720
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017634159313758915,
-      "loss": 0.6978,
-      "step": 8730
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001763143180547411,
-      "loss": 0.738,
-      "step": 8740
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017628704297189304,
-      "loss": 0.8653,
-      "step": 8750
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017625976788904497,
-      "loss": 0.7657,
-      "step": 8760
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001762324928061969,
-      "loss": 0.7356,
-      "step": 8770
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017620521772334884,
-      "loss": 0.8597,
-      "step": 8780
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017617794264050078,
-      "loss": 0.7535,
-      "step": 8790
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001761506675576527,
-      "loss": 0.7529,
-      "step": 8800
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017612339247480464,
-      "loss": 0.7705,
-      "step": 8810
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001760961173919566,
-      "loss": 0.7348,
-      "step": 8820
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001760688423091085,
-      "loss": 0.7152,
-      "step": 8830
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017604156722626047,
-      "loss": 0.7681,
-      "step": 8840
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017601429214341238,
-      "loss": 0.778,
-      "step": 8850
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017598701706056434,
-      "loss": 0.7452,
-      "step": 8860
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017595974197771627,
-      "loss": 0.7449,
-      "step": 8870
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.0001759324668948682,
-      "loss": 0.7182,
-      "step": 8880
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017590519181202014,
-      "loss": 0.721,
-      "step": 8890
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017587791672917208,
-      "loss": 0.7464,
-      "step": 8900
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.000175850641646324,
-      "loss": 0.7211,
-      "step": 8910
-    },
-    {
-      "epoch": 0.36,
-      "learning_rate": 0.00017582336656347594,
-      "loss": 0.7178,
-      "step": 8920
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017579609148062788,
-      "loss": 0.7636,
-      "step": 8930
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001757688163977798,
-      "loss": 0.766,
-      "step": 8940
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017574154131493175,
-      "loss": 0.6817,
-      "step": 8950
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017571426623208368,
-      "loss": 0.8706,
-      "step": 8960
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001756869911492356,
-      "loss": 0.7172,
-      "step": 8970
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017565971606638755,
-      "loss": 0.7251,
-      "step": 8980
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001756324409835395,
-      "loss": 0.7021,
-      "step": 8990
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017560516590069142,
-      "loss": 0.7773,
-      "step": 9000
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017557789081784338,
-      "loss": 0.7687,
-      "step": 9010
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017555061573499528,
-      "loss": 0.7669,
-      "step": 9020
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017552334065214724,
-      "loss": 0.7225,
-      "step": 9030
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017549606556929918,
-      "loss": 0.7207,
-      "step": 9040
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001754687904864511,
-      "loss": 0.7417,
-      "step": 9050
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017544151540360305,
-      "loss": 0.7467,
-      "step": 9060
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017541424032075498,
-      "loss": 0.747,
-      "step": 9070
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017538696523790691,
-      "loss": 0.7182,
-      "step": 9080
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017535969015505887,
-      "loss": 0.7104,
-      "step": 9090
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017533241507221078,
-      "loss": 0.6549,
-      "step": 9100
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017530513998936274,
-      "loss": 0.7331,
-      "step": 9110
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017527786490651465,
-      "loss": 0.6157,
-      "step": 9120
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001752505898236666,
-      "loss": 0.6651,
-      "step": 9130
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017522331474081852,
-      "loss": 0.7328,
-      "step": 9140
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017519603965797048,
-      "loss": 0.8033,
-      "step": 9150
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.0001751687645751224,
-      "loss": 0.7846,
-      "step": 9160
-    },
-    {
-      "epoch": 0.37,
-      "learning_rate": 0.00017514148949227435,
-      "loss": 0.7483,
-      "step": 9170
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017511421440942628,
-      "loss": 0.6703,
-      "step": 9180
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017508693932657821,
-      "loss": 0.8176,
-      "step": 9190
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017505966424373015,
-      "loss": 0.7352,
-      "step": 9200
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017503238916088208,
-      "loss": 0.7076,
-      "step": 9210
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017500511407803402,
-      "loss": 0.6832,
-      "step": 9220
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017497783899518595,
-      "loss": 0.6542,
-      "step": 9230
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017495056391233788,
-      "loss": 0.8063,
-      "step": 9240
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017492328882948982,
-      "loss": 0.8097,
-      "step": 9250
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017489601374664178,
-      "loss": 0.6901,
-      "step": 9260
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017486873866379369,
-      "loss": 0.7013,
-      "step": 9270
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017484146358094565,
-      "loss": 0.7137,
-      "step": 9280
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017481418849809755,
-      "loss": 0.8118,
-      "step": 9290
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017478691341524951,
-      "loss": 0.7204,
-      "step": 9300
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017475963833240145,
-      "loss": 0.7525,
-      "step": 9310
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017473236324955338,
-      "loss": 0.8205,
-      "step": 9320
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017470508816670532,
-      "loss": 0.6828,
-      "step": 9330
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017467781308385725,
-      "loss": 0.6661,
-      "step": 9340
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017465053800100918,
-      "loss": 0.6799,
-      "step": 9350
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017462326291816112,
-      "loss": 0.7523,
-      "step": 9360
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017459598783531305,
-      "loss": 0.7566,
-      "step": 9370
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.000174568712752465,
-      "loss": 0.7847,
-      "step": 9380
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017454143766961692,
-      "loss": 0.7617,
-      "step": 9390
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.00017451416258676888,
-      "loss": 0.7648,
-      "step": 9400
-    },
-    {
-      "epoch": 0.38,
-      "learning_rate": 0.0001744868875039208,
-      "loss": 0.7454,
-      "step": 9410
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017445961242107275,
-      "loss": 0.6576,
-      "step": 9420
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017443233733822468,
-      "loss": 0.699,
-      "step": 9430
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017440506225537662,
-      "loss": 0.7219,
-      "step": 9440
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017437778717252855,
-      "loss": 0.6954,
-      "step": 9450
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017435051208968048,
-      "loss": 0.7065,
-      "step": 9460
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017432323700683242,
-      "loss": 0.8538,
-      "step": 9470
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017429596192398435,
-      "loss": 0.8284,
-      "step": 9480
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017426868684113629,
-      "loss": 0.7501,
-      "step": 9490
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017424141175828822,
-      "loss": 0.7786,
-      "step": 9500
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017421413667544015,
-      "loss": 0.7549,
-      "step": 9510
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.0001741868615925921,
-      "loss": 0.7595,
-      "step": 9520
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017415958650974402,
-      "loss": 0.7935,
-      "step": 9530
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017413231142689596,
-      "loss": 0.743,
-      "step": 9540
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017410503634404792,
-      "loss": 0.6588,
-      "step": 9550
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017407776126119982,
-      "loss": 0.7206,
-      "step": 9560
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017405048617835178,
-      "loss": 0.7517,
-      "step": 9570
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.0001740232110955037,
-      "loss": 0.741,
-      "step": 9580
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017399593601265565,
-      "loss": 0.7312,
-      "step": 9590
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017396866092980759,
-      "loss": 0.7457,
-      "step": 9600
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017394138584695952,
-      "loss": 0.7571,
-      "step": 9610
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017391411076411145,
-      "loss": 0.7214,
-      "step": 9620
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.0001738868356812634,
-      "loss": 0.7766,
-      "step": 9630
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017385956059841532,
-      "loss": 0.7315,
-      "step": 9640
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.00017383228551556728,
-      "loss": 0.7097,
-      "step": 9650
-    },
-    {
-      "epoch": 0.39,
-      "learning_rate": 0.0001738050104327192,
-      "loss": 0.647,
-      "step": 9660
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017377773534987115,
-      "loss": 0.6916,
-      "step": 9670
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017375046026702306,
-      "loss": 0.8147,
-      "step": 9680
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017372318518417502,
-      "loss": 0.7019,
-      "step": 9690
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017369591010132693,
-      "loss": 0.8481,
-      "step": 9700
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017366863501847889,
-      "loss": 0.7104,
-      "step": 9710
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017364135993563082,
-      "loss": 0.8464,
-      "step": 9720
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017361408485278275,
-      "loss": 0.777,
-      "step": 9730
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.0001735868097699347,
-      "loss": 0.7667,
-      "step": 9740
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017355953468708662,
-      "loss": 0.728,
-      "step": 9750
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017353225960423856,
-      "loss": 0.7443,
-      "step": 9760
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.0001735049845213905,
-      "loss": 0.8501,
-      "step": 9770
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017347770943854242,
-      "loss": 0.817,
-      "step": 9780
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017345043435569436,
-      "loss": 0.7557,
-      "step": 9790
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.0001734231592728463,
-      "loss": 0.7954,
-      "step": 9800
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017339588418999823,
-      "loss": 0.8079,
-      "step": 9810
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017336860910715019,
-      "loss": 0.7053,
-      "step": 9820
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.0001733413340243021,
-      "loss": 0.7581,
-      "step": 9830
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017331405894145405,
-      "loss": 0.7399,
-      "step": 9840
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017328678385860596,
-      "loss": 0.6399,
-      "step": 9850
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017325950877575792,
-      "loss": 0.731,
-      "step": 9860
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017323223369290983,
-      "loss": 0.6951,
-      "step": 9870
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.0001732049586100618,
-      "loss": 0.7492,
-      "step": 9880
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017317768352721372,
-      "loss": 0.7712,
-      "step": 9890
-    },
-    {
-      "epoch": 0.4,
-      "learning_rate": 0.00017315040844436566,
-      "loss": 0.8102,
-      "step": 9900
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001731231333615176,
-      "loss": 0.7756,
-      "step": 9910
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017309585827866953,
-      "loss": 0.7889,
-      "step": 9920
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017306858319582146,
-      "loss": 0.7363,
-      "step": 9930
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017304130811297342,
-      "loss": 0.7117,
-      "step": 9940
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017301403303012533,
-      "loss": 0.7368,
-      "step": 9950
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001729867579472773,
-      "loss": 0.7641,
-      "step": 9960
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001729594828644292,
-      "loss": 0.6961,
-      "step": 9970
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017293220778158116,
-      "loss": 0.7581,
-      "step": 9980
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001729049326987331,
-      "loss": 0.7933,
-      "step": 9990
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017287765761588502,
-      "loss": 0.7365,
-      "step": 10000
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017285038253303696,
-      "loss": 0.7072,
-      "step": 10010
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001728231074501889,
-      "loss": 0.671,
-      "step": 10020
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017279583236734083,
-      "loss": 0.6972,
-      "step": 10030
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017276855728449276,
-      "loss": 0.7829,
-      "step": 10040
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001727412822016447,
-      "loss": 0.735,
-      "step": 10050
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017271400711879663,
-      "loss": 0.672,
-      "step": 10060
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017268673203594856,
-      "loss": 0.769,
-      "step": 10070
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001726594569531005,
-      "loss": 0.7252,
-      "step": 10080
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017263218187025243,
-      "loss": 0.7882,
-      "step": 10090
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017260490678740436,
-      "loss": 0.7748,
-      "step": 10100
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017257763170455632,
-      "loss": 0.6785,
-      "step": 10110
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017255035662170823,
-      "loss": 0.7546,
-      "step": 10120
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001725230815388602,
-      "loss": 0.7522,
-      "step": 10130
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.0001724958064560121,
-      "loss": 0.8466,
-      "step": 10140
-    },
-    {
-      "epoch": 0.41,
-      "learning_rate": 0.00017246853137316406,
-      "loss": 0.7151,
-      "step": 10150
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.000172441256290316,
-      "loss": 0.7667,
-      "step": 10160
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017241398120746793,
-      "loss": 0.6881,
-      "step": 10170
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017238670612461986,
-      "loss": 0.8029,
-      "step": 10180
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001723594310417718,
-      "loss": 0.7247,
-      "step": 10190
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017233215595892373,
-      "loss": 0.7573,
-      "step": 10200
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017230488087607566,
-      "loss": 0.7151,
-      "step": 10210
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001722776057932276,
-      "loss": 0.6939,
-      "step": 10220
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017225033071037956,
-      "loss": 0.7474,
-      "step": 10230
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017222305562753147,
-      "loss": 0.7681,
-      "step": 10240
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017219578054468343,
-      "loss": 0.7174,
-      "step": 10250
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017216850546183533,
-      "loss": 0.7404,
-      "step": 10260
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001721412303789873,
-      "loss": 0.7297,
-      "step": 10270
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017211395529613923,
-      "loss": 0.7241,
-      "step": 10280
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017208668021329116,
-      "loss": 0.7693,
-      "step": 10290
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001720594051304431,
-      "loss": 0.7276,
-      "step": 10300
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017203213004759503,
-      "loss": 0.7506,
-      "step": 10310
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017200485496474696,
-      "loss": 0.8007,
-      "step": 10320
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001719775798818989,
-      "loss": 0.7668,
-      "step": 10330
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017195030479905083,
-      "loss": 0.7672,
-      "step": 10340
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017192302971620277,
-      "loss": 0.6876,
-      "step": 10350
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001718957546333547,
-      "loss": 0.7881,
-      "step": 10360
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.00017186847955050663,
-      "loss": 0.8103,
-      "step": 10370
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001718412044676586,
-      "loss": 0.7961,
-      "step": 10380
-    },
-    {
-      "epoch": 0.42,
-      "learning_rate": 0.0001718139293848105,
-      "loss": 0.7283,
-      "step": 10390
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017178665430196246,
-      "loss": 0.7786,
-      "step": 10400
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017175937921911437,
-      "loss": 0.7046,
-      "step": 10410
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017173210413626633,
-      "loss": 0.6408,
-      "step": 10420
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017170482905341824,
-      "loss": 0.7695,
-      "step": 10430
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001716775539705702,
-      "loss": 0.7027,
-      "step": 10440
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017165027888772213,
-      "loss": 0.7744,
-      "step": 10450
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017162300380487407,
-      "loss": 0.7269,
-      "step": 10460
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.000171595728722026,
-      "loss": 0.7368,
-      "step": 10470
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017156845363917793,
-      "loss": 0.7411,
-      "step": 10480
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017154117855632987,
-      "loss": 0.7351,
-      "step": 10490
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017151390347348183,
-      "loss": 0.768,
-      "step": 10500
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017148662839063374,
-      "loss": 0.624,
-      "step": 10510
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001714593533077857,
-      "loss": 0.7836,
-      "step": 10520
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001714320782249376,
-      "loss": 0.7845,
-      "step": 10530
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017140480314208956,
-      "loss": 0.7887,
-      "step": 10540
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001713775280592415,
-      "loss": 0.7651,
-      "step": 10550
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017135025297639343,
-      "loss": 0.6953,
-      "step": 10560
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017132297789354537,
-      "loss": 0.6843,
-      "step": 10570
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001712957028106973,
-      "loss": 0.6862,
-      "step": 10580
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017126842772784923,
-      "loss": 0.6569,
-      "step": 10590
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017124115264500117,
-      "loss": 0.8058,
-      "step": 10600
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.0001712138775621531,
-      "loss": 0.8112,
-      "step": 10610
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017118660247930504,
-      "loss": 0.7938,
-      "step": 10620
-    },
-    {
-      "epoch": 0.43,
-      "learning_rate": 0.00017115932739645697,
-      "loss": 0.7299,
-      "step": 10630
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001711320523136089,
-      "loss": 0.7772,
-      "step": 10640
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017110477723076084,
-      "loss": 0.7629,
-      "step": 10650
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017107750214791277,
-      "loss": 0.888,
-      "step": 10660
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017105022706506473,
-      "loss": 0.722,
-      "step": 10670
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017102295198221664,
-      "loss": 0.7445,
-      "step": 10680
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001709956768993686,
-      "loss": 0.7041,
-      "step": 10690
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001709684018165205,
-      "loss": 0.8301,
-      "step": 10700
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017094112673367247,
-      "loss": 0.8537,
-      "step": 10710
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001709138516508244,
-      "loss": 0.6915,
-      "step": 10720
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017088657656797634,
-      "loss": 0.8135,
-      "step": 10730
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017085930148512827,
-      "loss": 0.7464,
-      "step": 10740
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001708320264022802,
-      "loss": 0.7924,
-      "step": 10750
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017080475131943214,
-      "loss": 0.8201,
-      "step": 10760
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017077747623658407,
-      "loss": 0.6306,
-      "step": 10770
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.000170750201153736,
-      "loss": 0.7053,
-      "step": 10780
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017072292607088797,
-      "loss": 0.8318,
-      "step": 10790
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017069565098803987,
-      "loss": 0.8826,
-      "step": 10800
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017066837590519183,
-      "loss": 0.8356,
-      "step": 10810
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017064110082234374,
-      "loss": 0.7542,
-      "step": 10820
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001706138257394957,
-      "loss": 0.6621,
-      "step": 10830
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017058655065664764,
-      "loss": 0.9733,
-      "step": 10840
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017055927557379957,
-      "loss": 0.8173,
-      "step": 10850
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.0001705320004909515,
-      "loss": 0.7323,
-      "step": 10860
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017050472540810344,
-      "loss": 0.7958,
-      "step": 10870
-    },
-    {
-      "epoch": 0.44,
-      "learning_rate": 0.00017047745032525537,
-      "loss": 0.6722,
-      "step": 10880
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001704501752424073,
-      "loss": 0.7308,
-      "step": 10890
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017042290015955924,
-      "loss": 0.7135,
-      "step": 10900
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017039562507671117,
-      "loss": 0.7846,
-      "step": 10910
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001703683499938631,
-      "loss": 0.7496,
-      "step": 10920
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017034107491101504,
-      "loss": 0.6861,
-      "step": 10930
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017031379982816698,
-      "loss": 0.6847,
-      "step": 10940
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001702865247453189,
-      "loss": 0.7122,
-      "step": 10950
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017025924966247087,
-      "loss": 0.715,
-      "step": 10960
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017023197457962278,
-      "loss": 0.6798,
-      "step": 10970
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017020469949677474,
-      "loss": 0.7661,
-      "step": 10980
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017017742441392664,
-      "loss": 0.7034,
-      "step": 10990
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001701501493310786,
-      "loss": 0.6606,
-      "step": 11000
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017012287424823054,
-      "loss": 0.771,
-      "step": 11010
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017009559916538247,
-      "loss": 0.694,
-      "step": 11020
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001700683240825344,
-      "loss": 0.8144,
-      "step": 11030
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017004104899968634,
-      "loss": 0.7428,
-      "step": 11040
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00017001377391683828,
-      "loss": 0.701,
-      "step": 11050
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00016998649883399024,
-      "loss": 0.7241,
-      "step": 11060
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00016995922375114214,
-      "loss": 0.6784,
-      "step": 11070
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.0001699319486682941,
-      "loss": 0.706,
-      "step": 11080
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.000169904673585446,
-      "loss": 0.6501,
-      "step": 11090
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00016987739850259797,
-      "loss": 0.8005,
-      "step": 11100
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00016985012341974988,
-      "loss": 0.6106,
-      "step": 11110
-    },
-    {
-      "epoch": 0.45,
-      "learning_rate": 0.00016982284833690184,
-      "loss": 0.7506,
-      "step": 11120
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016979557325405377,
-      "loss": 0.7911,
-      "step": 11130
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001697682981712057,
-      "loss": 0.8406,
-      "step": 11140
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016974102308835764,
-      "loss": 0.6644,
-      "step": 11150
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016971374800550958,
-      "loss": 0.8487,
-      "step": 11160
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001696864729226615,
-      "loss": 0.7108,
-      "step": 11170
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016965919783981344,
-      "loss": 0.669,
-      "step": 11180
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016963192275696538,
-      "loss": 0.7807,
-      "step": 11190
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001696046476741173,
-      "loss": 0.8057,
-      "step": 11200
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016957737259126925,
-      "loss": 0.6849,
-      "step": 11210
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016955009750842118,
-      "loss": 0.7503,
-      "step": 11220
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016952282242557314,
-      "loss": 0.7862,
-      "step": 11230
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016949554734272505,
-      "loss": 0.6987,
-      "step": 11240
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.000169468272259877,
-      "loss": 0.7468,
-      "step": 11250
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016944099717702891,
-      "loss": 0.762,
-      "step": 11260
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016941372209418088,
-      "loss": 0.7181,
-      "step": 11270
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001693864470113328,
-      "loss": 0.832,
-      "step": 11280
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016935917192848474,
-      "loss": 0.6933,
-      "step": 11290
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016933189684563668,
-      "loss": 0.7325,
-      "step": 11300
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001693046217627886,
-      "loss": 0.8011,
-      "step": 11310
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016927734667994055,
-      "loss": 0.7034,
-      "step": 11320
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016925007159709248,
-      "loss": 0.6772,
-      "step": 11330
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.0001692227965142444,
-      "loss": 0.7877,
-      "step": 11340
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016919552143139637,
-      "loss": 0.6723,
-      "step": 11350
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016916824634854828,
-      "loss": 0.6684,
-      "step": 11360
-    },
-    {
-      "epoch": 0.46,
-      "learning_rate": 0.00016914097126570024,
-      "loss": 0.8067,
-      "step": 11370
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016911369618285215,
-      "loss": 0.8226,
-      "step": 11380
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.0001690864211000041,
-      "loss": 0.7053,
-      "step": 11390
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016905914601715604,
-      "loss": 0.7296,
-      "step": 11400
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016903187093430798,
-      "loss": 0.7104,
-      "step": 11410
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.0001690045958514599,
-      "loss": 0.6877,
-      "step": 11420
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016897732076861185,
-      "loss": 0.732,
-      "step": 11430
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016895004568576378,
-      "loss": 0.7831,
-      "step": 11440
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.0001689227706029157,
-      "loss": 0.7803,
-      "step": 11450
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016889549552006765,
-      "loss": 0.6635,
-      "step": 11460
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016886822043721958,
-      "loss": 0.7061,
-      "step": 11470
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016884094535437152,
-      "loss": 0.8193,
-      "step": 11480
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016881367027152345,
-      "loss": 0.6728,
-      "step": 11490
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016878639518867538,
-      "loss": 0.8079,
-      "step": 11500
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016875912010582732,
-      "loss": 0.7231,
-      "step": 11510
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016873184502297928,
-      "loss": 0.776,
-      "step": 11520
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016870456994013118,
-      "loss": 0.7501,
-      "step": 11530
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016867729485728315,
-      "loss": 0.7053,
-      "step": 11540
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016865001977443505,
-      "loss": 0.7972,
-      "step": 11550
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016862274469158701,
-      "loss": 0.6683,
-      "step": 11560
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016859546960873895,
-      "loss": 0.77,
-      "step": 11570
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016856819452589088,
-      "loss": 0.6886,
-      "step": 11580
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016854091944304282,
-      "loss": 0.733,
-      "step": 11590
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016851364436019475,
-      "loss": 0.7761,
-      "step": 11600
-    },
-    {
-      "epoch": 0.47,
-      "learning_rate": 0.00016848636927734668,
-      "loss": 0.7189,
-      "step": 11610
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016845909419449864,
-      "loss": 0.734,
-      "step": 11620
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016843181911165055,
-      "loss": 0.728,
-      "step": 11630
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.0001684045440288025,
-      "loss": 0.7437,
-      "step": 11640
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016837726894595442,
-      "loss": 0.7554,
-      "step": 11650
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016834999386310638,
-      "loss": 0.7174,
-      "step": 11660
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.0001683227187802583,
-      "loss": 0.8031,
-      "step": 11670
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016829544369741025,
-      "loss": 0.6501,
-      "step": 11680
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016826816861456218,
-      "loss": 0.7378,
-      "step": 11690
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016824089353171412,
-      "loss": 0.7201,
-      "step": 11700
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016821361844886605,
-      "loss": 0.6518,
-      "step": 11710
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016818634336601798,
-      "loss": 0.8387,
-      "step": 11720
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016815906828316992,
-      "loss": 0.7224,
-      "step": 11730
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016813179320032185,
-      "loss": 0.8272,
-      "step": 11740
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016810451811747379,
-      "loss": 0.6646,
-      "step": 11750
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016807724303462572,
-      "loss": 0.6643,
-      "step": 11760
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016804996795177765,
-      "loss": 0.7452,
-      "step": 11770
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.0001680226928689296,
-      "loss": 0.709,
-      "step": 11780
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016799541778608155,
-      "loss": 0.6293,
-      "step": 11790
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016796814270323345,
-      "loss": 0.6654,
-      "step": 11800
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016794086762038542,
-      "loss": 0.7117,
-      "step": 11810
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016791359253753732,
-      "loss": 0.838,
-      "step": 11820
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016788631745468928,
-      "loss": 0.714,
-      "step": 11830
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.0001678590423718412,
-      "loss": 0.7409,
-      "step": 11840
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016783176728899315,
-      "loss": 0.6685,
-      "step": 11850
-    },
-    {
-      "epoch": 0.48,
-      "learning_rate": 0.00016780449220614509,
-      "loss": 0.6835,
-      "step": 11860
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016777721712329702,
-      "loss": 0.7163,
-      "step": 11870
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016774994204044895,
-      "loss": 0.696,
-      "step": 11880
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.0001677226669576009,
-      "loss": 0.8786,
-      "step": 11890
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016769539187475282,
-      "loss": 0.7739,
-      "step": 11900
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016766811679190478,
-      "loss": 0.6761,
-      "step": 11910
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.0001676408417090567,
-      "loss": 0.732,
-      "step": 11920
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016761356662620865,
-      "loss": 0.7268,
-      "step": 11930
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016758629154336056,
-      "loss": 0.7325,
-      "step": 11940
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016755901646051252,
-      "loss": 0.7951,
-      "step": 11950
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016753174137766445,
-      "loss": 0.8462,
-      "step": 11960
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016750446629481639,
-      "loss": 0.6878,
-      "step": 11970
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016747719121196832,
-      "loss": 0.6927,
-      "step": 11980
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016744991612912025,
-      "loss": 0.6552,
-      "step": 11990
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.0001674226410462722,
-      "loss": 0.6429,
-      "step": 12000
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016739536596342412,
-      "loss": 0.7024,
-      "step": 12010
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016736809088057606,
-      "loss": 0.8114,
-      "step": 12020
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.000167340815797728,
-      "loss": 0.717,
-      "step": 12030
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016731354071487992,
-      "loss": 0.7127,
-      "step": 12040
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016728626563203186,
-      "loss": 0.7555,
-      "step": 12050
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.0001672589905491838,
-      "loss": 0.7205,
-      "step": 12060
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016723171546633572,
-      "loss": 0.7402,
-      "step": 12070
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016720444038348769,
-      "loss": 0.8193,
-      "step": 12080
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.0001671771653006396,
-      "loss": 0.7543,
-      "step": 12090
-    },
-    {
-      "epoch": 0.49,
-      "learning_rate": 0.00016714989021779155,
-      "loss": 0.7513,
-      "step": 12100
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016712261513494346,
-      "loss": 0.7816,
-      "step": 12110
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016709534005209542,
-      "loss": 0.7395,
-      "step": 12120
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016706806496924736,
-      "loss": 0.7744,
-      "step": 12130
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001670407898863993,
-      "loss": 0.7445,
-      "step": 12140
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016701351480355122,
-      "loss": 0.7988,
-      "step": 12150
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016698623972070316,
-      "loss": 0.6938,
-      "step": 12160
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001669589646378551,
-      "loss": 0.6721,
-      "step": 12170
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016693168955500703,
-      "loss": 0.6683,
-      "step": 12180
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016690441447215896,
-      "loss": 0.8129,
-      "step": 12190
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016687713938931092,
-      "loss": 0.7866,
-      "step": 12200
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016684986430646283,
-      "loss": 0.6863,
-      "step": 12210
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001668225892236148,
-      "loss": 0.735,
-      "step": 12220
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001667953141407667,
-      "loss": 0.689,
-      "step": 12230
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016676803905791866,
-      "loss": 0.778,
-      "step": 12240
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001667407639750706,
-      "loss": 0.706,
-      "step": 12250
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016671348889222252,
-      "loss": 0.6689,
-      "step": 12260
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016668621380937446,
-      "loss": 0.8262,
-      "step": 12270
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001666589387265264,
-      "loss": 0.8251,
-      "step": 12280
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016663166364367833,
-      "loss": 0.7307,
-      "step": 12290
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016660438856083026,
-      "loss": 0.7572,
-      "step": 12300
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.0001665771134779822,
-      "loss": 0.6313,
-      "step": 12310
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016654983839513413,
-      "loss": 0.7543,
-      "step": 12320
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016652256331228606,
-      "loss": 0.6723,
-      "step": 12330
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.000166495288229438,
-      "loss": 0.7157,
-      "step": 12340
-    },
-    {
-      "epoch": 0.5,
-      "learning_rate": 0.00016646801314658996,
-      "loss": 0.7558,
-      "step": 12350
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016644073806374186,
-      "loss": 0.7304,
-      "step": 12360
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016641346298089382,
-      "loss": 0.6949,
-      "step": 12370
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016638618789804573,
-      "loss": 0.7574,
-      "step": 12380
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001663589128151977,
-      "loss": 0.7725,
-      "step": 12390
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001663316377323496,
-      "loss": 0.6592,
-      "step": 12400
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016630436264950156,
-      "loss": 0.8204,
-      "step": 12410
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001662770875666535,
-      "loss": 0.663,
-      "step": 12420
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016624981248380543,
-      "loss": 0.7181,
-      "step": 12430
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016622253740095736,
-      "loss": 0.7638,
-      "step": 12440
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001661952623181093,
-      "loss": 0.865,
-      "step": 12450
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016616798723526123,
-      "loss": 0.6977,
-      "step": 12460
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001661407121524132,
-      "loss": 0.6307,
-      "step": 12470
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001661134370695651,
-      "loss": 0.7025,
-      "step": 12480
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016608616198671706,
-      "loss": 0.7555,
-      "step": 12490
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016605888690386896,
-      "loss": 0.8146,
-      "step": 12500
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016603161182102093,
-      "loss": 0.7358,
-      "step": 12510
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016600433673817286,
-      "loss": 0.6151,
-      "step": 12520
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001659770616553248,
-      "loss": 0.7456,
-      "step": 12530
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016594978657247673,
-      "loss": 0.7272,
-      "step": 12540
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016592251148962866,
-      "loss": 0.7483,
-      "step": 12550
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001658952364067806,
-      "loss": 0.8566,
-      "step": 12560
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016586796132393253,
-      "loss": 0.6715,
-      "step": 12570
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.00016584068624108446,
-      "loss": 0.768,
-      "step": 12580
-    },
-    {
-      "epoch": 0.51,
-      "learning_rate": 0.0001658134111582364,
-      "loss": 0.7655,
-      "step": 12590
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016578613607538833,
-      "loss": 0.7568,
-      "step": 12600
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016575886099254027,
-      "loss": 0.7644,
-      "step": 12610
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001657315859096922,
-      "loss": 0.7307,
-      "step": 12620
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016570431082684413,
-      "loss": 0.7028,
-      "step": 12630
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001656770357439961,
-      "loss": 0.7862,
-      "step": 12640
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.000165649760661148,
-      "loss": 0.6771,
-      "step": 12650
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016562248557829996,
-      "loss": 0.7594,
-      "step": 12660
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016559521049545187,
-      "loss": 0.7173,
-      "step": 12670
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016556793541260383,
-      "loss": 0.7258,
-      "step": 12680
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016554066032975576,
-      "loss": 0.7076,
-      "step": 12690
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001655133852469077,
-      "loss": 0.6868,
-      "step": 12700
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016548611016405963,
-      "loss": 0.6584,
-      "step": 12710
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016545883508121157,
-      "loss": 0.7005,
-      "step": 12720
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001654315599983635,
-      "loss": 0.6864,
-      "step": 12730
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016540428491551543,
-      "loss": 0.7575,
-      "step": 12740
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016537700983266737,
-      "loss": 0.8042,
-      "step": 12750
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016534973474981933,
-      "loss": 0.7534,
-      "step": 12760
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016532245966697123,
-      "loss": 0.5861,
-      "step": 12770
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001652951845841232,
-      "loss": 0.6951,
-      "step": 12780
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001652679095012751,
-      "loss": 0.6927,
-      "step": 12790
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016524063441842706,
-      "loss": 0.8114,
-      "step": 12800
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.000165213359335579,
-      "loss": 0.6582,
-      "step": 12810
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016518608425273093,
-      "loss": 0.7388,
-      "step": 12820
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.00016515880916988287,
-      "loss": 0.8282,
-      "step": 12830
-    },
-    {
-      "epoch": 0.52,
-      "learning_rate": 0.0001651315340870348,
-      "loss": 0.8309,
-      "step": 12840
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016510425900418673,
-      "loss": 0.6529,
-      "step": 12850
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016507698392133867,
-      "loss": 0.731,
-      "step": 12860
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001650497088384906,
-      "loss": 0.776,
-      "step": 12870
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016502243375564254,
-      "loss": 0.7334,
-      "step": 12880
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016499515867279447,
-      "loss": 0.6963,
-      "step": 12890
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001649678835899464,
-      "loss": 0.8315,
-      "step": 12900
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016494060850709834,
-      "loss": 0.798,
-      "step": 12910
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016491333342425027,
-      "loss": 0.6982,
-      "step": 12920
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016488605834140223,
-      "loss": 0.7523,
-      "step": 12930
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016485878325855414,
-      "loss": 0.7196,
-      "step": 12940
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001648315081757061,
-      "loss": 0.7171,
-      "step": 12950
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.000164804233092858,
-      "loss": 0.8153,
-      "step": 12960
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016477695801000997,
-      "loss": 0.6359,
-      "step": 12970
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001647496829271619,
-      "loss": 0.6891,
-      "step": 12980
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016472240784431384,
-      "loss": 0.7077,
-      "step": 12990
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016469513276146577,
-      "loss": 0.6459,
-      "step": 13000
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001646678576786177,
-      "loss": 0.826,
-      "step": 13010
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016464058259576964,
-      "loss": 0.6456,
-      "step": 13020
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001646133075129216,
-      "loss": 0.6665,
-      "step": 13030
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.0001645860324300735,
-      "loss": 0.7081,
-      "step": 13040
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016455875734722547,
-      "loss": 0.8899,
-      "step": 13050
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016453148226437737,
-      "loss": 0.7663,
-      "step": 13060
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016450420718152933,
-      "loss": 0.6745,
-      "step": 13070
-    },
-    {
-      "epoch": 0.53,
-      "learning_rate": 0.00016447693209868124,
-      "loss": 0.7391,
-      "step": 13080
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001644496570158332,
-      "loss": 0.7858,
-      "step": 13090
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016442238193298514,
-      "loss": 0.7052,
-      "step": 13100
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016439510685013707,
-      "loss": 0.7436,
-      "step": 13110
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.000164367831767289,
-      "loss": 0.6955,
-      "step": 13120
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016434055668444094,
-      "loss": 0.7256,
-      "step": 13130
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016431328160159287,
-      "loss": 0.7355,
-      "step": 13140
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001642860065187448,
-      "loss": 0.7227,
-      "step": 13150
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016425873143589674,
-      "loss": 0.7409,
-      "step": 13160
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016423145635304867,
-      "loss": 0.7708,
-      "step": 13170
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001642041812702006,
-      "loss": 0.7996,
-      "step": 13180
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016417690618735254,
-      "loss": 0.6474,
-      "step": 13190
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001641496311045045,
-      "loss": 0.7906,
-      "step": 13200
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001641223560216564,
-      "loss": 0.6648,
-      "step": 13210
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016409508093880837,
-      "loss": 0.7577,
-      "step": 13220
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016406780585596028,
-      "loss": 0.6878,
-      "step": 13230
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016404053077311224,
-      "loss": 0.6956,
-      "step": 13240
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016401325569026414,
-      "loss": 0.7672,
-      "step": 13250
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001639859806074161,
-      "loss": 0.762,
-      "step": 13260
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016395870552456804,
-      "loss": 0.6665,
-      "step": 13270
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016393143044171997,
-      "loss": 0.8091,
-      "step": 13280
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.0001639041553588719,
-      "loss": 0.6739,
-      "step": 13290
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016387688027602384,
-      "loss": 0.7732,
-      "step": 13300
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016384960519317578,
-      "loss": 0.745,
-      "step": 13310
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016382233011032774,
-      "loss": 0.7656,
-      "step": 13320
-    },
-    {
-      "epoch": 0.54,
-      "learning_rate": 0.00016379505502747964,
-      "loss": 0.7873,
-      "step": 13330
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001637677799446316,
-      "loss": 0.8576,
-      "step": 13340
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001637405048617835,
-      "loss": 0.7738,
-      "step": 13350
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016371322977893547,
-      "loss": 0.7276,
-      "step": 13360
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001636859546960874,
-      "loss": 0.784,
-      "step": 13370
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016365867961323934,
-      "loss": 0.8401,
-      "step": 13380
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016363140453039127,
-      "loss": 0.7491,
-      "step": 13390
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001636041294475432,
-      "loss": 0.6576,
-      "step": 13400
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016357685436469514,
-      "loss": 0.7453,
-      "step": 13410
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016354957928184708,
-      "loss": 0.7648,
-      "step": 13420
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.000163522304198999,
-      "loss": 0.6725,
-      "step": 13430
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016349502911615094,
-      "loss": 0.7783,
-      "step": 13440
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016346775403330288,
-      "loss": 0.7444,
-      "step": 13450
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001634404789504548,
-      "loss": 0.7914,
-      "step": 13460
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016341320386760674,
-      "loss": 0.7197,
-      "step": 13470
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016338592878475868,
-      "loss": 0.7027,
-      "step": 13480
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016335865370191064,
-      "loss": 0.8391,
-      "step": 13490
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016333137861906255,
-      "loss": 0.7334,
-      "step": 13500
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001633041035362145,
-      "loss": 0.7269,
-      "step": 13510
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016327682845336641,
-      "loss": 0.6591,
-      "step": 13520
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016324955337051838,
-      "loss": 0.7188,
-      "step": 13530
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001632222782876703,
-      "loss": 0.7462,
-      "step": 13540
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016319500320482224,
-      "loss": 0.646,
-      "step": 13550
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.00016316772812197418,
-      "loss": 0.8132,
-      "step": 13560
-    },
-    {
-      "epoch": 0.55,
-      "learning_rate": 0.0001631404530391261,
-      "loss": 0.736,
-      "step": 13570
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016311317795627805,
-      "loss": 0.7134,
-      "step": 13580
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016308590287343,
-      "loss": 0.7161,
-      "step": 13590
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.0001630586277905819,
-      "loss": 0.727,
-      "step": 13600
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016303135270773387,
-      "loss": 0.8414,
-      "step": 13610
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016300407762488578,
-      "loss": 0.7477,
-      "step": 13620
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016297680254203774,
-      "loss": 0.7021,
-      "step": 13630
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016294952745918965,
-      "loss": 0.8004,
-      "step": 13640
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.0001629222523763416,
-      "loss": 0.7698,
-      "step": 13650
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016289497729349354,
-      "loss": 0.8907,
-      "step": 13660
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016286770221064548,
-      "loss": 0.7116,
-      "step": 13670
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.0001628404271277974,
-      "loss": 0.75,
-      "step": 13680
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016281315204494935,
-      "loss": 0.7809,
-      "step": 13690
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016278587696210128,
-      "loss": 0.7028,
-      "step": 13700
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.0001627586018792532,
-      "loss": 0.7097,
-      "step": 13710
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016273132679640515,
-      "loss": 0.7471,
-      "step": 13720
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016270405171355708,
-      "loss": 0.6898,
-      "step": 13730
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016267677663070901,
-      "loss": 0.7527,
-      "step": 13740
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016264950154786095,
-      "loss": 0.7201,
-      "step": 13750
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.0001626222264650129,
-      "loss": 0.7359,
-      "step": 13760
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016259495138216482,
-      "loss": 0.6306,
-      "step": 13770
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016256767629931678,
-      "loss": 0.6521,
-      "step": 13780
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016254040121646868,
-      "loss": 0.7796,
-      "step": 13790
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016251312613362065,
-      "loss": 0.7979,
-      "step": 13800
-    },
-    {
-      "epoch": 0.56,
-      "learning_rate": 0.00016248585105077255,
-      "loss": 0.7481,
-      "step": 13810
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.0001624585759679245,
-      "loss": 0.762,
-      "step": 13820
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016243130088507645,
-      "loss": 0.6946,
-      "step": 13830
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016240402580222838,
-      "loss": 0.6664,
-      "step": 13840
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016237675071938032,
-      "loss": 0.7065,
-      "step": 13850
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016234947563653225,
-      "loss": 0.7535,
-      "step": 13860
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016232220055368418,
-      "loss": 0.7626,
-      "step": 13870
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016229492547083614,
-      "loss": 0.7711,
-      "step": 13880
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016226765038798805,
-      "loss": 0.7428,
-      "step": 13890
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016224037530514,
-      "loss": 0.7157,
-      "step": 13900
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016221310022229192,
-      "loss": 0.6563,
-      "step": 13910
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016218582513944388,
-      "loss": 0.8019,
-      "step": 13920
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016215855005659581,
-      "loss": 0.709,
-      "step": 13930
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016213127497374775,
-      "loss": 0.6864,
-      "step": 13940
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016210399989089968,
-      "loss": 0.6924,
-      "step": 13950
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016207672480805162,
-      "loss": 0.7601,
-      "step": 13960
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016204944972520355,
-      "loss": 0.6401,
-      "step": 13970
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016202217464235548,
-      "loss": 0.7251,
-      "step": 13980
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016199489955950742,
-      "loss": 0.6555,
-      "step": 13990
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016196762447665935,
-      "loss": 0.6909,
-      "step": 14000
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016194034939381128,
-      "loss": 0.6992,
-      "step": 14010
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016191307431096322,
-      "loss": 0.6763,
-      "step": 14020
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016188579922811515,
-      "loss": 0.6777,
-      "step": 14030
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.0001618585241452671,
-      "loss": 0.6414,
-      "step": 14040
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016183124906241905,
-      "loss": 0.7635,
-      "step": 14050
-    },
-    {
-      "epoch": 0.57,
-      "learning_rate": 0.00016180397397957095,
-      "loss": 0.7689,
-      "step": 14060
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016177669889672292,
-      "loss": 0.6628,
-      "step": 14070
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016174942381387482,
-      "loss": 0.6743,
-      "step": 14080
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016172214873102678,
-      "loss": 0.7105,
-      "step": 14090
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016169487364817872,
-      "loss": 0.6585,
-      "step": 14100
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016166759856533065,
-      "loss": 0.8455,
-      "step": 14110
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016164032348248259,
-      "loss": 0.7503,
-      "step": 14120
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016161304839963452,
-      "loss": 0.8407,
-      "step": 14130
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016158577331678645,
-      "loss": 0.7039,
-      "step": 14140
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.0001615584982339384,
-      "loss": 0.6839,
-      "step": 14150
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016153122315109032,
-      "loss": 0.7551,
-      "step": 14160
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016150394806824228,
-      "loss": 0.7312,
-      "step": 14170
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.0001614766729853942,
-      "loss": 0.6186,
-      "step": 14180
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016144939790254615,
-      "loss": 0.8464,
-      "step": 14190
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016142212281969806,
-      "loss": 0.7465,
-      "step": 14200
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016139484773685002,
-      "loss": 0.8432,
-      "step": 14210
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016136757265400195,
-      "loss": 0.6799,
-      "step": 14220
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016134029757115389,
-      "loss": 0.7944,
-      "step": 14230
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016131302248830582,
-      "loss": 0.7127,
-      "step": 14240
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016128574740545775,
-      "loss": 0.6804,
-      "step": 14250
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.0001612584723226097,
-      "loss": 0.7987,
-      "step": 14260
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016123119723976162,
-      "loss": 0.6791,
-      "step": 14270
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016120392215691355,
-      "loss": 0.73,
-      "step": 14280
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.0001611766470740655,
-      "loss": 0.6821,
-      "step": 14290
-    },
-    {
-      "epoch": 0.58,
-      "learning_rate": 0.00016114937199121742,
-      "loss": 0.7868,
-      "step": 14300
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016112209690836936,
-      "loss": 0.6863,
-      "step": 14310
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016109482182552132,
-      "loss": 0.7532,
-      "step": 14320
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016106754674267322,
-      "loss": 0.7395,
-      "step": 14330
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016104027165982519,
-      "loss": 0.7826,
-      "step": 14340
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001610129965769771,
-      "loss": 0.7414,
-      "step": 14350
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016098572149412905,
-      "loss": 0.7257,
-      "step": 14360
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016095844641128096,
-      "loss": 0.7469,
-      "step": 14370
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016093117132843292,
-      "loss": 0.7608,
-      "step": 14380
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016090389624558486,
-      "loss": 0.6497,
-      "step": 14390
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001608766211627368,
-      "loss": 0.709,
-      "step": 14400
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016084934607988872,
-      "loss": 0.7378,
-      "step": 14410
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016082207099704066,
-      "loss": 0.7516,
-      "step": 14420
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001607947959141926,
-      "loss": 0.7067,
-      "step": 14430
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016076752083134455,
-      "loss": 0.7643,
-      "step": 14440
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016074024574849646,
-      "loss": 0.6636,
-      "step": 14450
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016071297066564842,
-      "loss": 0.8085,
-      "step": 14460
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016068569558280033,
-      "loss": 0.7118,
-      "step": 14470
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001606584204999523,
-      "loss": 0.8108,
-      "step": 14480
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016063114541710422,
-      "loss": 0.6912,
-      "step": 14490
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016060387033425616,
-      "loss": 0.7383,
-      "step": 14500
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001605765952514081,
-      "loss": 0.6578,
-      "step": 14510
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016054932016856002,
-      "loss": 0.7104,
-      "step": 14520
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016052204508571196,
-      "loss": 0.7005,
-      "step": 14530
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.0001604947700028639,
-      "loss": 0.7697,
-      "step": 14540
-    },
-    {
-      "epoch": 0.59,
-      "learning_rate": 0.00016046749492001583,
-      "loss": 0.7601,
-      "step": 14550
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016044021983716776,
-      "loss": 0.6624,
-      "step": 14560
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001604129447543197,
-      "loss": 0.6851,
-      "step": 14570
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016038566967147163,
-      "loss": 0.6885,
-      "step": 14580
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016035839458862356,
-      "loss": 0.7428,
-      "step": 14590
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001603311195057755,
-      "loss": 0.7335,
-      "step": 14600
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016030384442292746,
-      "loss": 0.7388,
-      "step": 14610
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016027656934007936,
-      "loss": 0.7332,
-      "step": 14620
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016024929425723132,
-      "loss": 0.6882,
-      "step": 14630
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016022201917438323,
-      "loss": 0.7381,
-      "step": 14640
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001601947440915352,
-      "loss": 0.7177,
-      "step": 14650
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016016746900868713,
-      "loss": 0.7355,
-      "step": 14660
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016014019392583906,
-      "loss": 0.7548,
-      "step": 14670
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.000160112918842991,
-      "loss": 0.6784,
-      "step": 14680
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016008564376014293,
-      "loss": 0.7836,
-      "step": 14690
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016005836867729486,
-      "loss": 0.778,
-      "step": 14700
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001600310935944468,
-      "loss": 0.6539,
-      "step": 14710
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00016000381851159873,
-      "loss": 0.9174,
-      "step": 14720
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001599765434287507,
-      "loss": 0.8509,
-      "step": 14730
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001599492683459026,
-      "loss": 0.7605,
-      "step": 14740
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00015992199326305456,
-      "loss": 0.6748,
-      "step": 14750
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00015989471818020646,
-      "loss": 0.7872,
-      "step": 14760
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00015986744309735843,
-      "loss": 0.6765,
-      "step": 14770
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.00015984016801451036,
-      "loss": 0.7101,
-      "step": 14780
-    },
-    {
-      "epoch": 0.6,
-      "learning_rate": 0.0001598128929316623,
-      "loss": 0.737,
-      "step": 14790
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015978561784881423,
-      "loss": 0.6066,
-      "step": 14800
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015975834276596616,
-      "loss": 0.7474,
-      "step": 14810
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001597310676831181,
-      "loss": 0.6513,
-      "step": 14820
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015970379260027003,
-      "loss": 0.7095,
-      "step": 14830
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015967651751742196,
-      "loss": 0.6932,
-      "step": 14840
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001596492424345739,
-      "loss": 0.7661,
-      "step": 14850
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015962196735172583,
-      "loss": 0.7204,
-      "step": 14860
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015959469226887776,
-      "loss": 0.6514,
-      "step": 14870
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001595674171860297,
-      "loss": 0.7599,
-      "step": 14880
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015954014210318163,
-      "loss": 0.7514,
-      "step": 14890
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001595128670203336,
-      "loss": 0.6555,
-      "step": 14900
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001594855919374855,
-      "loss": 0.7836,
-      "step": 14910
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015945831685463746,
-      "loss": 0.6842,
-      "step": 14920
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015943104177178937,
-      "loss": 0.7898,
-      "step": 14930
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015940376668894133,
-      "loss": 0.7939,
-      "step": 14940
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015937649160609326,
-      "loss": 0.7504,
-      "step": 14950
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001593492165232452,
-      "loss": 0.6913,
-      "step": 14960
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015932194144039713,
-      "loss": 0.7128,
-      "step": 14970
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015929466635754906,
-      "loss": 0.7692,
-      "step": 14980
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.000159267391274701,
-      "loss": 0.7826,
-      "step": 14990
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015924011619185296,
-      "loss": 0.8714,
-      "step": 15000
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015921284110900487,
-      "loss": 0.8244,
-      "step": 15010
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015918556602615683,
-      "loss": 0.7036,
-      "step": 15020
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.00015915829094330873,
-      "loss": 0.608,
-      "step": 15030
-    },
-    {
-      "epoch": 0.61,
-      "learning_rate": 0.0001591310158604607,
-      "loss": 0.692,
-      "step": 15040
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001591037407776126,
-      "loss": 0.6793,
-      "step": 15050
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015907646569476456,
-      "loss": 0.6979,
-      "step": 15060
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001590491906119165,
-      "loss": 0.6286,
-      "step": 15070
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015902191552906843,
-      "loss": 0.8094,
-      "step": 15080
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015899464044622037,
-      "loss": 0.7609,
-      "step": 15090
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001589673653633723,
-      "loss": 0.6964,
-      "step": 15100
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015894009028052423,
-      "loss": 0.7948,
-      "step": 15110
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015891281519767617,
-      "loss": 0.646,
-      "step": 15120
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001588855401148281,
-      "loss": 0.7292,
-      "step": 15130
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015885826503198003,
-      "loss": 0.7915,
-      "step": 15140
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015883098994913197,
-      "loss": 0.6453,
-      "step": 15150
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001588037148662839,
-      "loss": 0.7762,
-      "step": 15160
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015877643978343586,
-      "loss": 0.6296,
-      "step": 15170
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015874916470058777,
-      "loss": 0.789,
-      "step": 15180
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015872188961773973,
-      "loss": 0.6845,
-      "step": 15190
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015869461453489164,
-      "loss": 0.8686,
-      "step": 15200
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001586673394520436,
-      "loss": 0.697,
-      "step": 15210
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001586400643691955,
-      "loss": 0.7019,
-      "step": 15220
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015861278928634747,
-      "loss": 0.6996,
-      "step": 15230
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001585855142034994,
-      "loss": 0.7509,
-      "step": 15240
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015855823912065133,
-      "loss": 0.7452,
-      "step": 15250
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015853096403780327,
-      "loss": 0.7953,
-      "step": 15260
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.0001585036889549552,
-      "loss": 0.784,
-      "step": 15270
-    },
-    {
-      "epoch": 0.62,
-      "learning_rate": 0.00015847641387210714,
-      "loss": 0.8448,
-      "step": 15280
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001584491387892591,
-      "loss": 0.7353,
-      "step": 15290
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.000158421863706411,
-      "loss": 0.764,
-      "step": 15300
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015839458862356297,
-      "loss": 0.6522,
-      "step": 15310
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015836731354071487,
-      "loss": 0.6964,
-      "step": 15320
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015834003845786683,
-      "loss": 0.7595,
-      "step": 15330
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015831276337501877,
-      "loss": 0.8126,
-      "step": 15340
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001582854882921707,
-      "loss": 0.6704,
-      "step": 15350
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015825821320932264,
-      "loss": 0.663,
-      "step": 15360
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015823093812647457,
-      "loss": 0.6471,
-      "step": 15370
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001582036630436265,
-      "loss": 0.8409,
-      "step": 15380
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015817638796077844,
-      "loss": 0.659,
-      "step": 15390
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015814911287793037,
-      "loss": 0.6894,
-      "step": 15400
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001581218377950823,
-      "loss": 0.759,
-      "step": 15410
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015809456271223424,
-      "loss": 0.8249,
-      "step": 15420
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015806728762938617,
-      "loss": 0.6454,
-      "step": 15430
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001580400125465381,
-      "loss": 0.7419,
-      "step": 15440
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015801273746369004,
-      "loss": 0.7235,
-      "step": 15450
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.000157985462380842,
-      "loss": 0.7503,
-      "step": 15460
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001579581872979939,
-      "loss": 0.8264,
-      "step": 15470
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015793091221514587,
-      "loss": 0.7077,
-      "step": 15480
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015790363713229778,
-      "loss": 0.8071,
-      "step": 15490
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015787636204944974,
-      "loss": 0.6754,
-      "step": 15500
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015784908696660167,
-      "loss": 0.7276,
-      "step": 15510
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.0001578218118837536,
-      "loss": 0.7928,
-      "step": 15520
-    },
-    {
-      "epoch": 0.63,
-      "learning_rate": 0.00015779453680090554,
-      "loss": 0.8141,
-      "step": 15530
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015776726171805747,
-      "loss": 0.719,
-      "step": 15540
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001577399866352094,
-      "loss": 0.7938,
-      "step": 15550
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015771271155236137,
-      "loss": 0.7299,
-      "step": 15560
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015768543646951327,
-      "loss": 0.6747,
-      "step": 15570
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015765816138666524,
-      "loss": 0.6771,
-      "step": 15580
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015763088630381714,
-      "loss": 0.7579,
-      "step": 15590
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001576036112209691,
-      "loss": 0.6548,
-      "step": 15600
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.000157576336138121,
-      "loss": 0.7686,
-      "step": 15610
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015754906105527297,
-      "loss": 0.7387,
-      "step": 15620
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001575217859724249,
-      "loss": 0.7443,
-      "step": 15630
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015749451088957684,
-      "loss": 0.6802,
-      "step": 15640
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015746723580672877,
-      "loss": 0.7195,
-      "step": 15650
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001574399607238807,
-      "loss": 0.7289,
-      "step": 15660
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015741268564103264,
-      "loss": 0.7718,
-      "step": 15670
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015738541055818457,
-      "loss": 0.6023,
-      "step": 15680
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001573581354753365,
-      "loss": 0.7369,
-      "step": 15690
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015733086039248844,
-      "loss": 0.7521,
-      "step": 15700
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015730358530964038,
-      "loss": 0.794,
-      "step": 15710
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.0001572763102267923,
-      "loss": 0.7348,
-      "step": 15720
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015724903514394427,
-      "loss": 0.7115,
-      "step": 15730
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015722176006109618,
-      "loss": 0.7667,
-      "step": 15740
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015719448497824814,
-      "loss": 0.7593,
-      "step": 15750
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.00015716720989540005,
-      "loss": 0.6902,
-      "step": 15760
-    },
-    {
-      "epoch": 0.64,
-      "learning_rate": 0.000157139934812552,
-      "loss": 0.8187,
-      "step": 15770
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015711265972970391,
-      "loss": 0.6655,
-      "step": 15780
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015708538464685588,
-      "loss": 0.7303,
-      "step": 15790
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001570581095640078,
-      "loss": 0.7247,
-      "step": 15800
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015703083448115974,
-      "loss": 0.7357,
-      "step": 15810
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015700355939831168,
-      "loss": 0.7158,
-      "step": 15820
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001569762843154636,
-      "loss": 0.6481,
-      "step": 15830
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015694900923261554,
-      "loss": 0.7263,
-      "step": 15840
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001569217341497675,
-      "loss": 0.7764,
-      "step": 15850
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001568944590669194,
-      "loss": 0.7978,
-      "step": 15860
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015686718398407137,
-      "loss": 0.8003,
-      "step": 15870
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015683990890122328,
-      "loss": 0.7913,
-      "step": 15880
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015681263381837524,
-      "loss": 0.8312,
-      "step": 15890
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015678535873552718,
-      "loss": 0.8242,
-      "step": 15900
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001567580836526791,
-      "loss": 0.6646,
-      "step": 15910
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015673080856983104,
-      "loss": 0.7493,
-      "step": 15920
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015670353348698298,
-      "loss": 0.6784,
-      "step": 15930
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001566762584041349,
-      "loss": 0.7355,
-      "step": 15940
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015664898332128684,
-      "loss": 0.8048,
-      "step": 15950
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015662170823843878,
-      "loss": 0.6915,
-      "step": 15960
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001565944331555907,
-      "loss": 0.759,
-      "step": 15970
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015656715807274265,
-      "loss": 0.7607,
-      "step": 15980
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015653988298989458,
-      "loss": 0.7123,
-      "step": 15990
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015651260790704651,
-      "loss": 0.7979,
-      "step": 16000
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.00015648533282419845,
-      "loss": 0.7695,
-      "step": 16010
-    },
-    {
-      "epoch": 0.65,
-      "learning_rate": 0.0001564580577413504,
-      "loss": 0.7093,
-      "step": 16020
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015643078265850232,
-      "loss": 0.7746,
-      "step": 16030
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015640350757565428,
-      "loss": 0.7797,
-      "step": 16040
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015637623249280618,
-      "loss": 0.6744,
-      "step": 16050
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015634895740995815,
-      "loss": 0.7474,
-      "step": 16060
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015632168232711008,
-      "loss": 0.8192,
-      "step": 16070
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.000156294407244262,
-      "loss": 0.7901,
-      "step": 16080
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015626713216141395,
-      "loss": 0.6892,
-      "step": 16090
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015623985707856588,
-      "loss": 0.7011,
-      "step": 16100
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015621258199571781,
-      "loss": 0.7932,
-      "step": 16110
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015618530691286975,
-      "loss": 0.7019,
-      "step": 16120
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015615803183002168,
-      "loss": 0.6904,
-      "step": 16130
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015613075674717364,
-      "loss": 0.7034,
-      "step": 16140
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015610348166432555,
-      "loss": 0.7735,
-      "step": 16150
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.0001560762065814775,
-      "loss": 0.8221,
-      "step": 16160
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015604893149862942,
-      "loss": 0.8069,
-      "step": 16170
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015602165641578138,
-      "loss": 0.6877,
-      "step": 16180
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.0001559943813329333,
-      "loss": 0.7357,
-      "step": 16190
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015596710625008525,
-      "loss": 0.7293,
-      "step": 16200
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015593983116723718,
-      "loss": 0.6491,
-      "step": 16210
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015591255608438911,
-      "loss": 0.8723,
-      "step": 16220
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015588528100154105,
-      "loss": 0.7285,
-      "step": 16230
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015585800591869298,
-      "loss": 0.7187,
-      "step": 16240
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015583073083584492,
-      "loss": 0.748,
-      "step": 16250
-    },
-    {
-      "epoch": 0.66,
-      "learning_rate": 0.00015580345575299685,
-      "loss": 0.7334,
-      "step": 16260
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015577618067014878,
-      "loss": 0.704,
-      "step": 16270
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015574890558730072,
-      "loss": 0.6556,
-      "step": 16280
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015572163050445268,
-      "loss": 0.7791,
-      "step": 16290
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015569435542160459,
-      "loss": 0.8125,
-      "step": 16300
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015566708033875655,
-      "loss": 0.711,
-      "step": 16310
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015563980525590845,
-      "loss": 0.7581,
-      "step": 16320
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015561253017306042,
-      "loss": 0.6835,
-      "step": 16330
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015558525509021232,
-      "loss": 0.7391,
-      "step": 16340
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015555798000736428,
-      "loss": 0.7347,
-      "step": 16350
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015553070492451622,
-      "loss": 0.7668,
-      "step": 16360
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015550342984166815,
-      "loss": 0.8037,
-      "step": 16370
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015547615475882008,
-      "loss": 0.7003,
-      "step": 16380
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015544887967597202,
-      "loss": 0.7145,
-      "step": 16390
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015542160459312395,
-      "loss": 0.6823,
-      "step": 16400
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015539432951027591,
-      "loss": 0.6675,
-      "step": 16410
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015536705442742782,
-      "loss": 0.7394,
-      "step": 16420
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015533977934457978,
-      "loss": 0.7918,
-      "step": 16430
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.0001553125042617317,
-      "loss": 0.6519,
-      "step": 16440
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015528522917888365,
-      "loss": 0.7321,
-      "step": 16450
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015525795409603558,
-      "loss": 0.7262,
-      "step": 16460
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015523067901318752,
-      "loss": 0.7034,
-      "step": 16470
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015520340393033945,
-      "loss": 0.6764,
-      "step": 16480
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015517612884749138,
-      "loss": 0.7839,
-      "step": 16490
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015514885376464332,
-      "loss": 0.6395,
-      "step": 16500
-    },
-    {
-      "epoch": 0.67,
-      "learning_rate": 0.00015512157868179525,
-      "loss": 0.7425,
-      "step": 16510
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001550943035989472,
-      "loss": 0.6868,
-      "step": 16520
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015506702851609912,
-      "loss": 0.7494,
-      "step": 16530
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015503975343325105,
-      "loss": 0.7003,
-      "step": 16540
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.000155012478350403,
-      "loss": 0.7181,
-      "step": 16550
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015498520326755492,
-      "loss": 0.7077,
-      "step": 16560
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015495792818470686,
-      "loss": 0.7845,
-      "step": 16570
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015493065310185882,
-      "loss": 0.7715,
-      "step": 16580
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015490337801901072,
-      "loss": 0.6853,
-      "step": 16590
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015487610293616269,
-      "loss": 0.6554,
-      "step": 16600
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001548488278533146,
-      "loss": 0.7987,
-      "step": 16610
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015482155277046655,
-      "loss": 0.6671,
-      "step": 16620
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001547942776876185,
-      "loss": 0.6671,
-      "step": 16630
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015476700260477042,
-      "loss": 0.7374,
-      "step": 16640
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015473972752192235,
-      "loss": 0.6347,
-      "step": 16650
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001547124524390743,
-      "loss": 0.7549,
-      "step": 16660
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015468517735622622,
-      "loss": 0.7376,
-      "step": 16670
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015465790227337816,
-      "loss": 0.6205,
-      "step": 16680
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001546306271905301,
-      "loss": 0.5919,
-      "step": 16690
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015460335210768205,
-      "loss": 0.8279,
-      "step": 16700
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015457607702483396,
-      "loss": 0.673,
-      "step": 16710
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015454880194198592,
-      "loss": 0.7835,
-      "step": 16720
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015452152685913783,
-      "loss": 0.8632,
-      "step": 16730
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.0001544942517762898,
-      "loss": 0.8187,
-      "step": 16740
-    },
-    {
-      "epoch": 0.68,
-      "learning_rate": 0.00015446697669344172,
-      "loss": 0.7756,
-      "step": 16750
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015443970161059366,
-      "loss": 0.6758,
-      "step": 16760
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.0001544124265277456,
-      "loss": 0.7641,
-      "step": 16770
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015438515144489752,
-      "loss": 0.7624,
-      "step": 16780
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015435787636204946,
-      "loss": 0.6512,
-      "step": 16790
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.0001543306012792014,
-      "loss": 0.8512,
-      "step": 16800
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015430332619635332,
-      "loss": 0.771,
-      "step": 16810
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015427605111350526,
-      "loss": 0.813,
-      "step": 16820
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.0001542487760306572,
-      "loss": 0.7269,
-      "step": 16830
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015422150094780913,
-      "loss": 0.657,
-      "step": 16840
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015419422586496106,
-      "loss": 0.7938,
-      "step": 16850
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.000154166950782113,
-      "loss": 0.8683,
-      "step": 16860
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015413967569926496,
-      "loss": 0.7107,
-      "step": 16870
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015411240061641686,
-      "loss": 0.8317,
-      "step": 16880
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015408512553356882,
-      "loss": 0.6824,
-      "step": 16890
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015406057795900555,
-      "loss": 0.6715,
-      "step": 16900
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015403330287615748,
-      "loss": 0.6945,
-      "step": 16910
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015400602779330945,
-      "loss": 0.7244,
-      "step": 16920
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015397875271046135,
-      "loss": 0.7008,
-      "step": 16930
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.0001539514776276133,
-      "loss": 0.8177,
-      "step": 16940
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015392420254476522,
-      "loss": 0.7434,
-      "step": 16950
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015389692746191718,
-      "loss": 0.7349,
-      "step": 16960
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015386965237906911,
-      "loss": 0.6761,
-      "step": 16970
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015384237729622105,
-      "loss": 0.6346,
-      "step": 16980
-    },
-    {
-      "epoch": 0.69,
-      "learning_rate": 0.00015381510221337298,
-      "loss": 0.5948,
-      "step": 16990
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015378782713052492,
-      "loss": 0.727,
-      "step": 17000
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015376055204767685,
-      "loss": 0.689,
-      "step": 17010
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.0001537332769648288,
-      "loss": 0.6917,
-      "step": 17020
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015370600188198072,
-      "loss": 0.7057,
-      "step": 17030
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015367872679913268,
-      "loss": 0.8051,
-      "step": 17040
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015365145171628459,
-      "loss": 0.7088,
-      "step": 17050
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015362417663343655,
-      "loss": 0.8042,
-      "step": 17060
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015359690155058845,
-      "loss": 0.7789,
-      "step": 17070
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015356962646774041,
-      "loss": 0.7838,
-      "step": 17080
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015354235138489235,
-      "loss": 0.7274,
-      "step": 17090
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015351507630204428,
-      "loss": 0.7557,
-      "step": 17100
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015348780121919622,
-      "loss": 0.7853,
-      "step": 17110
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015346052613634815,
-      "loss": 0.6625,
-      "step": 17120
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015343325105350008,
-      "loss": 0.675,
-      "step": 17130
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015340597597065202,
-      "loss": 0.7464,
-      "step": 17140
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015337870088780395,
-      "loss": 0.7056,
-      "step": 17150
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015335142580495589,
-      "loss": 0.6987,
-      "step": 17160
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015332415072210782,
-      "loss": 0.7282,
-      "step": 17170
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015329687563925975,
-      "loss": 0.7068,
-      "step": 17180
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015326960055641172,
-      "loss": 0.704,
-      "step": 17190
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015324232547356362,
-      "loss": 0.6726,
-      "step": 17200
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015321505039071558,
-      "loss": 0.8013,
-      "step": 17210
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.0001531877753078675,
-      "loss": 0.6417,
-      "step": 17220
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015316050022501945,
-      "loss": 0.6909,
-      "step": 17230
-    },
-    {
-      "epoch": 0.7,
-      "learning_rate": 0.00015313322514217136,
-      "loss": 0.7073,
-      "step": 17240
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015310595005932332,
-      "loss": 0.6489,
-      "step": 17250
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015307867497647525,
-      "loss": 0.7406,
-      "step": 17260
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015305139989362719,
-      "loss": 0.763,
-      "step": 17270
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015302412481077912,
-      "loss": 0.7797,
-      "step": 17280
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015299684972793105,
-      "loss": 0.7496,
-      "step": 17290
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.000152969574645083,
-      "loss": 0.7595,
-      "step": 17300
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015294229956223495,
-      "loss": 0.6548,
-      "step": 17310
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015291502447938686,
-      "loss": 0.7826,
-      "step": 17320
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015288774939653882,
-      "loss": 0.7032,
-      "step": 17330
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015286047431369072,
-      "loss": 0.7885,
-      "step": 17340
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015283319923084268,
-      "loss": 0.8244,
-      "step": 17350
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015280592414799462,
-      "loss": 0.8216,
-      "step": 17360
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015277864906514655,
-      "loss": 0.6614,
-      "step": 17370
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.0001527513739822985,
-      "loss": 0.7333,
-      "step": 17380
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015272409889945042,
-      "loss": 0.7814,
-      "step": 17390
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015269682381660235,
-      "loss": 0.7738,
-      "step": 17400
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.0001526695487337543,
-      "loss": 0.6814,
-      "step": 17410
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015264227365090622,
-      "loss": 0.8202,
-      "step": 17420
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015261499856805816,
-      "loss": 0.707,
-      "step": 17430
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.0001525877234852101,
-      "loss": 0.7408,
-      "step": 17440
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015256044840236202,
-      "loss": 0.7694,
-      "step": 17450
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015253317331951396,
-      "loss": 0.7723,
-      "step": 17460
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.0001525058982366659,
-      "loss": 0.6474,
-      "step": 17470
-    },
-    {
-      "epoch": 0.71,
-      "learning_rate": 0.00015247862315381785,
-      "loss": 0.713,
-      "step": 17480
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015245134807096976,
-      "loss": 0.7433,
-      "step": 17490
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015242407298812172,
-      "loss": 0.7309,
-      "step": 17500
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015239679790527363,
-      "loss": 0.7339,
-      "step": 17510
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001523695228224256,
-      "loss": 0.647,
-      "step": 17520
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015234224773957752,
-      "loss": 0.7171,
-      "step": 17530
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015231497265672946,
-      "loss": 0.7666,
-      "step": 17540
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001522876975738814,
-      "loss": 0.744,
-      "step": 17550
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015226042249103332,
-      "loss": 0.6781,
-      "step": 17560
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015223314740818526,
-      "loss": 0.7109,
-      "step": 17570
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015220587232533722,
-      "loss": 0.7629,
-      "step": 17580
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015217859724248913,
-      "loss": 0.7538,
-      "step": 17590
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001521513221596411,
-      "loss": 0.8057,
-      "step": 17600
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.000152124047076793,
-      "loss": 0.7191,
-      "step": 17610
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015209677199394495,
-      "loss": 0.7395,
-      "step": 17620
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015206949691109686,
-      "loss": 0.7545,
-      "step": 17630
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015204222182824882,
-      "loss": 0.721,
-      "step": 17640
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015201494674540076,
-      "loss": 0.7007,
-      "step": 17650
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001519876716625527,
-      "loss": 0.8453,
-      "step": 17660
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015196039657970462,
-      "loss": 0.6695,
-      "step": 17670
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015193312149685656,
-      "loss": 0.844,
-      "step": 17680
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001519058464140085,
-      "loss": 0.7657,
-      "step": 17690
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015187857133116043,
-      "loss": 0.727,
-      "step": 17700
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015185129624831236,
-      "loss": 0.7059,
-      "step": 17710
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.0001518240211654643,
-      "loss": 0.726,
-      "step": 17720
-    },
-    {
-      "epoch": 0.72,
-      "learning_rate": 0.00015179674608261623,
-      "loss": 0.7527,
-      "step": 17730
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015176947099976816,
-      "loss": 0.7735,
-      "step": 17740
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015174219591692012,
-      "loss": 0.6531,
-      "step": 17750
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015171492083407203,
-      "loss": 0.799,
-      "step": 17760
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.000151687645751224,
-      "loss": 0.7077,
-      "step": 17770
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001516603706683759,
-      "loss": 0.7112,
-      "step": 17780
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015163309558552786,
-      "loss": 0.7547,
-      "step": 17790
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015160582050267977,
-      "loss": 0.7848,
-      "step": 17800
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015157854541983173,
-      "loss": 0.6983,
-      "step": 17810
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015155127033698366,
-      "loss": 0.784,
-      "step": 17820
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001515239952541356,
-      "loss": 0.7814,
-      "step": 17830
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015149672017128753,
-      "loss": 0.6989,
-      "step": 17840
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015146944508843946,
-      "loss": 0.7057,
-      "step": 17850
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001514421700055914,
-      "loss": 0.6621,
-      "step": 17860
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015141489492274336,
-      "loss": 0.7439,
-      "step": 17870
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015138761983989526,
-      "loss": 0.8142,
-      "step": 17880
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015136034475704723,
-      "loss": 0.6518,
-      "step": 17890
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015133306967419913,
-      "loss": 0.6888,
-      "step": 17900
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001513057945913511,
-      "loss": 0.7798,
-      "step": 17910
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015127851950850303,
-      "loss": 0.8026,
-      "step": 17920
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015125124442565496,
-      "loss": 0.7742,
-      "step": 17930
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001512239693428069,
-      "loss": 0.7127,
-      "step": 17940
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015119669425995883,
-      "loss": 0.74,
-      "step": 17950
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.00015116941917711076,
-      "loss": 0.8706,
-      "step": 17960
-    },
-    {
-      "epoch": 0.73,
-      "learning_rate": 0.0001511421440942627,
-      "loss": 0.8217,
-      "step": 17970
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015111486901141463,
-      "loss": 0.648,
-      "step": 17980
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015108759392856656,
-      "loss": 0.7907,
-      "step": 17990
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001510603188457185,
-      "loss": 0.6666,
-      "step": 18000
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015103304376287043,
-      "loss": 0.6278,
-      "step": 18010
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015100576868002237,
-      "loss": 0.7282,
-      "step": 18020
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001509784935971743,
-      "loss": 0.7618,
-      "step": 18030
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015095121851432626,
-      "loss": 0.6412,
-      "step": 18040
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015092394343147817,
-      "loss": 0.6673,
-      "step": 18050
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015089666834863013,
-      "loss": 0.7847,
-      "step": 18060
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015086939326578204,
-      "loss": 0.7141,
-      "step": 18070
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.000150842118182934,
-      "loss": 0.8098,
-      "step": 18080
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015081484310008593,
-      "loss": 0.7345,
-      "step": 18090
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015078756801723786,
-      "loss": 0.7402,
-      "step": 18100
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001507602929343898,
-      "loss": 0.8067,
-      "step": 18110
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015073301785154173,
-      "loss": 0.7915,
-      "step": 18120
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015070574276869367,
-      "loss": 0.6441,
-      "step": 18130
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001506784676858456,
-      "loss": 0.7171,
-      "step": 18140
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015065119260299753,
-      "loss": 0.7246,
-      "step": 18150
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001506239175201495,
-      "loss": 0.7707,
-      "step": 18160
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001505966424373014,
-      "loss": 0.7629,
-      "step": 18170
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015056936735445336,
-      "loss": 0.7569,
-      "step": 18180
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015054209227160527,
-      "loss": 0.7003,
-      "step": 18190
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015051481718875723,
-      "loss": 0.8308,
-      "step": 18200
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.00015048754210590916,
-      "loss": 0.8145,
-      "step": 18210
-    },
-    {
-      "epoch": 0.74,
-      "learning_rate": 0.0001504602670230611,
-      "loss": 0.6925,
-      "step": 18220
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015043299194021303,
-      "loss": 0.727,
-      "step": 18230
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015040571685736497,
-      "loss": 0.7037,
-      "step": 18240
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001503784417745169,
-      "loss": 0.7558,
-      "step": 18250
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015035116669166883,
-      "loss": 0.6799,
-      "step": 18260
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015032389160882077,
-      "loss": 0.6909,
-      "step": 18270
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001502966165259727,
-      "loss": 0.7625,
-      "step": 18280
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015026934144312464,
-      "loss": 0.7085,
-      "step": 18290
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015024206636027657,
-      "loss": 0.7257,
-      "step": 18300
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001502147912774285,
-      "loss": 0.752,
-      "step": 18310
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015018751619458044,
-      "loss": 0.7113,
-      "step": 18320
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001501602411117324,
-      "loss": 0.7269,
-      "step": 18330
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001501329660288843,
-      "loss": 0.736,
-      "step": 18340
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015010569094603627,
-      "loss": 0.8128,
-      "step": 18350
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015007841586318817,
-      "loss": 0.6542,
-      "step": 18360
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015005114078034013,
-      "loss": 0.8337,
-      "step": 18370
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00015002386569749207,
-      "loss": 0.6377,
-      "step": 18380
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.000149996590614644,
-      "loss": 0.7268,
-      "step": 18390
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014996931553179594,
-      "loss": 0.682,
-      "step": 18400
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014994204044894787,
-      "loss": 0.6954,
-      "step": 18410
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.0001499147653660998,
-      "loss": 0.802,
-      "step": 18420
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014988749028325177,
-      "loss": 0.7913,
-      "step": 18430
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014986021520040367,
-      "loss": 0.712,
-      "step": 18440
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014983294011755563,
-      "loss": 0.6997,
-      "step": 18450
-    },
-    {
-      "epoch": 0.75,
-      "learning_rate": 0.00014980566503470754,
-      "loss": 0.7574,
-      "step": 18460
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001497783899518595,
-      "loss": 0.7378,
-      "step": 18470
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001497511148690114,
-      "loss": 0.7143,
-      "step": 18480
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014972383978616337,
-      "loss": 0.7691,
-      "step": 18490
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001496965647033153,
-      "loss": 0.7045,
-      "step": 18500
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014966928962046724,
-      "loss": 0.6385,
-      "step": 18510
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014964201453761917,
-      "loss": 0.8171,
-      "step": 18520
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001496147394547711,
-      "loss": 0.7269,
-      "step": 18530
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014958746437192304,
-      "loss": 0.679,
-      "step": 18540
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014956018928907497,
-      "loss": 0.7094,
-      "step": 18550
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001495329142062269,
-      "loss": 0.6408,
-      "step": 18560
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014950563912337884,
-      "loss": 0.8107,
-      "step": 18570
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014947836404053077,
-      "loss": 0.7246,
-      "step": 18580
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001494510889576827,
-      "loss": 0.7292,
-      "step": 18590
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014942381387483467,
-      "loss": 0.7087,
-      "step": 18600
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014939653879198658,
-      "loss": 0.733,
-      "step": 18610
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014936926370913854,
-      "loss": 0.7144,
-      "step": 18620
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014934198862629044,
-      "loss": 0.6281,
-      "step": 18630
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001493147135434424,
-      "loss": 0.6901,
-      "step": 18640
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014928743846059434,
-      "loss": 0.7098,
-      "step": 18650
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014926016337774627,
-      "loss": 0.8178,
-      "step": 18660
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.0001492328882948982,
-      "loss": 0.7664,
-      "step": 18670
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014920561321205014,
-      "loss": 0.7186,
-      "step": 18680
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014917833812920207,
-      "loss": 0.7013,
-      "step": 18690
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.000149151063046354,
-      "loss": 0.7441,
-      "step": 18700
-    },
-    {
-      "epoch": 0.76,
-      "learning_rate": 0.00014912378796350594,
-      "loss": 0.7602,
-      "step": 18710
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001490965128806579,
-      "loss": 0.8591,
-      "step": 18720
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001490692377978098,
-      "loss": 0.8006,
-      "step": 18730
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014904196271496177,
-      "loss": 0.7021,
-      "step": 18740
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014901468763211368,
-      "loss": 0.6865,
-      "step": 18750
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014898741254926564,
-      "loss": 0.7456,
-      "step": 18760
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014896013746641757,
-      "loss": 0.7929,
-      "step": 18770
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001489328623835695,
-      "loss": 0.7526,
-      "step": 18780
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014890558730072144,
-      "loss": 0.7754,
-      "step": 18790
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014887831221787337,
-      "loss": 0.7892,
-      "step": 18800
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001488510371350253,
-      "loss": 0.7683,
-      "step": 18810
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014882376205217724,
-      "loss": 0.7226,
-      "step": 18820
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014879648696932918,
-      "loss": 0.6825,
-      "step": 18830
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001487692118864811,
-      "loss": 0.8509,
-      "step": 18840
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014874193680363304,
-      "loss": 0.7351,
-      "step": 18850
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014871466172078498,
-      "loss": 0.7915,
-      "step": 18860
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001486873866379369,
-      "loss": 0.7335,
-      "step": 18870
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014866011155508885,
-      "loss": 0.6575,
-      "step": 18880
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001486328364722408,
-      "loss": 0.6636,
-      "step": 18890
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001486055613893927,
-      "loss": 0.731,
-      "step": 18900
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014857828630654467,
-      "loss": 0.7194,
-      "step": 18910
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014855101122369658,
-      "loss": 0.7155,
-      "step": 18920
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014852373614084854,
-      "loss": 0.7731,
-      "step": 18930
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.00014849646105800048,
-      "loss": 0.673,
-      "step": 18940
-    },
-    {
-      "epoch": 0.77,
-      "learning_rate": 0.0001484691859751524,
-      "loss": 0.8003,
-      "step": 18950
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014844191089230434,
-      "loss": 0.6806,
-      "step": 18960
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014841463580945628,
-      "loss": 0.7765,
-      "step": 18970
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.0001483873607266082,
-      "loss": 0.7774,
-      "step": 18980
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014836008564376017,
-      "loss": 0.7581,
-      "step": 18990
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014833281056091208,
-      "loss": 0.7496,
-      "step": 19000
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014830553547806404,
-      "loss": 0.7212,
-      "step": 19010
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014827826039521595,
-      "loss": 0.6967,
-      "step": 19020
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.0001482509853123679,
-      "loss": 0.7427,
-      "step": 19030
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014822371022951982,
-      "loss": 0.7656,
-      "step": 19040
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014819643514667178,
-      "loss": 0.7844,
-      "step": 19050
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.0001481691600638237,
-      "loss": 0.7346,
-      "step": 19060
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014814188498097564,
-      "loss": 0.7003,
-      "step": 19070
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014811460989812758,
-      "loss": 0.7488,
-      "step": 19080
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.0001480873348152795,
-      "loss": 0.7939,
-      "step": 19090
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014806005973243145,
-      "loss": 0.7511,
-      "step": 19100
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014803278464958338,
-      "loss": 0.6774,
-      "step": 19110
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014800550956673531,
-      "loss": 0.7242,
-      "step": 19120
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014797823448388725,
-      "loss": 0.6666,
-      "step": 19130
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014795095940103918,
-      "loss": 0.6828,
-      "step": 19140
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014792368431819112,
-      "loss": 0.6673,
-      "step": 19150
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014789640923534308,
-      "loss": 0.6954,
-      "step": 19160
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014786913415249498,
-      "loss": 0.7081,
-      "step": 19170
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014784185906964694,
-      "loss": 0.6791,
-      "step": 19180
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.00014781458398679885,
-      "loss": 0.8634,
-      "step": 19190
-    },
-    {
-      "epoch": 0.78,
-      "learning_rate": 0.0001477900364122356,
-      "loss": 0.7821,
-      "step": 19200
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014776276132938757,
-      "loss": 0.7641,
-      "step": 19210
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014773548624653947,
-      "loss": 0.6815,
-      "step": 19220
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014770821116369143,
-      "loss": 0.6379,
-      "step": 19230
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014768093608084334,
-      "loss": 0.7646,
-      "step": 19240
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001476536609979953,
-      "loss": 0.733,
-      "step": 19250
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001476263859151472,
-      "loss": 0.6661,
-      "step": 19260
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014759911083229917,
-      "loss": 0.6948,
-      "step": 19270
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001475718357494511,
-      "loss": 0.7549,
-      "step": 19280
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014754456066660304,
-      "loss": 0.7902,
-      "step": 19290
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014751728558375497,
-      "loss": 0.7751,
-      "step": 19300
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001474900105009069,
-      "loss": 0.7495,
-      "step": 19310
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014746273541805884,
-      "loss": 0.6827,
-      "step": 19320
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014743546033521077,
-      "loss": 0.7478,
-      "step": 19330
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001474081852523627,
-      "loss": 0.7824,
-      "step": 19340
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014738091016951464,
-      "loss": 0.748,
-      "step": 19350
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014735363508666658,
-      "loss": 0.6375,
-      "step": 19360
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001473263600038185,
-      "loss": 0.6242,
-      "step": 19370
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014729908492097047,
-      "loss": 0.6303,
-      "step": 19380
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014727180983812238,
-      "loss": 0.7386,
-      "step": 19390
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014724453475527434,
-      "loss": 0.7988,
-      "step": 19400
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014721725967242624,
-      "loss": 0.7597,
-      "step": 19410
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001471899845895782,
-      "loss": 0.6665,
-      "step": 19420
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.0001471627095067301,
-      "loss": 0.731,
-      "step": 19430
-    },
-    {
-      "epoch": 0.79,
-      "learning_rate": 0.00014713543442388207,
-      "loss": 0.7377,
-      "step": 19440
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.000147108159341034,
-      "loss": 0.735,
-      "step": 19450
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014708088425818594,
-      "loss": 0.6892,
-      "step": 19460
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014705360917533788,
-      "loss": 0.699,
-      "step": 19470
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001470263340924898,
-      "loss": 0.6353,
-      "step": 19480
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014699905900964174,
-      "loss": 0.7556,
-      "step": 19490
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001469717839267937,
-      "loss": 0.741,
-      "step": 19500
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001469445088439456,
-      "loss": 0.8859,
-      "step": 19510
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014691723376109757,
-      "loss": 0.6544,
-      "step": 19520
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001468926861865343,
-      "loss": 0.7687,
-      "step": 19530
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014686541110368623,
-      "loss": 0.8513,
-      "step": 19540
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014683813602083817,
-      "loss": 0.7144,
-      "step": 19550
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001468108609379901,
-      "loss": 0.777,
-      "step": 19560
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014678358585514203,
-      "loss": 0.7429,
-      "step": 19570
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014675631077229397,
-      "loss": 0.7243,
-      "step": 19580
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001467290356894459,
-      "loss": 0.8024,
-      "step": 19590
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014670176060659786,
-      "loss": 0.6133,
-      "step": 19600
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014667448552374977,
-      "loss": 0.739,
-      "step": 19610
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014664721044090173,
-      "loss": 0.7314,
-      "step": 19620
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014661993535805364,
-      "loss": 0.6962,
-      "step": 19630
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001465926602752056,
-      "loss": 0.6894,
-      "step": 19640
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001465653851923575,
-      "loss": 0.6629,
-      "step": 19650
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014653811010950947,
-      "loss": 0.682,
-      "step": 19660
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.0001465108350266614,
-      "loss": 0.7529,
-      "step": 19670
-    },
-    {
-      "epoch": 0.8,
-      "learning_rate": 0.00014648355994381334,
-      "loss": 0.7525,
-      "step": 19680
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014645628486096527,
-      "loss": 0.7124,
-      "step": 19690
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001464290097781172,
-      "loss": 0.7771,
-      "step": 19700
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014640173469526914,
-      "loss": 0.7654,
-      "step": 19710
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001463744596124211,
-      "loss": 0.7889,
-      "step": 19720
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.000146347184529573,
-      "loss": 0.7125,
-      "step": 19730
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014631990944672497,
-      "loss": 0.7429,
-      "step": 19740
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014629263436387687,
-      "loss": 0.7717,
-      "step": 19750
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014626535928102883,
-      "loss": 0.6751,
-      "step": 19760
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014623808419818077,
-      "loss": 0.6491,
-      "step": 19770
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001462108091153327,
-      "loss": 0.7669,
-      "step": 19780
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014618353403248464,
-      "loss": 0.7484,
-      "step": 19790
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014615625894963657,
-      "loss": 0.7401,
-      "step": 19800
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001461289838667885,
-      "loss": 0.6716,
-      "step": 19810
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014610170878394044,
-      "loss": 0.6846,
-      "step": 19820
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014607443370109237,
-      "loss": 0.7494,
-      "step": 19830
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001460471586182443,
-      "loss": 0.7572,
-      "step": 19840
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014601988353539624,
-      "loss": 0.7394,
-      "step": 19850
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014599260845254817,
-      "loss": 0.757,
-      "step": 19860
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001459653333697001,
-      "loss": 0.7334,
-      "step": 19870
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014593805828685204,
-      "loss": 0.7711,
-      "step": 19880
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.000145910783204004,
-      "loss": 0.7116,
-      "step": 19890
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.0001458835081211559,
-      "loss": 0.816,
-      "step": 19900
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014585623303830787,
-      "loss": 0.7182,
-      "step": 19910
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014582895795545978,
-      "loss": 0.7044,
-      "step": 19920
-    },
-    {
-      "epoch": 0.81,
-      "learning_rate": 0.00014580168287261174,
-      "loss": 0.6679,
-      "step": 19930
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014577440778976367,
-      "loss": 0.7552,
-      "step": 19940
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001457471327069156,
-      "loss": 0.7714,
-      "step": 19950
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014571985762406754,
-      "loss": 0.7095,
-      "step": 19960
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014569258254121947,
-      "loss": 0.7684,
-      "step": 19970
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001456653074583714,
-      "loss": 0.7959,
-      "step": 19980
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014563803237552334,
-      "loss": 0.697,
-      "step": 19990
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014561075729267527,
-      "loss": 0.7336,
-      "step": 20000
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014558348220982724,
-      "loss": 0.7255,
-      "step": 20010
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014555620712697914,
-      "loss": 0.7782,
-      "step": 20020
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001455289320441311,
-      "loss": 0.6511,
-      "step": 20030
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.000145501656961283,
-      "loss": 0.699,
-      "step": 20040
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014547438187843497,
-      "loss": 0.7263,
-      "step": 20050
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001454471067955869,
-      "loss": 0.7386,
-      "step": 20060
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014541983171273884,
-      "loss": 0.7941,
-      "step": 20070
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014539255662989077,
-      "loss": 0.77,
-      "step": 20080
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001453652815470427,
-      "loss": 0.7135,
-      "step": 20090
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014533800646419464,
-      "loss": 0.7674,
-      "step": 20100
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014531073138134658,
-      "loss": 0.7728,
-      "step": 20110
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001452834562984985,
-      "loss": 0.7308,
-      "step": 20120
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014525618121565044,
-      "loss": 0.7509,
-      "step": 20130
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014522890613280238,
-      "loss": 0.6203,
-      "step": 20140
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.0001452016310499543,
-      "loss": 0.6399,
-      "step": 20150
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014517435596710627,
-      "loss": 0.7363,
-      "step": 20160
-    },
-    {
-      "epoch": 0.82,
-      "learning_rate": 0.00014514708088425818,
-      "loss": 0.6899,
-      "step": 20170
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014511980580141014,
-      "loss": 0.7619,
-      "step": 20180
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014509253071856205,
-      "loss": 0.7009,
-      "step": 20190
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.000145065255635714,
-      "loss": 0.6868,
-      "step": 20200
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014503798055286591,
-      "loss": 0.6995,
-      "step": 20210
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014501070547001788,
-      "loss": 0.7199,
-      "step": 20220
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001449834303871698,
-      "loss": 0.7214,
-      "step": 20230
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014495615530432174,
-      "loss": 0.746,
-      "step": 20240
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014492888022147368,
-      "loss": 0.7994,
-      "step": 20250
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001449016051386256,
-      "loss": 0.7594,
-      "step": 20260
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014487433005577754,
-      "loss": 0.6672,
-      "step": 20270
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001448470549729295,
-      "loss": 0.7544,
-      "step": 20280
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001448197798900814,
-      "loss": 0.6976,
-      "step": 20290
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014479250480723337,
-      "loss": 0.7594,
-      "step": 20300
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014476522972438528,
-      "loss": 0.7067,
-      "step": 20310
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014473795464153724,
-      "loss": 0.7258,
-      "step": 20320
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014471067955868918,
-      "loss": 0.7381,
-      "step": 20330
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001446834044758411,
-      "loss": 0.7488,
-      "step": 20340
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014465612939299304,
-      "loss": 0.6686,
-      "step": 20350
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014462885431014498,
-      "loss": 0.7769,
-      "step": 20360
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001446015792272969,
-      "loss": 0.7329,
-      "step": 20370
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014457430414444885,
-      "loss": 0.703,
-      "step": 20380
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014454702906160078,
-      "loss": 0.7329,
-      "step": 20390
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.0001445197539787527,
-      "loss": 0.7128,
-      "step": 20400
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014449247889590465,
-      "loss": 0.6994,
-      "step": 20410
-    },
-    {
-      "epoch": 0.83,
-      "learning_rate": 0.00014446520381305658,
-      "loss": 0.7932,
-      "step": 20420
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014443792873020851,
-      "loss": 0.7274,
-      "step": 20430
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014441065364736045,
-      "loss": 0.6917,
-      "step": 20440
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.0001443833785645124,
-      "loss": 0.7369,
-      "step": 20450
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014435610348166432,
-      "loss": 0.7235,
-      "step": 20460
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014432882839881628,
-      "loss": 0.6707,
-      "step": 20470
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014430155331596818,
-      "loss": 0.7591,
-      "step": 20480
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014427427823312015,
-      "loss": 0.7691,
-      "step": 20490
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014424700315027208,
-      "loss": 0.7677,
-      "step": 20500
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.000144219728067424,
-      "loss": 0.7226,
-      "step": 20510
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014419245298457595,
-      "loss": 0.7445,
-      "step": 20520
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014416517790172788,
-      "loss": 0.6678,
-      "step": 20530
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014413790281887981,
-      "loss": 0.7737,
-      "step": 20540
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014411062773603175,
-      "loss": 0.6504,
-      "step": 20550
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014408335265318368,
-      "loss": 0.7689,
-      "step": 20560
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014405607757033564,
-      "loss": 0.8477,
-      "step": 20570
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014402880248748755,
-      "loss": 0.8209,
-      "step": 20580
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.0001440015274046395,
-      "loss": 0.7375,
-      "step": 20590
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014397425232179142,
-      "loss": 0.6719,
-      "step": 20600
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014394697723894338,
-      "loss": 0.7257,
-      "step": 20610
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.0001439197021560953,
-      "loss": 0.76,
-      "step": 20620
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014389242707324725,
-      "loss": 0.804,
-      "step": 20630
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014386515199039918,
-      "loss": 0.8462,
-      "step": 20640
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014383787690755112,
-      "loss": 0.6463,
-      "step": 20650
-    },
-    {
-      "epoch": 0.84,
-      "learning_rate": 0.00014381060182470305,
-      "loss": 0.6938,
-      "step": 20660
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014378332674185498,
-      "loss": 0.7626,
-      "step": 20670
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014375605165900692,
-      "loss": 0.7149,
-      "step": 20680
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014372877657615885,
-      "loss": 0.7285,
-      "step": 20690
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014370150149331078,
-      "loss": 0.6752,
-      "step": 20700
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014367422641046272,
-      "loss": 0.7394,
-      "step": 20710
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014364695132761465,
-      "loss": 0.6739,
-      "step": 20720
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.0001436196762447666,
-      "loss": 0.7561,
-      "step": 20730
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014359240116191855,
-      "loss": 0.7682,
-      "step": 20740
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014356512607907045,
-      "loss": 0.849,
-      "step": 20750
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014353785099622242,
-      "loss": 0.7239,
-      "step": 20760
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014351057591337432,
-      "loss": 0.597,
-      "step": 20770
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014348330083052628,
-      "loss": 0.6394,
-      "step": 20780
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014345602574767822,
-      "loss": 0.7469,
-      "step": 20790
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014342875066483015,
-      "loss": 0.6883,
-      "step": 20800
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014340147558198208,
-      "loss": 0.7538,
-      "step": 20810
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014337420049913402,
-      "loss": 0.797,
-      "step": 20820
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014334692541628595,
-      "loss": 0.7382,
-      "step": 20830
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014331965033343791,
-      "loss": 0.6878,
-      "step": 20840
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014329237525058982,
-      "loss": 0.7795,
-      "step": 20850
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014326510016774178,
-      "loss": 0.6667,
-      "step": 20860
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.0001432378250848937,
-      "loss": 0.717,
-      "step": 20870
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014321327751033044,
-      "loss": 0.7661,
-      "step": 20880
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.0001431860024274824,
-      "loss": 0.7427,
-      "step": 20890
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.0001431587273446343,
-      "loss": 0.7246,
-      "step": 20900
-    },
-    {
-      "epoch": 0.85,
-      "learning_rate": 0.00014313145226178627,
-      "loss": 0.7294,
-      "step": 20910
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014310417717893818,
-      "loss": 0.8409,
-      "step": 20920
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014307690209609014,
-      "loss": 0.7379,
-      "step": 20930
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014304962701324205,
-      "loss": 0.6668,
-      "step": 20940
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.000143022351930394,
-      "loss": 0.6436,
-      "step": 20950
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014299507684754594,
-      "loss": 0.7902,
-      "step": 20960
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014296780176469787,
-      "loss": 0.6933,
-      "step": 20970
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001429405266818498,
-      "loss": 0.6706,
-      "step": 20980
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014291325159900174,
-      "loss": 0.7575,
-      "step": 20990
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014288597651615368,
-      "loss": 0.7768,
-      "step": 21000
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001428587014333056,
-      "loss": 0.7229,
-      "step": 21010
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014283142635045754,
-      "loss": 0.6877,
-      "step": 21020
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014280415126760948,
-      "loss": 0.721,
-      "step": 21030
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001427768761847614,
-      "loss": 0.7422,
-      "step": 21040
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014274960110191335,
-      "loss": 0.757,
-      "step": 21050
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001427223260190653,
-      "loss": 0.8106,
-      "step": 21060
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014269505093621721,
-      "loss": 0.7714,
-      "step": 21070
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014266777585336918,
-      "loss": 0.819,
-      "step": 21080
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014264050077052108,
-      "loss": 0.8464,
-      "step": 21090
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014261322568767304,
-      "loss": 0.7334,
-      "step": 21100
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014258595060482495,
-      "loss": 0.661,
-      "step": 21110
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001425586755219769,
-      "loss": 0.7677,
-      "step": 21120
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014253140043912884,
-      "loss": 0.7095,
-      "step": 21130
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.00014250412535628078,
-      "loss": 0.7676,
-      "step": 21140
-    },
-    {
-      "epoch": 0.86,
-      "learning_rate": 0.0001424768502734327,
-      "loss": 0.7423,
-      "step": 21150
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014244957519058465,
-      "loss": 0.7316,
-      "step": 21160
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014242230010773658,
-      "loss": 0.6743,
-      "step": 21170
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014239502502488854,
-      "loss": 0.7186,
-      "step": 21180
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014236774994204045,
-      "loss": 0.7331,
-      "step": 21190
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.0001423404748591924,
-      "loss": 0.7461,
-      "step": 21200
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014231319977634432,
-      "loss": 0.6531,
-      "step": 21210
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014228592469349628,
-      "loss": 0.6779,
-      "step": 21220
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.0001422586496106482,
-      "loss": 0.8439,
-      "step": 21230
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014223137452780014,
-      "loss": 0.8227,
-      "step": 21240
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014220409944495208,
-      "loss": 0.7956,
-      "step": 21250
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.000142176824362104,
-      "loss": 0.7431,
-      "step": 21260
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014214954927925595,
-      "loss": 0.7707,
-      "step": 21270
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014212227419640788,
-      "loss": 0.6576,
-      "step": 21280
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014209499911355981,
-      "loss": 0.6972,
-      "step": 21290
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014206772403071175,
-      "loss": 0.7248,
-      "step": 21300
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014204044894786368,
-      "loss": 0.7194,
-      "step": 21310
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014201317386501562,
-      "loss": 0.734,
-      "step": 21320
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014198589878216755,
-      "loss": 0.735,
-      "step": 21330
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014195862369931948,
-      "loss": 0.708,
-      "step": 21340
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014193134861647145,
-      "loss": 0.806,
-      "step": 21350
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014190407353362335,
-      "loss": 0.6618,
-      "step": 21360
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.0001418767984507753,
-      "loss": 0.8169,
-      "step": 21370
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014184952336792722,
-      "loss": 0.74,
-      "step": 21380
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014182224828507918,
-      "loss": 0.7216,
-      "step": 21390
-    },
-    {
-      "epoch": 0.87,
-      "learning_rate": 0.00014179497320223111,
-      "loss": 0.7066,
-      "step": 21400
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014176769811938305,
-      "loss": 0.6617,
-      "step": 21410
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014174042303653498,
-      "loss": 0.7299,
-      "step": 21420
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014171314795368692,
-      "loss": 0.7496,
-      "step": 21430
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014168587287083885,
-      "loss": 0.7381,
-      "step": 21440
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.0001416585977879908,
-      "loss": 0.7354,
-      "step": 21450
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014163132270514272,
-      "loss": 0.7125,
-      "step": 21460
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014160404762229468,
-      "loss": 0.6702,
-      "step": 21470
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014157677253944659,
-      "loss": 0.7514,
-      "step": 21480
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014154949745659855,
-      "loss": 0.7507,
-      "step": 21490
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014152222237375045,
-      "loss": 0.6776,
-      "step": 21500
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014149494729090242,
-      "loss": 0.773,
-      "step": 21510
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014146767220805435,
-      "loss": 0.7145,
-      "step": 21520
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014144039712520628,
-      "loss": 0.7115,
-      "step": 21530
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014141312204235822,
-      "loss": 0.6626,
-      "step": 21540
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014138584695951015,
-      "loss": 0.7372,
-      "step": 21550
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014135857187666208,
-      "loss": 0.6884,
-      "step": 21560
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014133129679381402,
-      "loss": 0.7024,
-      "step": 21570
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014130402171096595,
-      "loss": 0.7445,
-      "step": 21580
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014127674662811789,
-      "loss": 0.8385,
-      "step": 21590
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014124947154526982,
-      "loss": 0.7327,
-      "step": 21600
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014122219646242175,
-      "loss": 0.6905,
-      "step": 21610
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014119492137957372,
-      "loss": 0.6964,
-      "step": 21620
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014116764629672562,
-      "loss": 0.6944,
-      "step": 21630
-    },
-    {
-      "epoch": 0.88,
-      "learning_rate": 0.00014114037121387758,
-      "loss": 0.7298,
-      "step": 21640
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.0001411130961310295,
-      "loss": 0.8142,
-      "step": 21650
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014108582104818145,
-      "loss": 0.7516,
-      "step": 21660
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014105854596533336,
-      "loss": 0.7207,
-      "step": 21670
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014103127088248532,
-      "loss": 0.7178,
-      "step": 21680
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014100399579963725,
-      "loss": 0.7013,
-      "step": 21690
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.0001409767207167892,
-      "loss": 0.6956,
-      "step": 21700
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014094944563394112,
-      "loss": 0.8074,
-      "step": 21710
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014092217055109305,
-      "loss": 0.7125,
-      "step": 21720
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.000140894895468245,
-      "loss": 0.6997,
-      "step": 21730
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014086762038539695,
-      "loss": 0.6975,
-      "step": 21740
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014084034530254886,
-      "loss": 0.8493,
-      "step": 21750
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014081307021970082,
-      "loss": 0.7765,
-      "step": 21760
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014078579513685272,
-      "loss": 0.7095,
-      "step": 21770
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014075852005400469,
-      "loss": 0.7037,
-      "step": 21780
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014073124497115662,
-      "loss": 0.7047,
-      "step": 21790
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014070396988830855,
-      "loss": 0.7906,
-      "step": 21800
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.0001406766948054605,
-      "loss": 0.7032,
-      "step": 21810
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014064941972261242,
-      "loss": 0.6398,
-      "step": 21820
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014062214463976435,
-      "loss": 0.6891,
-      "step": 21830
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.0001405948695569163,
-      "loss": 0.6992,
-      "step": 21840
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014056759447406822,
-      "loss": 0.7337,
-      "step": 21850
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014054031939122016,
-      "loss": 0.6856,
-      "step": 21860
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.0001405130443083721,
-      "loss": 0.7268,
-      "step": 21870
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014048576922552402,
-      "loss": 0.7104,
-      "step": 21880
-    },
-    {
-      "epoch": 0.89,
-      "learning_rate": 0.00014045849414267596,
-      "loss": 0.7383,
-      "step": 21890
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001404312190598279,
-      "loss": 0.617,
-      "step": 21900
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014040394397697985,
-      "loss": 0.7452,
-      "step": 21910
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014037666889413176,
-      "loss": 0.7981,
-      "step": 21920
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014034939381128372,
-      "loss": 0.6603,
-      "step": 21930
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014032211872843563,
-      "loss": 0.6597,
-      "step": 21940
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001402948436455876,
-      "loss": 0.6243,
-      "step": 21950
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014026756856273952,
-      "loss": 0.7084,
-      "step": 21960
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014024029347989146,
-      "loss": 0.7477,
-      "step": 21970
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001402130183970434,
-      "loss": 0.6673,
-      "step": 21980
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014018574331419532,
-      "loss": 0.7737,
-      "step": 21990
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014015846823134726,
-      "loss": 0.6925,
-      "step": 22000
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001401311931484992,
-      "loss": 0.6877,
-      "step": 22010
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014010391806565113,
-      "loss": 0.7478,
-      "step": 22020
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001400766429828031,
-      "loss": 0.7577,
-      "step": 22030
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.000140049367899955,
-      "loss": 0.8086,
-      "step": 22040
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00014002209281710696,
-      "loss": 0.695,
-      "step": 22050
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013999481773425886,
-      "loss": 0.6813,
-      "step": 22060
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013996754265141082,
-      "loss": 0.685,
-      "step": 22070
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013994026756856276,
-      "loss": 0.7796,
-      "step": 22080
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001399129924857147,
-      "loss": 0.755,
-      "step": 22090
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013988571740286662,
-      "loss": 0.7492,
-      "step": 22100
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013985844232001856,
-      "loss": 0.6721,
-      "step": 22110
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.0001398311672371705,
-      "loss": 0.704,
-      "step": 22120
-    },
-    {
-      "epoch": 0.9,
-      "learning_rate": 0.00013980389215432243,
-      "loss": 0.7597,
-      "step": 22130
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013977661707147436,
-      "loss": 0.7774,
-      "step": 22140
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001397493419886263,
-      "loss": 0.7259,
-      "step": 22150
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013972206690577823,
-      "loss": 0.7171,
-      "step": 22160
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013969479182293016,
-      "loss": 0.738,
-      "step": 22170
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001396675167400821,
-      "loss": 0.7134,
-      "step": 22180
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013964024165723403,
-      "loss": 0.7298,
-      "step": 22190
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.000139612966574386,
-      "loss": 0.8089,
-      "step": 22200
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001395856914915379,
-      "loss": 0.6344,
-      "step": 22210
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013955841640868986,
-      "loss": 0.6582,
-      "step": 22220
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013953114132584177,
-      "loss": 0.6951,
-      "step": 22230
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013950386624299373,
-      "loss": 0.7891,
-      "step": 22240
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013947659116014566,
-      "loss": 0.7863,
-      "step": 22250
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001394493160772976,
-      "loss": 0.6737,
-      "step": 22260
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013942204099444953,
-      "loss": 0.8074,
-      "step": 22270
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013939476591160146,
-      "loss": 0.7694,
-      "step": 22280
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001393674908287534,
-      "loss": 0.7439,
-      "step": 22290
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013934021574590536,
-      "loss": 0.752,
-      "step": 22300
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013931294066305726,
-      "loss": 0.7592,
-      "step": 22310
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013928566558020923,
-      "loss": 0.6491,
-      "step": 22320
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013925839049736113,
-      "loss": 0.6798,
-      "step": 22330
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001392311154145131,
-      "loss": 0.754,
-      "step": 22340
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.000139203840331665,
-      "loss": 0.6845,
-      "step": 22350
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013917656524881696,
-      "loss": 0.6571,
-      "step": 22360
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.0001391492901659689,
-      "loss": 0.6341,
-      "step": 22370
-    },
-    {
-      "epoch": 0.91,
-      "learning_rate": 0.00013912201508312083,
-      "loss": 0.716,
-      "step": 22380
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013909474000027276,
-      "loss": 0.7439,
-      "step": 22390
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001390674649174247,
-      "loss": 0.6551,
-      "step": 22400
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013904018983457663,
-      "loss": 0.8119,
-      "step": 22410
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013901291475172856,
-      "loss": 0.7183,
-      "step": 22420
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001389856396688805,
-      "loss": 0.6832,
-      "step": 22430
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013895836458603243,
-      "loss": 0.6862,
-      "step": 22440
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013893108950318437,
-      "loss": 0.6608,
-      "step": 22450
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001389038144203363,
-      "loss": 0.6294,
-      "step": 22460
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013887653933748826,
-      "loss": 0.7427,
-      "step": 22470
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013884926425464017,
-      "loss": 0.7218,
-      "step": 22480
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013882198917179213,
-      "loss": 0.7803,
-      "step": 22490
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013879471408894404,
-      "loss": 0.8107,
-      "step": 22500
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.000138767439006096,
-      "loss": 0.728,
-      "step": 22510
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013874016392324793,
-      "loss": 0.6562,
-      "step": 22520
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013871288884039986,
-      "loss": 0.8057,
-      "step": 22530
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001386856137575518,
-      "loss": 0.7344,
-      "step": 22540
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013865833867470373,
-      "loss": 0.7776,
-      "step": 22550
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013863106359185567,
-      "loss": 0.7477,
-      "step": 22560
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001386037885090076,
-      "loss": 0.7914,
-      "step": 22570
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013857651342615953,
-      "loss": 0.7434,
-      "step": 22580
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001385492383433115,
-      "loss": 0.758,
-      "step": 22590
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.0001385219632604634,
-      "loss": 0.7063,
-      "step": 22600
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013849468817761536,
-      "loss": 0.7029,
-      "step": 22610
-    },
-    {
-      "epoch": 0.92,
-      "learning_rate": 0.00013846741309476727,
-      "loss": 0.7305,
-      "step": 22620
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013844013801191923,
-      "loss": 0.8083,
-      "step": 22630
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013841286292907116,
-      "loss": 0.7042,
-      "step": 22640
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001383855878462231,
-      "loss": 0.7144,
-      "step": 22650
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013835831276337503,
-      "loss": 0.8116,
-      "step": 22660
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013833103768052697,
-      "loss": 0.6899,
-      "step": 22670
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001383037625976789,
-      "loss": 0.694,
-      "step": 22680
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013827648751483083,
-      "loss": 0.7915,
-      "step": 22690
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013824921243198277,
-      "loss": 0.7137,
-      "step": 22700
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001382219373491347,
-      "loss": 0.7815,
-      "step": 22710
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013819466226628664,
-      "loss": 0.7412,
-      "step": 22720
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013816738718343857,
-      "loss": 0.6985,
-      "step": 22730
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001381401121005905,
-      "loss": 0.7487,
-      "step": 22740
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013811283701774244,
-      "loss": 0.7845,
-      "step": 22750
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001380855619348944,
-      "loss": 0.7292,
-      "step": 22760
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001380582868520463,
-      "loss": 0.768,
-      "step": 22770
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013803101176919827,
-      "loss": 0.6746,
-      "step": 22780
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013800373668635017,
-      "loss": 0.7899,
-      "step": 22790
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013797646160350213,
-      "loss": 0.716,
-      "step": 22800
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013794918652065407,
-      "loss": 0.767,
-      "step": 22810
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.000137921911437806,
-      "loss": 0.7329,
-      "step": 22820
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013789463635495794,
-      "loss": 0.7107,
-      "step": 22830
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013786736127210987,
-      "loss": 0.855,
-      "step": 22840
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.0001378400861892618,
-      "loss": 0.7723,
-      "step": 22850
-    },
-    {
-      "epoch": 0.93,
-      "learning_rate": 0.00013781281110641377,
-      "loss": 0.6317,
-      "step": 22860
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013778553602356567,
-      "loss": 0.8118,
-      "step": 22870
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013775826094071763,
-      "loss": 0.711,
-      "step": 22880
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013773098585786954,
-      "loss": 0.763,
-      "step": 22890
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001377037107750215,
-      "loss": 0.6204,
-      "step": 22900
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001376764356921734,
-      "loss": 0.6613,
-      "step": 22910
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013764916060932537,
-      "loss": 0.736,
-      "step": 22920
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001376218855264773,
-      "loss": 0.7365,
-      "step": 22930
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013759461044362924,
-      "loss": 0.7871,
-      "step": 22940
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013756733536078117,
-      "loss": 0.7343,
-      "step": 22950
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001375400602779331,
-      "loss": 0.6984,
-      "step": 22960
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013751278519508504,
-      "loss": 0.8129,
-      "step": 22970
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013748551011223697,
-      "loss": 0.7185,
-      "step": 22980
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001374582350293889,
-      "loss": 0.7154,
-      "step": 22990
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013743095994654084,
-      "loss": 0.7229,
-      "step": 23000
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013740368486369277,
-      "loss": 0.6333,
-      "step": 23010
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001373764097808447,
-      "loss": 0.7537,
-      "step": 23020
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013734913469799667,
-      "loss": 0.7867,
-      "step": 23030
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013732185961514858,
-      "loss": 0.7559,
-      "step": 23040
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013729458453230054,
-      "loss": 0.6951,
-      "step": 23050
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013726730944945244,
-      "loss": 0.7569,
-      "step": 23060
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001372400343666044,
-      "loss": 0.7794,
-      "step": 23070
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001372127592837563,
-      "loss": 0.6891,
-      "step": 23080
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013718548420090827,
-      "loss": 0.6972,
-      "step": 23090
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.0001371582091180602,
-      "loss": 0.6762,
-      "step": 23100
-    },
-    {
-      "epoch": 0.94,
-      "learning_rate": 0.00013713093403521214,
-      "loss": 0.6339,
-      "step": 23110
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013710365895236407,
-      "loss": 0.7814,
-      "step": 23120
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.000137076383869516,
-      "loss": 0.7033,
-      "step": 23130
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013704910878666794,
-      "loss": 0.8267,
-      "step": 23140
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001370218337038199,
-      "loss": 0.6972,
-      "step": 23150
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001369945586209718,
-      "loss": 0.7744,
-      "step": 23160
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013696728353812377,
-      "loss": 0.7815,
-      "step": 23170
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013694000845527568,
-      "loss": 0.6899,
-      "step": 23180
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013691273337242764,
-      "loss": 0.774,
-      "step": 23190
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013688545828957957,
-      "loss": 0.6363,
-      "step": 23200
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001368581832067315,
-      "loss": 0.6289,
-      "step": 23210
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013683090812388344,
-      "loss": 0.7584,
-      "step": 23220
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013680363304103537,
-      "loss": 0.6495,
-      "step": 23230
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001367763579581873,
-      "loss": 0.6941,
-      "step": 23240
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013674908287533924,
-      "loss": 0.6494,
-      "step": 23250
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013672180779249118,
-      "loss": 0.8119,
-      "step": 23260
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001366945327096431,
-      "loss": 0.658,
-      "step": 23270
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013666725762679504,
-      "loss": 0.695,
-      "step": 23280
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013663998254394698,
-      "loss": 0.6652,
-      "step": 23290
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001366127074610989,
-      "loss": 0.7618,
-      "step": 23300
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013658543237825085,
-      "loss": 0.6821,
-      "step": 23310
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.0001365581572954028,
-      "loss": 0.683,
-      "step": 23320
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013653088221255471,
-      "loss": 0.7451,
-      "step": 23330
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013650360712970667,
-      "loss": 0.7171,
-      "step": 23340
-    },
-    {
-      "epoch": 0.95,
-      "learning_rate": 0.00013647633204685858,
-      "loss": 0.6841,
-      "step": 23350
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013644905696401054,
-      "loss": 0.6918,
-      "step": 23360
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013642178188116248,
-      "loss": 0.7237,
-      "step": 23370
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.0001363945067983144,
-      "loss": 0.6396,
-      "step": 23380
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013636723171546634,
-      "loss": 0.6661,
-      "step": 23390
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013633995663261828,
-      "loss": 0.6914,
-      "step": 23400
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.0001363126815497702,
-      "loss": 0.7223,
-      "step": 23410
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013628540646692217,
-      "loss": 0.746,
-      "step": 23420
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013625813138407408,
-      "loss": 0.7635,
-      "step": 23430
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013623085630122604,
-      "loss": 0.629,
-      "step": 23440
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013620358121837795,
-      "loss": 0.7111,
-      "step": 23450
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.0001361763061355299,
-      "loss": 0.724,
-      "step": 23460
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013614903105268182,
-      "loss": 0.6681,
-      "step": 23470
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013612175596983378,
-      "loss": 0.6338,
-      "step": 23480
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.0001360944808869857,
-      "loss": 0.7451,
-      "step": 23490
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013606720580413764,
-      "loss": 0.746,
-      "step": 23500
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013603993072128958,
-      "loss": 0.709,
-      "step": 23510
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.0001360126556384415,
-      "loss": 0.6583,
-      "step": 23520
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013598538055559345,
-      "loss": 0.7146,
-      "step": 23530
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013595810547274538,
-      "loss": 0.666,
-      "step": 23540
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013593083038989731,
-      "loss": 0.7313,
-      "step": 23550
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013590355530704925,
-      "loss": 0.8368,
-      "step": 23560
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013587628022420118,
-      "loss": 0.9082,
-      "step": 23570
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013584900514135312,
-      "loss": 0.74,
-      "step": 23580
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013582173005850508,
-      "loss": 0.6901,
-      "step": 23590
-    },
-    {
-      "epoch": 0.96,
-      "learning_rate": 0.00013579445497565698,
-      "loss": 0.7958,
-      "step": 23600
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013576717989280894,
-      "loss": 0.7307,
-      "step": 23610
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013573990480996085,
-      "loss": 0.8025,
-      "step": 23620
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.0001357126297271128,
-      "loss": 0.7162,
-      "step": 23630
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013568535464426472,
-      "loss": 0.7359,
-      "step": 23640
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013565807956141668,
-      "loss": 0.6984,
-      "step": 23650
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013563080447856861,
-      "loss": 0.8048,
-      "step": 23660
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013560352939572055,
-      "loss": 0.8227,
-      "step": 23670
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013557625431287248,
-      "loss": 0.6428,
-      "step": 23680
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013554897923002442,
-      "loss": 0.7334,
-      "step": 23690
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013552170414717635,
-      "loss": 0.74,
-      "step": 23700
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.0001354944290643283,
-      "loss": 0.756,
-      "step": 23710
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013546715398148022,
-      "loss": 0.7026,
-      "step": 23720
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013543987889863218,
-      "loss": 0.6174,
-      "step": 23730
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013541260381578409,
-      "loss": 0.7034,
-      "step": 23740
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013538532873293605,
-      "loss": 0.6337,
-      "step": 23750
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013535805365008798,
-      "loss": 0.7339,
-      "step": 23760
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013533077856723991,
-      "loss": 0.7673,
-      "step": 23770
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013530350348439185,
-      "loss": 0.8011,
-      "step": 23780
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013527622840154378,
-      "loss": 0.7216,
-      "step": 23790
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013524895331869572,
-      "loss": 0.6725,
-      "step": 23800
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013522167823584765,
-      "loss": 0.738,
-      "step": 23810
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013519440315299958,
-      "loss": 0.7271,
-      "step": 23820
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013516712807015152,
-      "loss": 0.8051,
-      "step": 23830
-    },
-    {
-      "epoch": 0.97,
-      "learning_rate": 0.00013513985298730345,
-      "loss": 0.7117,
-      "step": 23840
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013511257790445539,
-      "loss": 0.7554,
-      "step": 23850
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013508530282160732,
-      "loss": 0.709,
-      "step": 23860
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013505802773875925,
-      "loss": 0.7855,
-      "step": 23870
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013503075265591121,
-      "loss": 0.6878,
-      "step": 23880
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013500347757306312,
-      "loss": 0.6576,
-      "step": 23890
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013497620249021508,
-      "loss": 0.7308,
-      "step": 23900
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.000134948927407367,
-      "loss": 0.7109,
-      "step": 23910
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013492165232451895,
-      "loss": 0.783,
-      "step": 23920
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013489437724167088,
-      "loss": 0.7585,
-      "step": 23930
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013486710215882282,
-      "loss": 0.7102,
-      "step": 23940
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013483982707597475,
-      "loss": 0.6607,
-      "step": 23950
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013481255199312669,
-      "loss": 0.7341,
-      "step": 23960
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013478527691027862,
-      "loss": 0.7197,
-      "step": 23970
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013475800182743055,
-      "loss": 0.7075,
-      "step": 23980
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.0001347307267445825,
-      "loss": 0.797,
-      "step": 23990
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013470345166173445,
-      "loss": 0.6882,
-      "step": 24000
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013467617657888636,
-      "loss": 0.7128,
-      "step": 24010
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013464890149603832,
-      "loss": 0.7426,
-      "step": 24020
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013462162641319022,
-      "loss": 0.7549,
-      "step": 24030
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013459435133034218,
-      "loss": 0.7085,
-      "step": 24040
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013456707624749412,
-      "loss": 0.6801,
-      "step": 24050
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013453980116464605,
-      "loss": 0.6663,
-      "step": 24060
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.000134512526081798,
-      "loss": 0.8245,
-      "step": 24070
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013448525099894992,
-      "loss": 0.7168,
-      "step": 24080
-    },
-    {
-      "epoch": 0.98,
-      "learning_rate": 0.00013445797591610185,
-      "loss": 0.6466,
-      "step": 24090
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001344307008332538,
-      "loss": 0.7438,
-      "step": 24100
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013440342575040572,
-      "loss": 0.6098,
-      "step": 24110
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013437615066755766,
-      "loss": 0.82,
-      "step": 24120
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001343488755847096,
-      "loss": 0.7117,
-      "step": 24130
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013432160050186152,
-      "loss": 0.7012,
-      "step": 24140
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013429432541901346,
-      "loss": 0.6816,
-      "step": 24150
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001342670503361654,
-      "loss": 0.7737,
-      "step": 24160
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013423977525331735,
-      "loss": 0.7387,
-      "step": 24170
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013421250017046926,
-      "loss": 0.7852,
-      "step": 24180
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013418522508762122,
-      "loss": 0.7538,
-      "step": 24190
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013415795000477313,
-      "loss": 0.7313,
-      "step": 24200
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001341306749219251,
-      "loss": 0.8403,
-      "step": 24210
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013410339983907702,
-      "loss": 0.6766,
-      "step": 24220
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013407612475622896,
-      "loss": 0.6955,
-      "step": 24230
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001340488496733809,
-      "loss": 0.7038,
-      "step": 24240
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013402157459053282,
-      "loss": 0.7561,
-      "step": 24250
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013399429950768476,
-      "loss": 0.6762,
-      "step": 24260
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013396702442483672,
-      "loss": 0.7992,
-      "step": 24270
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013393974934198863,
-      "loss": 0.729,
-      "step": 24280
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001339124742591406,
-      "loss": 0.7452,
-      "step": 24290
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.0001338851991762925,
-      "loss": 0.7208,
-      "step": 24300
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013385792409344445,
-      "loss": 0.7397,
-      "step": 24310
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013383064901059636,
-      "loss": 0.7172,
-      "step": 24320
-    },
-    {
-      "epoch": 0.99,
-      "learning_rate": 0.00013380337392774832,
-      "loss": 0.8183,
-      "step": 24330
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013377609884490026,
-      "loss": 0.7481,
-      "step": 24340
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001337488237620522,
-      "loss": 0.698,
-      "step": 24350
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013372154867920412,
-      "loss": 0.7886,
-      "step": 24360
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013369427359635606,
-      "loss": 0.7431,
-      "step": 24370
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.000133666998513508,
-      "loss": 0.776,
-      "step": 24380
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013363972343065993,
-      "loss": 0.6055,
-      "step": 24390
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013361244834781186,
-      "loss": 0.7069,
-      "step": 24400
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001335851732649638,
-      "loss": 0.6422,
-      "step": 24410
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013355789818211573,
-      "loss": 0.6974,
-      "step": 24420
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013353062309926766,
-      "loss": 0.7191,
-      "step": 24430
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013350334801641962,
-      "loss": 0.6522,
-      "step": 24440
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013347607293357153,
-      "loss": 0.6989,
-      "step": 24450
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001334487978507235,
-      "loss": 0.7897,
-      "step": 24460
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001334215227678754,
-      "loss": 0.7575,
-      "step": 24470
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013339424768502736,
-      "loss": 0.6857,
-      "step": 24480
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001333669726021793,
-      "loss": 0.803,
-      "step": 24490
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013333969751933123,
-      "loss": 0.6317,
-      "step": 24500
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013331242243648316,
-      "loss": 0.741,
-      "step": 24510
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001332851473536351,
-      "loss": 0.903,
-      "step": 24520
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013325787227078703,
-      "loss": 0.7078,
-      "step": 24530
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013323059718793896,
-      "loss": 0.6136,
-      "step": 24540
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.0001332033221050909,
-      "loss": 0.6962,
-      "step": 24550
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013317604702224286,
-      "loss": 0.6601,
-      "step": 24560
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013314877193939476,
-      "loss": 0.6173,
-      "step": 24570
-    },
-    {
-      "epoch": 1.0,
-      "learning_rate": 0.00013312149685654672,
-      "loss": 0.7196,
-      "step": 24580
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013309422177369863,
-      "loss": 0.6797,
-      "step": 24590
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001330669466908506,
-      "loss": 0.7028,
-      "step": 24600
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013303967160800253,
-      "loss": 0.7226,
-      "step": 24610
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013301239652515446,
-      "loss": 0.6322,
-      "step": 24620
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001329851214423064,
-      "loss": 0.6685,
-      "step": 24630
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013295784635945833,
-      "loss": 0.7257,
-      "step": 24640
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013293057127661026,
-      "loss": 0.6152,
-      "step": 24650
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001329032961937622,
-      "loss": 0.6791,
-      "step": 24660
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013287602111091413,
-      "loss": 0.6509,
-      "step": 24670
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013284874602806606,
-      "loss": 0.5824,
-      "step": 24680
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.000132821470945218,
-      "loss": 0.6316,
-      "step": 24690
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013279419586236993,
-      "loss": 0.6123,
-      "step": 24700
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013276692077952187,
-      "loss": 0.7582,
-      "step": 24710
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001327396456966738,
-      "loss": 0.6694,
-      "step": 24720
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013271237061382576,
-      "loss": 0.6397,
-      "step": 24730
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013268509553097767,
-      "loss": 0.6656,
-      "step": 24740
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013265782044812963,
-      "loss": 0.721,
-      "step": 24750
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013263054536528154,
-      "loss": 0.6168,
-      "step": 24760
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001326032702824335,
-      "loss": 0.676,
-      "step": 24770
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013257599519958543,
-      "loss": 0.6127,
-      "step": 24780
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013254872011673736,
-      "loss": 0.6602,
-      "step": 24790
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.0001325214450338893,
-      "loss": 0.7242,
-      "step": 24800
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013249416995104123,
-      "loss": 0.7563,
-      "step": 24810
-    },
-    {
-      "epoch": 1.01,
-      "learning_rate": 0.00013246689486819317,
-      "loss": 0.6183,
-      "step": 24820
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013243961978534513,
-      "loss": 0.7143,
-      "step": 24830
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013241234470249703,
-      "loss": 0.7336,
-      "step": 24840
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.000132385069619649,
-      "loss": 0.6599,
-      "step": 24850
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001323577945368009,
-      "loss": 0.6695,
-      "step": 24860
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013233051945395286,
-      "loss": 0.709,
-      "step": 24870
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013230324437110477,
-      "loss": 0.5845,
-      "step": 24880
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013227596928825673,
-      "loss": 0.681,
-      "step": 24890
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013224869420540866,
-      "loss": 0.6555,
-      "step": 24900
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001322214191225606,
-      "loss": 0.6382,
-      "step": 24910
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013219414403971253,
-      "loss": 0.71,
-      "step": 24920
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013216686895686447,
-      "loss": 0.5961,
-      "step": 24930
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001321395938740164,
-      "loss": 0.7115,
-      "step": 24940
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013211231879116833,
-      "loss": 0.6692,
-      "step": 24950
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013208504370832027,
-      "loss": 0.6744,
-      "step": 24960
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001320577686254722,
-      "loss": 0.645,
-      "step": 24970
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013203049354262414,
-      "loss": 0.6831,
-      "step": 24980
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013200321845977607,
-      "loss": 0.6372,
-      "step": 24990
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013197594337692803,
-      "loss": 0.7072,
-      "step": 25000
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013194866829407994,
-      "loss": 0.6725,
-      "step": 25010
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001319213932112319,
-      "loss": 0.6861,
-      "step": 25020
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.0001318941181283838,
-      "loss": 0.8274,
-      "step": 25030
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013186684304553577,
-      "loss": 0.7118,
-      "step": 25040
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013183956796268767,
-      "loss": 0.6882,
-      "step": 25050
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013181229287983963,
-      "loss": 0.6448,
-      "step": 25060
-    },
-    {
-      "epoch": 1.02,
-      "learning_rate": 0.00013178501779699157,
-      "loss": 0.7812,
-      "step": 25070
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001317577427141435,
-      "loss": 0.701,
-      "step": 25080
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013173046763129544,
-      "loss": 0.744,
-      "step": 25090
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013170319254844737,
-      "loss": 0.6859,
-      "step": 25100
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001316759174655993,
-      "loss": 0.6823,
-      "step": 25110
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013164864238275126,
-      "loss": 0.7431,
-      "step": 25120
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013162136729990317,
-      "loss": 0.682,
-      "step": 25130
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013159409221705513,
-      "loss": 0.7128,
-      "step": 25140
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013156681713420704,
-      "loss": 0.664,
-      "step": 25150
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.000131539542051359,
-      "loss": 0.7856,
-      "step": 25160
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013151226696851093,
-      "loss": 0.7518,
-      "step": 25170
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013148499188566287,
-      "loss": 0.7665,
-      "step": 25180
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001314577168028148,
-      "loss": 0.6907,
-      "step": 25190
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013143044171996674,
-      "loss": 0.6329,
-      "step": 25200
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013140316663711867,
-      "loss": 0.7235,
-      "step": 25210
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001313758915542706,
-      "loss": 0.5661,
-      "step": 25220
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013134861647142254,
-      "loss": 0.7023,
-      "step": 25230
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013132134138857447,
-      "loss": 0.7957,
-      "step": 25240
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001312940663057264,
-      "loss": 0.6531,
-      "step": 25250
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013126679122287834,
-      "loss": 0.6932,
-      "step": 25260
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013123951614003027,
-      "loss": 0.6624,
-      "step": 25270
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.0001312122410571822,
-      "loss": 0.6581,
-      "step": 25280
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013118496597433417,
-      "loss": 0.7378,
-      "step": 25290
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013115769089148608,
-      "loss": 0.7085,
-      "step": 25300
-    },
-    {
-      "epoch": 1.03,
-      "learning_rate": 0.00013113041580863804,
-      "loss": 0.6867,
-      "step": 25310
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013110314072578994,
-      "loss": 0.6676,
-      "step": 25320
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001310758656429419,
-      "loss": 0.7398,
-      "step": 25330
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013104859056009384,
-      "loss": 0.6442,
-      "step": 25340
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013102131547724577,
-      "loss": 0.766,
-      "step": 25350
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001309940403943977,
-      "loss": 0.7808,
-      "step": 25360
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013096676531154964,
-      "loss": 0.6698,
-      "step": 25370
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013093949022870157,
-      "loss": 0.6942,
-      "step": 25380
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013091221514585353,
-      "loss": 0.7006,
-      "step": 25390
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013088494006300544,
-      "loss": 0.6677,
-      "step": 25400
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001308576649801574,
-      "loss": 0.6784,
-      "step": 25410
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001308303898973093,
-      "loss": 0.7285,
-      "step": 25420
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013080311481446127,
-      "loss": 0.6591,
-      "step": 25430
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013077583973161318,
-      "loss": 0.6582,
-      "step": 25440
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013074856464876514,
-      "loss": 0.771,
-      "step": 25450
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013072128956591707,
-      "loss": 0.8023,
-      "step": 25460
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.000130694014483069,
-      "loss": 0.7202,
-      "step": 25470
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013066673940022094,
-      "loss": 0.7078,
-      "step": 25480
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013063946431737287,
-      "loss": 0.6955,
-      "step": 25490
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001306121892345248,
-      "loss": 0.6732,
-      "step": 25500
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013058491415167674,
-      "loss": 0.6972,
-      "step": 25510
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013055763906882868,
-      "loss": 0.7054,
-      "step": 25520
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.0001305303639859806,
-      "loss": 0.6951,
-      "step": 25530
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013050308890313254,
-      "loss": 0.6992,
-      "step": 25540
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013047581382028448,
-      "loss": 0.6551,
-      "step": 25550
-    },
-    {
-      "epoch": 1.04,
-      "learning_rate": 0.00013044853873743644,
-      "loss": 0.6976,
-      "step": 25560
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013042126365458835,
-      "loss": 0.7711,
-      "step": 25570
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001303939885717403,
-      "loss": 0.7252,
-      "step": 25580
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001303667134888922,
-      "loss": 0.7336,
-      "step": 25590
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013033943840604417,
-      "loss": 0.6985,
-      "step": 25600
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013031216332319608,
-      "loss": 0.6868,
-      "step": 25610
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013028488824034804,
-      "loss": 0.7456,
-      "step": 25620
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013025761315749998,
-      "loss": 0.7673,
-      "step": 25630
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001302303380746519,
-      "loss": 0.6717,
-      "step": 25640
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013020306299180384,
-      "loss": 0.7646,
-      "step": 25650
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013017578790895578,
-      "loss": 0.5754,
-      "step": 25660
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001301485128261077,
-      "loss": 0.7654,
-      "step": 25670
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013012123774325967,
-      "loss": 0.6859,
-      "step": 25680
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013009396266041158,
-      "loss": 0.5776,
-      "step": 25690
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013006668757756354,
-      "loss": 0.7116,
-      "step": 25700
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00013003941249471545,
-      "loss": 0.7135,
-      "step": 25710
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001300121374118674,
-      "loss": 0.741,
-      "step": 25720
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012998486232901934,
-      "loss": 0.7852,
-      "step": 25730
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012995758724617128,
-      "loss": 0.6993,
-      "step": 25740
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.0001299303121633232,
-      "loss": 0.7112,
-      "step": 25750
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012990303708047514,
-      "loss": 0.7291,
-      "step": 25760
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012987576199762708,
-      "loss": 0.6004,
-      "step": 25770
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.000129848486914779,
-      "loss": 0.6393,
-      "step": 25780
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012982121183193095,
-      "loss": 0.6928,
-      "step": 25790
-    },
-    {
-      "epoch": 1.05,
-      "learning_rate": 0.00012979393674908288,
-      "loss": 0.6153,
-      "step": 25800
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012976666166623481,
-      "loss": 0.6919,
-      "step": 25810
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012973938658338675,
-      "loss": 0.7024,
-      "step": 25820
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012971211150053868,
-      "loss": 0.6209,
-      "step": 25830
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012968483641769062,
-      "loss": 0.7204,
-      "step": 25840
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012965756133484258,
-      "loss": 0.6484,
-      "step": 25850
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012963028625199448,
-      "loss": 0.6348,
-      "step": 25860
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012960301116914644,
-      "loss": 0.6686,
-      "step": 25870
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012957573608629835,
-      "loss": 0.6674,
-      "step": 25880
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.0001295484610034503,
-      "loss": 0.6935,
-      "step": 25890
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012952118592060225,
-      "loss": 0.6432,
-      "step": 25900
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012949391083775418,
-      "loss": 0.6945,
-      "step": 25910
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012946663575490611,
-      "loss": 0.755,
-      "step": 25920
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012943936067205805,
-      "loss": 0.7094,
-      "step": 25930
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012941208558920998,
-      "loss": 0.6543,
-      "step": 25940
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012938481050636192,
-      "loss": 0.7393,
-      "step": 25950
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012935753542351385,
-      "loss": 0.6779,
-      "step": 25960
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.0001293302603406658,
-      "loss": 0.6909,
-      "step": 25970
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012930298525781772,
-      "loss": 0.8056,
-      "step": 25980
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012927571017496968,
-      "loss": 0.6972,
-      "step": 25990
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012924843509212159,
-      "loss": 0.7646,
-      "step": 26000
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012922116000927355,
-      "loss": 0.655,
-      "step": 26010
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012919388492642548,
-      "loss": 0.6734,
-      "step": 26020
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012916660984357741,
-      "loss": 0.6595,
-      "step": 26030
-    },
-    {
-      "epoch": 1.06,
-      "learning_rate": 0.00012913933476072935,
-      "loss": 0.7136,
-      "step": 26040
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012911205967788128,
-      "loss": 0.6889,
-      "step": 26050
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012908478459503322,
-      "loss": 0.7402,
-      "step": 26060
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012905750951218515,
-      "loss": 0.7153,
-      "step": 26070
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012903023442933708,
-      "loss": 0.7424,
-      "step": 26080
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012900295934648902,
-      "loss": 0.7159,
-      "step": 26090
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012897568426364095,
-      "loss": 0.6725,
-      "step": 26100
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012894840918079289,
-      "loss": 0.7034,
-      "step": 26110
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012892113409794482,
-      "loss": 0.645,
-      "step": 26120
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012889385901509675,
-      "loss": 0.6774,
-      "step": 26130
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012886658393224871,
-      "loss": 0.6966,
-      "step": 26140
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012883930884940062,
-      "loss": 0.6776,
-      "step": 26150
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012881203376655258,
-      "loss": 0.6337,
-      "step": 26160
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.0001287847586837045,
-      "loss": 0.693,
-      "step": 26170
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012875748360085645,
-      "loss": 0.6747,
-      "step": 26180
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012873020851800838,
-      "loss": 0.6615,
-      "step": 26190
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012870293343516032,
-      "loss": 0.7375,
-      "step": 26200
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012867565835231225,
-      "loss": 0.7074,
-      "step": 26210
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012864838326946419,
-      "loss": 0.6792,
-      "step": 26220
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012862110818661612,
-      "loss": 0.7465,
-      "step": 26230
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012859383310376808,
-      "loss": 0.6581,
-      "step": 26240
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012856655802092,
-      "loss": 0.6548,
-      "step": 26250
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012853928293807195,
-      "loss": 0.6776,
-      "step": 26260
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012851200785522386,
-      "loss": 0.6471,
-      "step": 26270
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012848473277237582,
-      "loss": 0.6981,
-      "step": 26280
-    },
-    {
-      "epoch": 1.07,
-      "learning_rate": 0.00012845745768952772,
-      "loss": 0.6148,
-      "step": 26290
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012843018260667968,
-      "loss": 0.7,
-      "step": 26300
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012840290752383162,
-      "loss": 0.6996,
-      "step": 26310
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012837563244098355,
-      "loss": 0.6436,
-      "step": 26320
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012834835735813549,
-      "loss": 0.6923,
-      "step": 26330
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012832108227528742,
-      "loss": 0.6882,
-      "step": 26340
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012829380719243935,
-      "loss": 0.6903,
-      "step": 26350
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.0001282665321095913,
-      "loss": 0.6585,
-      "step": 26360
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012823925702674322,
-      "loss": 0.7196,
-      "step": 26370
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012821198194389516,
-      "loss": 0.7095,
-      "step": 26380
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.0001281847068610471,
-      "loss": 0.6866,
-      "step": 26390
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012815743177819902,
-      "loss": 0.6458,
-      "step": 26400
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012813015669535098,
-      "loss": 0.6446,
-      "step": 26410
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.0001281028816125029,
-      "loss": 0.6714,
-      "step": 26420
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012807560652965485,
-      "loss": 0.7181,
-      "step": 26430
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012804833144680676,
-      "loss": 0.6592,
-      "step": 26440
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012802105636395872,
-      "loss": 0.6415,
-      "step": 26450
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012799378128111065,
-      "loss": 0.6552,
-      "step": 26460
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.0001279665061982626,
-      "loss": 0.6655,
-      "step": 26470
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012793923111541452,
-      "loss": 0.6736,
-      "step": 26480
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012791195603256646,
-      "loss": 0.6872,
-      "step": 26490
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.0001278846809497184,
-      "loss": 0.6569,
-      "step": 26500
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012785740586687032,
-      "loss": 0.6286,
-      "step": 26510
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012783013078402226,
-      "loss": 0.7081,
-      "step": 26520
-    },
-    {
-      "epoch": 1.08,
-      "learning_rate": 0.00012780285570117422,
-      "loss": 0.6593,
-      "step": 26530
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012777558061832613,
-      "loss": 0.738,
-      "step": 26540
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001277483055354781,
-      "loss": 0.6901,
-      "step": 26550
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012772103045263,
-      "loss": 0.6265,
-      "step": 26560
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012769375536978195,
-      "loss": 0.6819,
-      "step": 26570
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001276664802869339,
-      "loss": 0.6676,
-      "step": 26580
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012763920520408582,
-      "loss": 0.6785,
-      "step": 26590
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012761193012123776,
-      "loss": 0.6665,
-      "step": 26600
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001275846550383897,
-      "loss": 0.7488,
-      "step": 26610
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012755737995554162,
-      "loss": 0.63,
-      "step": 26620
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012753010487269356,
-      "loss": 0.7662,
-      "step": 26630
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001275028297898455,
-      "loss": 0.7355,
-      "step": 26640
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012747555470699743,
-      "loss": 0.7205,
-      "step": 26650
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012744827962414936,
-      "loss": 0.6482,
-      "step": 26660
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001274210045413013,
-      "loss": 0.607,
-      "step": 26670
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012739372945845323,
-      "loss": 0.7399,
-      "step": 26680
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012736645437560516,
-      "loss": 0.7872,
-      "step": 26690
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012733917929275712,
-      "loss": 0.6167,
-      "step": 26700
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012731190420990903,
-      "loss": 0.7348,
-      "step": 26710
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.000127284629127061,
-      "loss": 0.674,
-      "step": 26720
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001272573540442129,
-      "loss": 0.7236,
-      "step": 26730
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012723007896136486,
-      "loss": 0.7422,
-      "step": 26740
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001272028038785168,
-      "loss": 0.7232,
-      "step": 26750
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012717552879566873,
-      "loss": 0.7968,
-      "step": 26760
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.00012714825371282066,
-      "loss": 0.613,
-      "step": 26770
-    },
-    {
-      "epoch": 1.09,
-      "learning_rate": 0.0001271209786299726,
-      "loss": 0.6943,
-      "step": 26780
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012709370354712453,
-      "loss": 0.7809,
-      "step": 26790
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.0001270664284642765,
-      "loss": 0.6441,
-      "step": 26800
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.0001270391533814284,
-      "loss": 0.6672,
-      "step": 26810
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012701187829858036,
-      "loss": 0.7346,
-      "step": 26820
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012698460321573226,
-      "loss": 0.6764,
-      "step": 26830
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012695732813288422,
-      "loss": 0.7626,
-      "step": 26840
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012693005305003613,
-      "loss": 0.6597,
-      "step": 26850
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012690550547547289,
-      "loss": 0.6661,
-      "step": 26860
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012687823039262482,
-      "loss": 0.6063,
-      "step": 26870
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012685095530977675,
-      "loss": 0.7191,
-      "step": 26880
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.0001268236802269287,
-      "loss": 0.6317,
-      "step": 26890
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012679640514408062,
-      "loss": 0.7419,
-      "step": 26900
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012676913006123255,
-      "loss": 0.7441,
-      "step": 26910
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012674185497838452,
-      "loss": 0.6882,
-      "step": 26920
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012671457989553642,
-      "loss": 0.7676,
-      "step": 26930
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012668730481268838,
-      "loss": 0.6591,
-      "step": 26940
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.0001266600297298403,
-      "loss": 0.6549,
-      "step": 26950
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012663275464699225,
-      "loss": 0.7058,
-      "step": 26960
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012660547956414419,
-      "loss": 0.709,
-      "step": 26970
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012657820448129612,
-      "loss": 0.7011,
-      "step": 26980
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012655092939844805,
-      "loss": 0.7516,
-      "step": 26990
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.0001265236543156,
-      "loss": 0.6908,
-      "step": 27000
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012649637923275192,
-      "loss": 0.632,
-      "step": 27010
-    },
-    {
-      "epoch": 1.1,
-      "learning_rate": 0.00012646910414990388,
-      "loss": 0.5599,
-      "step": 27020
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001264418290670558,
-      "loss": 0.7122,
-      "step": 27030
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012641455398420775,
-      "loss": 0.7006,
-      "step": 27040
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012638727890135966,
-      "loss": 0.8002,
-      "step": 27050
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012636000381851162,
-      "loss": 0.6401,
-      "step": 27060
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012633272873566352,
-      "loss": 0.7363,
-      "step": 27070
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012630545365281549,
-      "loss": 0.673,
-      "step": 27080
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012627817856996742,
-      "loss": 0.6464,
-      "step": 27090
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012625090348711935,
-      "loss": 0.7146,
-      "step": 27100
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001262236284042713,
-      "loss": 0.6731,
-      "step": 27110
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012619635332142322,
-      "loss": 0.6323,
-      "step": 27120
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012616907823857516,
-      "loss": 0.6997,
-      "step": 27130
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001261418031557271,
-      "loss": 0.6898,
-      "step": 27140
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012611452807287902,
-      "loss": 0.6583,
-      "step": 27150
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012608725299003096,
-      "loss": 0.5569,
-      "step": 27160
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001260599779071829,
-      "loss": 0.8005,
-      "step": 27170
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012603270282433482,
-      "loss": 0.7631,
-      "step": 27180
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012600542774148679,
-      "loss": 0.6353,
-      "step": 27190
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001259781526586387,
-      "loss": 0.7275,
-      "step": 27200
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012595087757579065,
-      "loss": 0.7183,
-      "step": 27210
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012592360249294256,
-      "loss": 0.7896,
-      "step": 27220
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012589632741009452,
-      "loss": 0.731,
-      "step": 27230
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012586905232724643,
-      "loss": 0.7543,
-      "step": 27240
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.0001258417772443984,
-      "loss": 0.733,
-      "step": 27250
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012581450216155032,
-      "loss": 0.7301,
-      "step": 27260
-    },
-    {
-      "epoch": 1.11,
-      "learning_rate": 0.00012578722707870226,
-      "loss": 0.7143,
-      "step": 27270
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001257599519958542,
-      "loss": 0.8294,
-      "step": 27280
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012573267691300613,
-      "loss": 0.7248,
-      "step": 27290
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012570540183015806,
-      "loss": 0.7454,
-      "step": 27300
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012567812674731002,
-      "loss": 0.6998,
-      "step": 27310
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012565085166446193,
-      "loss": 0.6696,
-      "step": 27320
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001256235765816139,
-      "loss": 0.7649,
-      "step": 27330
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001255963014987658,
-      "loss": 0.6705,
-      "step": 27340
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012556902641591776,
-      "loss": 0.6878,
-      "step": 27350
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001255417513330697,
-      "loss": 0.6834,
-      "step": 27360
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012551447625022162,
-      "loss": 0.7429,
-      "step": 27370
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012548720116737356,
-      "loss": 0.6097,
-      "step": 27380
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001254599260845255,
-      "loss": 0.744,
-      "step": 27390
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012543265100167743,
-      "loss": 0.6893,
-      "step": 27400
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012540537591882936,
-      "loss": 0.6233,
-      "step": 27410
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001253781008359813,
-      "loss": 0.6891,
-      "step": 27420
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012535082575313323,
-      "loss": 0.7191,
-      "step": 27430
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012532355067028516,
-      "loss": 0.7,
-      "step": 27440
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001252962755874371,
-      "loss": 0.7476,
-      "step": 27450
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012526900050458903,
-      "loss": 0.65,
-      "step": 27460
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012524172542174096,
-      "loss": 0.7536,
-      "step": 27470
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012521445033889292,
-      "loss": 0.8043,
-      "step": 27480
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.00012518717525604483,
-      "loss": 0.6934,
-      "step": 27490
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001251599001731968,
-      "loss": 0.7716,
-      "step": 27500
-    },
-    {
-      "epoch": 1.12,
-      "learning_rate": 0.0001251326250903487,
-      "loss": 0.6359,
-      "step": 27510
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012510535000750066,
-      "loss": 0.6869,
-      "step": 27520
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001250780749246526,
-      "loss": 0.7089,
-      "step": 27530
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012505079984180453,
-      "loss": 0.6348,
-      "step": 27540
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012502352475895646,
-      "loss": 0.7036,
-      "step": 27550
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001249962496761084,
-      "loss": 0.7373,
-      "step": 27560
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012496897459326033,
-      "loss": 0.7194,
-      "step": 27570
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012494169951041226,
-      "loss": 0.7283,
-      "step": 27580
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001249144244275642,
-      "loss": 0.7708,
-      "step": 27590
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012488714934471616,
-      "loss": 0.8126,
-      "step": 27600
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012485987426186806,
-      "loss": 0.6351,
-      "step": 27610
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012483259917902003,
-      "loss": 0.6826,
-      "step": 27620
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012480532409617193,
-      "loss": 0.6293,
-      "step": 27630
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001247780490133239,
-      "loss": 0.6681,
-      "step": 27640
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012475077393047583,
-      "loss": 0.7689,
-      "step": 27650
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012472349884762776,
-      "loss": 0.6841,
-      "step": 27660
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001246962237647797,
-      "loss": 0.6634,
-      "step": 27670
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012466894868193163,
-      "loss": 0.726,
-      "step": 27680
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012464167359908356,
-      "loss": 0.6821,
-      "step": 27690
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001246143985162355,
-      "loss": 0.7627,
-      "step": 27700
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012458712343338743,
-      "loss": 0.7524,
-      "step": 27710
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012455984835053937,
-      "loss": 0.681,
-      "step": 27720
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001245325732676913,
-      "loss": 0.6424,
-      "step": 27730
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.00012450529818484323,
-      "loss": 0.6615,
-      "step": 27740
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001244780231019952,
-      "loss": 0.6727,
-      "step": 27750
-    },
-    {
-      "epoch": 1.13,
-      "learning_rate": 0.0001244507480191471,
-      "loss": 0.718,
-      "step": 27760
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012442347293629906,
-      "loss": 0.7121,
-      "step": 27770
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012439619785345097,
-      "loss": 0.6778,
-      "step": 27780
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012436892277060293,
-      "loss": 0.7634,
-      "step": 27790
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012434164768775484,
-      "loss": 0.6482,
-      "step": 27800
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001243143726049068,
-      "loss": 0.6003,
-      "step": 27810
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012428709752205873,
-      "loss": 0.7427,
-      "step": 27820
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012425982243921067,
-      "loss": 0.6397,
-      "step": 27830
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001242325473563626,
-      "loss": 0.643,
-      "step": 27840
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012420527227351453,
-      "loss": 0.7102,
-      "step": 27850
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012417799719066647,
-      "loss": 0.6359,
-      "step": 27860
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012415072210781843,
-      "loss": 0.7674,
-      "step": 27870
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012412344702497033,
-      "loss": 0.7271,
-      "step": 27880
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001240961719421223,
-      "loss": 0.6835,
-      "step": 27890
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001240688968592742,
-      "loss": 0.7094,
-      "step": 27900
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012404162177642616,
-      "loss": 0.736,
-      "step": 27910
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001240143466935781,
-      "loss": 0.6724,
-      "step": 27920
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012398707161073003,
-      "loss": 0.7031,
-      "step": 27930
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012395979652788197,
-      "loss": 0.7466,
-      "step": 27940
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001239325214450339,
-      "loss": 0.7879,
-      "step": 27950
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012390524636218583,
-      "loss": 0.6036,
-      "step": 27960
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012387797127933777,
-      "loss": 0.6847,
-      "step": 27970
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.0001238506961964897,
-      "loss": 0.7041,
-      "step": 27980
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012382342111364164,
-      "loss": 0.6485,
-      "step": 27990
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00012379614603079357,
-      "loss": 0.6483,
-      "step": 28000
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001237688709479455,
-      "loss": 0.7035,
-      "step": 28010
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012374159586509744,
-      "loss": 0.6539,
-      "step": 28020
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012371432078224937,
-      "loss": 0.7418,
-      "step": 28030
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012368704569940133,
-      "loss": 0.7159,
-      "step": 28040
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012365977061655324,
-      "loss": 0.7013,
-      "step": 28050
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001236324955337052,
-      "loss": 0.5667,
-      "step": 28060
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001236052204508571,
-      "loss": 0.6779,
-      "step": 28070
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012357794536800907,
-      "loss": 0.6459,
-      "step": 28080
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.000123550670285161,
-      "loss": 0.6433,
-      "step": 28090
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012352339520231294,
-      "loss": 0.7574,
-      "step": 28100
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012349612011946487,
-      "loss": 0.7874,
-      "step": 28110
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001234688450366168,
-      "loss": 0.7249,
-      "step": 28120
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012344156995376874,
-      "loss": 0.5934,
-      "step": 28130
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012341429487092067,
-      "loss": 0.678,
-      "step": 28140
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001233870197880726,
-      "loss": 0.6994,
-      "step": 28150
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012335974470522457,
-      "loss": 0.7108,
-      "step": 28160
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012333246962237647,
-      "loss": 0.6922,
-      "step": 28170
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012330519453952843,
-      "loss": 0.6731,
-      "step": 28180
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012327791945668034,
-      "loss": 0.6758,
-      "step": 28190
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001232506443738323,
-      "loss": 0.6531,
-      "step": 28200
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012322336929098424,
-      "loss": 0.7061,
-      "step": 28210
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012319609420813617,
-      "loss": 0.684,
-      "step": 28220
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.0001231688191252881,
-      "loss": 0.6811,
-      "step": 28230
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012314154404244004,
-      "loss": 0.6585,
-      "step": 28240
-    },
-    {
-      "epoch": 1.15,
-      "learning_rate": 0.00012311426895959197,
-      "loss": 0.6205,
-      "step": 28250
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001230869938767439,
-      "loss": 0.7208,
-      "step": 28260
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012305971879389584,
-      "loss": 0.625,
-      "step": 28270
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012303244371104777,
-      "loss": 0.6367,
-      "step": 28280
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001230051686281997,
-      "loss": 0.6244,
-      "step": 28290
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012297789354535164,
-      "loss": 0.6955,
-      "step": 28300
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012295061846250357,
-      "loss": 0.7363,
-      "step": 28310
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001229233433796555,
-      "loss": 0.6412,
-      "step": 28320
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012289606829680747,
-      "loss": 0.7591,
-      "step": 28330
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012286879321395938,
-      "loss": 0.6804,
-      "step": 28340
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012284151813111134,
-      "loss": 0.654,
-      "step": 28350
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012281424304826324,
-      "loss": 0.7578,
-      "step": 28360
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001227869679654152,
-      "loss": 0.6591,
-      "step": 28370
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012275969288256714,
-      "loss": 0.6334,
-      "step": 28380
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012273241779971907,
-      "loss": 0.7376,
-      "step": 28390
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.000122705142716871,
-      "loss": 0.6229,
-      "step": 28400
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012267786763402294,
-      "loss": 0.6063,
-      "step": 28410
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012265059255117487,
-      "loss": 0.6328,
-      "step": 28420
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012262331746832684,
-      "loss": 0.7966,
-      "step": 28430
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012259604238547874,
-      "loss": 0.686,
-      "step": 28440
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001225687673026307,
-      "loss": 0.6949,
-      "step": 28450
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.0001225414922197826,
-      "loss": 0.7803,
-      "step": 28460
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012251421713693457,
-      "loss": 0.6533,
-      "step": 28470
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012248694205408648,
-      "loss": 0.6384,
-      "step": 28480
-    },
-    {
-      "epoch": 1.16,
-      "learning_rate": 0.00012246239447952323,
-      "loss": 0.6953,
-      "step": 28490
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012243511939667517,
-      "loss": 0.6594,
-      "step": 28500
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.0001224078443138271,
-      "loss": 0.6002,
-      "step": 28510
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012238056923097903,
-      "loss": 0.708,
-      "step": 28520
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012235329414813097,
-      "loss": 0.665,
-      "step": 28530
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.0001223260190652829,
-      "loss": 0.6753,
-      "step": 28540
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012229874398243486,
-      "loss": 0.6629,
-      "step": 28550
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012227146889958677,
-      "loss": 0.6734,
-      "step": 28560
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012224419381673873,
-      "loss": 0.7622,
-      "step": 28570
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012221691873389064,
-      "loss": 0.6507,
-      "step": 28580
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.0001221896436510426,
-      "loss": 0.7613,
-      "step": 28590
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012216236856819453,
-      "loss": 0.6843,
-      "step": 28600
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012213509348534647,
-      "loss": 0.7017,
-      "step": 28610
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.0001221078184024984,
-      "loss": 0.6323,
-      "step": 28620
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012208054331965033,
-      "loss": 0.7453,
-      "step": 28630
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012205326823680228,
-      "loss": 0.7226,
-      "step": 28640
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012202599315395422,
-      "loss": 0.6484,
-      "step": 28650
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012199871807110615,
-      "loss": 0.6178,
-      "step": 28660
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012197144298825808,
-      "loss": 0.6579,
-      "step": 28670
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012194416790541002,
-      "loss": 0.7564,
-      "step": 28680
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012191689282256195,
-      "loss": 0.7136,
-      "step": 28690
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012188961773971389,
-      "loss": 0.6098,
-      "step": 28700
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012186234265686582,
-      "loss": 0.6926,
-      "step": 28710
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012183506757401777,
-      "loss": 0.7294,
-      "step": 28720
-    },
-    {
-      "epoch": 1.17,
-      "learning_rate": 0.00012180779249116969,
-      "loss": 0.8162,
-      "step": 28730
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012178051740832163,
-      "loss": 0.7238,
-      "step": 28740
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012175324232547356,
-      "loss": 0.7511,
-      "step": 28750
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.0001217259672426255,
-      "loss": 0.7267,
-      "step": 28760
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012169869215977745,
-      "loss": 0.6945,
-      "step": 28770
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012167141707692937,
-      "loss": 0.7227,
-      "step": 28780
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012164414199408132,
-      "loss": 0.591,
-      "step": 28790
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012161686691123324,
-      "loss": 0.6502,
-      "step": 28800
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012158959182838519,
-      "loss": 0.7758,
-      "step": 28810
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012156231674553713,
-      "loss": 0.7995,
-      "step": 28820
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012153504166268905,
-      "loss": 0.7346,
-      "step": 28830
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.000121507766579841,
-      "loss": 0.5983,
-      "step": 28840
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012148049149699292,
-      "loss": 0.6597,
-      "step": 28850
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012145321641414487,
-      "loss": 0.747,
-      "step": 28860
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012142594133129679,
-      "loss": 0.651,
-      "step": 28870
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012139866624844874,
-      "loss": 0.7095,
-      "step": 28880
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012137139116560068,
-      "loss": 0.7113,
-      "step": 28890
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.0001213441160827526,
-      "loss": 0.6687,
-      "step": 28900
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012131684099990455,
-      "loss": 0.8104,
-      "step": 28910
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012128956591705647,
-      "loss": 0.6248,
-      "step": 28920
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012126229083420842,
-      "loss": 0.5972,
-      "step": 28930
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012123501575136035,
-      "loss": 0.8084,
-      "step": 28940
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012120774066851229,
-      "loss": 0.6339,
-      "step": 28950
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012118046558566422,
-      "loss": 0.7292,
-      "step": 28960
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012115319050281616,
-      "loss": 0.746,
-      "step": 28970
-    },
-    {
-      "epoch": 1.18,
-      "learning_rate": 0.00012112591541996809,
-      "loss": 0.691,
-      "step": 28980
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012109864033712004,
-      "loss": 0.6429,
-      "step": 28990
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012107136525427196,
-      "loss": 0.7797,
-      "step": 29000
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.0001210440901714239,
-      "loss": 0.667,
-      "step": 29010
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012101681508857583,
-      "loss": 0.669,
-      "step": 29020
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012098954000572777,
-      "loss": 0.7591,
-      "step": 29030
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012096226492287972,
-      "loss": 0.6911,
-      "step": 29040
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012093498984003164,
-      "loss": 0.6761,
-      "step": 29050
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012090771475718359,
-      "loss": 0.7027,
-      "step": 29060
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012088043967433551,
-      "loss": 0.7277,
-      "step": 29070
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012085316459148746,
-      "loss": 0.7097,
-      "step": 29080
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012082588950863938,
-      "loss": 0.7434,
-      "step": 29090
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012079861442579132,
-      "loss": 0.6689,
-      "step": 29100
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012077133934294327,
-      "loss": 0.6218,
-      "step": 29110
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012074406426009519,
-      "loss": 0.6271,
-      "step": 29120
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012071678917724714,
-      "loss": 0.6968,
-      "step": 29130
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012068951409439906,
-      "loss": 0.748,
-      "step": 29140
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012066223901155101,
-      "loss": 0.6728,
-      "step": 29150
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012063496392870295,
-      "loss": 0.664,
-      "step": 29160
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012060768884585487,
-      "loss": 0.7166,
-      "step": 29170
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012058041376300682,
-      "loss": 0.7984,
-      "step": 29180
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012055313868015874,
-      "loss": 0.7149,
-      "step": 29190
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012052586359731069,
-      "loss": 0.6891,
-      "step": 29200
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012049858851446262,
-      "loss": 0.7241,
-      "step": 29210
-    },
-    {
-      "epoch": 1.19,
-      "learning_rate": 0.00012047131343161456,
-      "loss": 0.7083,
-      "step": 29220
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012044403834876649,
-      "loss": 0.7808,
-      "step": 29230
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012041676326591843,
-      "loss": 0.6696,
-      "step": 29240
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012038948818307036,
-      "loss": 0.6928,
-      "step": 29250
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.0001203622131002223,
-      "loss": 0.7209,
-      "step": 29260
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012033493801737423,
-      "loss": 0.7243,
-      "step": 29270
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012030766293452617,
-      "loss": 0.7699,
-      "step": 29280
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.0001202803878516781,
-      "loss": 0.7003,
-      "step": 29290
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012025311276883004,
-      "loss": 0.6593,
-      "step": 29300
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012022583768598196,
-      "loss": 0.6665,
-      "step": 29310
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012019856260313391,
-      "loss": 0.6665,
-      "step": 29320
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012017128752028586,
-      "loss": 0.6831,
-      "step": 29330
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012014401243743778,
-      "loss": 0.6318,
-      "step": 29340
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012011673735458973,
-      "loss": 0.6536,
-      "step": 29350
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012008946227174165,
-      "loss": 0.6414,
-      "step": 29360
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.0001200621871888936,
-      "loss": 0.6315,
-      "step": 29370
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012003491210604554,
-      "loss": 0.7247,
-      "step": 29380
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00012000763702319746,
-      "loss": 0.6474,
-      "step": 29390
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011998036194034941,
-      "loss": 0.6653,
-      "step": 29400
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011995308685750133,
-      "loss": 0.699,
-      "step": 29410
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011992581177465328,
-      "loss": 0.682,
-      "step": 29420
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.0001198985366918052,
-      "loss": 0.7699,
-      "step": 29430
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011987126160895714,
-      "loss": 0.7398,
-      "step": 29440
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011984398652610909,
-      "loss": 0.6697,
-      "step": 29450
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011981671144326101,
-      "loss": 0.6933,
-      "step": 29460
-    },
-    {
-      "epoch": 1.2,
-      "learning_rate": 0.00011978943636041296,
-      "loss": 0.6584,
-      "step": 29470
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011976216127756488,
-      "loss": 0.6059,
-      "step": 29480
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011973488619471683,
-      "loss": 0.7055,
-      "step": 29490
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011970761111186876,
-      "loss": 0.6618,
-      "step": 29500
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.0001196803360290207,
-      "loss": 0.7123,
-      "step": 29510
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011965306094617263,
-      "loss": 0.6137,
-      "step": 29520
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011962578586332456,
-      "loss": 0.7599,
-      "step": 29530
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.0001195985107804765,
-      "loss": 0.7851,
-      "step": 29540
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011957123569762844,
-      "loss": 0.7848,
-      "step": 29550
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011954396061478037,
-      "loss": 0.649,
-      "step": 29560
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011951668553193231,
-      "loss": 0.6456,
-      "step": 29570
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011948941044908423,
-      "loss": 0.7581,
-      "step": 29580
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011946213536623618,
-      "loss": 0.7502,
-      "step": 29590
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.0001194348602833881,
-      "loss": 0.7317,
-      "step": 29600
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011940758520054005,
-      "loss": 0.6935,
-      "step": 29610
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.000119380310117692,
-      "loss": 0.7184,
-      "step": 29620
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011935303503484392,
-      "loss": 0.795,
-      "step": 29630
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011932575995199586,
-      "loss": 0.6802,
-      "step": 29640
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011929848486914778,
-      "loss": 0.6969,
-      "step": 29650
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011927120978629973,
-      "loss": 0.7249,
-      "step": 29660
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011924393470345168,
-      "loss": 0.7138,
-      "step": 29670
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.0001192166596206036,
-      "loss": 0.6982,
-      "step": 29680
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011918938453775555,
-      "loss": 0.6909,
-      "step": 29690
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011916210945490747,
-      "loss": 0.7,
-      "step": 29700
-    },
-    {
-      "epoch": 1.21,
-      "learning_rate": 0.00011913483437205941,
-      "loss": 0.7535,
-      "step": 29710
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011910755928921136,
-      "loss": 0.6578,
-      "step": 29720
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011908028420636328,
-      "loss": 0.6402,
-      "step": 29730
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011905300912351523,
-      "loss": 0.5988,
-      "step": 29740
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011902573404066715,
-      "loss": 0.7739,
-      "step": 29750
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.0001189984589578191,
-      "loss": 0.7245,
-      "step": 29760
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011897118387497102,
-      "loss": 0.6349,
-      "step": 29770
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011894390879212297,
-      "loss": 0.7236,
-      "step": 29780
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.0001189166337092749,
-      "loss": 0.6532,
-      "step": 29790
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011888935862642683,
-      "loss": 0.7198,
-      "step": 29800
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011886208354357877,
-      "loss": 0.589,
-      "step": 29810
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.0001188348084607307,
-      "loss": 0.7562,
-      "step": 29820
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011880753337788264,
-      "loss": 0.7854,
-      "step": 29830
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011878025829503458,
-      "loss": 0.7001,
-      "step": 29840
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.0001187529832121865,
-      "loss": 0.6253,
-      "step": 29850
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011872570812933845,
-      "loss": 0.7227,
-      "step": 29860
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011869843304649037,
-      "loss": 0.6637,
-      "step": 29870
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011867115796364232,
-      "loss": 0.8029,
-      "step": 29880
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011864388288079427,
-      "loss": 0.7716,
-      "step": 29890
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011861660779794619,
-      "loss": 0.6971,
-      "step": 29900
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011858933271509813,
-      "loss": 0.7079,
-      "step": 29910
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011856205763225005,
-      "loss": 0.6714,
-      "step": 29920
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.000118534782549402,
-      "loss": 0.6675,
-      "step": 29930
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011850750746655395,
-      "loss": 0.7328,
-      "step": 29940
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011848023238370587,
-      "loss": 0.6677,
-      "step": 29950
-    },
-    {
-      "epoch": 1.22,
-      "learning_rate": 0.00011845295730085782,
-      "loss": 0.6166,
-      "step": 29960
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011842568221800974,
-      "loss": 0.7184,
-      "step": 29970
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011839840713516168,
-      "loss": 0.7419,
-      "step": 29980
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.0001183711320523136,
-      "loss": 0.6136,
-      "step": 29990
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011834385696946555,
-      "loss": 0.6557,
-      "step": 30000
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.0001183165818866175,
-      "loss": 0.6682,
-      "step": 30010
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011828930680376942,
-      "loss": 0.6574,
-      "step": 30020
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011826203172092137,
-      "loss": 0.5762,
-      "step": 30030
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011823475663807329,
-      "loss": 0.7133,
-      "step": 30040
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011820748155522524,
-      "loss": 0.7526,
-      "step": 30050
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011818020647237717,
-      "loss": 0.6728,
-      "step": 30060
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.0001181529313895291,
-      "loss": 0.6803,
-      "step": 30070
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011812565630668104,
-      "loss": 0.7144,
-      "step": 30080
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011809838122383297,
-      "loss": 0.7212,
-      "step": 30090
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.0001180711061409849,
-      "loss": 0.7467,
-      "step": 30100
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011804383105813685,
-      "loss": 0.6885,
-      "step": 30110
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011801655597528877,
-      "loss": 0.6942,
-      "step": 30120
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011798928089244072,
-      "loss": 0.6305,
-      "step": 30130
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011796200580959264,
-      "loss": 0.7101,
-      "step": 30140
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011793473072674459,
-      "loss": 0.6924,
-      "step": 30150
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011790745564389651,
-      "loss": 0.7279,
-      "step": 30160
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011788018056104846,
-      "loss": 0.8034,
-      "step": 30170
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.0001178529054782004,
-      "loss": 0.677,
-      "step": 30180
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011782563039535232,
-      "loss": 0.6252,
-      "step": 30190
-    },
-    {
-      "epoch": 1.23,
-      "learning_rate": 0.00011779835531250427,
-      "loss": 0.7189,
-      "step": 30200
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011777108022965619,
-      "loss": 0.7866,
-      "step": 30210
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011774380514680814,
-      "loss": 0.6971,
-      "step": 30220
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011771653006396009,
-      "loss": 0.5874,
-      "step": 30230
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011768925498111201,
-      "loss": 0.6803,
-      "step": 30240
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011766197989826395,
-      "loss": 0.6012,
-      "step": 30250
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011763470481541588,
-      "loss": 0.6172,
-      "step": 30260
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011760742973256782,
-      "loss": 0.6732,
-      "step": 30270
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011758015464971977,
-      "loss": 0.6761,
-      "step": 30280
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011755287956687169,
-      "loss": 0.6035,
-      "step": 30290
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011752560448402364,
-      "loss": 0.707,
-      "step": 30300
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011749832940117556,
-      "loss": 0.7022,
-      "step": 30310
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.0001174710543183275,
-      "loss": 0.632,
-      "step": 30320
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011744377923547943,
-      "loss": 0.7578,
-      "step": 30330
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011741650415263137,
-      "loss": 0.7757,
-      "step": 30340
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011738922906978331,
-      "loss": 0.7117,
-      "step": 30350
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011736195398693524,
-      "loss": 0.7192,
-      "step": 30360
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011733467890408718,
-      "loss": 0.7405,
-      "step": 30370
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011730740382123911,
-      "loss": 0.6777,
-      "step": 30380
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011728012873839104,
-      "loss": 0.6951,
-      "step": 30390
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011725285365554299,
-      "loss": 0.5793,
-      "step": 30400
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011722557857269491,
-      "loss": 0.7132,
-      "step": 30410
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011719830348984686,
-      "loss": 0.7623,
-      "step": 30420
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011717102840699878,
-      "loss": 0.6945,
-      "step": 30430
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011714375332415073,
-      "loss": 0.71,
-      "step": 30440
-    },
-    {
-      "epoch": 1.24,
-      "learning_rate": 0.00011711647824130267,
-      "loss": 0.6511,
-      "step": 30450
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.0001170892031584546,
-      "loss": 0.6939,
-      "step": 30460
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011706192807560654,
-      "loss": 0.7557,
-      "step": 30470
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011703465299275846,
-      "loss": 0.7246,
-      "step": 30480
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011700737790991041,
-      "loss": 0.7622,
-      "step": 30490
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011698010282706233,
-      "loss": 0.651,
-      "step": 30500
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011695282774421428,
-      "loss": 0.7694,
-      "step": 30510
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011692555266136622,
-      "loss": 0.7504,
-      "step": 30520
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011689827757851815,
-      "loss": 0.7325,
-      "step": 30530
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011687100249567009,
-      "loss": 0.7251,
-      "step": 30540
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011684372741282201,
-      "loss": 0.7515,
-      "step": 30550
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011681645232997396,
-      "loss": 0.6262,
-      "step": 30560
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011678917724712591,
-      "loss": 0.755,
-      "step": 30570
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011676190216427783,
-      "loss": 0.7872,
-      "step": 30580
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011673462708142978,
-      "loss": 0.682,
-      "step": 30590
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.0001167073519985817,
-      "loss": 0.6484,
-      "step": 30600
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011668007691573364,
-      "loss": 0.6746,
-      "step": 30610
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011665280183288558,
-      "loss": 0.7013,
-      "step": 30620
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011662552675003751,
-      "loss": 0.697,
-      "step": 30630
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011659825166718945,
-      "loss": 0.7367,
-      "step": 30640
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011657097658434138,
-      "loss": 0.6954,
-      "step": 30650
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011654370150149331,
-      "loss": 0.7988,
-      "step": 30660
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011651642641864525,
-      "loss": 0.7612,
-      "step": 30670
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011648915133579718,
-      "loss": 0.7421,
-      "step": 30680
-    },
-    {
-      "epoch": 1.25,
-      "learning_rate": 0.00011646187625294913,
-      "loss": 0.7751,
-      "step": 30690
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011643460117010105,
-      "loss": 0.7372,
-      "step": 30700
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.000116407326087253,
-      "loss": 0.7193,
-      "step": 30710
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011638005100440492,
-      "loss": 0.6493,
-      "step": 30720
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011635277592155686,
-      "loss": 0.7047,
-      "step": 30730
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011632550083870881,
-      "loss": 0.6897,
-      "step": 30740
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011629822575586073,
-      "loss": 0.7522,
-      "step": 30750
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011627095067301268,
-      "loss": 0.6521,
-      "step": 30760
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.0001162436755901646,
-      "loss": 0.6636,
-      "step": 30770
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011621640050731655,
-      "loss": 0.7022,
-      "step": 30780
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.0001161891254244685,
-      "loss": 0.7022,
-      "step": 30790
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011616185034162042,
-      "loss": 0.8098,
-      "step": 30800
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011613457525877236,
-      "loss": 0.6369,
-      "step": 30810
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011610730017592428,
-      "loss": 0.7302,
-      "step": 30820
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011608002509307623,
-      "loss": 0.6863,
-      "step": 30830
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011605275001022815,
-      "loss": 0.7105,
-      "step": 30840
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.0001160254749273801,
-      "loss": 0.6884,
-      "step": 30850
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011599819984453205,
-      "loss": 0.7112,
-      "step": 30860
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011597092476168397,
-      "loss": 0.6466,
-      "step": 30870
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011594364967883591,
-      "loss": 0.6664,
-      "step": 30880
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011591637459598783,
-      "loss": 0.7734,
-      "step": 30890
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011588909951313978,
-      "loss": 0.7703,
-      "step": 30900
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011586182443029172,
-      "loss": 0.6779,
-      "step": 30910
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011583454934744365,
-      "loss": 0.7033,
-      "step": 30920
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011580727426459558,
-      "loss": 0.7753,
-      "step": 30930
-    },
-    {
-      "epoch": 1.26,
-      "learning_rate": 0.00011577999918174752,
-      "loss": 0.6735,
-      "step": 30940
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011575272409889945,
-      "loss": 0.665,
-      "step": 30950
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.0001157254490160514,
-      "loss": 0.6457,
-      "step": 30960
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011569817393320332,
-      "loss": 0.6726,
-      "step": 30970
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011567089885035527,
-      "loss": 0.7993,
-      "step": 30980
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011564362376750719,
-      "loss": 0.6849,
-      "step": 30990
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011561634868465913,
-      "loss": 0.682,
-      "step": 31000
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011558907360181108,
-      "loss": 0.7518,
-      "step": 31010
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.000115561798518963,
-      "loss": 0.6647,
-      "step": 31020
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011553452343611495,
-      "loss": 0.6091,
-      "step": 31030
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011550724835326687,
-      "loss": 0.7391,
-      "step": 31040
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011547997327041882,
-      "loss": 0.6997,
-      "step": 31050
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011545269818757074,
-      "loss": 0.7121,
-      "step": 31060
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011542542310472269,
-      "loss": 0.7717,
-      "step": 31070
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011539814802187463,
-      "loss": 0.8039,
-      "step": 31080
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011537087293902655,
-      "loss": 0.7678,
-      "step": 31090
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.0001153435978561785,
-      "loss": 0.7761,
-      "step": 31100
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011531632277333042,
-      "loss": 0.5756,
-      "step": 31110
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011528904769048237,
-      "loss": 0.681,
-      "step": 31120
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011526177260763432,
-      "loss": 0.6888,
-      "step": 31130
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011523449752478624,
-      "loss": 0.7241,
-      "step": 31140
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011520722244193818,
-      "loss": 0.6278,
-      "step": 31150
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.0001151799473590901,
-      "loss": 0.7779,
-      "step": 31160
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011515267227624205,
-      "loss": 0.7089,
-      "step": 31170
-    },
-    {
-      "epoch": 1.27,
-      "learning_rate": 0.00011512539719339399,
-      "loss": 0.6664,
-      "step": 31180
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011509812211054592,
-      "loss": 0.6505,
-      "step": 31190
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011507084702769785,
-      "loss": 0.652,
-      "step": 31200
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011504357194484979,
-      "loss": 0.6415,
-      "step": 31210
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011501629686200172,
-      "loss": 0.6198,
-      "step": 31220
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011498902177915366,
-      "loss": 0.6831,
-      "step": 31230
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011496174669630559,
-      "loss": 0.7026,
-      "step": 31240
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011493447161345754,
-      "loss": 0.7203,
-      "step": 31250
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011490719653060946,
-      "loss": 0.7579,
-      "step": 31260
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.0001148799214477614,
-      "loss": 0.6699,
-      "step": 31270
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011485264636491332,
-      "loss": 0.7282,
-      "step": 31280
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011482537128206527,
-      "loss": 0.7009,
-      "step": 31290
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011479809619921722,
-      "loss": 0.6554,
-      "step": 31300
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011477082111636914,
-      "loss": 0.7604,
-      "step": 31310
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011474354603352109,
-      "loss": 0.6032,
-      "step": 31320
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011471627095067301,
-      "loss": 0.7052,
-      "step": 31330
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011468899586782496,
-      "loss": 0.712,
-      "step": 31340
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.0001146617207849769,
-      "loss": 0.6777,
-      "step": 31350
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011463444570212882,
-      "loss": 0.7213,
-      "step": 31360
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011460717061928077,
-      "loss": 0.6366,
-      "step": 31370
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011457989553643269,
-      "loss": 0.6917,
-      "step": 31380
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011455262045358464,
-      "loss": 0.6442,
-      "step": 31390
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011452534537073656,
-      "loss": 0.6716,
-      "step": 31400
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.0001144980702878885,
-      "loss": 0.7197,
-      "step": 31410
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011447079520504045,
-      "loss": 0.6539,
-      "step": 31420
-    },
-    {
-      "epoch": 1.28,
-      "learning_rate": 0.00011444352012219237,
-      "loss": 0.6442,
-      "step": 31430
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011441624503934432,
-      "loss": 0.7379,
-      "step": 31440
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011438896995649624,
-      "loss": 0.6724,
-      "step": 31450
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011436169487364819,
-      "loss": 0.7141,
-      "step": 31460
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011433441979080012,
-      "loss": 0.6761,
-      "step": 31470
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011430714470795206,
-      "loss": 0.6496,
-      "step": 31480
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011427986962510399,
-      "loss": 0.6545,
-      "step": 31490
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011425259454225593,
-      "loss": 0.7429,
-      "step": 31500
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011422531945940786,
-      "loss": 0.6817,
-      "step": 31510
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.0001141980443765598,
-      "loss": 0.7125,
-      "step": 31520
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011417076929371173,
-      "loss": 0.6789,
-      "step": 31530
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011414349421086367,
-      "loss": 0.6485,
-      "step": 31540
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.0001141162191280156,
-      "loss": 0.7426,
-      "step": 31550
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011408894404516754,
-      "loss": 0.6827,
-      "step": 31560
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011406166896231946,
-      "loss": 0.7049,
-      "step": 31570
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011403439387947141,
-      "loss": 0.6483,
-      "step": 31580
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011400711879662336,
-      "loss": 0.6443,
-      "step": 31590
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011397984371377528,
-      "loss": 0.7444,
-      "step": 31600
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011395256863092723,
-      "loss": 0.7047,
-      "step": 31610
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011392529354807915,
-      "loss": 0.7058,
-      "step": 31620
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.0001138980184652311,
-      "loss": 0.7415,
-      "step": 31630
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011387074338238304,
-      "loss": 0.7359,
-      "step": 31640
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011384346829953496,
-      "loss": 0.6789,
-      "step": 31650
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011381619321668691,
-      "loss": 0.6482,
-      "step": 31660
-    },
-    {
-      "epoch": 1.29,
-      "learning_rate": 0.00011378891813383883,
-      "loss": 0.689,
-      "step": 31670
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011376164305099078,
-      "loss": 0.5996,
-      "step": 31680
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011373436796814272,
-      "loss": 0.7195,
-      "step": 31690
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011370709288529464,
-      "loss": 0.6978,
-      "step": 31700
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011367981780244659,
-      "loss": 0.7154,
-      "step": 31710
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011365254271959851,
-      "loss": 0.5577,
-      "step": 31720
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011362526763675046,
-      "loss": 0.6543,
-      "step": 31730
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011359799255390238,
-      "loss": 0.8034,
-      "step": 31740
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011357071747105433,
-      "loss": 0.6604,
-      "step": 31750
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011354344238820626,
-      "loss": 0.7051,
-      "step": 31760
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.0001135161673053582,
-      "loss": 0.6474,
-      "step": 31770
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011348889222251013,
-      "loss": 0.6294,
-      "step": 31780
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011346161713966206,
-      "loss": 0.7124,
-      "step": 31790
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.000113434342056814,
-      "loss": 0.8107,
-      "step": 31800
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011340706697396594,
-      "loss": 0.6375,
-      "step": 31810
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011337979189111786,
-      "loss": 0.6876,
-      "step": 31820
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011335251680826981,
-      "loss": 0.6753,
-      "step": 31830
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011332524172542173,
-      "loss": 0.7379,
-      "step": 31840
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011329796664257368,
-      "loss": 0.697,
-      "step": 31850
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011327069155972563,
-      "loss": 0.5993,
-      "step": 31860
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011324341647687755,
-      "loss": 0.7695,
-      "step": 31870
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.0001132161413940295,
-      "loss": 0.7115,
-      "step": 31880
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011318886631118142,
-      "loss": 0.692,
-      "step": 31890
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011316159122833336,
-      "loss": 0.7502,
-      "step": 31900
-    },
-    {
-      "epoch": 1.3,
-      "learning_rate": 0.00011313431614548528,
-      "loss": 0.6768,
-      "step": 31910
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011310704106263723,
-      "loss": 0.6724,
-      "step": 31920
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011307976597978918,
-      "loss": 0.717,
-      "step": 31930
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.0001130524908969411,
-      "loss": 0.644,
-      "step": 31940
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011302521581409305,
-      "loss": 0.7347,
-      "step": 31950
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011299794073124497,
-      "loss": 0.7578,
-      "step": 31960
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011297066564839691,
-      "loss": 0.7082,
-      "step": 31970
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011294339056554886,
-      "loss": 0.6984,
-      "step": 31980
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011291611548270078,
-      "loss": 0.7,
-      "step": 31990
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011288884039985273,
-      "loss": 0.7426,
-      "step": 32000
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011286156531700465,
-      "loss": 0.6485,
-      "step": 32010
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.0001128342902341566,
-      "loss": 0.7414,
-      "step": 32020
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011280701515130853,
-      "loss": 0.6942,
-      "step": 32030
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011277974006846047,
-      "loss": 0.6577,
-      "step": 32040
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.0001127524649856124,
-      "loss": 0.6831,
-      "step": 32050
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011272518990276433,
-      "loss": 0.7384,
-      "step": 32060
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011269791481991627,
-      "loss": 0.6852,
-      "step": 32070
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011267063973706821,
-      "loss": 0.6342,
-      "step": 32080
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011264336465422013,
-      "loss": 0.838,
-      "step": 32090
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011261608957137208,
-      "loss": 0.6894,
-      "step": 32100
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.000112588814488524,
-      "loss": 0.7558,
-      "step": 32110
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011256153940567595,
-      "loss": 0.7668,
-      "step": 32120
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011253426432282787,
-      "loss": 0.7021,
-      "step": 32130
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011250698923997982,
-      "loss": 0.6959,
-      "step": 32140
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011247971415713177,
-      "loss": 0.7705,
-      "step": 32150
-    },
-    {
-      "epoch": 1.31,
-      "learning_rate": 0.00011245243907428369,
-      "loss": 0.7386,
-      "step": 32160
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011242516399143563,
-      "loss": 0.7293,
-      "step": 32170
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011239788890858755,
-      "loss": 0.7161,
-      "step": 32180
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.0001123706138257395,
-      "loss": 0.6831,
-      "step": 32190
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011234333874289145,
-      "loss": 0.697,
-      "step": 32200
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011231606366004337,
-      "loss": 0.6112,
-      "step": 32210
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011228878857719532,
-      "loss": 0.612,
-      "step": 32220
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011226151349434724,
-      "loss": 0.7571,
-      "step": 32230
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011223423841149918,
-      "loss": 0.7087,
-      "step": 32240
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011220696332865113,
-      "loss": 0.6905,
-      "step": 32250
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011217968824580305,
-      "loss": 0.7134,
-      "step": 32260
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.000112152413162955,
-      "loss": 0.6164,
-      "step": 32270
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011212513808010692,
-      "loss": 0.727,
-      "step": 32280
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011209786299725887,
-      "loss": 0.7312,
-      "step": 32290
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011207058791441079,
-      "loss": 0.7121,
-      "step": 32300
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011204331283156274,
-      "loss": 0.6867,
-      "step": 32310
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011201603774871467,
-      "loss": 0.6864,
-      "step": 32320
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.0001119887626658666,
-      "loss": 0.719,
-      "step": 32330
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011196148758301854,
-      "loss": 0.5989,
-      "step": 32340
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011193421250017047,
-      "loss": 0.7233,
-      "step": 32350
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.0001119069374173224,
-      "loss": 0.698,
-      "step": 32360
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011187966233447435,
-      "loss": 0.772,
-      "step": 32370
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011185238725162627,
-      "loss": 0.7136,
-      "step": 32380
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011182511216877822,
-      "loss": 0.6191,
-      "step": 32390
-    },
-    {
-      "epoch": 1.32,
-      "learning_rate": 0.00011179783708593014,
-      "loss": 0.7405,
-      "step": 32400
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011177056200308209,
-      "loss": 0.7163,
-      "step": 32410
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011174328692023404,
-      "loss": 0.6441,
-      "step": 32420
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011171601183738596,
-      "loss": 0.736,
-      "step": 32430
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.0001116887367545379,
-      "loss": 0.5884,
-      "step": 32440
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011166146167168982,
-      "loss": 0.6967,
-      "step": 32450
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011163418658884177,
-      "loss": 0.6592,
-      "step": 32460
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011160691150599369,
-      "loss": 0.6392,
-      "step": 32470
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011157963642314564,
-      "loss": 0.6597,
-      "step": 32480
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011155236134029759,
-      "loss": 0.6006,
-      "step": 32490
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011152508625744951,
-      "loss": 0.6218,
-      "step": 32500
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011149781117460145,
-      "loss": 0.615,
-      "step": 32510
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011147053609175337,
-      "loss": 0.6232,
-      "step": 32520
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011144326100890532,
-      "loss": 0.7522,
-      "step": 32530
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011141598592605727,
-      "loss": 0.7115,
-      "step": 32540
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011138871084320919,
-      "loss": 0.6754,
-      "step": 32550
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011136143576036114,
-      "loss": 0.7099,
-      "step": 32560
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011133416067751306,
-      "loss": 0.5947,
-      "step": 32570
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.000111306885594665,
-      "loss": 0.6627,
-      "step": 32580
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011127961051181694,
-      "loss": 0.7339,
-      "step": 32590
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011125233542896887,
-      "loss": 0.7069,
-      "step": 32600
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011122506034612081,
-      "loss": 0.6106,
-      "step": 32610
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011119778526327274,
-      "loss": 0.6763,
-      "step": 32620
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011117051018042468,
-      "loss": 0.7474,
-      "step": 32630
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011114323509757661,
-      "loss": 0.731,
-      "step": 32640
-    },
-    {
-      "epoch": 1.33,
-      "learning_rate": 0.00011111596001472854,
-      "loss": 0.7102,
-      "step": 32650
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011108868493188049,
-      "loss": 0.8084,
-      "step": 32660
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011106140984903241,
-      "loss": 0.6874,
-      "step": 32670
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011103413476618436,
-      "loss": 0.7645,
-      "step": 32680
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011100685968333628,
-      "loss": 0.6684,
-      "step": 32690
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011097958460048823,
-      "loss": 0.5879,
-      "step": 32700
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011095230951764017,
-      "loss": 0.7115,
-      "step": 32710
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.0001109250344347921,
-      "loss": 0.7208,
-      "step": 32720
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011089775935194404,
-      "loss": 0.6914,
-      "step": 32730
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011087048426909596,
-      "loss": 0.6386,
-      "step": 32740
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011084320918624791,
-      "loss": 0.68,
-      "step": 32750
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011081593410339986,
-      "loss": 0.6594,
-      "step": 32760
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011078865902055178,
-      "loss": 0.7098,
-      "step": 32770
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011076138393770372,
-      "loss": 0.7045,
-      "step": 32780
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011073410885485564,
-      "loss": 0.7122,
-      "step": 32790
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011070683377200759,
-      "loss": 0.7231,
-      "step": 32800
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011067955868915951,
-      "loss": 0.6728,
-      "step": 32810
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011065228360631146,
-      "loss": 0.8363,
-      "step": 32820
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011062500852346341,
-      "loss": 0.7154,
-      "step": 32830
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011059773344061533,
-      "loss": 0.7985,
-      "step": 32840
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011057045835776728,
-      "loss": 0.6532,
-      "step": 32850
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.0001105431832749192,
-      "loss": 0.704,
-      "step": 32860
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011051590819207114,
-      "loss": 0.6713,
-      "step": 32870
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011048863310922308,
-      "loss": 0.7075,
-      "step": 32880
-    },
-    {
-      "epoch": 1.34,
-      "learning_rate": 0.00011046135802637501,
-      "loss": 0.5987,
-      "step": 32890
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011043408294352695,
-      "loss": 0.6686,
-      "step": 32900
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011040680786067888,
-      "loss": 0.7106,
-      "step": 32910
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011037953277783081,
-      "loss": 0.6421,
-      "step": 32920
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011035225769498276,
-      "loss": 0.6992,
-      "step": 32930
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011032498261213468,
-      "loss": 0.6741,
-      "step": 32940
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011029770752928663,
-      "loss": 0.6198,
-      "step": 32950
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011027043244643855,
-      "loss": 0.687,
-      "step": 32960
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.0001102431573635905,
-      "loss": 0.7101,
-      "step": 32970
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011021588228074244,
-      "loss": 0.7608,
-      "step": 32980
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011018860719789436,
-      "loss": 0.6443,
-      "step": 32990
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011016133211504631,
-      "loss": 0.6824,
-      "step": 33000
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011013405703219823,
-      "loss": 0.6898,
-      "step": 33010
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011010678194935018,
-      "loss": 0.7287,
-      "step": 33020
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.0001100795068665021,
-      "loss": 0.7165,
-      "step": 33030
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00011005223178365405,
-      "loss": 0.7512,
-      "step": 33040
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.000110024956700806,
-      "loss": 0.6549,
-      "step": 33050
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010999768161795791,
-      "loss": 0.7603,
-      "step": 33060
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010997040653510986,
-      "loss": 0.6642,
-      "step": 33070
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010994313145226178,
-      "loss": 0.6689,
-      "step": 33080
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010991585636941373,
-      "loss": 0.6708,
-      "step": 33090
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010988858128656568,
-      "loss": 0.8193,
-      "step": 33100
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.0001098613062037176,
-      "loss": 0.6449,
-      "step": 33110
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010983403112086955,
-      "loss": 0.7139,
-      "step": 33120
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010980675603802147,
-      "loss": 0.7108,
-      "step": 33130
-    },
-    {
-      "epoch": 1.35,
-      "learning_rate": 0.00010977948095517341,
-      "loss": 0.7578,
-      "step": 33140
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010975220587232535,
-      "loss": 0.6961,
-      "step": 33150
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010972493078947728,
-      "loss": 0.769,
-      "step": 33160
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010969765570662922,
-      "loss": 0.6824,
-      "step": 33170
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010967038062378115,
-      "loss": 0.7413,
-      "step": 33180
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010964310554093308,
-      "loss": 0.6916,
-      "step": 33190
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010961583045808502,
-      "loss": 0.7141,
-      "step": 33200
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010958855537523695,
-      "loss": 0.7115,
-      "step": 33210
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.0001095612802923889,
-      "loss": 0.6322,
-      "step": 33220
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010953400520954082,
-      "loss": 0.6738,
-      "step": 33230
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010950673012669277,
-      "loss": 0.6679,
-      "step": 33240
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010947945504384469,
-      "loss": 0.7132,
-      "step": 33250
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010945217996099663,
-      "loss": 0.7892,
-      "step": 33260
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010942490487814858,
-      "loss": 0.7244,
-      "step": 33270
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.0001093976297953005,
-      "loss": 0.757,
-      "step": 33280
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010937035471245245,
-      "loss": 0.7598,
-      "step": 33290
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010934307962960437,
-      "loss": 0.7594,
-      "step": 33300
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010931580454675632,
-      "loss": 0.7498,
-      "step": 33310
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010928852946390826,
-      "loss": 0.7228,
-      "step": 33320
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010926125438106018,
-      "loss": 0.7163,
-      "step": 33330
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010923397929821213,
-      "loss": 0.7156,
-      "step": 33340
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010920670421536405,
-      "loss": 0.6808,
-      "step": 33350
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.000109179429132516,
-      "loss": 0.6526,
-      "step": 33360
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010915215404966792,
-      "loss": 0.7653,
-      "step": 33370
-    },
-    {
-      "epoch": 1.36,
-      "learning_rate": 0.00010912487896681987,
-      "loss": 0.715,
-      "step": 33380
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010909760388397182,
-      "loss": 0.7756,
-      "step": 33390
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010907032880112374,
-      "loss": 0.7391,
-      "step": 33400
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010904305371827568,
-      "loss": 0.5822,
-      "step": 33410
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.0001090157786354276,
-      "loss": 0.774,
-      "step": 33420
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010898850355257955,
-      "loss": 0.7082,
-      "step": 33430
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010896122846973149,
-      "loss": 0.6618,
-      "step": 33440
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010893395338688342,
-      "loss": 0.8074,
-      "step": 33450
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010890667830403535,
-      "loss": 0.6403,
-      "step": 33460
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010887940322118729,
-      "loss": 0.6203,
-      "step": 33470
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010885212813833922,
-      "loss": 0.6431,
-      "step": 33480
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010882485305549117,
-      "loss": 0.6171,
-      "step": 33490
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010879757797264309,
-      "loss": 0.5004,
-      "step": 33500
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010877030288979504,
-      "loss": 0.678,
-      "step": 33510
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010874302780694696,
-      "loss": 0.6833,
-      "step": 33520
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.0001087157527240989,
-      "loss": 0.7838,
-      "step": 33530
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010868847764125082,
-      "loss": 0.7098,
-      "step": 33540
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010866120255840277,
-      "loss": 0.7729,
-      "step": 33550
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010863392747555472,
-      "loss": 0.6337,
-      "step": 33560
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010860665239270664,
-      "loss": 0.7802,
-      "step": 33570
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010857937730985859,
-      "loss": 0.6766,
-      "step": 33580
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010855210222701051,
-      "loss": 0.6251,
-      "step": 33590
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010852482714416246,
-      "loss": 0.6358,
-      "step": 33600
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.0001084975520613144,
-      "loss": 0.7832,
-      "step": 33610
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010847027697846632,
-      "loss": 0.5909,
-      "step": 33620
-    },
-    {
-      "epoch": 1.37,
-      "learning_rate": 0.00010844300189561827,
-      "loss": 0.7196,
-      "step": 33630
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010841572681277019,
-      "loss": 0.5909,
-      "step": 33640
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010838845172992214,
-      "loss": 0.7119,
-      "step": 33650
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010836117664707409,
-      "loss": 0.6653,
-      "step": 33660
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.000108333901564226,
-      "loss": 0.6966,
-      "step": 33670
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010830662648137795,
-      "loss": 0.7306,
-      "step": 33680
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010827935139852987,
-      "loss": 0.7649,
-      "step": 33690
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010825207631568182,
-      "loss": 0.6823,
-      "step": 33700
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010822480123283374,
-      "loss": 0.7013,
-      "step": 33710
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010819752614998569,
-      "loss": 0.6796,
-      "step": 33720
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010817025106713762,
-      "loss": 0.708,
-      "step": 33730
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010814297598428956,
-      "loss": 0.6493,
-      "step": 33740
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010811570090144149,
-      "loss": 0.5919,
-      "step": 33750
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010808842581859342,
-      "loss": 0.6656,
-      "step": 33760
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010806387824403017,
-      "loss": 0.7548,
-      "step": 33770
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010803660316118211,
-      "loss": 0.6414,
-      "step": 33780
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010800932807833403,
-      "loss": 0.6465,
-      "step": 33790
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010798205299548598,
-      "loss": 0.7825,
-      "step": 33800
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.0001079547779126379,
-      "loss": 0.7194,
-      "step": 33810
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010792750282978985,
-      "loss": 0.6502,
-      "step": 33820
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.0001079002277469418,
-      "loss": 0.6849,
-      "step": 33830
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010787295266409372,
-      "loss": 0.6313,
-      "step": 33840
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010784567758124566,
-      "loss": 0.6862,
-      "step": 33850
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010781840249839758,
-      "loss": 0.6562,
-      "step": 33860
-    },
-    {
-      "epoch": 1.38,
-      "learning_rate": 0.00010779112741554953,
-      "loss": 0.6058,
-      "step": 33870
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010776385233270148,
-      "loss": 0.7113,
-      "step": 33880
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.0001077365772498534,
-      "loss": 0.6413,
-      "step": 33890
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010770930216700535,
-      "loss": 0.6685,
-      "step": 33900
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010768202708415727,
-      "loss": 0.6894,
-      "step": 33910
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010765475200130921,
-      "loss": 0.6641,
-      "step": 33920
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010762747691846114,
-      "loss": 0.6825,
-      "step": 33930
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010760020183561308,
-      "loss": 0.6985,
-      "step": 33940
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010757292675276503,
-      "loss": 0.7031,
-      "step": 33950
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010754565166991695,
-      "loss": 0.6711,
-      "step": 33960
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.0001075183765870689,
-      "loss": 0.6353,
-      "step": 33970
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010749110150422082,
-      "loss": 0.727,
-      "step": 33980
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010746382642137277,
-      "loss": 0.608,
-      "step": 33990
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.0001074365513385247,
-      "loss": 0.7336,
-      "step": 34000
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010740927625567663,
-      "loss": 0.6742,
-      "step": 34010
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010738200117282857,
-      "loss": 0.6307,
-      "step": 34020
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.0001073547260899805,
-      "loss": 0.6192,
-      "step": 34030
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010732745100713244,
-      "loss": 0.6797,
-      "step": 34040
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010730017592428438,
-      "loss": 0.6301,
-      "step": 34050
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.0001072729008414363,
-      "loss": 0.653,
-      "step": 34060
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010724562575858825,
-      "loss": 0.6918,
-      "step": 34070
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010721835067574017,
-      "loss": 0.7034,
-      "step": 34080
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010719107559289212,
-      "loss": 0.7372,
-      "step": 34090
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010716380051004404,
-      "loss": 0.7315,
-      "step": 34100
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010713652542719599,
-      "loss": 0.7184,
-      "step": 34110
-    },
-    {
-      "epoch": 1.39,
-      "learning_rate": 0.00010710925034434793,
-      "loss": 0.6363,
-      "step": 34120
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010708197526149985,
-      "loss": 0.7747,
-      "step": 34130
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.0001070547001786518,
-      "loss": 0.7384,
-      "step": 34140
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010702742509580372,
-      "loss": 0.6794,
-      "step": 34150
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010700015001295567,
-      "loss": 0.6957,
-      "step": 34160
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010697287493010762,
-      "loss": 0.7193,
-      "step": 34170
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010694559984725954,
-      "loss": 0.7366,
-      "step": 34180
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010691832476441148,
-      "loss": 0.7268,
-      "step": 34190
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.0001068910496815634,
-      "loss": 0.6832,
-      "step": 34200
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010686377459871535,
-      "loss": 0.7387,
-      "step": 34210
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.0001068364995158673,
-      "loss": 0.6306,
-      "step": 34220
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010680922443301922,
-      "loss": 0.6212,
-      "step": 34230
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010678194935017117,
-      "loss": 0.6512,
-      "step": 34240
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010675467426732309,
-      "loss": 0.6661,
-      "step": 34250
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010672739918447504,
-      "loss": 0.6896,
-      "step": 34260
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010670012410162697,
-      "loss": 0.6596,
-      "step": 34270
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.0001066728490187789,
-      "loss": 0.6822,
-      "step": 34280
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010664557393593084,
-      "loss": 0.6633,
-      "step": 34290
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010661829885308277,
-      "loss": 0.6561,
-      "step": 34300
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.0001065910237702347,
-      "loss": 0.6717,
-      "step": 34310
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010656374868738664,
-      "loss": 0.7112,
-      "step": 34320
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010653647360453857,
-      "loss": 0.6711,
-      "step": 34330
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010650919852169052,
-      "loss": 0.7254,
-      "step": 34340
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010648192343884244,
-      "loss": 0.6472,
-      "step": 34350
-    },
-    {
-      "epoch": 1.4,
-      "learning_rate": 0.00010645464835599439,
-      "loss": 0.6295,
-      "step": 34360
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010642737327314631,
-      "loss": 0.6828,
-      "step": 34370
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010640009819029826,
-      "loss": 0.7304,
-      "step": 34380
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.0001063728231074502,
-      "loss": 0.7082,
-      "step": 34390
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010634554802460212,
-      "loss": 0.7815,
-      "step": 34400
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010631827294175407,
-      "loss": 0.7571,
-      "step": 34410
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010629099785890599,
-      "loss": 0.6693,
-      "step": 34420
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010626372277605794,
-      "loss": 0.6815,
-      "step": 34430
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010623644769320989,
-      "loss": 0.6288,
-      "step": 34440
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010620917261036181,
-      "loss": 0.6793,
-      "step": 34450
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010618189752751375,
-      "loss": 0.8543,
-      "step": 34460
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010615462244466568,
-      "loss": 0.751,
-      "step": 34470
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010612734736181762,
-      "loss": 0.766,
-      "step": 34480
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010610007227896954,
-      "loss": 0.6164,
-      "step": 34490
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010607279719612149,
-      "loss": 0.6774,
-      "step": 34500
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010604552211327344,
-      "loss": 0.5674,
-      "step": 34510
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010601824703042536,
-      "loss": 0.6808,
-      "step": 34520
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.0001059909719475773,
-      "loss": 0.708,
-      "step": 34530
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010596369686472923,
-      "loss": 0.7916,
-      "step": 34540
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010593642178188117,
-      "loss": 0.7095,
-      "step": 34550
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010590914669903311,
-      "loss": 0.686,
-      "step": 34560
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010588187161618504,
-      "loss": 0.655,
-      "step": 34570
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010585459653333698,
-      "loss": 0.6171,
-      "step": 34580
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010582732145048891,
-      "loss": 0.6376,
-      "step": 34590
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010580004636764084,
-      "loss": 0.6669,
-      "step": 34600
-    },
-    {
-      "epoch": 1.41,
-      "learning_rate": 0.00010577277128479279,
-      "loss": 0.699,
-      "step": 34610
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010574549620194471,
-      "loss": 0.6783,
-      "step": 34620
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010571822111909666,
-      "loss": 0.6823,
-      "step": 34630
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010569094603624858,
-      "loss": 0.6501,
-      "step": 34640
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010566367095340053,
-      "loss": 0.647,
-      "step": 34650
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010563639587055245,
-      "loss": 0.6843,
-      "step": 34660
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.0001056091207877044,
-      "loss": 0.6949,
-      "step": 34670
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010558184570485634,
-      "loss": 0.727,
-      "step": 34680
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010555457062200826,
-      "loss": 0.7836,
-      "step": 34690
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010552729553916021,
-      "loss": 0.6706,
-      "step": 34700
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010550002045631213,
-      "loss": 0.6712,
-      "step": 34710
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010547274537346408,
-      "loss": 0.7162,
-      "step": 34720
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010544547029061602,
-      "loss": 0.7025,
-      "step": 34730
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010541819520776795,
-      "loss": 0.7207,
-      "step": 34740
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010539092012491989,
-      "loss": 0.6934,
-      "step": 34750
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010536364504207181,
-      "loss": 0.6586,
-      "step": 34760
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010533636995922376,
-      "loss": 0.7153,
-      "step": 34770
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010530909487637571,
-      "loss": 0.6615,
-      "step": 34780
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010528181979352763,
-      "loss": 0.6589,
-      "step": 34790
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010525454471067958,
-      "loss": 0.838,
-      "step": 34800
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.0001052272696278315,
-      "loss": 0.6923,
-      "step": 34810
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010519999454498344,
-      "loss": 0.6385,
-      "step": 34820
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010517271946213536,
-      "loss": 0.7683,
-      "step": 34830
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010514544437928731,
-      "loss": 0.6278,
-      "step": 34840
-    },
-    {
-      "epoch": 1.42,
-      "learning_rate": 0.00010511816929643925,
-      "loss": 0.7451,
-      "step": 34850
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010509089421359118,
-      "loss": 0.6595,
-      "step": 34860
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010506361913074311,
-      "loss": 0.7073,
-      "step": 34870
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010503634404789505,
-      "loss": 0.6595,
-      "step": 34880
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010500906896504698,
-      "loss": 0.8015,
-      "step": 34890
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010498179388219893,
-      "loss": 0.779,
-      "step": 34900
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010495451879935085,
-      "loss": 0.7916,
-      "step": 34910
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.0001049272437165028,
-      "loss": 0.7816,
-      "step": 34920
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010489996863365472,
-      "loss": 0.7139,
-      "step": 34930
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010487269355080666,
-      "loss": 0.6469,
-      "step": 34940
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010484541846795861,
-      "loss": 0.6585,
-      "step": 34950
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010481814338511053,
-      "loss": 0.7075,
-      "step": 34960
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010479086830226248,
-      "loss": 0.7074,
-      "step": 34970
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.0001047635932194144,
-      "loss": 0.6316,
-      "step": 34980
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010473631813656635,
-      "loss": 0.7258,
-      "step": 34990
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010470904305371827,
-      "loss": 0.7223,
-      "step": 35000
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010468176797087022,
-      "loss": 0.6956,
-      "step": 35010
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010465449288802216,
-      "loss": 0.6587,
-      "step": 35020
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010462721780517408,
-      "loss": 0.6852,
-      "step": 35030
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010459994272232603,
-      "loss": 0.6962,
-      "step": 35040
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010457266763947795,
-      "loss": 0.6742,
-      "step": 35050
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.0001045453925566299,
-      "loss": 0.6678,
-      "step": 35060
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010451811747378185,
-      "loss": 0.7378,
-      "step": 35070
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010449084239093377,
-      "loss": 0.6788,
-      "step": 35080
-    },
-    {
-      "epoch": 1.43,
-      "learning_rate": 0.00010446356730808571,
-      "loss": 0.7031,
-      "step": 35090
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010443629222523763,
-      "loss": 0.6719,
-      "step": 35100
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010440901714238958,
-      "loss": 0.6531,
-      "step": 35110
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010438174205954152,
-      "loss": 0.694,
-      "step": 35120
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010435446697669345,
-      "loss": 0.6856,
-      "step": 35130
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010432719189384538,
-      "loss": 0.6361,
-      "step": 35140
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010429991681099732,
-      "loss": 0.7916,
-      "step": 35150
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010427264172814925,
-      "loss": 0.7482,
-      "step": 35160
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010424536664530119,
-      "loss": 0.6754,
-      "step": 35170
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010421809156245312,
-      "loss": 0.7119,
-      "step": 35180
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010419081647960507,
-      "loss": 0.7222,
-      "step": 35190
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010416354139675699,
-      "loss": 0.8065,
-      "step": 35200
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010413626631390893,
-      "loss": 0.7475,
-      "step": 35210
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010410899123106085,
-      "loss": 0.6481,
-      "step": 35220
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.0001040817161482128,
-      "loss": 0.7353,
-      "step": 35230
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010405444106536475,
-      "loss": 0.5901,
-      "step": 35240
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010402716598251667,
-      "loss": 0.5693,
-      "step": 35250
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010399989089966862,
-      "loss": 0.8342,
-      "step": 35260
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010397261581682054,
-      "loss": 0.7551,
-      "step": 35270
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010394534073397249,
-      "loss": 0.7306,
-      "step": 35280
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010391806565112443,
-      "loss": 0.67,
-      "step": 35290
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010389079056827635,
-      "loss": 0.6503,
-      "step": 35300
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.0001038635154854283,
-      "loss": 0.7003,
-      "step": 35310
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010383624040258022,
-      "loss": 0.647,
-      "step": 35320
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010380896531973217,
-      "loss": 0.784,
-      "step": 35330
-    },
-    {
-      "epoch": 1.44,
-      "learning_rate": 0.00010378169023688412,
-      "loss": 0.6832,
-      "step": 35340
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010375441515403604,
-      "loss": 0.6708,
-      "step": 35350
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010372714007118798,
-      "loss": 0.6367,
-      "step": 35360
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.0001036998649883399,
-      "loss": 0.7005,
-      "step": 35370
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010367258990549185,
-      "loss": 0.6504,
-      "step": 35380
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010364531482264377,
-      "loss": 0.6768,
-      "step": 35390
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010361803973979572,
-      "loss": 0.7091,
-      "step": 35400
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010359076465694765,
-      "loss": 0.5929,
-      "step": 35410
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010356348957409959,
-      "loss": 0.6812,
-      "step": 35420
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010353621449125152,
-      "loss": 0.7018,
-      "step": 35430
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010350893940840346,
-      "loss": 0.6624,
-      "step": 35440
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010348166432555539,
-      "loss": 0.639,
-      "step": 35450
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010345438924270734,
-      "loss": 0.698,
-      "step": 35460
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010342711415985926,
-      "loss": 0.7512,
-      "step": 35470
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.0001033998390770112,
-      "loss": 0.6661,
-      "step": 35480
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010337256399416312,
-      "loss": 0.5785,
-      "step": 35490
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010334528891131507,
-      "loss": 0.7851,
-      "step": 35500
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010331801382846702,
-      "loss": 0.6602,
-      "step": 35510
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010329073874561894,
-      "loss": 0.7315,
-      "step": 35520
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010326346366277089,
-      "loss": 0.7191,
-      "step": 35530
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010323618857992281,
-      "loss": 0.7813,
-      "step": 35540
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010320891349707476,
-      "loss": 0.7151,
-      "step": 35550
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010318163841422668,
-      "loss": 0.6065,
-      "step": 35560
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010315436333137862,
-      "loss": 0.6499,
-      "step": 35570
-    },
-    {
-      "epoch": 1.45,
-      "learning_rate": 0.00010312708824853057,
-      "loss": 0.7191,
-      "step": 35580
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010309981316568249,
-      "loss": 0.661,
-      "step": 35590
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010307253808283444,
-      "loss": 0.5544,
-      "step": 35600
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010304526299998636,
-      "loss": 0.6601,
-      "step": 35610
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.0001030179879171383,
-      "loss": 0.6574,
-      "step": 35620
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010299071283429025,
-      "loss": 0.6737,
-      "step": 35630
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010296343775144217,
-      "loss": 0.6894,
-      "step": 35640
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010293616266859412,
-      "loss": 0.6217,
-      "step": 35650
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010290888758574604,
-      "loss": 0.7864,
-      "step": 35660
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010288161250289799,
-      "loss": 0.6222,
-      "step": 35670
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010285433742004992,
-      "loss": 0.6832,
-      "step": 35680
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010282706233720186,
-      "loss": 0.7128,
-      "step": 35690
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010279978725435379,
-      "loss": 0.7886,
-      "step": 35700
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010277251217150573,
-      "loss": 0.6504,
-      "step": 35710
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010274523708865766,
-      "loss": 0.7733,
-      "step": 35720
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010271796200580959,
-      "loss": 0.5697,
-      "step": 35730
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010269068692296153,
-      "loss": 0.6933,
-      "step": 35740
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010266341184011347,
-      "loss": 0.7667,
-      "step": 35750
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.0001026361367572654,
-      "loss": 0.63,
-      "step": 35760
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010260886167441734,
-      "loss": 0.7243,
-      "step": 35770
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010258158659156926,
-      "loss": 0.7057,
-      "step": 35780
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010255431150872121,
-      "loss": 0.7261,
-      "step": 35790
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010252703642587316,
-      "loss": 0.6714,
-      "step": 35800
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010249976134302508,
-      "loss": 0.6834,
-      "step": 35810
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010247248626017703,
-      "loss": 0.6866,
-      "step": 35820
-    },
-    {
-      "epoch": 1.46,
-      "learning_rate": 0.00010244521117732895,
-      "loss": 0.8004,
-      "step": 35830
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.0001024179360944809,
-      "loss": 0.6379,
-      "step": 35840
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010239066101163284,
-      "loss": 0.7089,
-      "step": 35850
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010236338592878476,
-      "loss": 0.7064,
-      "step": 35860
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010233611084593671,
-      "loss": 0.6669,
-      "step": 35870
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010230883576308863,
-      "loss": 0.6514,
-      "step": 35880
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010228156068024058,
-      "loss": 0.7144,
-      "step": 35890
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.0001022542855973925,
-      "loss": 0.7839,
-      "step": 35900
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010222701051454444,
-      "loss": 0.6291,
-      "step": 35910
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010219973543169639,
-      "loss": 0.6849,
-      "step": 35920
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010217246034884831,
-      "loss": 0.6563,
-      "step": 35930
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010214518526600026,
-      "loss": 0.6899,
-      "step": 35940
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010211791018315218,
-      "loss": 0.7085,
-      "step": 35950
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010209063510030413,
-      "loss": 0.6927,
-      "step": 35960
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010206336001745606,
-      "loss": 0.6982,
-      "step": 35970
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.000102036084934608,
-      "loss": 0.667,
-      "step": 35980
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010200880985175993,
-      "loss": 0.6842,
-      "step": 35990
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010198153476891186,
-      "loss": 0.812,
-      "step": 36000
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.0001019542596860638,
-      "loss": 0.7448,
-      "step": 36010
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010192698460321574,
-      "loss": 0.6957,
-      "step": 36020
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010189970952036766,
-      "loss": 0.6665,
-      "step": 36030
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010187243443751961,
-      "loss": 0.7062,
-      "step": 36040
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010184515935467153,
-      "loss": 0.679,
-      "step": 36050
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.00010181788427182348,
-      "loss": 0.6579,
-      "step": 36060
-    },
-    {
-      "epoch": 1.47,
-      "learning_rate": 0.0001017906091889754,
-      "loss": 0.7033,
-      "step": 36070
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010176333410612735,
-      "loss": 0.719,
-      "step": 36080
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.0001017360590232793,
-      "loss": 0.6792,
-      "step": 36090
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010170878394043122,
-      "loss": 0.6544,
-      "step": 36100
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010168150885758316,
-      "loss": 0.6839,
-      "step": 36110
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010165423377473508,
-      "loss": 0.7224,
-      "step": 36120
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010162695869188703,
-      "loss": 0.6766,
-      "step": 36130
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010159968360903898,
-      "loss": 0.8952,
-      "step": 36140
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.0001015724085261909,
-      "loss": 0.6633,
-      "step": 36150
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010154513344334285,
-      "loss": 0.7446,
-      "step": 36160
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010151785836049477,
-      "loss": 0.6931,
-      "step": 36170
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010149058327764671,
-      "loss": 0.6435,
-      "step": 36180
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010146330819479866,
-      "loss": 0.6722,
-      "step": 36190
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010143603311195058,
-      "loss": 0.7139,
-      "step": 36200
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010140875802910253,
-      "loss": 0.6393,
-      "step": 36210
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010138148294625445,
-      "loss": 0.7179,
-      "step": 36220
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.0001013542078634064,
-      "loss": 0.6998,
-      "step": 36230
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010132693278055833,
-      "loss": 0.7232,
-      "step": 36240
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010129965769771027,
-      "loss": 0.627,
-      "step": 36250
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.0001012723826148622,
-      "loss": 0.6817,
-      "step": 36260
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010124510753201413,
-      "loss": 0.7269,
-      "step": 36270
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010121783244916607,
-      "loss": 0.6858,
-      "step": 36280
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.000101190557366318,
-      "loss": 0.6755,
-      "step": 36290
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010116328228346994,
-      "loss": 0.724,
-      "step": 36300
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.00010113600720062188,
-      "loss": 0.7797,
-      "step": 36310
-    },
-    {
-      "epoch": 1.48,
-      "learning_rate": 0.0001011087321177738,
-      "loss": 0.7088,
-      "step": 36320
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010108145703492575,
-      "loss": 0.6431,
-      "step": 36330
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010105418195207767,
-      "loss": 0.7454,
-      "step": 36340
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010102690686922962,
-      "loss": 0.7053,
-      "step": 36350
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010099963178638157,
-      "loss": 0.6902,
-      "step": 36360
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010097235670353349,
-      "loss": 0.6439,
-      "step": 36370
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010094508162068543,
-      "loss": 0.6086,
-      "step": 36380
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010091780653783735,
-      "loss": 0.6716,
-      "step": 36390
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.0001008905314549893,
-      "loss": 0.8344,
-      "step": 36400
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010086325637214125,
-      "loss": 0.679,
-      "step": 36410
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010083598128929317,
-      "loss": 0.6781,
-      "step": 36420
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010080870620644512,
-      "loss": 0.7685,
-      "step": 36430
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010078143112359704,
-      "loss": 0.6203,
-      "step": 36440
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010075415604074898,
-      "loss": 0.6991,
-      "step": 36450
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.0001007268809579009,
-      "loss": 0.6522,
-      "step": 36460
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010069960587505285,
-      "loss": 0.7443,
-      "step": 36470
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.0001006723307922048,
-      "loss": 0.626,
-      "step": 36480
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010064505570935672,
-      "loss": 0.7148,
-      "step": 36490
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010061778062650867,
-      "loss": 0.6116,
-      "step": 36500
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010059050554366059,
-      "loss": 0.7652,
-      "step": 36510
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010056323046081254,
-      "loss": 0.6407,
-      "step": 36520
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010053595537796447,
-      "loss": 0.6977,
-      "step": 36530
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.0001005086802951164,
-      "loss": 0.702,
-      "step": 36540
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010048140521226834,
-      "loss": 0.7372,
-      "step": 36550
-    },
-    {
-      "epoch": 1.49,
-      "learning_rate": 0.00010045413012942027,
-      "loss": 0.7253,
-      "step": 36560
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.0001004268550465722,
-      "loss": 0.7878,
-      "step": 36570
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010039957996372415,
-      "loss": 0.6613,
-      "step": 36580
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010037230488087607,
-      "loss": 0.7589,
-      "step": 36590
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010034502979802802,
-      "loss": 0.6561,
-      "step": 36600
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010031775471517994,
-      "loss": 0.7002,
-      "step": 36610
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010029047963233189,
-      "loss": 0.6683,
-      "step": 36620
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010026320454948381,
-      "loss": 0.708,
-      "step": 36630
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010023592946663576,
-      "loss": 0.5984,
-      "step": 36640
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.0001002086543837877,
-      "loss": 0.7128,
-      "step": 36650
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010018137930093962,
-      "loss": 0.6967,
-      "step": 36660
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010015410421809157,
-      "loss": 0.6451,
-      "step": 36670
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010012682913524349,
-      "loss": 0.7426,
-      "step": 36680
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010009955405239544,
-      "loss": 0.6991,
-      "step": 36690
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010007227896954739,
-      "loss": 0.6486,
-      "step": 36700
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010004500388669931,
-      "loss": 0.6891,
-      "step": 36710
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 0.00010001772880385125,
-      "loss": 0.7802,
-      "step": 36720
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.999045372100319e-05,
-      "loss": 0.7572,
-      "step": 36730
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.996317863815512e-05,
-      "loss": 0.6438,
-      "step": 36740
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.993590355530706e-05,
-      "loss": 0.7008,
-      "step": 36750
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.990862847245899e-05,
-      "loss": 0.7034,
-      "step": 36760
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.988135338961092e-05,
-      "loss": 0.6995,
-      "step": 36770
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.985407830676287e-05,
-      "loss": 0.6183,
-      "step": 36780
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.98268032239148e-05,
-      "loss": 0.6377,
-      "step": 36790
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.979952814106674e-05,
-      "loss": 0.6759,
-      "step": 36800
-    },
-    {
-      "epoch": 1.5,
-      "learning_rate": 9.977225305821867e-05,
-      "loss": 0.6671,
-      "step": 36810
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.974497797537061e-05,
-      "loss": 0.6818,
-      "step": 36820
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.971770289252254e-05,
-      "loss": 0.6754,
-      "step": 36830
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.969042780967448e-05,
-      "loss": 0.7648,
-      "step": 36840
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.966315272682641e-05,
-      "loss": 0.7246,
-      "step": 36850
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.963587764397834e-05,
-      "loss": 0.6564,
-      "step": 36860
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.960860256113028e-05,
-      "loss": 0.7617,
-      "step": 36870
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.958132747828221e-05,
-      "loss": 0.689,
-      "step": 36880
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.955677990371896e-05,
-      "loss": 0.6201,
-      "step": 36890
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.95295048208709e-05,
-      "loss": 0.6217,
-      "step": 36900
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.950222973802283e-05,
-      "loss": 0.6916,
-      "step": 36910
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.947495465517477e-05,
-      "loss": 0.6637,
-      "step": 36920
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.94476795723267e-05,
-      "loss": 0.5932,
-      "step": 36930
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.942040448947865e-05,
-      "loss": 0.7064,
-      "step": 36940
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.939312940663058e-05,
-      "loss": 0.6244,
-      "step": 36950
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.936585432378252e-05,
-      "loss": 0.7187,
-      "step": 36960
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.933857924093445e-05,
-      "loss": 0.6611,
-      "step": 36970
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.931130415808638e-05,
-      "loss": 0.6637,
-      "step": 36980
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.928402907523832e-05,
-      "loss": 0.6302,
-      "step": 36990
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.925675399239027e-05,
-      "loss": 0.7338,
-      "step": 37000
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.92294789095422e-05,
-      "loss": 0.6462,
-      "step": 37010
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.920220382669413e-05,
-      "loss": 0.5851,
-      "step": 37020
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.917492874384607e-05,
-      "loss": 0.7222,
-      "step": 37030
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.9147653660998e-05,
-      "loss": 0.6501,
-      "step": 37040
-    },
-    {
-      "epoch": 1.51,
-      "learning_rate": 9.912037857814993e-05,
-      "loss": 0.6681,
-      "step": 37050
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.909310349530188e-05,
-      "loss": 0.7032,
-      "step": 37060
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.906582841245382e-05,
-      "loss": 0.6741,
-      "step": 37070
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.903855332960575e-05,
-      "loss": 0.6603,
-      "step": 37080
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.901127824675768e-05,
-      "loss": 0.7209,
-      "step": 37090
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.89840031639096e-05,
-      "loss": 0.673,
-      "step": 37100
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.895672808106155e-05,
-      "loss": 0.6912,
-      "step": 37110
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.892945299821349e-05,
-      "loss": 0.749,
-      "step": 37120
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.890217791536542e-05,
-      "loss": 0.6136,
-      "step": 37130
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.887490283251735e-05,
-      "loss": 0.659,
-      "step": 37140
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.884762774966929e-05,
-      "loss": 0.7513,
-      "step": 37150
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.882035266682122e-05,
-      "loss": 0.7342,
-      "step": 37160
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.879307758397317e-05,
-      "loss": 0.6039,
-      "step": 37170
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.87658025011251e-05,
-      "loss": 0.7183,
-      "step": 37180
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.873852741827704e-05,
-      "loss": 0.7454,
-      "step": 37190
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.871125233542897e-05,
-      "loss": 0.5986,
-      "step": 37200
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.86839772525809e-05,
-      "loss": 0.6405,
-      "step": 37210
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.865670216973284e-05,
-      "loss": 0.7049,
-      "step": 37220
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.862942708688479e-05,
-      "loss": 0.7014,
-      "step": 37230
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.860215200403672e-05,
-      "loss": 0.6792,
-      "step": 37240
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.857487692118865e-05,
-      "loss": 0.7422,
-      "step": 37250
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.854760183834059e-05,
-      "loss": 0.6254,
-      "step": 37260
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.852032675549252e-05,
-      "loss": 0.6595,
-      "step": 37270
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.849305167264447e-05,
-      "loss": 0.7126,
-      "step": 37280
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.84657765897964e-05,
-      "loss": 0.7197,
-      "step": 37290
-    },
-    {
-      "epoch": 1.52,
-      "learning_rate": 9.843850150694834e-05,
-      "loss": 0.7539,
-      "step": 37300
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.841122642410027e-05,
-      "loss": 0.6418,
-      "step": 37310
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.83839513412522e-05,
-      "loss": 0.7068,
-      "step": 37320
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.835667625840414e-05,
-      "loss": 0.7175,
-      "step": 37330
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.832940117555609e-05,
-      "loss": 0.7152,
-      "step": 37340
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.830212609270802e-05,
-      "loss": 0.6163,
-      "step": 37350
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.827485100985995e-05,
-      "loss": 0.6158,
-      "step": 37360
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.824757592701189e-05,
-      "loss": 0.6691,
-      "step": 37370
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.822030084416382e-05,
-      "loss": 0.6945,
-      "step": 37380
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.819302576131576e-05,
-      "loss": 0.6112,
-      "step": 37390
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.816575067846769e-05,
-      "loss": 0.6718,
-      "step": 37400
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.813847559561962e-05,
-      "loss": 0.6937,
-      "step": 37410
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.811120051277156e-05,
-      "loss": 0.7517,
-      "step": 37420
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.808392542992349e-05,
-      "loss": 0.7894,
-      "step": 37430
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.805665034707543e-05,
-      "loss": 0.6594,
-      "step": 37440
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.802937526422737e-05,
-      "loss": 0.6894,
-      "step": 37450
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.80021001813793e-05,
-      "loss": 0.652,
-      "step": 37460
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.797482509853124e-05,
-      "loss": 0.6792,
-      "step": 37470
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.794755001568317e-05,
-      "loss": 0.6598,
-      "step": 37480
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.792027493283511e-05,
-      "loss": 0.7476,
-      "step": 37490
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.789299984998704e-05,
-      "loss": 0.6272,
-      "step": 37500
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.786572476713899e-05,
-      "loss": 0.7447,
-      "step": 37510
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.783844968429092e-05,
-      "loss": 0.7434,
-      "step": 37520
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.781117460144286e-05,
-      "loss": 0.5763,
-      "step": 37530
-    },
-    {
-      "epoch": 1.53,
-      "learning_rate": 9.778389951859479e-05,
-      "loss": 0.6455,
-      "step": 37540
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.775662443574673e-05,
-      "loss": 0.7171,
-      "step": 37550
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.772934935289867e-05,
-      "loss": 0.6391,
-      "step": 37560
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.770207427005061e-05,
-      "loss": 0.6704,
-      "step": 37570
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.767479918720254e-05,
-      "loss": 0.7222,
-      "step": 37580
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.764752410435447e-05,
-      "loss": 0.8152,
-      "step": 37590
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.762024902150641e-05,
-      "loss": 0.7543,
-      "step": 37600
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.759297393865834e-05,
-      "loss": 0.7237,
-      "step": 37610
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.756569885581029e-05,
-      "loss": 0.6644,
-      "step": 37620
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.753842377296222e-05,
-      "loss": 0.6283,
-      "step": 37630
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.751114869011416e-05,
-      "loss": 0.7454,
-      "step": 37640
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.748387360726609e-05,
-      "loss": 0.7171,
-      "step": 37650
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.745659852441803e-05,
-      "loss": 0.6513,
-      "step": 37660
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.742932344156996e-05,
-      "loss": 0.5974,
-      "step": 37670
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.74020483587219e-05,
-      "loss": 0.7099,
-      "step": 37680
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.737477327587383e-05,
-      "loss": 0.7508,
-      "step": 37690
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.734749819302576e-05,
-      "loss": 0.7034,
-      "step": 37700
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.73202231101777e-05,
-      "loss": 0.652,
-      "step": 37710
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.729294802732963e-05,
-      "loss": 0.6571,
-      "step": 37720
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.726567294448158e-05,
-      "loss": 0.6806,
-      "step": 37730
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.723839786163351e-05,
-      "loss": 0.7354,
-      "step": 37740
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.721112277878544e-05,
-      "loss": 0.7051,
-      "step": 37750
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.718384769593738e-05,
-      "loss": 0.6576,
-      "step": 37760
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.715657261308931e-05,
-      "loss": 0.7414,
-      "step": 37770
-    },
-    {
-      "epoch": 1.54,
-      "learning_rate": 9.712929753024125e-05,
-      "loss": 0.6665,
-      "step": 37780
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.71020224473932e-05,
-      "loss": 0.6848,
-      "step": 37790
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.707474736454513e-05,
-      "loss": 0.739,
-      "step": 37800
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.704747228169706e-05,
-      "loss": 0.6215,
-      "step": 37810
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.7020197198849e-05,
-      "loss": 0.6843,
-      "step": 37820
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.699292211600093e-05,
-      "loss": 0.7402,
-      "step": 37830
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.696564703315288e-05,
-      "loss": 0.6509,
-      "step": 37840
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.693837195030481e-05,
-      "loss": 0.7073,
-      "step": 37850
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.691109686745674e-05,
-      "loss": 0.7059,
-      "step": 37860
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.688382178460868e-05,
-      "loss": 0.7469,
-      "step": 37870
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.685654670176061e-05,
-      "loss": 0.7078,
-      "step": 37880
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.682927161891255e-05,
-      "loss": 0.7005,
-      "step": 37890
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.68019965360645e-05,
-      "loss": 0.7306,
-      "step": 37900
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.677472145321643e-05,
-      "loss": 0.7175,
-      "step": 37910
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.674744637036836e-05,
-      "loss": 0.813,
-      "step": 37920
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.67201712875203e-05,
-      "loss": 0.6438,
-      "step": 37930
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.669289620467223e-05,
-      "loss": 0.6546,
-      "step": 37940
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.666562112182416e-05,
-      "loss": 0.7797,
-      "step": 37950
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.66383460389761e-05,
-      "loss": 0.604,
-      "step": 37960
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.661107095612803e-05,
-      "loss": 0.7009,
-      "step": 37970
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.658379587327997e-05,
-      "loss": 0.7388,
-      "step": 37980
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.65565207904319e-05,
-      "loss": 0.7362,
-      "step": 37990
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.652924570758383e-05,
-      "loss": 0.7067,
-      "step": 38000
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.650197062473578e-05,
-      "loss": 0.6996,
-      "step": 38010
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.647469554188771e-05,
-      "loss": 0.6685,
-      "step": 38020
-    },
-    {
-      "epoch": 1.55,
-      "learning_rate": 9.644742045903965e-05,
-      "loss": 0.641,
-      "step": 38030
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.642014537619158e-05,
-      "loss": 0.689,
-      "step": 38040
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.639287029334352e-05,
-      "loss": 0.6,
-      "step": 38050
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.636559521049545e-05,
-      "loss": 0.696,
-      "step": 38060
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.63383201276474e-05,
-      "loss": 0.7449,
-      "step": 38070
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.631104504479933e-05,
-      "loss": 0.7374,
-      "step": 38080
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.628376996195127e-05,
-      "loss": 0.7697,
-      "step": 38090
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.62564948791032e-05,
-      "loss": 0.7221,
-      "step": 38100
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.622921979625513e-05,
-      "loss": 0.7618,
-      "step": 38110
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.620194471340707e-05,
-      "loss": 0.585,
-      "step": 38120
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.617466963055901e-05,
-      "loss": 0.6425,
-      "step": 38130
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.614739454771095e-05,
-      "loss": 0.7821,
-      "step": 38140
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.612011946486288e-05,
-      "loss": 0.6639,
-      "step": 38150
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.609284438201482e-05,
-      "loss": 0.6185,
-      "step": 38160
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.606556929916675e-05,
-      "loss": 0.7697,
-      "step": 38170
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.60382942163187e-05,
-      "loss": 0.6435,
-      "step": 38180
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.601101913347063e-05,
-      "loss": 0.5991,
-      "step": 38190
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.598374405062257e-05,
-      "loss": 0.707,
-      "step": 38200
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.59564689677745e-05,
-      "loss": 0.7253,
-      "step": 38210
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.592919388492643e-05,
-      "loss": 0.625,
-      "step": 38220
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.590191880207837e-05,
-      "loss": 0.6751,
-      "step": 38230
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.58746437192303e-05,
-      "loss": 0.7351,
-      "step": 38240
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.584736863638224e-05,
-      "loss": 0.6161,
-      "step": 38250
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.582009355353417e-05,
-      "loss": 0.7035,
-      "step": 38260
-    },
-    {
-      "epoch": 1.56,
-      "learning_rate": 9.57928184706861e-05,
-      "loss": 0.6068,
-      "step": 38270
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.576554338783804e-05,
-      "loss": 0.657,
-      "step": 38280
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.573826830498998e-05,
-      "loss": 0.7285,
-      "step": 38290
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.571099322214192e-05,
-      "loss": 0.622,
-      "step": 38300
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.568371813929385e-05,
-      "loss": 0.744,
-      "step": 38310
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.565644305644579e-05,
-      "loss": 0.7174,
-      "step": 38320
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.562916797359772e-05,
-      "loss": 0.6699,
-      "step": 38330
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.560189289074965e-05,
-      "loss": 0.6798,
-      "step": 38340
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.55746178079016e-05,
-      "loss": 0.7331,
-      "step": 38350
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.554734272505354e-05,
-      "loss": 0.646,
-      "step": 38360
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.552006764220547e-05,
-      "loss": 0.6654,
-      "step": 38370
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.54927925593574e-05,
-      "loss": 0.6945,
-      "step": 38380
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.546551747650934e-05,
-      "loss": 0.7369,
-      "step": 38390
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.543824239366127e-05,
-      "loss": 0.6083,
-      "step": 38400
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.541096731081322e-05,
-      "loss": 0.7575,
-      "step": 38410
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.538369222796515e-05,
-      "loss": 0.6306,
-      "step": 38420
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.535641714511709e-05,
-      "loss": 0.6695,
-      "step": 38430
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.532914206226902e-05,
-      "loss": 0.6921,
-      "step": 38440
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.530186697942095e-05,
-      "loss": 0.6057,
-      "step": 38450
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.52745918965729e-05,
-      "loss": 0.7206,
-      "step": 38460
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.524731681372484e-05,
-      "loss": 0.6964,
-      "step": 38470
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.522004173087677e-05,
-      "loss": 0.7107,
-      "step": 38480
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.51927666480287e-05,
-      "loss": 0.7038,
-      "step": 38490
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.516549156518064e-05,
-      "loss": 0.7706,
-      "step": 38500
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.513821648233257e-05,
-      "loss": 0.6766,
-      "step": 38510
-    },
-    {
-      "epoch": 1.57,
-      "learning_rate": 9.51109413994845e-05,
-      "loss": 0.8288,
-      "step": 38520
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.508366631663644e-05,
-      "loss": 0.6477,
-      "step": 38530
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.505639123378837e-05,
-      "loss": 0.6732,
-      "step": 38540
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.502911615094031e-05,
-      "loss": 0.7285,
-      "step": 38550
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.500184106809224e-05,
-      "loss": 0.8041,
-      "step": 38560
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.497456598524418e-05,
-      "loss": 0.6914,
-      "step": 38570
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.494729090239612e-05,
-      "loss": 0.6917,
-      "step": 38580
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.492001581954806e-05,
-      "loss": 0.6621,
-      "step": 38590
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.489274073669999e-05,
-      "loss": 0.6154,
-      "step": 38600
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.486546565385192e-05,
-      "loss": 0.6548,
-      "step": 38610
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.483819057100386e-05,
-      "loss": 0.7727,
-      "step": 38620
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.48109154881558e-05,
-      "loss": 0.6261,
-      "step": 38630
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.478364040530774e-05,
-      "loss": 0.639,
-      "step": 38640
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.475636532245967e-05,
-      "loss": 0.7037,
-      "step": 38650
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.472909023961161e-05,
-      "loss": 0.6872,
-      "step": 38660
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.470181515676354e-05,
-      "loss": 0.6222,
-      "step": 38670
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.467454007391548e-05,
-      "loss": 0.6676,
-      "step": 38680
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.464726499106742e-05,
-      "loss": 0.6694,
-      "step": 38690
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.461998990821936e-05,
-      "loss": 0.6873,
-      "step": 38700
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.459271482537129e-05,
-      "loss": 0.7044,
-      "step": 38710
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.456543974252322e-05,
-      "loss": 0.7048,
-      "step": 38720
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.453816465967516e-05,
-      "loss": 0.6555,
-      "step": 38730
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.45108895768271e-05,
-      "loss": 0.694,
-      "step": 38740
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.448361449397904e-05,
-      "loss": 0.7011,
-      "step": 38750
-    },
-    {
-      "epoch": 1.58,
-      "learning_rate": 9.445633941113097e-05,
-      "loss": 0.6077,
-      "step": 38760
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.442906432828291e-05,
-      "loss": 0.7743,
-      "step": 38770
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.440178924543484e-05,
-      "loss": 0.6598,
-      "step": 38780
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.437451416258678e-05,
-      "loss": 0.7474,
-      "step": 38790
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.434723907973871e-05,
-      "loss": 0.7004,
-      "step": 38800
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.431996399689064e-05,
-      "loss": 0.6044,
-      "step": 38810
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.429268891404258e-05,
-      "loss": 0.6308,
-      "step": 38820
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.426541383119451e-05,
-      "loss": 0.6578,
-      "step": 38830
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.423813874834645e-05,
-      "loss": 0.7153,
-      "step": 38840
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.421086366549838e-05,
-      "loss": 0.6392,
-      "step": 38850
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.418358858265033e-05,
-      "loss": 0.6034,
-      "step": 38860
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.415631349980226e-05,
-      "loss": 0.6438,
-      "step": 38870
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.41290384169542e-05,
-      "loss": 0.612,
-      "step": 38880
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.410176333410613e-05,
-      "loss": 0.6189,
-      "step": 38890
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.407448825125806e-05,
-      "loss": 0.6339,
-      "step": 38900
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.404721316841001e-05,
-      "loss": 0.7405,
-      "step": 38910
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.401993808556194e-05,
-      "loss": 0.6793,
-      "step": 38920
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.399266300271388e-05,
-      "loss": 0.6959,
-      "step": 38930
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.396538791986581e-05,
-      "loss": 0.7485,
-      "step": 38940
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.393811283701775e-05,
-      "loss": 0.636,
-      "step": 38950
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.391083775416968e-05,
-      "loss": 0.6483,
-      "step": 38960
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.388356267132163e-05,
-      "loss": 0.6647,
-      "step": 38970
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.385628758847356e-05,
-      "loss": 0.7806,
-      "step": 38980
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.38290125056255e-05,
-      "loss": 0.6424,
-      "step": 38990
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.380173742277743e-05,
-      "loss": 0.7239,
-      "step": 39000
-    },
-    {
-      "epoch": 1.59,
-      "learning_rate": 9.377446233992936e-05,
-      "loss": 0.7678,
-      "step": 39010
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.37471872570813e-05,
-      "loss": 0.6797,
-      "step": 39020
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.371991217423324e-05,
-      "loss": 0.6893,
-      "step": 39030
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.369263709138518e-05,
-      "loss": 0.7272,
-      "step": 39040
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.366536200853711e-05,
-      "loss": 0.7499,
-      "step": 39050
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.363808692568905e-05,
-      "loss": 0.6203,
-      "step": 39060
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.361081184284098e-05,
-      "loss": 0.6711,
-      "step": 39070
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.358353675999291e-05,
-      "loss": 0.6475,
-      "step": 39080
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.355626167714485e-05,
-      "loss": 0.6974,
-      "step": 39090
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.352898659429678e-05,
-      "loss": 0.615,
-      "step": 39100
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.350171151144872e-05,
-      "loss": 0.7495,
-      "step": 39110
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.347443642860065e-05,
-      "loss": 0.7354,
-      "step": 39120
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.344716134575258e-05,
-      "loss": 0.6918,
-      "step": 39130
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.341988626290453e-05,
-      "loss": 0.8137,
-      "step": 39140
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.339261118005646e-05,
-      "loss": 0.8551,
-      "step": 39150
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.33653360972084e-05,
-      "loss": 0.6542,
-      "step": 39160
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.333806101436033e-05,
-      "loss": 0.6411,
-      "step": 39170
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.331078593151227e-05,
-      "loss": 0.7251,
-      "step": 39180
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.32835108486642e-05,
-      "loss": 0.6998,
-      "step": 39190
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.325623576581615e-05,
-      "loss": 0.6765,
-      "step": 39200
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.322896068296808e-05,
-      "loss": 0.5852,
-      "step": 39210
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.320168560012002e-05,
-      "loss": 0.6203,
-      "step": 39220
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.317441051727195e-05,
-      "loss": 0.6497,
-      "step": 39230
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.314713543442388e-05,
-      "loss": 0.729,
-      "step": 39240
-    },
-    {
-      "epoch": 1.6,
-      "learning_rate": 9.311986035157583e-05,
-      "loss": 0.7076,
-      "step": 39250
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.309258526872776e-05,
-      "loss": 0.6753,
-      "step": 39260
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.30653101858797e-05,
-      "loss": 0.6424,
-      "step": 39270
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.303803510303163e-05,
-      "loss": 0.7749,
-      "step": 39280
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.301076002018357e-05,
-      "loss": 0.7297,
-      "step": 39290
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.29834849373355e-05,
-      "loss": 0.699,
-      "step": 39300
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.295620985448745e-05,
-      "loss": 0.6564,
-      "step": 39310
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.292893477163938e-05,
-      "loss": 0.694,
-      "step": 39320
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.290165968879132e-05,
-      "loss": 0.63,
-      "step": 39330
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.287438460594325e-05,
-      "loss": 0.7045,
-      "step": 39340
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.284710952309518e-05,
-      "loss": 0.5647,
-      "step": 39350
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.281983444024712e-05,
-      "loss": 0.6854,
-      "step": 39360
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.279255935739905e-05,
-      "loss": 0.7217,
-      "step": 39370
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.276528427455099e-05,
-      "loss": 0.7428,
-      "step": 39380
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.273800919170292e-05,
-      "loss": 0.6855,
-      "step": 39390
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.271073410885485e-05,
-      "loss": 0.6731,
-      "step": 39400
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.268345902600679e-05,
-      "loss": 0.6883,
-      "step": 39410
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.265618394315873e-05,
-      "loss": 0.6318,
-      "step": 39420
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.262890886031067e-05,
-      "loss": 0.6083,
-      "step": 39430
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.26016337774626e-05,
-      "loss": 0.6436,
-      "step": 39440
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.257435869461454e-05,
-      "loss": 0.6523,
-      "step": 39450
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.254708361176647e-05,
-      "loss": 0.7312,
-      "step": 39460
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.25198085289184e-05,
-      "loss": 0.7402,
-      "step": 39470
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.249253344607035e-05,
-      "loss": 0.7184,
-      "step": 39480
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.246525836322229e-05,
-      "loss": 0.6949,
-      "step": 39490
-    },
-    {
-      "epoch": 1.61,
-      "learning_rate": 9.243798328037422e-05,
-      "loss": 0.6714,
-      "step": 39500
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.241070819752615e-05,
-      "loss": 0.6742,
-      "step": 39510
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.238343311467809e-05,
-      "loss": 0.7094,
-      "step": 39520
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.235615803183003e-05,
-      "loss": 0.7143,
-      "step": 39530
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.232888294898197e-05,
-      "loss": 0.6863,
-      "step": 39540
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.23016078661339e-05,
-      "loss": 0.6647,
-      "step": 39550
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.227433278328584e-05,
-      "loss": 0.6492,
-      "step": 39560
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.224705770043777e-05,
-      "loss": 0.5496,
-      "step": 39570
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.22197826175897e-05,
-      "loss": 0.7552,
-      "step": 39580
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.219250753474165e-05,
-      "loss": 0.6284,
-      "step": 39590
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.216523245189359e-05,
-      "loss": 0.6918,
-      "step": 39600
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.213795736904552e-05,
-      "loss": 0.7037,
-      "step": 39610
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.211068228619745e-05,
-      "loss": 0.707,
-      "step": 39620
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.208340720334939e-05,
-      "loss": 0.6296,
-      "step": 39630
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.205613212050132e-05,
-      "loss": 0.6553,
-      "step": 39640
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.202885703765326e-05,
-      "loss": 0.6823,
-      "step": 39650
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.200158195480519e-05,
-      "loss": 0.6603,
-      "step": 39660
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.197430687195712e-05,
-      "loss": 0.7257,
-      "step": 39670
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.194703178910906e-05,
-      "loss": 0.6357,
-      "step": 39680
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.191975670626099e-05,
-      "loss": 0.7361,
-      "step": 39690
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.189248162341294e-05,
-      "loss": 0.6861,
-      "step": 39700
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.186520654056487e-05,
-      "loss": 0.6572,
-      "step": 39710
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.18379314577168e-05,
-      "loss": 0.7265,
-      "step": 39720
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.181065637486874e-05,
-      "loss": 0.6158,
-      "step": 39730
-    },
-    {
-      "epoch": 1.62,
-      "learning_rate": 9.178338129202067e-05,
-      "loss": 0.7462,
-      "step": 39740
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.175610620917261e-05,
-      "loss": 0.6588,
-      "step": 39750
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.172883112632456e-05,
-      "loss": 0.7201,
-      "step": 39760
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.170155604347649e-05,
-      "loss": 0.6607,
-      "step": 39770
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.167428096062842e-05,
-      "loss": 0.678,
-      "step": 39780
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.164700587778036e-05,
-      "loss": 0.6471,
-      "step": 39790
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.161973079493229e-05,
-      "loss": 0.6558,
-      "step": 39800
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.159245571208424e-05,
-      "loss": 0.7321,
-      "step": 39810
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.156518062923617e-05,
-      "loss": 0.7497,
-      "step": 39820
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.15379055463881e-05,
-      "loss": 0.6377,
-      "step": 39830
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.151063046354004e-05,
-      "loss": 0.6612,
-      "step": 39840
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.148335538069197e-05,
-      "loss": 0.8174,
-      "step": 39850
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.145608029784391e-05,
-      "loss": 0.6225,
-      "step": 39860
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.142880521499586e-05,
-      "loss": 0.6945,
-      "step": 39870
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.140153013214779e-05,
-      "loss": 0.6616,
-      "step": 39880
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.137425504929972e-05,
-      "loss": 0.8027,
-      "step": 39890
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.134697996645166e-05,
-      "loss": 0.7343,
-      "step": 39900
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.131970488360359e-05,
-      "loss": 0.6054,
-      "step": 39910
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.129242980075553e-05,
-      "loss": 0.6057,
-      "step": 39920
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.126515471790746e-05,
-      "loss": 0.7036,
-      "step": 39930
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.12378796350594e-05,
-      "loss": 0.675,
-      "step": 39940
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.121333206049613e-05,
-      "loss": 0.7507,
-      "step": 39950
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.118605697764807e-05,
-      "loss": 0.6294,
-      "step": 39960
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.11587818948e-05,
-      "loss": 0.758,
-      "step": 39970
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.113150681195195e-05,
-      "loss": 0.7296,
-      "step": 39980
-    },
-    {
-      "epoch": 1.63,
-      "learning_rate": 9.110423172910388e-05,
-      "loss": 0.7157,
-      "step": 39990
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.107695664625582e-05,
-      "loss": 0.7788,
-      "step": 40000
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.104968156340775e-05,
-      "loss": 0.7211,
-      "step": 40010
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.102240648055968e-05,
-      "loss": 0.7015,
-      "step": 40020
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.099513139771163e-05,
-      "loss": 0.7513,
-      "step": 40030
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.096785631486357e-05,
-      "loss": 0.6918,
-      "step": 40040
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.09405812320155e-05,
-      "loss": 0.7437,
-      "step": 40050
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.091330614916743e-05,
-      "loss": 0.7633,
-      "step": 40060
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.088603106631937e-05,
-      "loss": 0.6435,
-      "step": 40070
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.08587559834713e-05,
-      "loss": 0.6839,
-      "step": 40080
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.083148090062325e-05,
-      "loss": 0.7493,
-      "step": 40090
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.080420581777518e-05,
-      "loss": 0.738,
-      "step": 40100
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.077693073492712e-05,
-      "loss": 0.5646,
-      "step": 40110
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.074965565207905e-05,
-      "loss": 0.7595,
-      "step": 40120
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.072238056923098e-05,
-      "loss": 0.7072,
-      "step": 40130
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.069510548638292e-05,
-      "loss": 0.7927,
-      "step": 40140
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.066783040353485e-05,
-      "loss": 0.6036,
-      "step": 40150
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.064055532068679e-05,
-      "loss": 0.6703,
-      "step": 40160
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.061328023783872e-05,
-      "loss": 0.6676,
-      "step": 40170
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.058600515499065e-05,
-      "loss": 0.7697,
-      "step": 40180
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.055873007214259e-05,
-      "loss": 0.72,
-      "step": 40190
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.053145498929454e-05,
-      "loss": 0.6961,
-      "step": 40200
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.050417990644647e-05,
-      "loss": 0.6128,
-      "step": 40210
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.04769048235984e-05,
-      "loss": 0.7126,
-      "step": 40220
-    },
-    {
-      "epoch": 1.64,
-      "learning_rate": 9.044962974075034e-05,
-      "loss": 0.7553,
-      "step": 40230
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.042235465790227e-05,
-      "loss": 0.7474,
-      "step": 40240
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.03950795750542e-05,
-      "loss": 0.583,
-      "step": 40250
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.036780449220615e-05,
-      "loss": 0.7414,
-      "step": 40260
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.034052940935809e-05,
-      "loss": 0.6963,
-      "step": 40270
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.031325432651002e-05,
-      "loss": 0.7055,
-      "step": 40280
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.028597924366195e-05,
-      "loss": 0.6578,
-      "step": 40290
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.025870416081389e-05,
-      "loss": 0.6214,
-      "step": 40300
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.023142907796582e-05,
-      "loss": 0.8496,
-      "step": 40310
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.020415399511777e-05,
-      "loss": 0.7295,
-      "step": 40320
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.01768789122697e-05,
-      "loss": 0.6484,
-      "step": 40330
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.014960382942164e-05,
-      "loss": 0.7074,
-      "step": 40340
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.012232874657357e-05,
-      "loss": 0.7346,
-      "step": 40350
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.00950536637255e-05,
-      "loss": 0.6356,
-      "step": 40360
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.006777858087745e-05,
-      "loss": 0.7126,
-      "step": 40370
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.004050349802939e-05,
-      "loss": 0.6302,
-      "step": 40380
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 9.001322841518132e-05,
-      "loss": 0.7242,
-      "step": 40390
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.998595333233326e-05,
-      "loss": 0.7824,
-      "step": 40400
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.995867824948519e-05,
-      "loss": 0.667,
-      "step": 40410
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.993140316663712e-05,
-      "loss": 0.7026,
-      "step": 40420
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.990412808378906e-05,
-      "loss": 0.7434,
-      "step": 40430
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.987685300094099e-05,
-      "loss": 0.7026,
-      "step": 40440
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.984957791809292e-05,
-      "loss": 0.6669,
-      "step": 40450
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.982230283524486e-05,
-      "loss": 0.6744,
-      "step": 40460
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.979502775239679e-05,
-      "loss": 0.745,
-      "step": 40470
-    },
-    {
-      "epoch": 1.65,
-      "learning_rate": 8.976775266954873e-05,
-      "loss": 0.7421,
-      "step": 40480
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.974047758670067e-05,
-      "loss": 0.7138,
-      "step": 40490
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.971320250385261e-05,
-      "loss": 0.6473,
-      "step": 40500
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.968592742100454e-05,
-      "loss": 0.7447,
-      "step": 40510
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.965865233815648e-05,
-      "loss": 0.6928,
-      "step": 40520
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.963137725530841e-05,
-      "loss": 0.6559,
-      "step": 40530
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.960410217246036e-05,
-      "loss": 0.7271,
-      "step": 40540
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.957682708961229e-05,
-      "loss": 0.6741,
-      "step": 40550
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.954955200676422e-05,
-      "loss": 0.7017,
-      "step": 40560
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.952227692391616e-05,
-      "loss": 0.7268,
-      "step": 40570
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.949500184106809e-05,
-      "loss": 0.6727,
-      "step": 40580
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.946772675822003e-05,
-      "loss": 0.7516,
-      "step": 40590
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.944045167537197e-05,
-      "loss": 0.7122,
-      "step": 40600
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.941317659252391e-05,
-      "loss": 0.6934,
-      "step": 40610
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.938590150967584e-05,
-      "loss": 0.6631,
-      "step": 40620
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.935862642682778e-05,
-      "loss": 0.6943,
-      "step": 40630
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.933135134397971e-05,
-      "loss": 0.6657,
-      "step": 40640
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.930407626113166e-05,
-      "loss": 0.7502,
-      "step": 40650
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.927680117828359e-05,
-      "loss": 0.6507,
-      "step": 40660
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.924952609543553e-05,
-      "loss": 0.7207,
-      "step": 40670
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.922225101258746e-05,
-      "loss": 0.8606,
-      "step": 40680
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.919497592973939e-05,
-      "loss": 0.7148,
-      "step": 40690
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.916770084689133e-05,
-      "loss": 0.7375,
-      "step": 40700
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.914042576404326e-05,
-      "loss": 0.6107,
-      "step": 40710
-    },
-    {
-      "epoch": 1.66,
-      "learning_rate": 8.91131506811952e-05,
-      "loss": 0.6686,
-      "step": 40720
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.908587559834713e-05,
-      "loss": 0.668,
-      "step": 40730
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.905860051549906e-05,
-      "loss": 0.7284,
-      "step": 40740
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.9031325432651e-05,
-      "loss": 0.6127,
-      "step": 40750
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.900405034980293e-05,
-      "loss": 0.7988,
-      "step": 40760
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.897677526695488e-05,
-      "loss": 0.7593,
-      "step": 40770
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.894950018410681e-05,
-      "loss": 0.7659,
-      "step": 40780
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.892222510125875e-05,
-      "loss": 0.6721,
-      "step": 40790
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.889495001841068e-05,
-      "loss": 0.7618,
-      "step": 40800
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.886767493556261e-05,
-      "loss": 0.6883,
-      "step": 40810
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.884039985271456e-05,
-      "loss": 0.6182,
-      "step": 40820
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.88131247698665e-05,
-      "loss": 0.7265,
-      "step": 40830
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.878584968701843e-05,
-      "loss": 0.5516,
-      "step": 40840
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.875857460417036e-05,
-      "loss": 0.7249,
-      "step": 40850
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.87312995213223e-05,
-      "loss": 0.6585,
-      "step": 40860
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.870402443847423e-05,
-      "loss": 0.6973,
-      "step": 40870
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.867674935562618e-05,
-      "loss": 0.7235,
-      "step": 40880
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.864947427277811e-05,
-      "loss": 0.6732,
-      "step": 40890
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.862219918993005e-05,
-      "loss": 0.6857,
-      "step": 40900
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.859492410708198e-05,
-      "loss": 0.7328,
-      "step": 40910
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.856764902423391e-05,
-      "loss": 0.7646,
-      "step": 40920
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.854037394138585e-05,
-      "loss": 0.6644,
-      "step": 40930
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.85130988585378e-05,
-      "loss": 0.7569,
-      "step": 40940
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.848582377568973e-05,
-      "loss": 0.648,
-      "step": 40950
-    },
-    {
-      "epoch": 1.67,
-      "learning_rate": 8.845854869284166e-05,
-      "loss": 0.7327,
-      "step": 40960
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.84312736099936e-05,
-      "loss": 0.7537,
-      "step": 40970
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.840399852714553e-05,
-      "loss": 0.7521,
-      "step": 40980
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.837672344429746e-05,
-      "loss": 0.7238,
-      "step": 40990
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.83494483614494e-05,
-      "loss": 0.7117,
-      "step": 41000
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.832217327860133e-05,
-      "loss": 0.7795,
-      "step": 41010
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.829489819575327e-05,
-      "loss": 0.7955,
-      "step": 41020
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.82676231129052e-05,
-      "loss": 0.6211,
-      "step": 41030
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.824034803005713e-05,
-      "loss": 0.6554,
-      "step": 41040
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.821307294720908e-05,
-      "loss": 0.6575,
-      "step": 41050
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.818579786436102e-05,
-      "loss": 0.6782,
-      "step": 41060
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.815852278151295e-05,
-      "loss": 0.7275,
-      "step": 41070
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.813124769866488e-05,
-      "loss": 0.7079,
-      "step": 41080
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.810397261581682e-05,
-      "loss": 0.6501,
-      "step": 41090
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.807669753296876e-05,
-      "loss": 0.7118,
-      "step": 41100
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.80494224501207e-05,
-      "loss": 0.6719,
-      "step": 41110
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.802214736727263e-05,
-      "loss": 0.6798,
-      "step": 41120
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.799487228442457e-05,
-      "loss": 0.7037,
-      "step": 41130
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.79675972015765e-05,
-      "loss": 0.7515,
-      "step": 41140
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.794032211872843e-05,
-      "loss": 0.6809,
-      "step": 41150
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.791304703588038e-05,
-      "loss": 0.7499,
-      "step": 41160
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.788577195303232e-05,
-      "loss": 0.6875,
-      "step": 41170
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.785849687018425e-05,
-      "loss": 0.6146,
-      "step": 41180
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.783122178733618e-05,
-      "loss": 0.7612,
-      "step": 41190
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.780394670448812e-05,
-      "loss": 0.6864,
-      "step": 41200
-    },
-    {
-      "epoch": 1.68,
-      "learning_rate": 8.777667162164005e-05,
-      "loss": 0.7077,
-      "step": 41210
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.7749396538792e-05,
-      "loss": 0.7331,
-      "step": 41220
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.772212145594393e-05,
-      "loss": 0.7618,
-      "step": 41230
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.769484637309587e-05,
-      "loss": 0.6681,
-      "step": 41240
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.76675712902478e-05,
-      "loss": 0.6609,
-      "step": 41250
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.764029620739973e-05,
-      "loss": 0.6476,
-      "step": 41260
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.761302112455167e-05,
-      "loss": 0.7017,
-      "step": 41270
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.75857460417036e-05,
-      "loss": 0.7131,
-      "step": 41280
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.755847095885554e-05,
-      "loss": 0.6806,
-      "step": 41290
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.753119587600747e-05,
-      "loss": 0.7288,
-      "step": 41300
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.75039207931594e-05,
-      "loss": 0.6681,
-      "step": 41310
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.747664571031134e-05,
-      "loss": 0.7441,
-      "step": 41320
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.744937062746329e-05,
-      "loss": 0.6851,
-      "step": 41330
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.742209554461522e-05,
-      "loss": 0.7392,
-      "step": 41340
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.739482046176715e-05,
-      "loss": 0.726,
-      "step": 41350
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.736754537891909e-05,
-      "loss": 0.7583,
-      "step": 41360
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.734027029607102e-05,
-      "loss": 0.7908,
-      "step": 41370
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.731299521322296e-05,
-      "loss": 0.6062,
-      "step": 41380
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.72857201303749e-05,
-      "loss": 0.5837,
-      "step": 41390
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.725844504752684e-05,
-      "loss": 0.6233,
-      "step": 41400
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.723116996467877e-05,
-      "loss": 0.6409,
-      "step": 41410
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.72038948818307e-05,
-      "loss": 0.706,
-      "step": 41420
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.717661979898264e-05,
-      "loss": 0.6973,
-      "step": 41430
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.714934471613459e-05,
-      "loss": 0.7278,
-      "step": 41440
-    },
-    {
-      "epoch": 1.69,
-      "learning_rate": 8.712206963328652e-05,
-      "loss": 0.7098,
-      "step": 41450
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.709479455043845e-05,
-      "loss": 0.6057,
-      "step": 41460
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.706751946759039e-05,
-      "loss": 0.6916,
-      "step": 41470
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.704024438474232e-05,
-      "loss": 0.6842,
-      "step": 41480
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.701296930189426e-05,
-      "loss": 0.7204,
-      "step": 41490
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.69856942190462e-05,
-      "loss": 0.6639,
-      "step": 41500
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.695841913619814e-05,
-      "loss": 0.7269,
-      "step": 41510
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.693114405335007e-05,
-      "loss": 0.6652,
-      "step": 41520
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.6903868970502e-05,
-      "loss": 0.6521,
-      "step": 41530
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.687659388765394e-05,
-      "loss": 0.6544,
-      "step": 41540
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.684931880480587e-05,
-      "loss": 0.6959,
-      "step": 41550
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.68220437219578e-05,
-      "loss": 0.6914,
-      "step": 41560
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.679476863910974e-05,
-      "loss": 0.6688,
-      "step": 41570
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.676749355626167e-05,
-      "loss": 0.6831,
-      "step": 41580
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.674021847341361e-05,
-      "loss": 0.6814,
-      "step": 41590
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.671294339056554e-05,
-      "loss": 0.7563,
-      "step": 41600
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.668566830771749e-05,
-      "loss": 0.7222,
-      "step": 41610
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.665839322486942e-05,
-      "loss": 0.6352,
-      "step": 41620
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.663111814202136e-05,
-      "loss": 0.7811,
-      "step": 41630
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.660384305917329e-05,
-      "loss": 0.7275,
-      "step": 41640
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.657656797632523e-05,
-      "loss": 0.7057,
-      "step": 41650
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.654929289347716e-05,
-      "loss": 0.7,
-      "step": 41660
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.652201781062911e-05,
-      "loss": 0.6514,
-      "step": 41670
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.649474272778104e-05,
-      "loss": 0.6351,
-      "step": 41680
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.646746764493297e-05,
-      "loss": 0.6992,
-      "step": 41690
-    },
-    {
-      "epoch": 1.7,
-      "learning_rate": 8.644019256208491e-05,
-      "loss": 0.6895,
-      "step": 41700
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.641291747923684e-05,
-      "loss": 0.6469,
-      "step": 41710
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.638564239638879e-05,
-      "loss": 0.6867,
-      "step": 41720
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.635836731354072e-05,
-      "loss": 0.6292,
-      "step": 41730
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.633109223069266e-05,
-      "loss": 0.6619,
-      "step": 41740
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.630381714784459e-05,
-      "loss": 0.6683,
-      "step": 41750
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.627654206499653e-05,
-      "loss": 0.6721,
-      "step": 41760
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.624926698214846e-05,
-      "loss": 0.6629,
-      "step": 41770
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.622199189930041e-05,
-      "loss": 0.684,
-      "step": 41780
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.619471681645234e-05,
-      "loss": 0.7934,
-      "step": 41790
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.616744173360427e-05,
-      "loss": 0.6565,
-      "step": 41800
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.614016665075621e-05,
-      "loss": 0.6748,
-      "step": 41810
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.611289156790814e-05,
-      "loss": 0.7711,
-      "step": 41820
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.608561648506008e-05,
-      "loss": 0.6154,
-      "step": 41830
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.605834140221201e-05,
-      "loss": 0.7486,
-      "step": 41840
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.603106631936394e-05,
-      "loss": 0.7164,
-      "step": 41850
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.600379123651588e-05,
-      "loss": 0.6165,
-      "step": 41860
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.597651615366781e-05,
-      "loss": 0.6556,
-      "step": 41870
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.594924107081975e-05,
-      "loss": 0.7361,
-      "step": 41880
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.59219659879717e-05,
-      "loss": 0.7537,
-      "step": 41890
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.589469090512363e-05,
-      "loss": 0.783,
-      "step": 41900
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.586741582227556e-05,
-      "loss": 0.6804,
-      "step": 41910
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.58401407394275e-05,
-      "loss": 0.6632,
-      "step": 41920
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.581286565657943e-05,
-      "loss": 0.7231,
-      "step": 41930
-    },
-    {
-      "epoch": 1.71,
-      "learning_rate": 8.578559057373136e-05,
-      "loss": 0.8189,
-      "step": 41940
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.575831549088331e-05,
-      "loss": 0.7769,
-      "step": 41950
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.573104040803524e-05,
-      "loss": 0.7278,
-      "step": 41960
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.570376532518718e-05,
-      "loss": 0.7833,
-      "step": 41970
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.567649024233911e-05,
-      "loss": 0.611,
-      "step": 41980
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.564921515949105e-05,
-      "loss": 0.7953,
-      "step": 41990
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.5621940076643e-05,
-      "loss": 0.6429,
-      "step": 42000
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.559466499379493e-05,
-      "loss": 0.7285,
-      "step": 42010
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.556738991094686e-05,
-      "loss": 0.7256,
-      "step": 42020
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.55401148280988e-05,
-      "loss": 0.6621,
-      "step": 42030
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.551283974525073e-05,
-      "loss": 0.6418,
-      "step": 42040
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.548556466240266e-05,
-      "loss": 0.6194,
-      "step": 42050
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.545828957955461e-05,
-      "loss": 0.6159,
-      "step": 42060
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.543101449670654e-05,
-      "loss": 0.6388,
-      "step": 42070
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.540373941385848e-05,
-      "loss": 0.6191,
-      "step": 42080
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.537646433101041e-05,
-      "loss": 0.6686,
-      "step": 42090
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.534918924816235e-05,
-      "loss": 0.6717,
-      "step": 42100
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.532191416531428e-05,
-      "loss": 0.7346,
-      "step": 42110
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.529463908246621e-05,
-      "loss": 0.7753,
-      "step": 42120
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.526736399961815e-05,
-      "loss": 0.6808,
-      "step": 42130
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.524008891677008e-05,
-      "loss": 0.7455,
-      "step": 42140
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.521281383392202e-05,
-      "loss": 0.7417,
-      "step": 42150
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.518553875107395e-05,
-      "loss": 0.6257,
-      "step": 42160
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.51582636682259e-05,
-      "loss": 0.6952,
-      "step": 42170
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.513098858537783e-05,
-      "loss": 0.735,
-      "step": 42180
-    },
-    {
-      "epoch": 1.72,
-      "learning_rate": 8.510371350252977e-05,
-      "loss": 0.6913,
-      "step": 42190
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.50764384196817e-05,
-      "loss": 0.7715,
-      "step": 42200
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.504916333683363e-05,
-      "loss": 0.7061,
-      "step": 42210
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.502188825398557e-05,
-      "loss": 0.7185,
-      "step": 42220
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.499461317113751e-05,
-      "loss": 0.7331,
-      "step": 42230
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.496733808828945e-05,
-      "loss": 0.6538,
-      "step": 42240
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.494006300544138e-05,
-      "loss": 0.6658,
-      "step": 42250
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.491278792259332e-05,
-      "loss": 0.6898,
-      "step": 42260
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.488551283974525e-05,
-      "loss": 0.7441,
-      "step": 42270
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.485823775689718e-05,
-      "loss": 0.7142,
-      "step": 42280
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.483096267404913e-05,
-      "loss": 0.6448,
-      "step": 42290
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.480368759120107e-05,
-      "loss": 0.7268,
-      "step": 42300
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.4776412508353e-05,
-      "loss": 0.7129,
-      "step": 42310
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.474913742550493e-05,
-      "loss": 0.662,
-      "step": 42320
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.472186234265687e-05,
-      "loss": 0.6289,
-      "step": 42330
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.469458725980881e-05,
-      "loss": 0.6488,
-      "step": 42340
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.466731217696075e-05,
-      "loss": 0.6266,
-      "step": 42350
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.464003709411268e-05,
-      "loss": 0.7116,
-      "step": 42360
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.461276201126462e-05,
-      "loss": 0.6299,
-      "step": 42370
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.458548692841655e-05,
-      "loss": 0.6871,
-      "step": 42380
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.455821184556848e-05,
-      "loss": 0.7018,
-      "step": 42390
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.453093676272042e-05,
-      "loss": 0.6854,
-      "step": 42400
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.450366167987235e-05,
-      "loss": 0.6853,
-      "step": 42410
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.447638659702429e-05,
-      "loss": 0.6943,
-      "step": 42420
-    },
-    {
-      "epoch": 1.73,
-      "learning_rate": 8.444911151417622e-05,
-      "loss": 0.6638,
-      "step": 42430
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.442183643132815e-05,
-      "loss": 0.6752,
-      "step": 42440
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.439456134848009e-05,
-      "loss": 0.7521,
-      "step": 42450
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.436728626563204e-05,
-      "loss": 0.7924,
-      "step": 42460
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.434001118278397e-05,
-      "loss": 0.6405,
-      "step": 42470
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.43127360999359e-05,
-      "loss": 0.7427,
-      "step": 42480
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.428546101708784e-05,
-      "loss": 0.7352,
-      "step": 42490
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.425818593423977e-05,
-      "loss": 0.7146,
-      "step": 42500
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.423091085139172e-05,
-      "loss": 0.6461,
-      "step": 42510
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.420363576854365e-05,
-      "loss": 0.6672,
-      "step": 42520
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.417636068569559e-05,
-      "loss": 0.6622,
-      "step": 42530
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.414908560284752e-05,
-      "loss": 0.7256,
-      "step": 42540
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.412181051999945e-05,
-      "loss": 0.7682,
-      "step": 42550
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.409453543715139e-05,
-      "loss": 0.7162,
-      "step": 42560
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.406726035430334e-05,
-      "loss": 0.6639,
-      "step": 42570
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.403998527145527e-05,
-      "loss": 0.8275,
-      "step": 42580
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.40127101886072e-05,
-      "loss": 0.6941,
-      "step": 42590
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.398543510575914e-05,
-      "loss": 0.7077,
-      "step": 42600
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.395816002291107e-05,
-      "loss": 0.7204,
-      "step": 42610
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.393088494006302e-05,
-      "loss": 0.7555,
-      "step": 42620
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.390360985721495e-05,
-      "loss": 0.6613,
-      "step": 42630
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.387633477436689e-05,
-      "loss": 0.5489,
-      "step": 42640
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.384905969151882e-05,
-      "loss": 0.7062,
-      "step": 42650
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.382178460867075e-05,
-      "loss": 0.7789,
-      "step": 42660
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.379450952582269e-05,
-      "loss": 0.7053,
-      "step": 42670
-    },
-    {
-      "epoch": 1.74,
-      "learning_rate": 8.376723444297462e-05,
-      "loss": 0.7319,
-      "step": 42680
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.373995936012656e-05,
-      "loss": 0.6981,
-      "step": 42690
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.371268427727849e-05,
-      "loss": 0.7265,
-      "step": 42700
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.368540919443042e-05,
-      "loss": 0.7914,
-      "step": 42710
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.365813411158236e-05,
-      "loss": 0.6082,
-      "step": 42720
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.363085902873429e-05,
-      "loss": 0.7178,
-      "step": 42730
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.360358394588624e-05,
-      "loss": 0.7538,
-      "step": 42740
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.357630886303817e-05,
-      "loss": 0.6662,
-      "step": 42750
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.354903378019011e-05,
-      "loss": 0.6394,
-      "step": 42760
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.352175869734204e-05,
-      "loss": 0.6989,
-      "step": 42770
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.349448361449398e-05,
-      "loss": 0.8149,
-      "step": 42780
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.346720853164592e-05,
-      "loss": 0.6679,
-      "step": 42790
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.343993344879786e-05,
-      "loss": 0.7022,
-      "step": 42800
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.341265836594979e-05,
-      "loss": 0.7485,
-      "step": 42810
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.338538328310172e-05,
-      "loss": 0.5774,
-      "step": 42820
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.335810820025366e-05,
-      "loss": 0.7608,
-      "step": 42830
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.333083311740559e-05,
-      "loss": 0.6618,
-      "step": 42840
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.330355803455754e-05,
-      "loss": 0.728,
-      "step": 42850
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.327628295170947e-05,
-      "loss": 0.6937,
-      "step": 42860
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.324900786886141e-05,
-      "loss": 0.6578,
-      "step": 42870
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.322173278601334e-05,
-      "loss": 0.6325,
-      "step": 42880
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.319445770316528e-05,
-      "loss": 0.7125,
-      "step": 42890
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.316718262031721e-05,
-      "loss": 0.7677,
-      "step": 42900
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.313990753746916e-05,
-      "loss": 0.7777,
-      "step": 42910
-    },
-    {
-      "epoch": 1.75,
-      "learning_rate": 8.311263245462109e-05,
-      "loss": 0.7086,
-      "step": 42920
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.308535737177302e-05,
-      "loss": 0.6614,
-      "step": 42930
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.305808228892496e-05,
-      "loss": 0.6292,
-      "step": 42940
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.303080720607689e-05,
-      "loss": 0.8701,
-      "step": 42950
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.300353212322883e-05,
-      "loss": 0.6316,
-      "step": 42960
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.297625704038076e-05,
-      "loss": 0.6798,
-      "step": 42970
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.29489819575327e-05,
-      "loss": 0.7128,
-      "step": 42980
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.292170687468463e-05,
-      "loss": 0.7833,
-      "step": 42990
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.289443179183656e-05,
-      "loss": 0.7065,
-      "step": 43000
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.28671567089885e-05,
-      "loss": 0.6709,
-      "step": 43010
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.283988162614044e-05,
-      "loss": 0.6725,
-      "step": 43020
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.281260654329238e-05,
-      "loss": 0.6535,
-      "step": 43030
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.278533146044431e-05,
-      "loss": 0.694,
-      "step": 43040
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.275805637759625e-05,
-      "loss": 0.6351,
-      "step": 43050
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.273078129474818e-05,
-      "loss": 0.701,
-      "step": 43060
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.270350621190013e-05,
-      "loss": 0.7158,
-      "step": 43070
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.267623112905206e-05,
-      "loss": 0.6209,
-      "step": 43080
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.2648956046204e-05,
-      "loss": 0.7537,
-      "step": 43090
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.262168096335593e-05,
-      "loss": 0.6609,
-      "step": 43100
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.259440588050786e-05,
-      "loss": 0.6455,
-      "step": 43110
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.25671307976598e-05,
-      "loss": 0.6512,
-      "step": 43120
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.253985571481174e-05,
-      "loss": 0.7387,
-      "step": 43130
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.251258063196368e-05,
-      "loss": 0.6802,
-      "step": 43140
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.248530554911561e-05,
-      "loss": 0.6753,
-      "step": 43150
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.245803046626755e-05,
-      "loss": 0.7375,
-      "step": 43160
-    },
-    {
-      "epoch": 1.76,
-      "learning_rate": 8.243075538341948e-05,
-      "loss": 0.7115,
-      "step": 43170
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.240348030057141e-05,
-      "loss": 0.6331,
-      "step": 43180
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.237620521772336e-05,
-      "loss": 0.6465,
-      "step": 43190
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.23489301348753e-05,
-      "loss": 0.7319,
-      "step": 43200
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.232165505202723e-05,
-      "loss": 0.7484,
-      "step": 43210
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.229437996917916e-05,
-      "loss": 0.6612,
-      "step": 43220
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.22671048863311e-05,
-      "loss": 0.6146,
-      "step": 43230
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.223982980348304e-05,
-      "loss": 0.7663,
-      "step": 43240
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.221255472063498e-05,
-      "loss": 0.7018,
-      "step": 43250
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.21852796377869e-05,
-      "loss": 0.7782,
-      "step": 43260
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.215800455493883e-05,
-      "loss": 0.6388,
-      "step": 43270
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.213072947209077e-05,
-      "loss": 0.7146,
-      "step": 43280
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.21034543892427e-05,
-      "loss": 0.6146,
-      "step": 43290
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.207617930639465e-05,
-      "loss": 0.7889,
-      "step": 43300
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.204890422354658e-05,
-      "loss": 0.7755,
-      "step": 43310
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.202162914069852e-05,
-      "loss": 0.6576,
-      "step": 43320
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.199435405785045e-05,
-      "loss": 0.752,
-      "step": 43330
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.196707897500238e-05,
-      "loss": 0.6905,
-      "step": 43340
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.193980389215432e-05,
-      "loss": 0.7338,
-      "step": 43350
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.191252880930626e-05,
-      "loss": 0.7102,
-      "step": 43360
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.18852537264582e-05,
-      "loss": 0.6202,
-      "step": 43370
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.185797864361013e-05,
-      "loss": 0.7102,
-      "step": 43380
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.183070356076207e-05,
-      "loss": 0.7588,
-      "step": 43390
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.1803428477914e-05,
-      "loss": 0.6674,
-      "step": 43400
-    },
-    {
-      "epoch": 1.77,
-      "learning_rate": 8.177615339506595e-05,
-      "loss": 0.5854,
-      "step": 43410
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.174887831221788e-05,
-      "loss": 0.689,
-      "step": 43420
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.172160322936982e-05,
-      "loss": 0.6798,
-      "step": 43430
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.169705565480656e-05,
-      "loss": 0.7333,
-      "step": 43440
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.166978057195849e-05,
-      "loss": 0.6664,
-      "step": 43450
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.164250548911044e-05,
-      "loss": 0.6204,
-      "step": 43460
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.161523040626237e-05,
-      "loss": 0.6823,
-      "step": 43470
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.15879553234143e-05,
-      "loss": 0.6919,
-      "step": 43480
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.156068024056624e-05,
-      "loss": 0.6194,
-      "step": 43490
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.153340515771817e-05,
-      "loss": 0.7573,
-      "step": 43500
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.150613007487011e-05,
-      "loss": 0.7249,
-      "step": 43510
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.147885499202204e-05,
-      "loss": 0.7212,
-      "step": 43520
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.145157990917397e-05,
-      "loss": 0.7175,
-      "step": 43530
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.142430482632591e-05,
-      "loss": 0.6879,
-      "step": 43540
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.139702974347784e-05,
-      "loss": 0.7594,
-      "step": 43550
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.136975466062978e-05,
-      "loss": 0.7212,
-      "step": 43560
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.134247957778171e-05,
-      "loss": 0.6353,
-      "step": 43570
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.131520449493366e-05,
-      "loss": 0.6056,
-      "step": 43580
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.128792941208559e-05,
-      "loss": 0.7406,
-      "step": 43590
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.126065432923753e-05,
-      "loss": 0.6926,
-      "step": 43600
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.123337924638946e-05,
-      "loss": 0.6869,
-      "step": 43610
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.12061041635414e-05,
-      "loss": 0.7349,
-      "step": 43620
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.117882908069334e-05,
-      "loss": 0.7185,
-      "step": 43630
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.115155399784528e-05,
-      "loss": 0.7152,
-      "step": 43640
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.112427891499721e-05,
-      "loss": 0.8057,
-      "step": 43650
-    },
-    {
-      "epoch": 1.78,
-      "learning_rate": 8.109700383214914e-05,
-      "loss": 0.6609,
-      "step": 43660
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.106972874930108e-05,
-      "loss": 0.6393,
-      "step": 43670
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.104245366645301e-05,
-      "loss": 0.6713,
-      "step": 43680
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.101517858360496e-05,
-      "loss": 0.7168,
-      "step": 43690
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.098790350075689e-05,
-      "loss": 0.6223,
-      "step": 43700
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.096062841790883e-05,
-      "loss": 0.7651,
-      "step": 43710
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.093335333506076e-05,
-      "loss": 0.7193,
-      "step": 43720
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.09060782522127e-05,
-      "loss": 0.8043,
-      "step": 43730
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.087880316936463e-05,
-      "loss": 0.6712,
-      "step": 43740
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.085152808651658e-05,
-      "loss": 0.8616,
-      "step": 43750
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.082425300366851e-05,
-      "loss": 0.6544,
-      "step": 43760
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.079697792082044e-05,
-      "loss": 0.723,
-      "step": 43770
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.076970283797238e-05,
-      "loss": 0.7661,
-      "step": 43780
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.074242775512431e-05,
-      "loss": 0.6517,
-      "step": 43790
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.071515267227624e-05,
-      "loss": 0.6651,
-      "step": 43800
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.068787758942818e-05,
-      "loss": 0.7241,
-      "step": 43810
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.066060250658011e-05,
-      "loss": 0.7314,
-      "step": 43820
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.063332742373205e-05,
-      "loss": 0.7155,
-      "step": 43830
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.060605234088398e-05,
-      "loss": 0.6676,
-      "step": 43840
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.057877725803591e-05,
-      "loss": 0.7596,
-      "step": 43850
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.055150217518786e-05,
-      "loss": 0.6814,
-      "step": 43860
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.05242270923398e-05,
-      "loss": 0.6716,
-      "step": 43870
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.049695200949173e-05,
-      "loss": 0.655,
-      "step": 43880
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.046967692664366e-05,
-      "loss": 0.6952,
-      "step": 43890
-    },
-    {
-      "epoch": 1.79,
-      "learning_rate": 8.04424018437956e-05,
-      "loss": 0.7382,
-      "step": 43900
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.041512676094755e-05,
-      "loss": 0.7413,
-      "step": 43910
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.038785167809948e-05,
-      "loss": 0.6708,
-      "step": 43920
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.036057659525141e-05,
-      "loss": 0.7135,
-      "step": 43930
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.033330151240335e-05,
-      "loss": 0.5909,
-      "step": 43940
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.030602642955528e-05,
-      "loss": 0.7028,
-      "step": 43950
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.027875134670721e-05,
-      "loss": 0.659,
-      "step": 43960
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.025147626385916e-05,
-      "loss": 0.6525,
-      "step": 43970
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.02242011810111e-05,
-      "loss": 0.7584,
-      "step": 43980
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.019692609816303e-05,
-      "loss": 0.7287,
-      "step": 43990
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.016965101531496e-05,
-      "loss": 0.6532,
-      "step": 44000
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.01423759324669e-05,
-      "loss": 0.674,
-      "step": 44010
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.011510084961883e-05,
-      "loss": 0.6743,
-      "step": 44020
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.008782576677078e-05,
-      "loss": 0.7263,
-      "step": 44030
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.006055068392271e-05,
-      "loss": 0.6288,
-      "step": 44040
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.003327560107465e-05,
-      "loss": 0.6416,
-      "step": 44050
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 8.000600051822658e-05,
-      "loss": 0.6678,
-      "step": 44060
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.997872543537852e-05,
-      "loss": 0.7681,
-      "step": 44070
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.995145035253045e-05,
-      "loss": 0.653,
-      "step": 44080
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.992417526968238e-05,
-      "loss": 0.6418,
-      "step": 44090
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.989690018683432e-05,
-      "loss": 0.6828,
-      "step": 44100
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.986962510398625e-05,
-      "loss": 0.733,
-      "step": 44110
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.984235002113818e-05,
-      "loss": 0.6844,
-      "step": 44120
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.981507493829012e-05,
-      "loss": 0.7234,
-      "step": 44130
-    },
-    {
-      "epoch": 1.8,
-      "learning_rate": 7.978779985544207e-05,
-      "loss": 0.7127,
-      "step": 44140
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.9760524772594e-05,
-      "loss": 0.6997,
-      "step": 44150
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.973324968974593e-05,
-      "loss": 0.6959,
-      "step": 44160
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.970597460689787e-05,
-      "loss": 0.7135,
-      "step": 44170
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.96786995240498e-05,
-      "loss": 0.7201,
-      "step": 44180
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.965142444120174e-05,
-      "loss": 0.6541,
-      "step": 44190
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.962414935835368e-05,
-      "loss": 0.6438,
-      "step": 44200
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.959687427550562e-05,
-      "loss": 0.6603,
-      "step": 44210
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.956959919265755e-05,
-      "loss": 0.6729,
-      "step": 44220
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.954232410980948e-05,
-      "loss": 0.6774,
-      "step": 44230
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.951504902696142e-05,
-      "loss": 0.6502,
-      "step": 44240
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.948777394411337e-05,
-      "loss": 0.7089,
-      "step": 44250
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.94604988612653e-05,
-      "loss": 0.6277,
-      "step": 44260
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.943322377841723e-05,
-      "loss": 0.6592,
-      "step": 44270
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.940594869556917e-05,
-      "loss": 0.6842,
-      "step": 44280
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.93786736127211e-05,
-      "loss": 0.7809,
-      "step": 44290
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.935139852987304e-05,
-      "loss": 0.6918,
-      "step": 44300
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.932412344702498e-05,
-      "loss": 0.6554,
-      "step": 44310
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.929684836417692e-05,
-      "loss": 0.6896,
-      "step": 44320
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.926957328132885e-05,
-      "loss": 0.6655,
-      "step": 44330
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.924229819848079e-05,
-      "loss": 0.6797,
-      "step": 44340
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.921502311563272e-05,
-      "loss": 0.659,
-      "step": 44350
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.918774803278465e-05,
-      "loss": 0.6814,
-      "step": 44360
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.916047294993659e-05,
-      "loss": 0.6889,
-      "step": 44370
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.913319786708852e-05,
-      "loss": 0.7576,
-      "step": 44380
-    },
-    {
-      "epoch": 1.81,
-      "learning_rate": 7.910592278424045e-05,
-      "loss": 0.658,
-      "step": 44390
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.907864770139239e-05,
-      "loss": 0.6871,
-      "step": 44400
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.905137261854432e-05,
-      "loss": 0.6549,
-      "step": 44410
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.902409753569627e-05,
-      "loss": 0.68,
-      "step": 44420
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.89968224528482e-05,
-      "loss": 0.6766,
-      "step": 44430
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.896954737000014e-05,
-      "loss": 0.842,
-      "step": 44440
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.894227228715207e-05,
-      "loss": 0.6399,
-      "step": 44450
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.8914997204304e-05,
-      "loss": 0.6731,
-      "step": 44460
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.888772212145594e-05,
-      "loss": 0.761,
-      "step": 44470
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.886044703860789e-05,
-      "loss": 0.6387,
-      "step": 44480
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.883317195575982e-05,
-      "loss": 0.7326,
-      "step": 44490
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.880589687291175e-05,
-      "loss": 0.7958,
-      "step": 44500
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.877862179006369e-05,
-      "loss": 0.6179,
-      "step": 44510
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.875134670721562e-05,
-      "loss": 0.7361,
-      "step": 44520
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.872407162436757e-05,
-      "loss": 0.6808,
-      "step": 44530
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.86967965415195e-05,
-      "loss": 0.6859,
-      "step": 44540
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.866952145867144e-05,
-      "loss": 0.7336,
-      "step": 44550
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.864224637582337e-05,
-      "loss": 0.5979,
-      "step": 44560
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.86149712929753e-05,
-      "loss": 0.6579,
-      "step": 44570
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.858769621012724e-05,
-      "loss": 0.6423,
-      "step": 44580
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.856042112727919e-05,
-      "loss": 0.6385,
-      "step": 44590
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.853314604443112e-05,
-      "loss": 0.6325,
-      "step": 44600
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.850587096158306e-05,
-      "loss": 0.6837,
-      "step": 44610
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.847859587873499e-05,
-      "loss": 0.6789,
-      "step": 44620
-    },
-    {
-      "epoch": 1.82,
-      "learning_rate": 7.845132079588692e-05,
-      "loss": 0.8135,
-      "step": 44630
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.842404571303886e-05,
-      "loss": 0.6499,
-      "step": 44640
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.839677063019079e-05,
-      "loss": 0.6716,
-      "step": 44650
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.836949554734272e-05,
-      "loss": 0.6492,
-      "step": 44660
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.834222046449466e-05,
-      "loss": 0.6927,
-      "step": 44670
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.831494538164659e-05,
-      "loss": 0.6897,
-      "step": 44680
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.828767029879853e-05,
-      "loss": 0.6581,
-      "step": 44690
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.826039521595047e-05,
-      "loss": 0.7057,
-      "step": 44700
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.823312013310241e-05,
-      "loss": 0.7216,
-      "step": 44710
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.820584505025434e-05,
-      "loss": 0.6822,
-      "step": 44720
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.817856996740628e-05,
-      "loss": 0.7499,
-      "step": 44730
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.815129488455821e-05,
-      "loss": 0.6256,
-      "step": 44740
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.812401980171014e-05,
-      "loss": 0.7767,
-      "step": 44750
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.809674471886209e-05,
-      "loss": 0.6611,
-      "step": 44760
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.806946963601402e-05,
-      "loss": 0.7609,
-      "step": 44770
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.804219455316596e-05,
-      "loss": 0.5938,
-      "step": 44780
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.801491947031789e-05,
-      "loss": 0.6444,
-      "step": 44790
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.798764438746983e-05,
-      "loss": 0.7051,
-      "step": 44800
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.796036930462177e-05,
-      "loss": 0.7187,
-      "step": 44810
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.793309422177371e-05,
-      "loss": 0.7205,
-      "step": 44820
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.790581913892564e-05,
-      "loss": 0.6971,
-      "step": 44830
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.787854405607758e-05,
-      "loss": 0.6185,
-      "step": 44840
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.785126897322951e-05,
-      "loss": 0.7053,
-      "step": 44850
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.782399389038144e-05,
-      "loss": 0.6778,
-      "step": 44860
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.779671880753339e-05,
-      "loss": 0.7092,
-      "step": 44870
-    },
-    {
-      "epoch": 1.83,
-      "learning_rate": 7.776944372468533e-05,
-      "loss": 0.6865,
-      "step": 44880
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.774216864183726e-05,
-      "loss": 0.5754,
-      "step": 44890
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.771489355898919e-05,
-      "loss": 0.6069,
-      "step": 44900
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.768761847614113e-05,
-      "loss": 0.7503,
-      "step": 44910
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.766034339329306e-05,
-      "loss": 0.7544,
-      "step": 44920
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.7633068310445e-05,
-      "loss": 0.6882,
-      "step": 44930
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.760579322759693e-05,
-      "loss": 0.6939,
-      "step": 44940
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.757851814474886e-05,
-      "loss": 0.6341,
-      "step": 44950
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.75512430619008e-05,
-      "loss": 0.5857,
-      "step": 44960
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.752396797905273e-05,
-      "loss": 0.673,
-      "step": 44970
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.749669289620468e-05,
-      "loss": 0.6992,
-      "step": 44980
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.746941781335661e-05,
-      "loss": 0.6788,
-      "step": 44990
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.744214273050855e-05,
-      "loss": 0.7029,
-      "step": 45000
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.741486764766048e-05,
-      "loss": 0.6651,
-      "step": 45010
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.738759256481241e-05,
-      "loss": 0.6636,
-      "step": 45020
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.736031748196435e-05,
-      "loss": 0.6617,
-      "step": 45030
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.73330423991163e-05,
-      "loss": 0.6717,
-      "step": 45040
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.730576731626823e-05,
-      "loss": 0.7364,
-      "step": 45050
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.727849223342016e-05,
-      "loss": 0.7323,
-      "step": 45060
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.72512171505721e-05,
-      "loss": 0.6318,
-      "step": 45070
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.722394206772403e-05,
-      "loss": 0.6074,
-      "step": 45080
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.719666698487596e-05,
-      "loss": 0.6819,
-      "step": 45090
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.716939190202791e-05,
-      "loss": 0.7709,
-      "step": 45100
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.714211681917985e-05,
-      "loss": 0.6454,
-      "step": 45110
-    },
-    {
-      "epoch": 1.84,
-      "learning_rate": 7.711484173633178e-05,
-      "loss": 0.61,
-      "step": 45120
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.708756665348371e-05,
-      "loss": 0.7534,
-      "step": 45130
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.706029157063565e-05,
-      "loss": 0.6936,
-      "step": 45140
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.70330164877876e-05,
-      "loss": 0.6691,
-      "step": 45150
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.700574140493953e-05,
-      "loss": 0.6822,
-      "step": 45160
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.697846632209146e-05,
-      "loss": 0.7587,
-      "step": 45170
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.69511912392434e-05,
-      "loss": 0.7236,
-      "step": 45180
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.692391615639533e-05,
-      "loss": 0.6571,
-      "step": 45190
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.689664107354726e-05,
-      "loss": 0.6289,
-      "step": 45200
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.68693659906992e-05,
-      "loss": 0.6206,
-      "step": 45210
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.684209090785113e-05,
-      "loss": 0.5897,
-      "step": 45220
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.681481582500307e-05,
-      "loss": 0.7447,
-      "step": 45230
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.6787540742155e-05,
-      "loss": 0.6929,
-      "step": 45240
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.676026565930693e-05,
-      "loss": 0.6596,
-      "step": 45250
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.673299057645887e-05,
-      "loss": 0.5958,
-      "step": 45260
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.670571549361082e-05,
-      "loss": 0.7444,
-      "step": 45270
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.667844041076275e-05,
-      "loss": 0.7894,
-      "step": 45280
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.665116532791468e-05,
-      "loss": 0.7366,
-      "step": 45290
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.662389024506662e-05,
-      "loss": 0.6362,
-      "step": 45300
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.659661516221855e-05,
-      "loss": 0.8179,
-      "step": 45310
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.65693400793705e-05,
-      "loss": 0.6768,
-      "step": 45320
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.654206499652243e-05,
-      "loss": 0.7286,
-      "step": 45330
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.651478991367437e-05,
-      "loss": 0.6806,
-      "step": 45340
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.64875148308263e-05,
-      "loss": 0.7529,
-      "step": 45350
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.646023974797823e-05,
-      "loss": 0.7087,
-      "step": 45360
-    },
-    {
-      "epoch": 1.85,
-      "learning_rate": 7.643296466513017e-05,
-      "loss": 0.5728,
-      "step": 45370
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.640568958228212e-05,
-      "loss": 0.7992,
-      "step": 45380
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.637841449943405e-05,
-      "loss": 0.7347,
-      "step": 45390
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.635113941658598e-05,
-      "loss": 0.6485,
-      "step": 45400
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.632386433373792e-05,
-      "loss": 0.6222,
-      "step": 45410
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.629658925088985e-05,
-      "loss": 0.7281,
-      "step": 45420
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.62693141680418e-05,
-      "loss": 0.6216,
-      "step": 45430
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.624203908519373e-05,
-      "loss": 0.7776,
-      "step": 45440
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.621476400234567e-05,
-      "loss": 0.6563,
-      "step": 45450
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.61874889194976e-05,
-      "loss": 0.6161,
-      "step": 45460
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.616021383664953e-05,
-      "loss": 0.6784,
-      "step": 45470
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.613293875380147e-05,
-      "loss": 0.651,
-      "step": 45480
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.61056636709534e-05,
-      "loss": 0.6632,
-      "step": 45490
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.607838858810534e-05,
-      "loss": 0.6334,
-      "step": 45500
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.605111350525727e-05,
-      "loss": 0.6777,
-      "step": 45510
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.60238384224092e-05,
-      "loss": 0.7003,
-      "step": 45520
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.599656333956114e-05,
-      "loss": 0.6579,
-      "step": 45530
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.596928825671307e-05,
-      "loss": 0.5699,
-      "step": 45540
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.594201317386502e-05,
-      "loss": 0.7497,
-      "step": 45550
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.591473809101695e-05,
-      "loss": 0.6389,
-      "step": 45560
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.588746300816889e-05,
-      "loss": 0.6796,
-      "step": 45570
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.586018792532082e-05,
-      "loss": 0.5864,
-      "step": 45580
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.583291284247276e-05,
-      "loss": 0.677,
-      "step": 45590
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.58056377596247e-05,
-      "loss": 0.6375,
-      "step": 45600
-    },
-    {
-      "epoch": 1.86,
-      "learning_rate": 7.577836267677664e-05,
-      "loss": 0.7446,
-      "step": 45610
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.575108759392857e-05,
-      "loss": 0.5791,
-      "step": 45620
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.57238125110805e-05,
-      "loss": 0.7169,
-      "step": 45630
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.569653742823244e-05,
-      "loss": 0.6884,
-      "step": 45640
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.566926234538437e-05,
-      "loss": 0.7224,
-      "step": 45650
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.564198726253632e-05,
-      "loss": 0.6834,
-      "step": 45660
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.561471217968825e-05,
-      "loss": 0.6342,
-      "step": 45670
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.558743709684019e-05,
-      "loss": 0.6504,
-      "step": 45680
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.556016201399212e-05,
-      "loss": 0.6899,
-      "step": 45690
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.553288693114406e-05,
-      "loss": 0.6883,
-      "step": 45700
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.550561184829599e-05,
-      "loss": 0.6305,
-      "step": 45710
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.547833676544794e-05,
-      "loss": 0.6861,
-      "step": 45720
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.545106168259987e-05,
-      "loss": 0.8033,
-      "step": 45730
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.54237865997518e-05,
-      "loss": 0.6531,
-      "step": 45740
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.539651151690374e-05,
-      "loss": 0.7433,
-      "step": 45750
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.536923643405567e-05,
-      "loss": 0.715,
-      "step": 45760
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.53419613512076e-05,
-      "loss": 0.6348,
-      "step": 45770
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.531468626835954e-05,
-      "loss": 0.6503,
-      "step": 45780
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.528741118551147e-05,
-      "loss": 0.6792,
-      "step": 45790
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.526013610266341e-05,
-      "loss": 0.6993,
-      "step": 45800
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.523286101981534e-05,
-      "loss": 0.6877,
-      "step": 45810
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.520558593696728e-05,
-      "loss": 0.6954,
-      "step": 45820
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.517831085411922e-05,
-      "loss": 0.7853,
-      "step": 45830
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.515103577127116e-05,
-      "loss": 0.7478,
-      "step": 45840
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.512376068842309e-05,
-      "loss": 0.8127,
-      "step": 45850
-    },
-    {
-      "epoch": 1.87,
-      "learning_rate": 7.509648560557503e-05,
-      "loss": 0.7713,
-      "step": 45860
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.506921052272696e-05,
-      "loss": 0.7165,
-      "step": 45870
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.504193543987891e-05,
-      "loss": 0.6904,
-      "step": 45880
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.501466035703084e-05,
-      "loss": 0.7267,
-      "step": 45890
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.498738527418277e-05,
-      "loss": 0.7246,
-      "step": 45900
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.496011019133471e-05,
-      "loss": 0.6187,
-      "step": 45910
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.493283510848664e-05,
-      "loss": 0.8256,
-      "step": 45920
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.490556002563858e-05,
-      "loss": 0.7578,
-      "step": 45930
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.487828494279052e-05,
-      "loss": 0.6764,
-      "step": 45940
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.485100985994246e-05,
-      "loss": 0.6363,
-      "step": 45950
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.482373477709439e-05,
-      "loss": 0.6498,
-      "step": 45960
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.479645969424633e-05,
-      "loss": 0.7512,
-      "step": 45970
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.476918461139826e-05,
-      "loss": 0.649,
-      "step": 45980
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.47419095285502e-05,
-      "loss": 0.7356,
-      "step": 45990
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.471463444570214e-05,
-      "loss": 0.6843,
-      "step": 46000
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.468735936285407e-05,
-      "loss": 0.6715,
-      "step": 46010
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.466008428000601e-05,
-      "loss": 0.7449,
-      "step": 46020
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.463280919715794e-05,
-      "loss": 0.7221,
-      "step": 46030
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.460553411430988e-05,
-      "loss": 0.6764,
-      "step": 46040
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.457825903146181e-05,
-      "loss": 0.5416,
-      "step": 46050
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.455098394861374e-05,
-      "loss": 0.7652,
-      "step": 46060
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.452370886576568e-05,
-      "loss": 0.6816,
-      "step": 46070
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.449643378291761e-05,
-      "loss": 0.7805,
-      "step": 46080
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.446915870006955e-05,
-      "loss": 0.6756,
-      "step": 46090
-    },
-    {
-      "epoch": 1.88,
-      "learning_rate": 7.444188361722148e-05,
-      "loss": 0.7897,
-      "step": 46100
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.441460853437343e-05,
-      "loss": 0.7093,
-      "step": 46110
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.438733345152536e-05,
-      "loss": 0.6968,
-      "step": 46120
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.43600583686773e-05,
-      "loss": 0.7262,
-      "step": 46130
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.433278328582923e-05,
-      "loss": 0.641,
-      "step": 46140
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.430550820298116e-05,
-      "loss": 0.6159,
-      "step": 46150
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.42782331201331e-05,
-      "loss": 0.7135,
-      "step": 46160
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.425095803728504e-05,
-      "loss": 0.6682,
-      "step": 46170
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.422368295443698e-05,
-      "loss": 0.7237,
-      "step": 46180
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.419640787158891e-05,
-      "loss": 0.6619,
-      "step": 46190
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.416913278874085e-05,
-      "loss": 0.7796,
-      "step": 46200
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.414185770589278e-05,
-      "loss": 0.7094,
-      "step": 46210
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.411458262304473e-05,
-      "loss": 0.7082,
-      "step": 46220
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.408730754019666e-05,
-      "loss": 0.5704,
-      "step": 46230
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.40600324573486e-05,
-      "loss": 0.7155,
-      "step": 46240
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.403275737450053e-05,
-      "loss": 0.6391,
-      "step": 46250
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.400548229165246e-05,
-      "loss": 0.8321,
-      "step": 46260
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.39782072088044e-05,
-      "loss": 0.6934,
-      "step": 46270
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.395093212595635e-05,
-      "loss": 0.6939,
-      "step": 46280
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.392365704310828e-05,
-      "loss": 0.6711,
-      "step": 46290
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.389638196026021e-05,
-      "loss": 0.6791,
-      "step": 46300
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.386910687741215e-05,
-      "loss": 0.677,
-      "step": 46310
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.384183179456408e-05,
-      "loss": 0.6978,
-      "step": 46320
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.381455671171601e-05,
-      "loss": 0.7263,
-      "step": 46330
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.378728162886795e-05,
-      "loss": 0.7339,
-      "step": 46340
-    },
-    {
-      "epoch": 1.89,
-      "learning_rate": 7.376000654601988e-05,
-      "loss": 0.6543,
-      "step": 46350
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.373273146317182e-05,
-      "loss": 0.6861,
-      "step": 46360
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.370545638032375e-05,
-      "loss": 0.706,
-      "step": 46370
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.367818129747568e-05,
-      "loss": 0.6672,
-      "step": 46380
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.365090621462763e-05,
-      "loss": 0.6855,
-      "step": 46390
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.362363113177957e-05,
-      "loss": 0.5889,
-      "step": 46400
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.35963560489315e-05,
-      "loss": 0.7114,
-      "step": 46410
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.356908096608343e-05,
-      "loss": 0.6851,
-      "step": 46420
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.354180588323537e-05,
-      "loss": 0.711,
-      "step": 46430
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.35145308003873e-05,
-      "loss": 0.6123,
-      "step": 46440
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.348725571753925e-05,
-      "loss": 0.7919,
-      "step": 46450
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.345998063469118e-05,
-      "loss": 0.6584,
-      "step": 46460
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.343270555184312e-05,
-      "loss": 0.7178,
-      "step": 46470
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.340543046899505e-05,
-      "loss": 0.7371,
-      "step": 46480
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.337815538614698e-05,
-      "loss": 0.6523,
-      "step": 46490
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.335088030329893e-05,
-      "loss": 0.6545,
-      "step": 46500
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.332360522045087e-05,
-      "loss": 0.6956,
-      "step": 46510
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.32963301376028e-05,
-      "loss": 0.6468,
-      "step": 46520
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.326905505475473e-05,
-      "loss": 0.7144,
-      "step": 46530
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.324177997190667e-05,
-      "loss": 0.6571,
-      "step": 46540
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.32145048890586e-05,
-      "loss": 0.8274,
-      "step": 46550
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.318722980621055e-05,
-      "loss": 0.6926,
-      "step": 46560
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.315995472336248e-05,
-      "loss": 0.6735,
-      "step": 46570
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.313267964051442e-05,
-      "loss": 0.6585,
-      "step": 46580
-    },
-    {
-      "epoch": 1.9,
-      "learning_rate": 7.310540455766635e-05,
-      "loss": 0.6628,
-      "step": 46590
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.307812947481828e-05,
-      "loss": 0.707,
-      "step": 46600
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.305085439197022e-05,
-      "loss": 0.6976,
-      "step": 46610
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.302357930912215e-05,
-      "loss": 0.7045,
-      "step": 46620
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.299630422627409e-05,
-      "loss": 0.557,
-      "step": 46630
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.296902914342602e-05,
-      "loss": 0.6753,
-      "step": 46640
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.294175406057795e-05,
-      "loss": 0.7254,
-      "step": 46650
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.291447897772989e-05,
-      "loss": 0.6909,
-      "step": 46660
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.288720389488184e-05,
-      "loss": 0.604,
-      "step": 46670
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.285992881203377e-05,
-      "loss": 0.6739,
-      "step": 46680
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.28326537291857e-05,
-      "loss": 0.7207,
-      "step": 46690
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.280537864633764e-05,
-      "loss": 0.7261,
-      "step": 46700
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.277810356348957e-05,
-      "loss": 0.7133,
-      "step": 46710
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.27508284806415e-05,
-      "loss": 0.6106,
-      "step": 46720
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.272355339779345e-05,
-      "loss": 0.6646,
-      "step": 46730
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.269627831494539e-05,
-      "loss": 0.6403,
-      "step": 46740
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.266900323209732e-05,
-      "loss": 0.884,
-      "step": 46750
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.264172814924925e-05,
-      "loss": 0.6527,
-      "step": 46760
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.261445306640119e-05,
-      "loss": 0.6899,
-      "step": 46770
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.258717798355314e-05,
-      "loss": 0.674,
-      "step": 46780
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.255990290070507e-05,
-      "loss": 0.8404,
-      "step": 46790
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.2532627817857e-05,
-      "loss": 0.7703,
-      "step": 46800
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.250535273500894e-05,
-      "loss": 0.6255,
-      "step": 46810
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.247807765216087e-05,
-      "loss": 0.6839,
-      "step": 46820
-    },
-    {
-      "epoch": 1.91,
-      "learning_rate": 7.24508025693128e-05,
-      "loss": 0.7091,
-      "step": 46830
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.242352748646475e-05,
-      "loss": 0.6288,
-      "step": 46840
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.239625240361669e-05,
-      "loss": 0.6786,
-      "step": 46850
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.236897732076862e-05,
-      "loss": 0.7698,
-      "step": 46860
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.234170223792055e-05,
-      "loss": 0.6762,
-      "step": 46870
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.231442715507249e-05,
-      "loss": 0.6551,
-      "step": 46880
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.228715207222442e-05,
-      "loss": 0.7419,
-      "step": 46890
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.225987698937636e-05,
-      "loss": 0.7515,
-      "step": 46900
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.223260190652829e-05,
-      "loss": 0.7362,
-      "step": 46910
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.220532682368022e-05,
-      "loss": 0.6991,
-      "step": 46920
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.217805174083216e-05,
-      "loss": 0.7527,
-      "step": 46930
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.215077665798409e-05,
-      "loss": 0.6252,
-      "step": 46940
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.212350157513604e-05,
-      "loss": 0.7505,
-      "step": 46950
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.209622649228797e-05,
-      "loss": 0.6194,
-      "step": 46960
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.206895140943991e-05,
-      "loss": 0.6362,
-      "step": 46970
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.204167632659184e-05,
-      "loss": 0.7137,
-      "step": 46980
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.201440124374378e-05,
-      "loss": 0.6057,
-      "step": 46990
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.198712616089571e-05,
-      "loss": 0.6346,
-      "step": 47000
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.195985107804766e-05,
-      "loss": 0.7332,
-      "step": 47010
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.193257599519959e-05,
-      "loss": 0.7277,
-      "step": 47020
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.190530091235152e-05,
-      "loss": 0.6905,
-      "step": 47030
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.187802582950346e-05,
-      "loss": 0.6809,
-      "step": 47040
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.185075074665539e-05,
-      "loss": 0.6852,
-      "step": 47050
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.182347566380733e-05,
-      "loss": 0.6668,
-      "step": 47060
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.179620058095927e-05,
-      "loss": 0.6166,
-      "step": 47070
-    },
-    {
-      "epoch": 1.92,
-      "learning_rate": 7.176892549811121e-05,
-      "loss": 0.6788,
-      "step": 47080
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.174165041526314e-05,
-      "loss": 0.6771,
-      "step": 47090
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.171437533241508e-05,
-      "loss": 0.5856,
-      "step": 47100
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.168710024956701e-05,
-      "loss": 0.6449,
-      "step": 47110
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.165982516671896e-05,
-      "loss": 0.6881,
-      "step": 47120
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.163255008387089e-05,
-      "loss": 0.6329,
-      "step": 47130
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.160527500102282e-05,
-      "loss": 0.7333,
-      "step": 47140
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.157799991817476e-05,
-      "loss": 0.6386,
-      "step": 47150
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.155072483532669e-05,
-      "loss": 0.6058,
-      "step": 47160
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.152344975247863e-05,
-      "loss": 0.6406,
-      "step": 47170
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.149617466963056e-05,
-      "loss": 0.7035,
-      "step": 47180
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.14688995867825e-05,
-      "loss": 0.761,
-      "step": 47190
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.144162450393443e-05,
-      "loss": 0.7088,
-      "step": 47200
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.141434942108636e-05,
-      "loss": 0.6664,
-      "step": 47210
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.13870743382383e-05,
-      "loss": 0.6271,
-      "step": 47220
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.135979925539023e-05,
-      "loss": 0.6852,
-      "step": 47230
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.133252417254218e-05,
-      "loss": 0.6958,
-      "step": 47240
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.130524908969411e-05,
-      "loss": 0.6971,
-      "step": 47250
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.127797400684605e-05,
-      "loss": 0.6556,
-      "step": 47260
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.125069892399798e-05,
-      "loss": 0.6183,
-      "step": 47270
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.122342384114991e-05,
-      "loss": 0.635,
-      "step": 47280
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.119614875830186e-05,
-      "loss": 0.8215,
-      "step": 47290
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.11688736754538e-05,
-      "loss": 0.733,
-      "step": 47300
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.114159859260573e-05,
-      "loss": 0.7313,
-      "step": 47310
-    },
-    {
-      "epoch": 1.93,
-      "learning_rate": 7.111432350975766e-05,
-      "loss": 0.8026,
-      "step": 47320
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.10870484269096e-05,
-      "loss": 0.6153,
-      "step": 47330
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.105977334406153e-05,
-      "loss": 0.657,
-      "step": 47340
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.103249826121348e-05,
-      "loss": 0.5836,
-      "step": 47350
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.100522317836541e-05,
-      "loss": 0.6504,
-      "step": 47360
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.097794809551735e-05,
-      "loss": 0.616,
-      "step": 47370
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.095067301266928e-05,
-      "loss": 0.682,
-      "step": 47380
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.092339792982121e-05,
-      "loss": 0.716,
-      "step": 47390
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.089612284697316e-05,
-      "loss": 0.7462,
-      "step": 47400
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.08688477641251e-05,
-      "loss": 0.6852,
-      "step": 47410
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.084157268127703e-05,
-      "loss": 0.763,
-      "step": 47420
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.081429759842896e-05,
-      "loss": 0.71,
-      "step": 47430
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.07870225155809e-05,
-      "loss": 0.6724,
-      "step": 47440
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.075974743273283e-05,
-      "loss": 0.5909,
-      "step": 47450
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.073247234988476e-05,
-      "loss": 0.5667,
-      "step": 47460
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.07079247753215e-05,
-      "loss": 0.6525,
-      "step": 47470
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.068064969247344e-05,
-      "loss": 0.7448,
-      "step": 47480
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.065337460962537e-05,
-      "loss": 0.7132,
-      "step": 47490
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.06260995267773e-05,
-      "loss": 0.6311,
-      "step": 47500
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.059882444392925e-05,
-      "loss": 0.8348,
-      "step": 47510
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.057154936108119e-05,
-      "loss": 0.6047,
-      "step": 47520
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.054427427823312e-05,
-      "loss": 0.6555,
-      "step": 47530
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.051699919538506e-05,
-      "loss": 0.7598,
-      "step": 47540
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.048972411253699e-05,
-      "loss": 0.6182,
-      "step": 47550
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.046244902968892e-05,
-      "loss": 0.7013,
-      "step": 47560
-    },
-    {
-      "epoch": 1.94,
-      "learning_rate": 7.043517394684087e-05,
-      "loss": 0.717,
-      "step": 47570
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.04078988639928e-05,
-      "loss": 0.6296,
-      "step": 47580
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.038062378114474e-05,
-      "loss": 0.7149,
-      "step": 47590
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.035334869829667e-05,
-      "loss": 0.664,
-      "step": 47600
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.032607361544861e-05,
-      "loss": 0.7659,
-      "step": 47610
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.029879853260055e-05,
-      "loss": 0.7117,
-      "step": 47620
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.027152344975249e-05,
-      "loss": 0.772,
-      "step": 47630
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.024424836690442e-05,
-      "loss": 0.7264,
-      "step": 47640
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.021697328405636e-05,
-      "loss": 0.6171,
-      "step": 47650
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.018969820120829e-05,
-      "loss": 0.6351,
-      "step": 47660
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.016242311836022e-05,
-      "loss": 0.6514,
-      "step": 47670
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.013514803551217e-05,
-      "loss": 0.7472,
-      "step": 47680
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.01078729526641e-05,
-      "loss": 0.7343,
-      "step": 47690
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.008059786981604e-05,
-      "loss": 0.6311,
-      "step": 47700
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.005332278696797e-05,
-      "loss": 0.7247,
-      "step": 47710
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 7.002604770411991e-05,
-      "loss": 0.7142,
-      "step": 47720
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.999877262127184e-05,
-      "loss": 0.8288,
-      "step": 47730
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.997149753842377e-05,
-      "loss": 0.7275,
-      "step": 47740
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.994422245557571e-05,
-      "loss": 0.7228,
-      "step": 47750
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.991694737272764e-05,
-      "loss": 0.7091,
-      "step": 47760
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.988967228987958e-05,
-      "loss": 0.721,
-      "step": 47770
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.986239720703151e-05,
-      "loss": 0.7753,
-      "step": 47780
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.983512212418346e-05,
-      "loss": 0.6489,
-      "step": 47790
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.980784704133539e-05,
-      "loss": 0.7805,
-      "step": 47800
-    },
-    {
-      "epoch": 1.95,
-      "learning_rate": 6.978057195848733e-05,
-      "loss": 0.644,
-      "step": 47810
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.975329687563926e-05,
-      "loss": 0.6397,
-      "step": 47820
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.97260217927912e-05,
-      "loss": 0.7178,
-      "step": 47830
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.969874670994313e-05,
-      "loss": 0.6858,
-      "step": 47840
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.967147162709508e-05,
-      "loss": 0.7045,
-      "step": 47850
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.964419654424701e-05,
-      "loss": 0.6097,
-      "step": 47860
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.961692146139894e-05,
-      "loss": 0.6733,
-      "step": 47870
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.958964637855088e-05,
-      "loss": 0.6118,
-      "step": 47880
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.956237129570281e-05,
-      "loss": 0.7065,
-      "step": 47890
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.953509621285474e-05,
-      "loss": 0.5881,
-      "step": 47900
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.950782113000669e-05,
-      "loss": 0.6532,
-      "step": 47910
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.948054604715863e-05,
-      "loss": 0.7417,
-      "step": 47920
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.945327096431056e-05,
-      "loss": 0.6981,
-      "step": 47930
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.94259958814625e-05,
-      "loss": 0.6934,
-      "step": 47940
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.939872079861443e-05,
-      "loss": 0.6809,
-      "step": 47950
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.937144571576638e-05,
-      "loss": 0.6806,
-      "step": 47960
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.934417063291831e-05,
-      "loss": 0.7193,
-      "step": 47970
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.931689555007024e-05,
-      "loss": 0.6735,
-      "step": 47980
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.928962046722218e-05,
-      "loss": 0.613,
-      "step": 47990
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.926234538437411e-05,
-      "loss": 0.6367,
-      "step": 48000
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.923507030152605e-05,
-      "loss": 0.7603,
-      "step": 48010
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.920779521867798e-05,
-      "loss": 0.7445,
-      "step": 48020
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.918052013582991e-05,
-      "loss": 0.7541,
-      "step": 48030
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.915324505298185e-05,
-      "loss": 0.5959,
-      "step": 48040
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.912596997013378e-05,
-      "loss": 0.6155,
-      "step": 48050
-    },
-    {
-      "epoch": 1.96,
-      "learning_rate": 6.909869488728571e-05,
-      "loss": 0.6626,
-      "step": 48060
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.907141980443766e-05,
-      "loss": 0.6471,
-      "step": 48070
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.90441447215896e-05,
-      "loss": 0.6769,
-      "step": 48080
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.901686963874153e-05,
-      "loss": 0.6433,
-      "step": 48090
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.898959455589346e-05,
-      "loss": 0.6362,
-      "step": 48100
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.89623194730454e-05,
-      "loss": 0.7081,
-      "step": 48110
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.893504439019733e-05,
-      "loss": 0.6671,
-      "step": 48120
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.890776930734928e-05,
-      "loss": 0.6654,
-      "step": 48130
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.888049422450121e-05,
-      "loss": 0.6383,
-      "step": 48140
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.885321914165315e-05,
-      "loss": 0.6957,
-      "step": 48150
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.882594405880508e-05,
-      "loss": 0.6375,
-      "step": 48160
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.879866897595701e-05,
-      "loss": 0.747,
-      "step": 48170
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.877139389310895e-05,
-      "loss": 0.7843,
-      "step": 48180
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.87441188102609e-05,
-      "loss": 0.7591,
-      "step": 48190
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.871684372741283e-05,
-      "loss": 0.7036,
-      "step": 48200
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.868956864456476e-05,
-      "loss": 0.5962,
-      "step": 48210
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.86622935617167e-05,
-      "loss": 0.5941,
-      "step": 48220
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.863501847886863e-05,
-      "loss": 0.7045,
-      "step": 48230
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.860774339602058e-05,
-      "loss": 0.6813,
-      "step": 48240
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.858046831317251e-05,
-      "loss": 0.6258,
-      "step": 48250
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.855319323032445e-05,
-      "loss": 0.7163,
-      "step": 48260
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.852591814747638e-05,
-      "loss": 0.6769,
-      "step": 48270
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.849864306462832e-05,
-      "loss": 0.6129,
-      "step": 48280
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.847136798178025e-05,
-      "loss": 0.5939,
-      "step": 48290
-    },
-    {
-      "epoch": 1.97,
-      "learning_rate": 6.844409289893218e-05,
-      "loss": 0.6149,
-      "step": 48300
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.841681781608412e-05,
-      "loss": 0.6932,
-      "step": 48310
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.838954273323605e-05,
-      "loss": 0.6414,
-      "step": 48320
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.836226765038798e-05,
-      "loss": 0.6843,
-      "step": 48330
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.833499256753992e-05,
-      "loss": 0.6688,
-      "step": 48340
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.830771748469185e-05,
-      "loss": 0.6595,
-      "step": 48350
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.82804424018438e-05,
-      "loss": 0.6353,
-      "step": 48360
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.825316731899573e-05,
-      "loss": 0.7241,
-      "step": 48370
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.822589223614767e-05,
-      "loss": 0.7538,
-      "step": 48380
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.81986171532996e-05,
-      "loss": 0.7743,
-      "step": 48390
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.817134207045154e-05,
-      "loss": 0.6762,
-      "step": 48400
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.814406698760348e-05,
-      "loss": 0.6966,
-      "step": 48410
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.811679190475542e-05,
-      "loss": 0.6397,
-      "step": 48420
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.808951682190735e-05,
-      "loss": 0.7121,
-      "step": 48430
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.806224173905928e-05,
-      "loss": 0.6195,
-      "step": 48440
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.803496665621122e-05,
-      "loss": 0.7842,
-      "step": 48450
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.800769157336315e-05,
-      "loss": 0.6608,
-      "step": 48460
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.79804164905151e-05,
-      "loss": 0.7709,
-      "step": 48470
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.795314140766703e-05,
-      "loss": 0.6206,
-      "step": 48480
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.792586632481897e-05,
-      "loss": 0.6865,
-      "step": 48490
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.78985912419709e-05,
-      "loss": 0.6514,
-      "step": 48500
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.787131615912284e-05,
-      "loss": 0.6757,
-      "step": 48510
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.784404107627477e-05,
-      "loss": 0.7179,
-      "step": 48520
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.781676599342672e-05,
-      "loss": 0.6873,
-      "step": 48530
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.778949091057865e-05,
-      "loss": 0.6054,
-      "step": 48540
-    },
-    {
-      "epoch": 1.98,
-      "learning_rate": 6.776221582773059e-05,
-      "loss": 0.7375,
-      "step": 48550
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.773494074488252e-05,
-      "loss": 0.6109,
-      "step": 48560
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.770766566203445e-05,
-      "loss": 0.6879,
-      "step": 48570
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.768039057918639e-05,
-      "loss": 0.6868,
-      "step": 48580
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.765311549633832e-05,
-      "loss": 0.7537,
-      "step": 48590
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.762584041349025e-05,
-      "loss": 0.6718,
-      "step": 48600
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.759856533064219e-05,
-      "loss": 0.6847,
-      "step": 48610
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.757129024779412e-05,
-      "loss": 0.5923,
-      "step": 48620
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.754401516494606e-05,
-      "loss": 0.6187,
-      "step": 48630
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.7516740082098e-05,
-      "loss": 0.6272,
-      "step": 48640
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.748946499924994e-05,
-      "loss": 0.6386,
-      "step": 48650
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.746218991640187e-05,
-      "loss": 0.7015,
-      "step": 48660
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.74349148335538e-05,
-      "loss": 0.7105,
-      "step": 48670
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.740763975070574e-05,
-      "loss": 0.6291,
-      "step": 48680
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.738036466785769e-05,
-      "loss": 0.7036,
-      "step": 48690
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.735308958500962e-05,
-      "loss": 0.7193,
-      "step": 48700
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.732581450216155e-05,
-      "loss": 0.6992,
-      "step": 48710
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.729853941931349e-05,
-      "loss": 0.6475,
-      "step": 48720
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.727126433646542e-05,
-      "loss": 0.6414,
-      "step": 48730
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.724398925361736e-05,
-      "loss": 0.7239,
-      "step": 48740
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.72167141707693e-05,
-      "loss": 0.6597,
-      "step": 48750
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.718943908792124e-05,
-      "loss": 0.7478,
-      "step": 48760
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.716216400507317e-05,
-      "loss": 0.6668,
-      "step": 48770
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.71348889222251e-05,
-      "loss": 0.6317,
-      "step": 48780
-    },
-    {
-      "epoch": 1.99,
-      "learning_rate": 6.710761383937704e-05,
-      "loss": 0.7851,
-      "step": 48790
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.708033875652897e-05,
-      "loss": 0.69,
-      "step": 48800
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.705306367368092e-05,
-      "loss": 0.6559,
-      "step": 48810
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.702578859083286e-05,
-      "loss": 0.7253,
-      "step": 48820
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.699851350798479e-05,
-      "loss": 0.7104,
-      "step": 48830
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.697123842513672e-05,
-      "loss": 0.6637,
-      "step": 48840
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.694396334228866e-05,
-      "loss": 0.7175,
-      "step": 48850
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.691668825944059e-05,
-      "loss": 0.6707,
-      "step": 48860
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.688941317659252e-05,
-      "loss": 0.7645,
-      "step": 48870
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.686213809374446e-05,
-      "loss": 0.691,
-      "step": 48880
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.683486301089639e-05,
-      "loss": 0.619,
-      "step": 48890
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.680758792804833e-05,
-      "loss": 0.695,
-      "step": 48900
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.678031284520026e-05,
-      "loss": 0.8007,
-      "step": 48910
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.675303776235221e-05,
-      "loss": 0.6445,
-      "step": 48920
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.672576267950414e-05,
-      "loss": 0.6524,
-      "step": 48930
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.669848759665608e-05,
-      "loss": 0.697,
-      "step": 48940
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.667121251380801e-05,
-      "loss": 0.6343,
-      "step": 48950
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.664393743095994e-05,
-      "loss": 0.6885,
-      "step": 48960
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.661666234811188e-05,
-      "loss": 0.6426,
-      "step": 48970
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.658938726526383e-05,
-      "loss": 0.6679,
-      "step": 48980
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.656211218241576e-05,
-      "loss": 0.6246,
-      "step": 48990
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.653483709956769e-05,
-      "loss": 0.6078,
-      "step": 49000
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.650756201671963e-05,
-      "loss": 0.7978,
-      "step": 49010
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.648028693387156e-05,
-      "loss": 0.6885,
-      "step": 49020
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.645301185102351e-05,
-      "loss": 0.596,
-      "step": 49030
-    },
-    {
-      "epoch": 2.0,
-      "learning_rate": 6.642573676817544e-05,
-      "loss": 0.6758,
-      "step": 49040
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.639846168532738e-05,
-      "loss": 0.6598,
-      "step": 49050
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.637118660247931e-05,
-      "loss": 0.5806,
-      "step": 49060
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.634391151963124e-05,
-      "loss": 0.6399,
-      "step": 49070
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.631663643678318e-05,
-      "loss": 0.623,
-      "step": 49080
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.628936135393513e-05,
-      "loss": 0.6767,
-      "step": 49090
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.626208627108706e-05,
-      "loss": 0.6382,
-      "step": 49100
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.623481118823899e-05,
-      "loss": 0.671,
-      "step": 49110
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.620753610539093e-05,
-      "loss": 0.6468,
-      "step": 49120
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.618026102254286e-05,
-      "loss": 0.5934,
-      "step": 49130
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.61529859396948e-05,
-      "loss": 0.6846,
-      "step": 49140
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.612571085684673e-05,
-      "loss": 0.6337,
-      "step": 49150
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.609843577399866e-05,
-      "loss": 0.67,
-      "step": 49160
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.60711606911506e-05,
-      "loss": 0.5594,
-      "step": 49170
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.604388560830253e-05,
-      "loss": 0.6035,
-      "step": 49180
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.601661052545446e-05,
-      "loss": 0.6661,
-      "step": 49190
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.598933544260641e-05,
-      "loss": 0.6188,
-      "step": 49200
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.596206035975835e-05,
-      "loss": 0.6596,
-      "step": 49210
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.593478527691028e-05,
-      "loss": 0.542,
-      "step": 49220
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.590751019406221e-05,
-      "loss": 0.6474,
-      "step": 49230
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.588023511121415e-05,
-      "loss": 0.6344,
-      "step": 49240
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.585296002836608e-05,
-      "loss": 0.6761,
-      "step": 49250
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.582568494551803e-05,
-      "loss": 0.6206,
-      "step": 49260
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.579840986266996e-05,
-      "loss": 0.6968,
-      "step": 49270
-    },
-    {
-      "epoch": 2.01,
-      "learning_rate": 6.57711347798219e-05,
-      "loss": 0.7082,
-      "step": 49280
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.574385969697383e-05,
-      "loss": 0.6108,
-      "step": 49290
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.571658461412576e-05,
-      "loss": 0.6234,
-      "step": 49300
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.568930953127771e-05,
-      "loss": 0.5903,
-      "step": 49310
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.566203444842965e-05,
-      "loss": 0.6983,
-      "step": 49320
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.563475936558158e-05,
-      "loss": 0.6047,
-      "step": 49330
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.560748428273351e-05,
-      "loss": 0.6653,
-      "step": 49340
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.558020919988545e-05,
-      "loss": 0.6304,
-      "step": 49350
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.555293411703738e-05,
-      "loss": 0.6483,
-      "step": 49360
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.552565903418933e-05,
-      "loss": 0.7629,
-      "step": 49370
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.549838395134126e-05,
-      "loss": 0.5609,
-      "step": 49380
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.54711088684932e-05,
-      "loss": 0.6087,
-      "step": 49390
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.544383378564513e-05,
-      "loss": 0.644,
-      "step": 49400
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.541655870279706e-05,
-      "loss": 0.6475,
-      "step": 49410
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.5389283619949e-05,
-      "loss": 0.6108,
-      "step": 49420
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.536473604538574e-05,
-      "loss": 0.656,
-      "step": 49430
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.533746096253767e-05,
-      "loss": 0.6751,
-      "step": 49440
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.531018587968961e-05,
-      "loss": 0.6385,
-      "step": 49450
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.528291079684154e-05,
-      "loss": 0.6812,
-      "step": 49460
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.525563571399348e-05,
-      "loss": 0.783,
-      "step": 49470
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.522836063114542e-05,
-      "loss": 0.7148,
-      "step": 49480
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.520108554829736e-05,
-      "loss": 0.5924,
-      "step": 49490
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.517381046544929e-05,
-      "loss": 0.6782,
-      "step": 49500
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.514653538260122e-05,
-      "loss": 0.7116,
-      "step": 49510
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.511926029975316e-05,
-      "loss": 0.647,
-      "step": 49520
-    },
-    {
-      "epoch": 2.02,
-      "learning_rate": 6.50919852169051e-05,
-      "loss": 0.5111,
-      "step": 49530
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.506471013405704e-05,
-      "loss": 0.7152,
-      "step": 49540
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.503743505120897e-05,
-      "loss": 0.646,
-      "step": 49550
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.501015996836091e-05,
-      "loss": 0.562,
-      "step": 49560
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.498288488551284e-05,
-      "loss": 0.6811,
-      "step": 49570
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.495560980266478e-05,
-      "loss": 0.5943,
-      "step": 49580
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.492833471981672e-05,
-      "loss": 0.6711,
-      "step": 49590
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.490105963696866e-05,
-      "loss": 0.6291,
-      "step": 49600
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.487378455412059e-05,
-      "loss": 0.6872,
-      "step": 49610
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.484650947127252e-05,
-      "loss": 0.5833,
-      "step": 49620
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.481923438842446e-05,
-      "loss": 0.7189,
-      "step": 49630
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.479195930557639e-05,
-      "loss": 0.6918,
-      "step": 49640
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.476468422272834e-05,
-      "loss": 0.6569,
-      "step": 49650
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.473740913988027e-05,
-      "loss": 0.6052,
-      "step": 49660
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.471013405703221e-05,
-      "loss": 0.6481,
-      "step": 49670
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.468285897418414e-05,
-      "loss": 0.7415,
-      "step": 49680
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.465558389133608e-05,
-      "loss": 0.7068,
-      "step": 49690
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.462830880848801e-05,
-      "loss": 0.7314,
-      "step": 49700
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.460103372563994e-05,
-      "loss": 0.6635,
-      "step": 49710
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.457375864279188e-05,
-      "loss": 0.6996,
-      "step": 49720
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.454648355994381e-05,
-      "loss": 0.659,
-      "step": 49730
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.451920847709575e-05,
-      "loss": 0.5947,
-      "step": 49740
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.449193339424768e-05,
-      "loss": 0.5997,
-      "step": 49750
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.446465831139963e-05,
-      "loss": 0.6826,
-      "step": 49760
-    },
-    {
-      "epoch": 2.03,
-      "learning_rate": 6.443738322855156e-05,
-      "loss": 0.6235,
-      "step": 49770
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.44101081457035e-05,
-      "loss": 0.5937,
-      "step": 49780
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.438283306285543e-05,
-      "loss": 0.7867,
-      "step": 49790
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.435555798000736e-05,
-      "loss": 0.6155,
-      "step": 49800
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.43282828971593e-05,
-      "loss": 0.6775,
-      "step": 49810
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.430100781431124e-05,
-      "loss": 0.62,
-      "step": 49820
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.427373273146318e-05,
-      "loss": 0.7047,
-      "step": 49830
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.424645764861511e-05,
-      "loss": 0.6498,
-      "step": 49840
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.421918256576705e-05,
-      "loss": 0.6672,
-      "step": 49850
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.419190748291898e-05,
-      "loss": 0.6165,
-      "step": 49860
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.416463240007093e-05,
-      "loss": 0.5588,
-      "step": 49870
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.413735731722286e-05,
-      "loss": 0.8178,
-      "step": 49880
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.41100822343748e-05,
-      "loss": 0.6651,
-      "step": 49890
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.408280715152673e-05,
-      "loss": 0.7012,
-      "step": 49900
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.405553206867866e-05,
-      "loss": 0.6864,
-      "step": 49910
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.40282569858306e-05,
-      "loss": 0.7148,
-      "step": 49920
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.400098190298254e-05,
-      "loss": 0.6313,
-      "step": 49930
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.397370682013448e-05,
-      "loss": 0.6253,
-      "step": 49940
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.394643173728641e-05,
-      "loss": 0.6915,
-      "step": 49950
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.391915665443835e-05,
-      "loss": 0.7789,
-      "step": 49960
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.389188157159028e-05,
-      "loss": 0.6353,
-      "step": 49970
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.386460648874221e-05,
-      "loss": 0.731,
-      "step": 49980
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.383733140589415e-05,
-      "loss": 0.6884,
-      "step": 49990
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.381005632304608e-05,
-      "loss": 0.7122,
-      "step": 50000
-    },
-    {
-      "epoch": 2.04,
-      "learning_rate": 6.378278124019802e-05,
-      "loss": 0.6694,
-      "step": 50010
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.375550615734995e-05,
-      "loss": 0.6129,
-      "step": 50020
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.372823107450188e-05,
-      "loss": 0.6649,
-      "step": 50030
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.370095599165383e-05,
-      "loss": 0.7563,
-      "step": 50040
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.367368090880576e-05,
-      "loss": 0.6592,
-      "step": 50050
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.36464058259577e-05,
-      "loss": 0.6074,
-      "step": 50060
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.361913074310963e-05,
-      "loss": 0.6243,
-      "step": 50070
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.359185566026157e-05,
-      "loss": 0.5613,
-      "step": 50080
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.35645805774135e-05,
-      "loss": 0.6359,
-      "step": 50090
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.353730549456545e-05,
-      "loss": 0.5999,
-      "step": 50100
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.351003041171738e-05,
-      "loss": 0.6285,
-      "step": 50110
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.348275532886932e-05,
-      "loss": 0.6257,
-      "step": 50120
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.345548024602125e-05,
-      "loss": 0.6572,
-      "step": 50130
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.342820516317318e-05,
-      "loss": 0.5815,
-      "step": 50140
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.340093008032513e-05,
-      "loss": 0.6019,
-      "step": 50150
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.337365499747706e-05,
-      "loss": 0.5976,
-      "step": 50160
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.3346379914629e-05,
-      "loss": 0.7421,
-      "step": 50170
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.331910483178093e-05,
-      "loss": 0.5912,
-      "step": 50180
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.329182974893287e-05,
-      "loss": 0.5911,
-      "step": 50190
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.32645546660848e-05,
-      "loss": 0.5722,
-      "step": 50200
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.323727958323675e-05,
-      "loss": 0.6074,
-      "step": 50210
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.321000450038868e-05,
-      "loss": 0.6773,
-      "step": 50220
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.318272941754062e-05,
-      "loss": 0.6749,
-      "step": 50230
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.315545433469255e-05,
-      "loss": 0.6834,
-      "step": 50240
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.312817925184448e-05,
-      "loss": 0.6305,
-      "step": 50250
-    },
-    {
-      "epoch": 2.05,
-      "learning_rate": 6.310090416899642e-05,
-      "loss": 0.7243,
-      "step": 50260
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.307362908614835e-05,
-      "loss": 0.6222,
-      "step": 50270
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.304635400330029e-05,
-      "loss": 0.7579,
-      "step": 50280
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.301907892045222e-05,
-      "loss": 0.7009,
-      "step": 50290
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.299180383760415e-05,
-      "loss": 0.678,
-      "step": 50300
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.296452875475609e-05,
-      "loss": 0.6667,
-      "step": 50310
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.293725367190803e-05,
-      "loss": 0.76,
-      "step": 50320
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.290997858905997e-05,
-      "loss": 0.6945,
-      "step": 50330
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.28827035062119e-05,
-      "loss": 0.5822,
-      "step": 50340
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.285542842336384e-05,
-      "loss": 0.5939,
-      "step": 50350
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.282815334051577e-05,
-      "loss": 0.6289,
-      "step": 50360
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.28008782576677e-05,
-      "loss": 0.609,
-      "step": 50370
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.277360317481965e-05,
-      "loss": 0.7188,
-      "step": 50380
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.274632809197159e-05,
-      "loss": 0.6831,
-      "step": 50390
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.271905300912352e-05,
-      "loss": 0.7273,
-      "step": 50400
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.269177792627545e-05,
-      "loss": 0.5593,
-      "step": 50410
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.266450284342739e-05,
-      "loss": 0.6656,
-      "step": 50420
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.263722776057933e-05,
-      "loss": 0.5877,
-      "step": 50430
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.260995267773127e-05,
-      "loss": 0.6804,
-      "step": 50440
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.25826775948832e-05,
-      "loss": 0.763,
-      "step": 50450
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.255540251203514e-05,
-      "loss": 0.6907,
-      "step": 50460
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.252812742918707e-05,
-      "loss": 0.6702,
-      "step": 50470
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.2500852346339e-05,
-      "loss": 0.5802,
-      "step": 50480
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.247357726349095e-05,
-      "loss": 0.6918,
-      "step": 50490
-    },
-    {
-      "epoch": 2.06,
-      "learning_rate": 6.244630218064289e-05,
-      "loss": 0.6466,
-      "step": 50500
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.241902709779482e-05,
-      "loss": 0.7259,
-      "step": 50510
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.239175201494675e-05,
-      "loss": 0.6121,
-      "step": 50520
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.236447693209869e-05,
-      "loss": 0.7166,
-      "step": 50530
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.233720184925062e-05,
-      "loss": 0.7233,
-      "step": 50540
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.230992676640256e-05,
-      "loss": 0.6342,
-      "step": 50550
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.228265168355449e-05,
-      "loss": 0.6838,
-      "step": 50560
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.225537660070642e-05,
-      "loss": 0.5854,
-      "step": 50570
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.222810151785836e-05,
-      "loss": 0.6177,
-      "step": 50580
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.220082643501029e-05,
-      "loss": 0.56,
-      "step": 50590
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.217355135216224e-05,
-      "loss": 0.6488,
-      "step": 50600
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.214627626931417e-05,
-      "loss": 0.6818,
-      "step": 50610
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.21190011864661e-05,
-      "loss": 0.6574,
-      "step": 50620
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.209172610361804e-05,
-      "loss": 0.6725,
-      "step": 50630
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.206445102076997e-05,
-      "loss": 0.5838,
-      "step": 50640
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.203717593792191e-05,
-      "loss": 0.5027,
-      "step": 50650
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.200990085507386e-05,
-      "loss": 0.6312,
-      "step": 50660
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.198262577222579e-05,
-      "loss": 0.7183,
-      "step": 50670
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.195535068937772e-05,
-      "loss": 0.6681,
-      "step": 50680
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.192807560652966e-05,
-      "loss": 0.673,
-      "step": 50690
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.190080052368159e-05,
-      "loss": 0.5536,
-      "step": 50700
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.187352544083353e-05,
-      "loss": 0.7764,
-      "step": 50710
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.184625035798547e-05,
-      "loss": 0.6427,
-      "step": 50720
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.18189752751374e-05,
-      "loss": 0.5884,
-      "step": 50730
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.179170019228934e-05,
-      "loss": 0.5243,
-      "step": 50740
-    },
-    {
-      "epoch": 2.07,
-      "learning_rate": 6.176442510944127e-05,
-      "loss": 0.6261,
-      "step": 50750
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.173715002659321e-05,
-      "loss": 0.6026,
-      "step": 50760
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.170987494374516e-05,
-      "loss": 0.625,
-      "step": 50770
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.168259986089709e-05,
-      "loss": 0.6409,
-      "step": 50780
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.165532477804902e-05,
-      "loss": 0.6563,
-      "step": 50790
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.162804969520096e-05,
-      "loss": 0.6688,
-      "step": 50800
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.160077461235289e-05,
-      "loss": 0.6848,
-      "step": 50810
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.157349952950483e-05,
-      "loss": 0.5679,
-      "step": 50820
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.154622444665676e-05,
-      "loss": 0.6375,
-      "step": 50830
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.151894936380869e-05,
-      "loss": 0.6664,
-      "step": 50840
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.149167428096063e-05,
-      "loss": 0.6288,
-      "step": 50850
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.146439919811256e-05,
-      "loss": 0.613,
-      "step": 50860
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.14371241152645e-05,
-      "loss": 0.6081,
-      "step": 50870
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.140984903241644e-05,
-      "loss": 0.6799,
-      "step": 50880
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.138257394956838e-05,
-      "loss": 0.5623,
-      "step": 50890
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.135529886672031e-05,
-      "loss": 0.581,
-      "step": 50900
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.132802378387224e-05,
-      "loss": 0.5891,
-      "step": 50910
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.130074870102418e-05,
-      "loss": 0.6903,
-      "step": 50920
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.127347361817611e-05,
-      "loss": 0.6798,
-      "step": 50930
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.124619853532806e-05,
-      "loss": 0.6754,
-      "step": 50940
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.121892345248e-05,
-      "loss": 0.6545,
-      "step": 50950
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.119164836963193e-05,
-      "loss": 0.6753,
-      "step": 50960
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.116437328678386e-05,
-      "loss": 0.5643,
-      "step": 50970
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.11370982039358e-05,
-      "loss": 0.7438,
-      "step": 50980
-    },
-    {
-      "epoch": 2.08,
-      "learning_rate": 6.110982312108773e-05,
-      "loss": 0.6485,
-      "step": 50990
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.108254803823968e-05,
-      "loss": 0.6721,
-      "step": 51000
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.105527295539161e-05,
-      "loss": 0.6957,
-      "step": 51010
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.102799787254354e-05,
-      "loss": 0.5795,
-      "step": 51020
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.100072278969547e-05,
-      "loss": 0.67,
-      "step": 51030
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0973447706847405e-05,
-      "loss": 0.7571,
-      "step": 51040
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.094617262399935e-05,
-      "loss": 0.6969,
-      "step": 51050
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.091889754115129e-05,
-      "loss": 0.6284,
-      "step": 51060
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.089162245830322e-05,
-      "loss": 0.6825,
-      "step": 51070
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0864347375455155e-05,
-      "loss": 0.7374,
-      "step": 51080
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.083707229260709e-05,
-      "loss": 0.6554,
-      "step": 51090
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.080979720975902e-05,
-      "loss": 0.6352,
-      "step": 51100
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.078252212691097e-05,
-      "loss": 0.7131,
-      "step": 51110
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0755247044062904e-05,
-      "loss": 0.774,
-      "step": 51120
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.072797196121484e-05,
-      "loss": 0.6126,
-      "step": 51130
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.070069687836677e-05,
-      "loss": 0.6258,
-      "step": 51140
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0673421795518706e-05,
-      "loss": 0.6076,
-      "step": 51150
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.064614671267064e-05,
-      "loss": 0.6583,
-      "step": 51160
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.061887162982258e-05,
-      "loss": 0.6415,
-      "step": 51170
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0591596546974514e-05,
-      "loss": 0.5805,
-      "step": 51180
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.056432146412645e-05,
-      "loss": 0.6441,
-      "step": 51190
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.053704638127838e-05,
-      "loss": 0.6488,
-      "step": 51200
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.0509771298430316e-05,
-      "loss": 0.7085,
-      "step": 51210
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.048249621558226e-05,
-      "loss": 0.6829,
-      "step": 51220
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.04552211327342e-05,
-      "loss": 0.5808,
-      "step": 51230
-    },
-    {
-      "epoch": 2.09,
-      "learning_rate": 6.042794604988613e-05,
-      "loss": 0.6496,
-      "step": 51240
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.040339847532287e-05,
-      "loss": 0.6746,
-      "step": 51250
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0376123392474806e-05,
-      "loss": 0.6533,
-      "step": 51260
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0348848309626746e-05,
-      "loss": 0.692,
-      "step": 51270
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.032157322677868e-05,
-      "loss": 0.6871,
-      "step": 51280
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0294298143930614e-05,
-      "loss": 0.6903,
-      "step": 51290
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.026702306108255e-05,
-      "loss": 0.5854,
-      "step": 51300
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.023974797823448e-05,
-      "loss": 0.7266,
-      "step": 51310
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0212472895386416e-05,
-      "loss": 0.5636,
-      "step": 51320
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0185197812538363e-05,
-      "loss": 0.6954,
-      "step": 51330
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.01579227296903e-05,
-      "loss": 0.623,
-      "step": 51340
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.013064764684223e-05,
-      "loss": 0.7023,
-      "step": 51350
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.0103372563994165e-05,
-      "loss": 0.7487,
-      "step": 51360
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.00760974811461e-05,
-      "loss": 0.7301,
-      "step": 51370
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.004882239829803e-05,
-      "loss": 0.6515,
-      "step": 51380
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 6.002154731544998e-05,
-      "loss": 0.6563,
-      "step": 51390
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.9994272232601914e-05,
-      "loss": 0.658,
-      "step": 51400
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.996699714975385e-05,
-      "loss": 0.6364,
-      "step": 51410
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.993972206690578e-05,
-      "loss": 0.5799,
-      "step": 51420
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.9912446984057716e-05,
-      "loss": 0.6521,
-      "step": 51430
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.988517190120966e-05,
-      "loss": 0.6305,
-      "step": 51440
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.985789681836159e-05,
-      "loss": 0.6959,
-      "step": 51450
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.9830621735513525e-05,
-      "loss": 0.6456,
-      "step": 51460
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.980334665266546e-05,
-      "loss": 0.6958,
-      "step": 51470
-    },
-    {
-      "epoch": 2.1,
-      "learning_rate": 5.977607156981739e-05,
-      "loss": 0.6391,
-      "step": 51480
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9748796486969326e-05,
-      "loss": 0.6324,
-      "step": 51490
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9721521404121274e-05,
-      "loss": 0.7362,
-      "step": 51500
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.969424632127321e-05,
-      "loss": 0.7695,
-      "step": 51510
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.966697123842514e-05,
-      "loss": 0.6319,
-      "step": 51520
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9639696155577076e-05,
-      "loss": 0.6859,
-      "step": 51530
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.961242107272901e-05,
-      "loss": 0.5794,
-      "step": 51540
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9585145989880944e-05,
-      "loss": 0.6387,
-      "step": 51550
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9557870907032884e-05,
-      "loss": 0.7503,
-      "step": 51560
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.953059582418482e-05,
-      "loss": 0.5691,
-      "step": 51570
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.950332074133675e-05,
-      "loss": 0.6528,
-      "step": 51580
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9476045658488686e-05,
-      "loss": 0.5659,
-      "step": 51590
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.944877057564062e-05,
-      "loss": 0.6473,
-      "step": 51600
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.942149549279257e-05,
-      "loss": 0.614,
-      "step": 51610
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.93942204099445e-05,
-      "loss": 0.6346,
-      "step": 51620
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9366945327096435e-05,
-      "loss": 0.6446,
-      "step": 51630
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.933967024424837e-05,
-      "loss": 0.5061,
-      "step": 51640
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.93123951614003e-05,
-      "loss": 0.7152,
-      "step": 51650
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.928512007855224e-05,
-      "loss": 0.6823,
-      "step": 51660
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9257844995704184e-05,
-      "loss": 0.5664,
-      "step": 51670
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.923056991285612e-05,
-      "loss": 0.6807,
-      "step": 51680
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.920329483000805e-05,
-      "loss": 0.568,
-      "step": 51690
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9176019747159986e-05,
-      "loss": 0.7288,
-      "step": 51700
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.914874466431192e-05,
-      "loss": 0.6609,
-      "step": 51710
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.912146958146386e-05,
-      "loss": 0.5593,
-      "step": 51720
-    },
-    {
-      "epoch": 2.11,
-      "learning_rate": 5.9094194498615795e-05,
-      "loss": 0.7587,
-      "step": 51730
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.906691941576773e-05,
-      "loss": 0.6906,
-      "step": 51740
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.903964433291966e-05,
-      "loss": 0.7054,
-      "step": 51750
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.9012369250071597e-05,
-      "loss": 0.6174,
-      "step": 51760
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.898509416722353e-05,
-      "loss": 0.6864,
-      "step": 51770
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.895781908437548e-05,
-      "loss": 0.6408,
-      "step": 51780
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.893054400152741e-05,
-      "loss": 0.5942,
-      "step": 51790
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8903268918679346e-05,
-      "loss": 0.7524,
-      "step": 51800
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.887599383583128e-05,
-      "loss": 0.5723,
-      "step": 51810
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8848718752983214e-05,
-      "loss": 0.657,
-      "step": 51820
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.882144367013515e-05,
-      "loss": 0.6618,
-      "step": 51830
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.879416858728709e-05,
-      "loss": 0.6379,
-      "step": 51840
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.876689350443902e-05,
-      "loss": 0.6537,
-      "step": 51850
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8739618421590956e-05,
-      "loss": 0.6986,
-      "step": 51860
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.871234333874289e-05,
-      "loss": 0.609,
-      "step": 51870
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8685068255894824e-05,
-      "loss": 0.6543,
-      "step": 51880
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.865779317304677e-05,
-      "loss": 0.7034,
-      "step": 51890
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8630518090198705e-05,
-      "loss": 0.6412,
-      "step": 51900
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.860324300735064e-05,
-      "loss": 0.6248,
-      "step": 51910
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.857596792450257e-05,
-      "loss": 0.5756,
-      "step": 51920
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.854869284165451e-05,
-      "loss": 0.5951,
-      "step": 51930
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.852141775880644e-05,
-      "loss": 0.6973,
-      "step": 51940
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.849414267595839e-05,
-      "loss": 0.6851,
-      "step": 51950
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.846686759311032e-05,
-      "loss": 0.6118,
-      "step": 51960
-    },
-    {
-      "epoch": 2.12,
-      "learning_rate": 5.8439592510262256e-05,
-      "loss": 0.648,
-      "step": 51970
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.841231742741419e-05,
-      "loss": 0.7066,
-      "step": 51980
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8385042344566124e-05,
-      "loss": 0.6834,
-      "step": 51990
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.835776726171806e-05,
-      "loss": 0.5822,
-      "step": 52000
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.833049217887e-05,
-      "loss": 0.6519,
-      "step": 52010
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.830321709602193e-05,
-      "loss": 0.5754,
-      "step": 52020
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8275942013173867e-05,
-      "loss": 0.6689,
-      "step": 52030
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.82486669303258e-05,
-      "loss": 0.5924,
-      "step": 52040
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8221391847477734e-05,
-      "loss": 0.7223,
-      "step": 52050
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.819411676462968e-05,
-      "loss": 0.6249,
-      "step": 52060
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8166841681781616e-05,
-      "loss": 0.654,
-      "step": 52070
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.813956659893355e-05,
-      "loss": 0.669,
-      "step": 52080
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8112291516085484e-05,
-      "loss": 0.6457,
-      "step": 52090
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.808501643323742e-05,
-      "loss": 0.6757,
-      "step": 52100
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.805774135038935e-05,
-      "loss": 0.6465,
-      "step": 52110
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.803046626754129e-05,
-      "loss": 0.6942,
-      "step": 52120
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.8003191184693226e-05,
-      "loss": 0.6579,
-      "step": 52130
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.797591610184516e-05,
-      "loss": 0.601,
-      "step": 52140
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.7948641018997094e-05,
-      "loss": 0.6921,
-      "step": 52150
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.792136593614903e-05,
-      "loss": 0.599,
-      "step": 52160
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.7894090853300975e-05,
-      "loss": 0.5959,
-      "step": 52170
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.786681577045291e-05,
-      "loss": 0.6554,
-      "step": 52180
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.783954068760484e-05,
-      "loss": 0.7014,
-      "step": 52190
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.781226560475678e-05,
-      "loss": 0.7041,
-      "step": 52200
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.778499052190871e-05,
-      "loss": 0.5688,
-      "step": 52210
-    },
-    {
-      "epoch": 2.13,
-      "learning_rate": 5.7757715439060645e-05,
-      "loss": 0.6503,
-      "step": 52220
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.773044035621259e-05,
-      "loss": 0.6205,
-      "step": 52230
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7703165273364526e-05,
-      "loss": 0.6625,
-      "step": 52240
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.767589019051646e-05,
-      "loss": 0.6137,
-      "step": 52250
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7648615107668394e-05,
-      "loss": 0.7225,
-      "step": 52260
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.762134002482033e-05,
-      "loss": 0.6144,
-      "step": 52270
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.759406494197226e-05,
-      "loss": 0.5464,
-      "step": 52280
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.75667898591242e-05,
-      "loss": 0.6112,
-      "step": 52290
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.753951477627614e-05,
-      "loss": 0.6823,
-      "step": 52300
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.751223969342807e-05,
-      "loss": 0.6185,
-      "step": 52310
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7484964610580004e-05,
-      "loss": 0.6035,
-      "step": 52320
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.745768952773194e-05,
-      "loss": 0.6572,
-      "step": 52330
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7430414444883886e-05,
-      "loss": 0.6072,
-      "step": 52340
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.740313936203582e-05,
-      "loss": 0.6091,
-      "step": 52350
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7375864279187754e-05,
-      "loss": 0.6278,
-      "step": 52360
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.734858919633969e-05,
-      "loss": 0.6508,
-      "step": 52370
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.732131411349162e-05,
-      "loss": 0.6469,
-      "step": 52380
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7294039030643555e-05,
-      "loss": 0.557,
-      "step": 52390
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7266763947795496e-05,
-      "loss": 0.7146,
-      "step": 52400
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.723948886494743e-05,
-      "loss": 0.6565,
-      "step": 52410
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7212213782099364e-05,
-      "loss": 0.6111,
-      "step": 52420
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.71849386992513e-05,
-      "loss": 0.6712,
-      "step": 52430
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.715766361640323e-05,
-      "loss": 0.6679,
-      "step": 52440
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.7130388533555166e-05,
-      "loss": 0.6461,
-      "step": 52450
-    },
-    {
-      "epoch": 2.14,
-      "learning_rate": 5.710311345070711e-05,
-      "loss": 0.6732,
-      "step": 52460
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.707583836785905e-05,
-      "loss": 0.7128,
-      "step": 52470
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.704856328501098e-05,
-      "loss": 0.6589,
-      "step": 52480
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.7021288202162915e-05,
-      "loss": 0.7306,
-      "step": 52490
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.699401311931485e-05,
-      "loss": 0.6414,
-      "step": 52500
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6966738036466796e-05,
-      "loss": 0.7171,
-      "step": 52510
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.693946295361873e-05,
-      "loss": 0.6435,
-      "step": 52520
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6912187870770664e-05,
-      "loss": 0.6384,
-      "step": 52530
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.68849127879226e-05,
-      "loss": 0.7577,
-      "step": 52540
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.685763770507453e-05,
-      "loss": 0.591,
-      "step": 52550
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6830362622226466e-05,
-      "loss": 0.6411,
-      "step": 52560
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.680308753937841e-05,
-      "loss": 0.6296,
-      "step": 52570
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.677581245653034e-05,
-      "loss": 0.5971,
-      "step": 52580
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6748537373682275e-05,
-      "loss": 0.6208,
-      "step": 52590
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.672126229083421e-05,
-      "loss": 0.7329,
-      "step": 52600
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.669398720798614e-05,
-      "loss": 0.6226,
-      "step": 52610
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.666671212513809e-05,
-      "loss": 0.6495,
-      "step": 52620
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6639437042290024e-05,
-      "loss": 0.7005,
-      "step": 52630
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.661216195944196e-05,
-      "loss": 0.6145,
-      "step": 52640
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.658488687659389e-05,
-      "loss": 0.6054,
-      "step": 52650
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6557611793745826e-05,
-      "loss": 0.7814,
-      "step": 52660
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.653033671089776e-05,
-      "loss": 0.6995,
-      "step": 52670
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.65030616280497e-05,
-      "loss": 0.6635,
-      "step": 52680
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.6475786545201634e-05,
-      "loss": 0.6044,
-      "step": 52690
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.644851146235357e-05,
-      "loss": 0.6392,
-      "step": 52700
-    },
-    {
-      "epoch": 2.15,
-      "learning_rate": 5.64212363795055e-05,
-      "loss": 0.6972,
-      "step": 52710
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6393961296657436e-05,
-      "loss": 0.6905,
-      "step": 52720
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.636668621380937e-05,
-      "loss": 0.5751,
-      "step": 52730
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.633941113096132e-05,
-      "loss": 0.6487,
-      "step": 52740
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.631213604811325e-05,
-      "loss": 0.5849,
-      "step": 52750
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6284860965265185e-05,
-      "loss": 0.7577,
-      "step": 52760
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.625758588241712e-05,
-      "loss": 0.6687,
-      "step": 52770
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.623031079956905e-05,
-      "loss": 0.6508,
-      "step": 52780
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6203035716721e-05,
-      "loss": 0.6023,
-      "step": 52790
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6175760633872934e-05,
-      "loss": 0.681,
-      "step": 52800
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.614848555102487e-05,
-      "loss": 0.6838,
-      "step": 52810
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.61212104681768e-05,
-      "loss": 0.6514,
-      "step": 52820
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6093935385328736e-05,
-      "loss": 0.6618,
-      "step": 52830
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.606666030248067e-05,
-      "loss": 0.6122,
-      "step": 52840
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.603938521963261e-05,
-      "loss": 0.6355,
-      "step": 52850
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.6012110136784545e-05,
-      "loss": 0.6922,
-      "step": 52860
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.598483505393648e-05,
-      "loss": 0.603,
-      "step": 52870
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.595755997108841e-05,
-      "loss": 0.6507,
-      "step": 52880
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.5930284888240346e-05,
-      "loss": 0.7026,
-      "step": 52890
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.590300980539228e-05,
-      "loss": 0.6915,
-      "step": 52900
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.587573472254423e-05,
-      "loss": 0.6399,
-      "step": 52910
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.584845963969616e-05,
-      "loss": 0.6402,
-      "step": 52920
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.5821184556848096e-05,
-      "loss": 0.7069,
-      "step": 52930
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.579390947400003e-05,
-      "loss": 0.6533,
-      "step": 52940
-    },
-    {
-      "epoch": 2.16,
-      "learning_rate": 5.5766634391151963e-05,
-      "loss": 0.6416,
-      "step": 52950
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5739359308303904e-05,
-      "loss": 0.6567,
-      "step": 52960
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.571208422545584e-05,
-      "loss": 0.6761,
-      "step": 52970
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.568480914260777e-05,
-      "loss": 0.5144,
-      "step": 52980
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5657534059759706e-05,
-      "loss": 0.5921,
-      "step": 52990
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.563025897691164e-05,
-      "loss": 0.695,
-      "step": 53000
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5602983894063574e-05,
-      "loss": 0.6035,
-      "step": 53010
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.557570881121552e-05,
-      "loss": 0.5844,
-      "step": 53020
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5548433728367455e-05,
-      "loss": 0.5773,
-      "step": 53030
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.552115864551939e-05,
-      "loss": 0.5767,
-      "step": 53040
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.549388356267132e-05,
-      "loss": 0.6735,
-      "step": 53050
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.546660847982326e-05,
-      "loss": 0.6382,
-      "step": 53060
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.543933339697519e-05,
-      "loss": 0.6319,
-      "step": 53070
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.541205831412714e-05,
-      "loss": 0.6742,
-      "step": 53080
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.538478323127907e-05,
-      "loss": 0.6812,
-      "step": 53090
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5357508148431006e-05,
-      "loss": 0.65,
-      "step": 53100
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.533023306558294e-05,
-      "loss": 0.6305,
-      "step": 53110
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5302957982734874e-05,
-      "loss": 0.7118,
-      "step": 53120
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5275682899886815e-05,
-      "loss": 0.7322,
-      "step": 53130
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.524840781703875e-05,
-      "loss": 0.6872,
-      "step": 53140
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.522113273419068e-05,
-      "loss": 0.6246,
-      "step": 53150
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5193857651342616e-05,
-      "loss": 0.6961,
-      "step": 53160
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.516658256849455e-05,
-      "loss": 0.5834,
-      "step": 53170
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.5139307485646484e-05,
-      "loss": 0.6991,
-      "step": 53180
-    },
-    {
-      "epoch": 2.17,
-      "learning_rate": 5.511203240279843e-05,
-      "loss": 0.6123,
-      "step": 53190
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.5084757319950366e-05,
-      "loss": 0.6613,
-      "step": 53200
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.50574822371023e-05,
-      "loss": 0.6379,
-      "step": 53210
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.5030207154254233e-05,
-      "loss": 0.5527,
-      "step": 53220
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.500293207140617e-05,
-      "loss": 0.5663,
-      "step": 53230
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.497565698855811e-05,
-      "loss": 0.7615,
-      "step": 53240
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.494838190571004e-05,
-      "loss": 0.7086,
-      "step": 53250
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4921106822861976e-05,
-      "loss": 0.63,
-      "step": 53260
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.489383174001391e-05,
-      "loss": 0.6101,
-      "step": 53270
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4866556657165844e-05,
-      "loss": 0.7082,
-      "step": 53280
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.483928157431778e-05,
-      "loss": 0.6777,
-      "step": 53290
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4812006491469725e-05,
-      "loss": 0.553,
-      "step": 53300
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.478473140862166e-05,
-      "loss": 0.6438,
-      "step": 53310
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.475745632577359e-05,
-      "loss": 0.6839,
-      "step": 53320
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.473018124292553e-05,
-      "loss": 0.6611,
-      "step": 53330
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.470290616007746e-05,
-      "loss": 0.642,
-      "step": 53340
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4675631077229395e-05,
-      "loss": 0.6196,
-      "step": 53350
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.464835599438134e-05,
-      "loss": 0.6914,
-      "step": 53360
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4621080911533276e-05,
-      "loss": 0.6794,
-      "step": 53370
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.459380582868521e-05,
-      "loss": 0.6215,
-      "step": 53380
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4566530745837144e-05,
-      "loss": 0.6447,
-      "step": 53390
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.453925566298908e-05,
-      "loss": 0.6659,
-      "step": 53400
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.451198058014102e-05,
-      "loss": 0.6272,
-      "step": 53410
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.448470549729295e-05,
-      "loss": 0.653,
-      "step": 53420
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.4457430414444886e-05,
-      "loss": 0.7104,
-      "step": 53430
-    },
-    {
-      "epoch": 2.18,
-      "learning_rate": 5.443015533159682e-05,
-      "loss": 0.6788,
-      "step": 53440
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4402880248748754e-05,
-      "loss": 0.7208,
-      "step": 53450
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.437560516590069e-05,
-      "loss": 0.6779,
-      "step": 53460
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4348330083052636e-05,
-      "loss": 0.6718,
-      "step": 53470
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.432105500020457e-05,
-      "loss": 0.6513,
-      "step": 53480
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4293779917356503e-05,
-      "loss": 0.6564,
-      "step": 53490
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.426650483450844e-05,
-      "loss": 0.6882,
-      "step": 53500
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.423922975166037e-05,
-      "loss": 0.6974,
-      "step": 53510
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4211954668812305e-05,
-      "loss": 0.6944,
-      "step": 53520
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4184679585964246e-05,
-      "loss": 0.6426,
-      "step": 53530
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.415740450311618e-05,
-      "loss": 0.6465,
-      "step": 53540
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.4130129420268114e-05,
-      "loss": 0.6783,
-      "step": 53550
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.410285433742005e-05,
-      "loss": 0.6181,
-      "step": 53560
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.407557925457198e-05,
-      "loss": 0.7296,
-      "step": 53570
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.404830417172393e-05,
-      "loss": 0.604,
-      "step": 53580
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.402102908887586e-05,
-      "loss": 0.5773,
-      "step": 53590
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.39937540060278e-05,
-      "loss": 0.6159,
-      "step": 53600
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.396647892317973e-05,
-      "loss": 0.5938,
-      "step": 53610
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.3939203840331665e-05,
-      "loss": 0.6699,
-      "step": 53620
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.39119287574836e-05,
-      "loss": 0.6155,
-      "step": 53630
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.3884653674635546e-05,
-      "loss": 0.5875,
-      "step": 53640
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.385737859178748e-05,
-      "loss": 0.6621,
-      "step": 53650
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.3830103508939414e-05,
-      "loss": 0.7594,
-      "step": 53660
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.380282842609135e-05,
-      "loss": 0.6048,
-      "step": 53670
-    },
-    {
-      "epoch": 2.19,
-      "learning_rate": 5.377555334324328e-05,
-      "loss": 0.6488,
-      "step": 53680
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.374827826039522e-05,
-      "loss": 0.6581,
-      "step": 53690
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3721003177547156e-05,
-      "loss": 0.5905,
-      "step": 53700
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.369372809469909e-05,
-      "loss": 0.546,
-      "step": 53710
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3666453011851024e-05,
-      "loss": 0.6782,
-      "step": 53720
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.363917792900296e-05,
-      "loss": 0.7064,
-      "step": 53730
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.361190284615489e-05,
-      "loss": 0.658,
-      "step": 53740
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.358462776330684e-05,
-      "loss": 0.6697,
-      "step": 53750
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3557352680458774e-05,
-      "loss": 0.6969,
-      "step": 53760
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.353007759761071e-05,
-      "loss": 0.6641,
-      "step": 53770
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.350280251476264e-05,
-      "loss": 0.6765,
-      "step": 53780
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3475527431914575e-05,
-      "loss": 0.6975,
-      "step": 53790
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.344825234906651e-05,
-      "loss": 0.6559,
-      "step": 53800
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.342097726621845e-05,
-      "loss": 0.6897,
-      "step": 53810
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3393702183370384e-05,
-      "loss": 0.6459,
-      "step": 53820
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.336642710052232e-05,
-      "loss": 0.5861,
-      "step": 53830
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.333915201767425e-05,
-      "loss": 0.5844,
-      "step": 53840
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3311876934826186e-05,
-      "loss": 0.6299,
-      "step": 53850
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.328460185197813e-05,
-      "loss": 0.6627,
-      "step": 53860
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.325732676913007e-05,
-      "loss": 0.6106,
-      "step": 53870
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3230051686282e-05,
-      "loss": 0.6636,
-      "step": 53880
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.320550411171874e-05,
-      "loss": 0.6522,
-      "step": 53890
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3178229028870675e-05,
-      "loss": 0.524,
-      "step": 53900
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3150953946022616e-05,
-      "loss": 0.5848,
-      "step": 53910
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.312367886317455e-05,
-      "loss": 0.6939,
-      "step": 53920
-    },
-    {
-      "epoch": 2.2,
-      "learning_rate": 5.3096403780326484e-05,
-      "loss": 0.7098,
-      "step": 53930
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.306912869747842e-05,
-      "loss": 0.6961,
-      "step": 53940
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.304185361463035e-05,
-      "loss": 0.7117,
-      "step": 53950
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.3014578531782286e-05,
-      "loss": 0.6528,
-      "step": 53960
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.298730344893423e-05,
-      "loss": 0.6267,
-      "step": 53970
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.296002836608617e-05,
-      "loss": 0.6673,
-      "step": 53980
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.29327532832381e-05,
-      "loss": 0.666,
-      "step": 53990
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2905478200390035e-05,
-      "loss": 0.6683,
-      "step": 54000
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.287820311754197e-05,
-      "loss": 0.6995,
-      "step": 54010
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.28509280346939e-05,
-      "loss": 0.6828,
-      "step": 54020
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.282365295184585e-05,
-      "loss": 0.5543,
-      "step": 54030
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2796377868997784e-05,
-      "loss": 0.6741,
-      "step": 54040
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.276910278614972e-05,
-      "loss": 0.6411,
-      "step": 54050
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.274182770330165e-05,
-      "loss": 0.6381,
-      "step": 54060
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2714552620453586e-05,
-      "loss": 0.6018,
-      "step": 54070
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2687277537605527e-05,
-      "loss": 0.76,
-      "step": 54080
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.266000245475746e-05,
-      "loss": 0.664,
-      "step": 54090
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2632727371909394e-05,
-      "loss": 0.6837,
-      "step": 54100
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.260545228906133e-05,
-      "loss": 0.723,
-      "step": 54110
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.257817720621326e-05,
-      "loss": 0.6023,
-      "step": 54120
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2550902123365196e-05,
-      "loss": 0.5844,
-      "step": 54130
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2523627040517144e-05,
-      "loss": 0.6723,
-      "step": 54140
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.249635195766908e-05,
-      "loss": 0.656,
-      "step": 54150
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.246907687482101e-05,
-      "loss": 0.7019,
-      "step": 54160
-    },
-    {
-      "epoch": 2.21,
-      "learning_rate": 5.2441801791972945e-05,
-      "loss": 0.7086,
-      "step": 54170
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.241452670912488e-05,
-      "loss": 0.6901,
-      "step": 54180
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.238725162627681e-05,
-      "loss": 0.6372,
-      "step": 54190
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.2359976543428754e-05,
-      "loss": 0.5926,
-      "step": 54200
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.233270146058069e-05,
-      "loss": 0.5856,
-      "step": 54210
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.230542637773262e-05,
-      "loss": 0.6823,
-      "step": 54220
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.2278151294884556e-05,
-      "loss": 0.6832,
-      "step": 54230
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.225087621203649e-05,
-      "loss": 0.6312,
-      "step": 54240
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.222360112918844e-05,
-      "loss": 0.6098,
-      "step": 54250
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.219632604634037e-05,
-      "loss": 0.6703,
-      "step": 54260
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.2169050963492305e-05,
-      "loss": 0.6042,
-      "step": 54270
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.214177588064424e-05,
-      "loss": 0.717,
-      "step": 54280
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.211450079779617e-05,
-      "loss": 0.6637,
-      "step": 54290
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.208722571494811e-05,
-      "loss": 0.6475,
-      "step": 54300
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.2059950632100054e-05,
-      "loss": 0.629,
-      "step": 54310
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.203267554925199e-05,
-      "loss": 0.6972,
-      "step": 54320
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.200540046640392e-05,
-      "loss": 0.7201,
-      "step": 54330
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.1978125383555856e-05,
-      "loss": 0.6211,
-      "step": 54340
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.195085030070779e-05,
-      "loss": 0.649,
-      "step": 54350
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.1923575217859724e-05,
-      "loss": 0.6827,
-      "step": 54360
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.1896300135011664e-05,
-      "loss": 0.6493,
-      "step": 54370
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.18690250521636e-05,
-      "loss": 0.6171,
-      "step": 54380
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.184174996931553e-05,
-      "loss": 0.7338,
-      "step": 54390
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.1814474886467466e-05,
-      "loss": 0.6402,
-      "step": 54400
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.17871998036194e-05,
-      "loss": 0.5785,
-      "step": 54410
-    },
-    {
-      "epoch": 2.22,
-      "learning_rate": 5.175992472077135e-05,
-      "loss": 0.7055,
-      "step": 54420
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.173264963792328e-05,
-      "loss": 0.6271,
-      "step": 54430
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1705374555075215e-05,
-      "loss": 0.7365,
-      "step": 54440
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.167809947222715e-05,
-      "loss": 0.7013,
-      "step": 54450
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.165082438937908e-05,
-      "loss": 0.6061,
-      "step": 54460
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.162354930653102e-05,
-      "loss": 0.7386,
-      "step": 54470
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.159627422368296e-05,
-      "loss": 0.6492,
-      "step": 54480
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.156899914083489e-05,
-      "loss": 0.6097,
-      "step": 54490
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1541724057986826e-05,
-      "loss": 0.6014,
-      "step": 54500
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.151444897513876e-05,
-      "loss": 0.6701,
-      "step": 54510
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1487173892290694e-05,
-      "loss": 0.6462,
-      "step": 54520
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.145989880944264e-05,
-      "loss": 0.6504,
-      "step": 54530
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1432623726594575e-05,
-      "loss": 0.6207,
-      "step": 54540
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.140534864374651e-05,
-      "loss": 0.7835,
-      "step": 54550
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.137807356089844e-05,
-      "loss": 0.5942,
-      "step": 54560
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.135079847805038e-05,
-      "loss": 0.6044,
-      "step": 54570
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.132352339520231e-05,
-      "loss": 0.6703,
-      "step": 54580
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.129624831235426e-05,
-      "loss": 0.6506,
-      "step": 54590
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.126897322950619e-05,
-      "loss": 0.585,
-      "step": 54600
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1241698146658126e-05,
-      "loss": 0.6071,
-      "step": 54610
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.121442306381006e-05,
-      "loss": 0.6469,
-      "step": 54620
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.1187147980961994e-05,
-      "loss": 0.8301,
-      "step": 54630
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.115987289811393e-05,
-      "loss": 0.5998,
-      "step": 54640
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.113259781526587e-05,
-      "loss": 0.6679,
-      "step": 54650
-    },
-    {
-      "epoch": 2.23,
-      "learning_rate": 5.11053227324178e-05,
-      "loss": 0.696,
-      "step": 54660
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.1078047649569736e-05,
-      "loss": 0.6355,
-      "step": 54670
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.105077256672167e-05,
-      "loss": 0.4961,
-      "step": 54680
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.1023497483873604e-05,
-      "loss": 0.6883,
-      "step": 54690
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.099622240102555e-05,
-      "loss": 0.6756,
-      "step": 54700
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0968947318177486e-05,
-      "loss": 0.5644,
-      "step": 54710
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.094167223532942e-05,
-      "loss": 0.6401,
-      "step": 54720
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.091439715248135e-05,
-      "loss": 0.6609,
-      "step": 54730
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.088712206963329e-05,
-      "loss": 0.7136,
-      "step": 54740
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.085984698678522e-05,
-      "loss": 0.6881,
-      "step": 54750
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.083257190393716e-05,
-      "loss": 0.6228,
-      "step": 54760
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0805296821089096e-05,
-      "loss": 0.6385,
-      "step": 54770
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.077802173824103e-05,
-      "loss": 0.6903,
-      "step": 54780
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0750746655392964e-05,
-      "loss": 0.6356,
-      "step": 54790
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.07234715725449e-05,
-      "loss": 0.7179,
-      "step": 54800
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.069619648969683e-05,
-      "loss": 0.7769,
-      "step": 54810
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.066892140684878e-05,
-      "loss": 0.6496,
-      "step": 54820
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.064164632400071e-05,
-      "loss": 0.6091,
-      "step": 54830
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.061437124115265e-05,
-      "loss": 0.6157,
-      "step": 54840
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.058709615830458e-05,
-      "loss": 0.7087,
-      "step": 54850
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0559821075456515e-05,
-      "loss": 0.5963,
-      "step": 54860
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.053254599260846e-05,
-      "loss": 0.6103,
-      "step": 54870
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0505270909760396e-05,
-      "loss": 0.5736,
-      "step": 54880
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.047799582691233e-05,
-      "loss": 0.6335,
-      "step": 54890
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.0450720744064264e-05,
-      "loss": 0.6229,
-      "step": 54900
-    },
-    {
-      "epoch": 2.24,
-      "learning_rate": 5.04234456612162e-05,
-      "loss": 0.6528,
-      "step": 54910
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.039617057836813e-05,
-      "loss": 0.7096,
-      "step": 54920
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.036889549552007e-05,
-      "loss": 0.672,
-      "step": 54930
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0341620412672006e-05,
-      "loss": 0.5543,
-      "step": 54940
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.031434532982394e-05,
-      "loss": 0.6711,
-      "step": 54950
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0287070246975874e-05,
-      "loss": 0.6687,
-      "step": 54960
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.025979516412781e-05,
-      "loss": 0.6505,
-      "step": 54970
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0232520081279756e-05,
-      "loss": 0.6545,
-      "step": 54980
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.020524499843169e-05,
-      "loss": 0.719,
-      "step": 54990
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0177969915583623e-05,
-      "loss": 0.6139,
-      "step": 55000
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.015069483273556e-05,
-      "loss": 0.618,
-      "step": 55010
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.012341974988749e-05,
-      "loss": 0.6383,
-      "step": 55020
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0096144667039425e-05,
-      "loss": 0.6627,
-      "step": 55030
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0068869584191366e-05,
-      "loss": 0.6815,
-      "step": 55040
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.00415945013433e-05,
-      "loss": 0.5368,
-      "step": 55050
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 5.0014319418495234e-05,
-      "loss": 0.646,
-      "step": 55060
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.998704433564717e-05,
-      "loss": 0.629,
-      "step": 55070
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.995976925279911e-05,
-      "loss": 0.6345,
-      "step": 55080
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.993249416995104e-05,
-      "loss": 0.6539,
-      "step": 55090
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.9905219087102976e-05,
-      "loss": 0.5375,
-      "step": 55100
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.987794400425492e-05,
-      "loss": 0.676,
-      "step": 55110
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.985066892140685e-05,
-      "loss": 0.7084,
-      "step": 55120
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.9823393838558785e-05,
-      "loss": 0.6064,
-      "step": 55130
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.9796118755710725e-05,
-      "loss": 0.6094,
-      "step": 55140
-    },
-    {
-      "epoch": 2.25,
-      "learning_rate": 4.976884367286266e-05,
-      "loss": 0.6697,
-      "step": 55150
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.974156859001459e-05,
-      "loss": 0.6122,
-      "step": 55160
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9714293507166534e-05,
-      "loss": 0.5444,
-      "step": 55170
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.968701842431847e-05,
-      "loss": 0.6236,
-      "step": 55180
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.96597433414704e-05,
-      "loss": 0.5987,
-      "step": 55190
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9632468258622336e-05,
-      "loss": 0.6928,
-      "step": 55200
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.960519317577427e-05,
-      "loss": 0.6263,
-      "step": 55210
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.957791809292621e-05,
-      "loss": 0.6466,
-      "step": 55220
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9550643010078144e-05,
-      "loss": 0.6612,
-      "step": 55230
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.952336792723008e-05,
-      "loss": 0.6819,
-      "step": 55240
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.949609284438202e-05,
-      "loss": 0.6043,
-      "step": 55250
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.946881776153395e-05,
-      "loss": 0.7024,
-      "step": 55260
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.944154267868589e-05,
-      "loss": 0.7214,
-      "step": 55270
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.941426759583783e-05,
-      "loss": 0.6456,
-      "step": 55280
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.938699251298976e-05,
-      "loss": 0.6001,
-      "step": 55290
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9359717430141695e-05,
-      "loss": 0.6776,
-      "step": 55300
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9332442347293636e-05,
-      "loss": 0.5742,
-      "step": 55310
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.930516726444557e-05,
-      "loss": 0.5878,
-      "step": 55320
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9277892181597504e-05,
-      "loss": 0.7144,
-      "step": 55330
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.925061709874944e-05,
-      "loss": 0.7325,
-      "step": 55340
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.922334201590137e-05,
-      "loss": 0.6884,
-      "step": 55350
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.919606693305331e-05,
-      "loss": 0.6512,
-      "step": 55360
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9168791850205246e-05,
-      "loss": 0.7047,
-      "step": 55370
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.914151676735718e-05,
-      "loss": 0.6745,
-      "step": 55380
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.911424168450912e-05,
-      "loss": 0.706,
-      "step": 55390
-    },
-    {
-      "epoch": 2.26,
-      "learning_rate": 4.9086966601661055e-05,
-      "loss": 0.6711,
-      "step": 55400
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.905969151881299e-05,
-      "loss": 0.6822,
-      "step": 55410
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.903241643596493e-05,
-      "loss": 0.7758,
-      "step": 55420
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.900514135311686e-05,
-      "loss": 0.5713,
-      "step": 55430
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.89778662702688e-05,
-      "loss": 0.7544,
-      "step": 55440
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.895059118742074e-05,
-      "loss": 0.7122,
-      "step": 55450
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.892331610457267e-05,
-      "loss": 0.783,
-      "step": 55460
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.8896041021724606e-05,
-      "loss": 0.6703,
-      "step": 55470
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.886876593887654e-05,
-      "loss": 0.6324,
-      "step": 55480
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.8841490856028474e-05,
-      "loss": 0.5741,
-      "step": 55490
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.881421577318041e-05,
-      "loss": 0.6657,
-      "step": 55500
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.878694069033235e-05,
-      "loss": 0.6444,
-      "step": 55510
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.875966560748428e-05,
-      "loss": 0.6051,
-      "step": 55520
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.873239052463622e-05,
-      "loss": 0.6473,
-      "step": 55530
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.870511544178816e-05,
-      "loss": 0.6937,
-      "step": 55540
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.867784035894009e-05,
-      "loss": 0.6101,
-      "step": 55550
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.865056527609203e-05,
-      "loss": 0.7703,
-      "step": 55560
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.8623290193243965e-05,
-      "loss": 0.6826,
-      "step": 55570
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.85960151103959e-05,
-      "loss": 0.5741,
-      "step": 55580
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.856874002754784e-05,
-      "loss": 0.6989,
-      "step": 55590
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.8541464944699774e-05,
-      "loss": 0.5958,
-      "step": 55600
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.851418986185171e-05,
-      "loss": 0.6595,
-      "step": 55610
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.848691477900364e-05,
-      "loss": 0.6205,
-      "step": 55620
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.8459639696155576e-05,
-      "loss": 0.6045,
-      "step": 55630
-    },
-    {
-      "epoch": 2.27,
-      "learning_rate": 4.843236461330751e-05,
-      "loss": 0.7258,
-      "step": 55640
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.840508953045945e-05,
-      "loss": 0.6506,
-      "step": 55650
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.8377814447611384e-05,
-      "loss": 0.6069,
-      "step": 55660
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.8350539364763325e-05,
-      "loss": 0.6857,
-      "step": 55670
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.832326428191526e-05,
-      "loss": 0.6801,
-      "step": 55680
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.829598919906719e-05,
-      "loss": 0.6872,
-      "step": 55690
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.826871411621913e-05,
-      "loss": 0.6157,
-      "step": 55700
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.824143903337107e-05,
-      "loss": 0.6354,
-      "step": 55710
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.8214163950523e-05,
-      "loss": 0.5811,
-      "step": 55720
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.818688886767494e-05,
-      "loss": 0.596,
-      "step": 55730
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.8159613784826876e-05,
-      "loss": 0.5983,
-      "step": 55740
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.813233870197881e-05,
-      "loss": 0.6806,
-      "step": 55750
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.8105063619130744e-05,
-      "loss": 0.6458,
-      "step": 55760
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.807778853628268e-05,
-      "loss": 0.6554,
-      "step": 55770
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.805051345343461e-05,
-      "loss": 0.6164,
-      "step": 55780
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.802323837058655e-05,
-      "loss": 0.5548,
-      "step": 55790
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.7995963287738486e-05,
-      "loss": 0.6089,
-      "step": 55800
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.796868820489043e-05,
-      "loss": 0.6319,
-      "step": 55810
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.794141312204236e-05,
-      "loss": 0.6867,
-      "step": 55820
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.7914138039194295e-05,
-      "loss": 0.6078,
-      "step": 55830
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.7886862956346235e-05,
-      "loss": 0.6191,
-      "step": 55840
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.785958787349817e-05,
-      "loss": 0.6547,
-      "step": 55850
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.78323127906501e-05,
-      "loss": 0.6805,
-      "step": 55860
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.7805037707802044e-05,
-      "loss": 0.6595,
-      "step": 55870
-    },
-    {
-      "epoch": 2.28,
-      "learning_rate": 4.777776262495398e-05,
-      "loss": 0.605,
-      "step": 55880
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.775048754210591e-05,
-      "loss": 0.5602,
-      "step": 55890
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7723212459257846e-05,
-      "loss": 0.5206,
-      "step": 55900
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.769593737640978e-05,
-      "loss": 0.5679,
-      "step": 55910
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7668662293561713e-05,
-      "loss": 0.7142,
-      "step": 55920
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7641387210713654e-05,
-      "loss": 0.6414,
-      "step": 55930
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.761411212786559e-05,
-      "loss": 0.6936,
-      "step": 55940
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.758683704501752e-05,
-      "loss": 0.5745,
-      "step": 55950
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.755956196216946e-05,
-      "loss": 0.6496,
-      "step": 55960
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7532286879321397e-05,
-      "loss": 0.6217,
-      "step": 55970
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.750501179647334e-05,
-      "loss": 0.6802,
-      "step": 55980
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.747773671362527e-05,
-      "loss": 0.7229,
-      "step": 55990
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7450461630777205e-05,
-      "loss": 0.6352,
-      "step": 56000
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7423186547929146e-05,
-      "loss": 0.6274,
-      "step": 56010
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.739591146508108e-05,
-      "loss": 0.6587,
-      "step": 56020
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7368636382233014e-05,
-      "loss": 0.7219,
-      "step": 56030
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.734136129938495e-05,
-      "loss": 0.7009,
-      "step": 56040
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.731408621653688e-05,
-      "loss": 0.5842,
-      "step": 56050
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7286811133688815e-05,
-      "loss": 0.6158,
-      "step": 56060
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7259536050840756e-05,
-      "loss": 0.6423,
-      "step": 56070
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.723226096799269e-05,
-      "loss": 0.6219,
-      "step": 56080
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7204985885144624e-05,
-      "loss": 0.7085,
-      "step": 56090
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.7177710802296565e-05,
-      "loss": 0.7836,
-      "step": 56100
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.71504357194485e-05,
-      "loss": 0.6497,
-      "step": 56110
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.712316063660044e-05,
-      "loss": 0.6561,
-      "step": 56120
-    },
-    {
-      "epoch": 2.29,
-      "learning_rate": 4.709588555375237e-05,
-      "loss": 0.58,
-      "step": 56130
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.706861047090431e-05,
-      "loss": 0.7341,
-      "step": 56140
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.704133538805625e-05,
-      "loss": 0.6339,
-      "step": 56150
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.701406030520818e-05,
-      "loss": 0.7278,
-      "step": 56160
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6986785222360116e-05,
-      "loss": 0.6963,
-      "step": 56170
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.695951013951205e-05,
-      "loss": 0.6374,
-      "step": 56180
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6932235056663983e-05,
-      "loss": 0.6814,
-      "step": 56190
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.690495997381592e-05,
-      "loss": 0.6012,
-      "step": 56200
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.687768489096786e-05,
-      "loss": 0.6609,
-      "step": 56210
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.685040980811979e-05,
-      "loss": 0.5758,
-      "step": 56220
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6823134725271726e-05,
-      "loss": 0.5915,
-      "step": 56230
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.679585964242367e-05,
-      "loss": 0.6657,
-      "step": 56240
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.67685845595756e-05,
-      "loss": 0.631,
-      "step": 56250
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.674130947672754e-05,
-      "loss": 0.7707,
-      "step": 56260
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6714034393879475e-05,
-      "loss": 0.7783,
-      "step": 56270
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.668675931103141e-05,
-      "loss": 0.6075,
-      "step": 56280
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.665948422818335e-05,
-      "loss": 0.6783,
-      "step": 56290
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6632209145335284e-05,
-      "loss": 0.5936,
-      "step": 56300
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.660493406248722e-05,
-      "loss": 0.6377,
-      "step": 56310
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.657765897963915e-05,
-      "loss": 0.7079,
-      "step": 56320
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6550383896791085e-05,
-      "loss": 0.5787,
-      "step": 56330
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.652310881394302e-05,
-      "loss": 0.7223,
-      "step": 56340
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.649583373109496e-05,
-      "loss": 0.6168,
-      "step": 56350
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.6468558648246894e-05,
-      "loss": 0.7932,
-      "step": 56360
-    },
-    {
-      "epoch": 2.3,
-      "learning_rate": 4.644128356539883e-05,
-      "loss": 0.5589,
-      "step": 56370
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.641400848255077e-05,
-      "loss": 0.7058,
-      "step": 56380
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.63867333997027e-05,
-      "loss": 0.628,
-      "step": 56390
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.6359458316854636e-05,
-      "loss": 0.621,
-      "step": 56400
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.633218323400658e-05,
-      "loss": 0.6572,
-      "step": 56410
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.630490815115851e-05,
-      "loss": 0.6177,
-      "step": 56420
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.627763306831045e-05,
-      "loss": 0.6077,
-      "step": 56430
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.6250357985462386e-05,
-      "loss": 0.6298,
-      "step": 56440
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.622308290261432e-05,
-      "loss": 0.7119,
-      "step": 56450
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.6195807819766254e-05,
-      "loss": 0.66,
-      "step": 56460
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.616853273691819e-05,
-      "loss": 0.5386,
-      "step": 56470
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.614125765407012e-05,
-      "loss": 0.7435,
-      "step": 56480
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.611398257122206e-05,
-      "loss": 0.6911,
-      "step": 56490
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.6086707488373996e-05,
-      "loss": 0.6407,
-      "step": 56500
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.605943240552593e-05,
-      "loss": 0.7453,
-      "step": 56510
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.603215732267787e-05,
-      "loss": 0.6717,
-      "step": 56520
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.6004882239829805e-05,
-      "loss": 0.6713,
-      "step": 56530
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.597760715698174e-05,
-      "loss": 0.6427,
-      "step": 56540
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.595033207413368e-05,
-      "loss": 0.6427,
-      "step": 56550
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.592305699128561e-05,
-      "loss": 0.6872,
-      "step": 56560
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.5895781908437554e-05,
-      "loss": 0.6196,
-      "step": 56570
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.586850682558949e-05,
-      "loss": 0.71,
-      "step": 56580
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.584123174274142e-05,
-      "loss": 0.7439,
-      "step": 56590
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.581395665989336e-05,
-      "loss": 0.7055,
-      "step": 56600
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.578668157704529e-05,
-      "loss": 0.5723,
-      "step": 56610
-    },
-    {
-      "epoch": 2.31,
-      "learning_rate": 4.575940649419722e-05,
-      "loss": 0.6749,
-      "step": 56620
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5732131411349164e-05,
-      "loss": 0.6512,
-      "step": 56630
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.57048563285011e-05,
-      "loss": 0.6219,
-      "step": 56640
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.567758124565303e-05,
-      "loss": 0.6403,
-      "step": 56650
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.565030616280497e-05,
-      "loss": 0.6027,
-      "step": 56660
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5623031079956907e-05,
-      "loss": 0.7226,
-      "step": 56670
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.559575599710884e-05,
-      "loss": 0.563,
-      "step": 56680
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.556848091426078e-05,
-      "loss": 0.6712,
-      "step": 56690
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5541205831412715e-05,
-      "loss": 0.7325,
-      "step": 56700
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.551393074856465e-05,
-      "loss": 0.7122,
-      "step": 56710
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.548665566571659e-05,
-      "loss": 0.6246,
-      "step": 56720
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5459380582868524e-05,
-      "loss": 0.6516,
-      "step": 56730
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5432105500020464e-05,
-      "loss": 0.6932,
-      "step": 56740
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.54048304171724e-05,
-      "loss": 0.7523,
-      "step": 56750
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5377555334324325e-05,
-      "loss": 0.6205,
-      "step": 56760
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5350280251476266e-05,
-      "loss": 0.643,
-      "step": 56770
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.53230051686282e-05,
-      "loss": 0.6051,
-      "step": 56780
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5295730085780134e-05,
-      "loss": 0.8165,
-      "step": 56790
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.5268455002932075e-05,
-      "loss": 0.7017,
-      "step": 56800
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.524117992008401e-05,
-      "loss": 0.7156,
-      "step": 56810
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.521390483723594e-05,
-      "loss": 0.5784,
-      "step": 56820
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.518662975438788e-05,
-      "loss": 0.6922,
-      "step": 56830
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.515935467153982e-05,
-      "loss": 0.6128,
-      "step": 56840
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.513207958869175e-05,
-      "loss": 0.6672,
-      "step": 56850
-    },
-    {
-      "epoch": 2.32,
-      "learning_rate": 4.510480450584369e-05,
-      "loss": 0.6645,
-      "step": 56860
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.5077529422995626e-05,
-      "loss": 0.5864,
-      "step": 56870
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.5050254340147566e-05,
-      "loss": 0.7242,
-      "step": 56880
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.50229792572995e-05,
-      "loss": 0.5983,
-      "step": 56890
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.499570417445143e-05,
-      "loss": 0.7503,
-      "step": 56900
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.496842909160337e-05,
-      "loss": 0.5977,
-      "step": 56910
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.49411540087553e-05,
-      "loss": 0.6426,
-      "step": 56920
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4913878925907236e-05,
-      "loss": 0.6637,
-      "step": 56930
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4886603843059177e-05,
-      "loss": 0.6486,
-      "step": 56940
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.485932876021111e-05,
-      "loss": 0.647,
-      "step": 56950
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4832053677363044e-05,
-      "loss": 0.7786,
-      "step": 56960
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4804778594514985e-05,
-      "loss": 0.6909,
-      "step": 56970
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.477750351166692e-05,
-      "loss": 0.5788,
-      "step": 56980
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.475022842881885e-05,
-      "loss": 0.6425,
-      "step": 56990
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4722953345970794e-05,
-      "loss": 0.6901,
-      "step": 57000
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.469567826312273e-05,
-      "loss": 0.6338,
-      "step": 57010
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.466840318027467e-05,
-      "loss": 0.6694,
-      "step": 57020
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.46411280974266e-05,
-      "loss": 0.641,
-      "step": 57030
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4613853014578536e-05,
-      "loss": 0.5272,
-      "step": 57040
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.458657793173047e-05,
-      "loss": 0.7443,
-      "step": 57050
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4559302848882404e-05,
-      "loss": 0.6036,
-      "step": 57060
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.453202776603434e-05,
-      "loss": 0.6426,
-      "step": 57070
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.450475268318628e-05,
-      "loss": 0.644,
-      "step": 57080
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.447747760033821e-05,
-      "loss": 0.732,
-      "step": 57090
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.4450202517490146e-05,
-      "loss": 0.5718,
-      "step": 57100
-    },
-    {
-      "epoch": 2.33,
-      "learning_rate": 4.442292743464209e-05,
-      "loss": 0.6609,
-      "step": 57110
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.439565235179402e-05,
-      "loss": 0.6242,
-      "step": 57120
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4368377268945955e-05,
-      "loss": 0.722,
-      "step": 57130
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4341102186097896e-05,
-      "loss": 0.6287,
-      "step": 57140
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.431382710324983e-05,
-      "loss": 0.6713,
-      "step": 57150
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4286552020401763e-05,
-      "loss": 0.6018,
-      "step": 57160
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4259276937553704e-05,
-      "loss": 0.6297,
-      "step": 57170
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.423200185470564e-05,
-      "loss": 0.6829,
-      "step": 57180
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.420472677185757e-05,
-      "loss": 0.7209,
-      "step": 57190
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4177451689009506e-05,
-      "loss": 0.6617,
-      "step": 57200
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.415017660616144e-05,
-      "loss": 0.6456,
-      "step": 57210
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.412290152331338e-05,
-      "loss": 0.7179,
-      "step": 57220
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.4095626440465314e-05,
-      "loss": 0.7816,
-      "step": 57230
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.406835135761725e-05,
-      "loss": 0.7214,
-      "step": 57240
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.404107627476919e-05,
-      "loss": 0.7261,
-      "step": 57250
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.401380119192112e-05,
-      "loss": 0.6374,
-      "step": 57260
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.398652610907306e-05,
-      "loss": 0.6522,
-      "step": 57270
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.3959251026225e-05,
-      "loss": 0.7762,
-      "step": 57280
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.393197594337693e-05,
-      "loss": 0.6491,
-      "step": 57290
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.3904700860528865e-05,
-      "loss": 0.761,
-      "step": 57300
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.3877425777680806e-05,
-      "loss": 0.6509,
-      "step": 57310
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.385015069483274e-05,
-      "loss": 0.6016,
-      "step": 57320
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.3822875611984674e-05,
-      "loss": 0.6309,
-      "step": 57330
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.379560052913661e-05,
-      "loss": 0.5895,
-      "step": 57340
-    },
-    {
-      "epoch": 2.34,
-      "learning_rate": 4.376832544628854e-05,
-      "loss": 0.7036,
-      "step": 57350
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.374105036344048e-05,
-      "loss": 0.7181,
-      "step": 57360
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3713775280592416e-05,
-      "loss": 0.5824,
-      "step": 57370
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.368650019774435e-05,
-      "loss": 0.6448,
-      "step": 57380
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.365922511489629e-05,
-      "loss": 0.6351,
-      "step": 57390
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3631950032048225e-05,
-      "loss": 0.604,
-      "step": 57400
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.360467494920016e-05,
-      "loss": 0.6468,
-      "step": 57410
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.35773998663521e-05,
-      "loss": 0.5501,
-      "step": 57420
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3550124783504034e-05,
-      "loss": 0.686,
-      "step": 57430
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.352284970065597e-05,
-      "loss": 0.7157,
-      "step": 57440
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.349557461780791e-05,
-      "loss": 0.7251,
-      "step": 57450
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.346829953495984e-05,
-      "loss": 0.6841,
-      "step": 57460
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3441024452111776e-05,
-      "loss": 0.7651,
-      "step": 57470
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.341374936926371e-05,
-      "loss": 0.691,
-      "step": 57480
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3386474286415644e-05,
-      "loss": 0.6849,
-      "step": 57490
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3359199203567585e-05,
-      "loss": 0.6946,
-      "step": 57500
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.333192412071952e-05,
-      "loss": 0.6778,
-      "step": 57510
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.330464903787145e-05,
-      "loss": 0.6896,
-      "step": 57520
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.327737395502339e-05,
-      "loss": 0.7037,
-      "step": 57530
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.325009887217533e-05,
-      "loss": 0.5638,
-      "step": 57540
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.322282378932726e-05,
-      "loss": 0.6145,
-      "step": 57550
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.31955487064792e-05,
-      "loss": 0.9163,
-      "step": 57560
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3168273623631136e-05,
-      "loss": 0.6595,
-      "step": 57570
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.314099854078307e-05,
-      "loss": 0.6337,
-      "step": 57580
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.311372345793501e-05,
-      "loss": 0.5075,
-      "step": 57590
-    },
-    {
-      "epoch": 2.35,
-      "learning_rate": 4.3086448375086944e-05,
-      "loss": 0.6295,
-      "step": 57600
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.305917329223888e-05,
-      "loss": 0.7273,
-      "step": 57610
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.303189820939081e-05,
-      "loss": 0.6952,
-      "step": 57620
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.3004623126542746e-05,
-      "loss": 0.6109,
-      "step": 57630
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2977348043694687e-05,
-      "loss": 0.6772,
-      "step": 57640
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.295007296084662e-05,
-      "loss": 0.6469,
-      "step": 57650
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2922797877998554e-05,
-      "loss": 0.7108,
-      "step": 57660
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2895522795150495e-05,
-      "loss": 0.8079,
-      "step": 57670
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.286824771230243e-05,
-      "loss": 0.7305,
-      "step": 57680
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.284097262945436e-05,
-      "loss": 0.62,
-      "step": 57690
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2813697546606304e-05,
-      "loss": 0.6291,
-      "step": 57700
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.278642246375824e-05,
-      "loss": 0.6459,
-      "step": 57710
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.275914738091017e-05,
-      "loss": 0.6116,
-      "step": 57720
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.273187229806211e-05,
-      "loss": 0.786,
-      "step": 57730
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2704597215214046e-05,
-      "loss": 0.669,
-      "step": 57740
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.267732213236598e-05,
-      "loss": 0.6299,
-      "step": 57750
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2650047049517914e-05,
-      "loss": 0.6144,
-      "step": 57760
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.262277196666985e-05,
-      "loss": 0.5634,
-      "step": 57770
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.259549688382179e-05,
-      "loss": 0.5495,
-      "step": 57780
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.256822180097372e-05,
-      "loss": 0.6322,
-      "step": 57790
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2540946718125656e-05,
-      "loss": 0.6252,
-      "step": 57800
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.25136716352776e-05,
-      "loss": 0.6741,
-      "step": 57810
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.248639655242953e-05,
-      "loss": 0.6797,
-      "step": 57820
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2459121469581465e-05,
-      "loss": 0.6368,
-      "step": 57830
-    },
-    {
-      "epoch": 2.36,
-      "learning_rate": 4.2431846386733406e-05,
-      "loss": 0.6487,
-      "step": 57840
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.240457130388534e-05,
-      "loss": 0.7539,
-      "step": 57850
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.2377296221037273e-05,
-      "loss": 0.6363,
-      "step": 57860
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.2350021138189214e-05,
-      "loss": 0.6395,
-      "step": 57870
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.232274605534115e-05,
-      "loss": 0.6126,
-      "step": 57880
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.229547097249308e-05,
-      "loss": 0.67,
-      "step": 57890
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.2268195889645016e-05,
-      "loss": 0.6042,
-      "step": 57900
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.224092080679695e-05,
-      "loss": 0.6168,
-      "step": 57910
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.2213645723948884e-05,
-      "loss": 0.6282,
-      "step": 57920
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.2186370641100824e-05,
-      "loss": 0.6092,
-      "step": 57930
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.215909555825276e-05,
-      "loss": 0.7742,
-      "step": 57940
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.21318204754047e-05,
-      "loss": 0.7448,
-      "step": 57950
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.210454539255663e-05,
-      "loss": 0.6597,
-      "step": 57960
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.207727030970857e-05,
-      "loss": 0.5673,
-      "step": 57970
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.204999522686051e-05,
-      "loss": 0.6214,
-      "step": 57980
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.202272014401244e-05,
-      "loss": 0.6668,
-      "step": 57990
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.1995445061164375e-05,
-      "loss": 0.7636,
-      "step": 58000
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.1968169978316316e-05,
-      "loss": 0.7325,
-      "step": 58010
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.194089489546825e-05,
-      "loss": 0.5879,
-      "step": 58020
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.1913619812620184e-05,
-      "loss": 0.6179,
-      "step": 58030
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.188634472977212e-05,
-      "loss": 0.6088,
-      "step": 58040
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.185906964692405e-05,
-      "loss": 0.667,
-      "step": 58050
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.1831794564075986e-05,
-      "loss": 0.7047,
-      "step": 58060
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.1804519481227926e-05,
-      "loss": 0.5817,
-      "step": 58070
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.177724439837986e-05,
-      "loss": 0.6949,
-      "step": 58080
-    },
-    {
-      "epoch": 2.37,
-      "learning_rate": 4.17499693155318e-05,
-      "loss": 0.5716,
-      "step": 58090
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1722694232683735e-05,
-      "loss": 0.5882,
-      "step": 58100
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.169541914983567e-05,
-      "loss": 0.6301,
-      "step": 58110
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.166814406698761e-05,
-      "loss": 0.6474,
-      "step": 58120
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1640868984139543e-05,
-      "loss": 0.6273,
-      "step": 58130
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.161359390129148e-05,
-      "loss": 0.6156,
-      "step": 58140
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.158631881844342e-05,
-      "loss": 0.6175,
-      "step": 58150
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.155904373559535e-05,
-      "loss": 0.7041,
-      "step": 58160
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1531768652747286e-05,
-      "loss": 0.6436,
-      "step": 58170
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.150449356989922e-05,
-      "loss": 0.5208,
-      "step": 58180
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1477218487051154e-05,
-      "loss": 0.6642,
-      "step": 58190
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.144994340420309e-05,
-      "loss": 0.5369,
-      "step": 58200
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.142266832135503e-05,
-      "loss": 0.6116,
-      "step": 58210
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.139539323850696e-05,
-      "loss": 0.6189,
-      "step": 58220
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1368118155658896e-05,
-      "loss": 0.6733,
-      "step": 58230
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.134084307281084e-05,
-      "loss": 0.6367,
-      "step": 58240
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.131356798996277e-05,
-      "loss": 0.7187,
-      "step": 58250
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.128629290711471e-05,
-      "loss": 0.7142,
-      "step": 58260
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1259017824266645e-05,
-      "loss": 0.6311,
-      "step": 58270
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.123174274141858e-05,
-      "loss": 0.5572,
-      "step": 58280
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.120446765857052e-05,
-      "loss": 0.6126,
-      "step": 58290
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1177192575722454e-05,
-      "loss": 0.6654,
-      "step": 58300
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.114991749287439e-05,
-      "loss": 0.7271,
-      "step": 58310
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.112264241002632e-05,
-      "loss": 0.5558,
-      "step": 58320
-    },
-    {
-      "epoch": 2.38,
-      "learning_rate": 4.1095367327178256e-05,
-      "loss": 0.7574,
-      "step": 58330
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.106809224433019e-05,
-      "loss": 0.6831,
-      "step": 58340
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.104081716148213e-05,
-      "loss": 0.6516,
-      "step": 58350
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.1013542078634064e-05,
-      "loss": 0.6179,
-      "step": 58360
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0986266995786e-05,
-      "loss": 0.5987,
-      "step": 58370
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.095899191293794e-05,
-      "loss": 0.6034,
-      "step": 58380
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.093171683008987e-05,
-      "loss": 0.6964,
-      "step": 58390
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0904441747241813e-05,
-      "loss": 0.6679,
-      "step": 58400
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.087716666439375e-05,
-      "loss": 0.6161,
-      "step": 58410
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.084989158154568e-05,
-      "loss": 0.7555,
-      "step": 58420
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.082261649869762e-05,
-      "loss": 0.6427,
-      "step": 58430
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0795341415849556e-05,
-      "loss": 0.6285,
-      "step": 58440
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.076806633300149e-05,
-      "loss": 0.6665,
-      "step": 58450
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0740791250153424e-05,
-      "loss": 0.6255,
-      "step": 58460
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.071351616730536e-05,
-      "loss": 0.6718,
-      "step": 58470
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.068624108445729e-05,
-      "loss": 0.5709,
-      "step": 58480
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.065896600160923e-05,
-      "loss": 0.6643,
-      "step": 58490
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0631690918761166e-05,
-      "loss": 0.5477,
-      "step": 58500
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.06044158359131e-05,
-      "loss": 0.6212,
-      "step": 58510
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.057714075306504e-05,
-      "loss": 0.6163,
-      "step": 58520
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0549865670216975e-05,
-      "loss": 0.602,
-      "step": 58530
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0522590587368915e-05,
-      "loss": 0.7672,
-      "step": 58540
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.049531550452085e-05,
-      "loss": 0.6531,
-      "step": 58550
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.046804042167278e-05,
-      "loss": 0.6351,
-      "step": 58560
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.0440765338824724e-05,
-      "loss": 0.6218,
-      "step": 58570
-    },
-    {
-      "epoch": 2.39,
-      "learning_rate": 4.041349025597666e-05,
-      "loss": 0.6086,
-      "step": 58580
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.038621517312859e-05,
-      "loss": 0.6376,
-      "step": 58590
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0358940090280526e-05,
-      "loss": 0.6803,
-      "step": 58600
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.033166500743246e-05,
-      "loss": 0.6135,
-      "step": 58610
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0304389924584394e-05,
-      "loss": 0.5668,
-      "step": 58620
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0277114841736334e-05,
-      "loss": 0.6004,
-      "step": 58630
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.024983975888827e-05,
-      "loss": 0.6882,
-      "step": 58640
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.02225646760402e-05,
-      "loss": 0.6189,
-      "step": 58650
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.019528959319214e-05,
-      "loss": 0.701,
-      "step": 58660
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.016801451034408e-05,
-      "loss": 0.6066,
-      "step": 58670
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.014073942749601e-05,
-      "loss": 0.6974,
-      "step": 58680
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.011346434464795e-05,
-      "loss": 0.6979,
-      "step": 58690
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0086189261799885e-05,
-      "loss": 0.5729,
-      "step": 58700
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0058914178951826e-05,
-      "loss": 0.5998,
-      "step": 58710
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.003163909610376e-05,
-      "loss": 0.5957,
-      "step": 58720
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 4.0004364013255694e-05,
-      "loss": 0.6903,
-      "step": 58730
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.997708893040763e-05,
-      "loss": 0.6214,
-      "step": 58740
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.994981384755956e-05,
-      "loss": 0.612,
-      "step": 58750
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.9922538764711496e-05,
-      "loss": 0.676,
-      "step": 58760
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.9895263681863436e-05,
-      "loss": 0.724,
-      "step": 58770
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.986798859901537e-05,
-      "loss": 0.6115,
-      "step": 58780
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.9840713516167304e-05,
-      "loss": 0.7045,
-      "step": 58790
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.9813438433319245e-05,
-      "loss": 0.6148,
-      "step": 58800
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.978616335047118e-05,
-      "loss": 0.6875,
-      "step": 58810
-    },
-    {
-      "epoch": 2.4,
-      "learning_rate": 3.975888826762311e-05,
-      "loss": 0.689,
-      "step": 58820
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.973161318477505e-05,
-      "loss": 0.6594,
-      "step": 58830
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.970433810192699e-05,
-      "loss": 0.8008,
-      "step": 58840
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.967706301907893e-05,
-      "loss": 0.5281,
-      "step": 58850
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.964978793623086e-05,
-      "loss": 0.5528,
-      "step": 58860
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9622512853382796e-05,
-      "loss": 0.6599,
-      "step": 58870
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.959523777053473e-05,
-      "loss": 0.6343,
-      "step": 58880
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9567962687686664e-05,
-      "loss": 0.6329,
-      "step": 58890
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.95406876048386e-05,
-      "loss": 0.6117,
-      "step": 58900
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.951341252199054e-05,
-      "loss": 0.6933,
-      "step": 58910
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.948613743914247e-05,
-      "loss": 0.632,
-      "step": 58920
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9458862356294406e-05,
-      "loss": 0.5763,
-      "step": 58930
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.943158727344635e-05,
-      "loss": 0.7598,
-      "step": 58940
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.940431219059828e-05,
-      "loss": 0.7034,
-      "step": 58950
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9377037107750215e-05,
-      "loss": 0.6682,
-      "step": 58960
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9349762024902155e-05,
-      "loss": 0.5598,
-      "step": 58970
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.932248694205409e-05,
-      "loss": 0.5548,
-      "step": 58980
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.929521185920603e-05,
-      "loss": 0.702,
-      "step": 58990
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9267936776357964e-05,
-      "loss": 0.6429,
-      "step": 59000
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.92433892017947e-05,
-      "loss": 0.624,
-      "step": 59010
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.921611411894664e-05,
-      "loss": 0.6761,
-      "step": 59020
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.918883903609857e-05,
-      "loss": 0.6251,
-      "step": 59030
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.9161563953250506e-05,
-      "loss": 0.6805,
-      "step": 59040
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.913428887040245e-05,
-      "loss": 0.6948,
-      "step": 59050
-    },
-    {
-      "epoch": 2.41,
-      "learning_rate": 3.910701378755438e-05,
-      "loss": 0.6341,
-      "step": 59060
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.907973870470632e-05,
-      "loss": 0.6182,
-      "step": 59070
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.9052463621858255e-05,
-      "loss": 0.671,
-      "step": 59080
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.902518853901019e-05,
-      "loss": 0.5877,
-      "step": 59090
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.899791345616213e-05,
-      "loss": 0.6837,
-      "step": 59100
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8970638373314064e-05,
-      "loss": 0.6252,
-      "step": 59110
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8943363290466e-05,
-      "loss": 0.6151,
-      "step": 59120
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.891608820761793e-05,
-      "loss": 0.6231,
-      "step": 59130
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8888813124769866e-05,
-      "loss": 0.6861,
-      "step": 59140
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.88615380419218e-05,
-      "loss": 0.6741,
-      "step": 59150
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.883426295907374e-05,
-      "loss": 0.5741,
-      "step": 59160
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8806987876225674e-05,
-      "loss": 0.6467,
-      "step": 59170
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.877971279337761e-05,
-      "loss": 0.6954,
-      "step": 59180
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.875243771052955e-05,
-      "loss": 0.6339,
-      "step": 59190
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.872516262768148e-05,
-      "loss": 0.6618,
-      "step": 59200
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.869788754483342e-05,
-      "loss": 0.6334,
-      "step": 59210
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.867061246198536e-05,
-      "loss": 0.6094,
-      "step": 59220
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.864333737913729e-05,
-      "loss": 0.5563,
-      "step": 59230
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.861606229628923e-05,
-      "loss": 0.6283,
-      "step": 59240
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8588787213441166e-05,
-      "loss": 0.6801,
-      "step": 59250
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.85615121305931e-05,
-      "loss": 0.6483,
-      "step": 59260
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8534237047745034e-05,
-      "loss": 0.6551,
-      "step": 59270
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.850696196489697e-05,
-      "loss": 0.7155,
-      "step": 59280
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.84796868820489e-05,
-      "loss": 0.611,
-      "step": 59290
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.845241179920084e-05,
-      "loss": 0.6291,
-      "step": 59300
-    },
-    {
-      "epoch": 2.42,
-      "learning_rate": 3.8425136716352776e-05,
-      "loss": 0.666,
-      "step": 59310
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.839786163350471e-05,
-      "loss": 0.5656,
-      "step": 59320
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.837058655065665e-05,
-      "loss": 0.6395,
-      "step": 59330
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.8343311467808585e-05,
-      "loss": 0.6707,
-      "step": 59340
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.831603638496052e-05,
-      "loss": 0.6368,
-      "step": 59350
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.828876130211246e-05,
-      "loss": 0.5846,
-      "step": 59360
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.826148621926439e-05,
-      "loss": 0.6504,
-      "step": 59370
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.8234211136416334e-05,
-      "loss": 0.621,
-      "step": 59380
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.820693605356827e-05,
-      "loss": 0.692,
-      "step": 59390
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.81796609707202e-05,
-      "loss": 0.6174,
-      "step": 59400
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.8152385887872136e-05,
-      "loss": 0.6923,
-      "step": 59410
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.812511080502407e-05,
-      "loss": 0.6898,
-      "step": 59420
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.8097835722176004e-05,
-      "loss": 0.6241,
-      "step": 59430
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.8070560639327944e-05,
-      "loss": 0.6909,
-      "step": 59440
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.804328555647988e-05,
-      "loss": 0.7291,
-      "step": 59450
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.801601047363181e-05,
-      "loss": 0.7697,
-      "step": 59460
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.798873539078375e-05,
-      "loss": 0.5944,
-      "step": 59470
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.796146030793569e-05,
-      "loss": 0.7538,
-      "step": 59480
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.793418522508762e-05,
-      "loss": 0.6696,
-      "step": 59490
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.790691014223956e-05,
-      "loss": 0.6114,
-      "step": 59500
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.7879635059391495e-05,
-      "loss": 0.6039,
-      "step": 59510
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.785235997654343e-05,
-      "loss": 0.6247,
-      "step": 59520
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.782508489369537e-05,
-      "loss": 0.5955,
-      "step": 59530
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.7797809810847304e-05,
-      "loss": 0.6802,
-      "step": 59540
-    },
-    {
-      "epoch": 2.43,
-      "learning_rate": 3.777053472799924e-05,
-      "loss": 0.6992,
-      "step": 59550
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.774325964515117e-05,
-      "loss": 0.6778,
-      "step": 59560
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7715984562303106e-05,
-      "loss": 0.6197,
-      "step": 59570
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7688709479455046e-05,
-      "loss": 0.5882,
-      "step": 59580
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.766143439660698e-05,
-      "loss": 0.5772,
-      "step": 59590
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7634159313758914e-05,
-      "loss": 0.5932,
-      "step": 59600
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7606884230910855e-05,
-      "loss": 0.6522,
-      "step": 59610
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.757960914806279e-05,
-      "loss": 0.6027,
-      "step": 59620
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.755233406521472e-05,
-      "loss": 0.7309,
-      "step": 59630
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.752505898236666e-05,
-      "loss": 0.6534,
-      "step": 59640
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.74977838995186e-05,
-      "loss": 0.6756,
-      "step": 59650
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.747050881667053e-05,
-      "loss": 0.6725,
-      "step": 59660
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.744323373382247e-05,
-      "loss": 0.7138,
-      "step": 59670
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7415958650974406e-05,
-      "loss": 0.5683,
-      "step": 59680
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.738868356812634e-05,
-      "loss": 0.6002,
-      "step": 59690
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7361408485278274e-05,
-      "loss": 0.6971,
-      "step": 59700
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.733413340243021e-05,
-      "loss": 0.6039,
-      "step": 59710
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.730685831958215e-05,
-      "loss": 0.6206,
-      "step": 59720
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.727958323673408e-05,
-      "loss": 0.5966,
-      "step": 59730
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7252308153886016e-05,
-      "loss": 0.7182,
-      "step": 59740
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.722503307103796e-05,
-      "loss": 0.6895,
-      "step": 59750
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.719775798818989e-05,
-      "loss": 0.6907,
-      "step": 59760
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7170482905341825e-05,
-      "loss": 0.6943,
-      "step": 59770
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.7143207822493765e-05,
-      "loss": 0.7043,
-      "step": 59780
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.71159327396457e-05,
-      "loss": 0.7205,
-      "step": 59790
-    },
-    {
-      "epoch": 2.44,
-      "learning_rate": 3.708865765679763e-05,
-      "loss": 0.8047,
-      "step": 59800
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.7061382573949574e-05,
-      "loss": 0.5912,
-      "step": 59810
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.703410749110151e-05,
-      "loss": 0.6917,
-      "step": 59820
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.700683240825344e-05,
-      "loss": 0.6478,
-      "step": 59830
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6979557325405376e-05,
-      "loss": 0.6145,
-      "step": 59840
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.695228224255731e-05,
-      "loss": 0.7015,
-      "step": 59850
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.692500715970925e-05,
-      "loss": 0.6672,
-      "step": 59860
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6897732076861184e-05,
-      "loss": 0.6323,
-      "step": 59870
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.687045699401312e-05,
-      "loss": 0.6246,
-      "step": 59880
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.684318191116506e-05,
-      "loss": 0.7698,
-      "step": 59890
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.681590682831699e-05,
-      "loss": 0.5108,
-      "step": 59900
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.678863174546893e-05,
-      "loss": 0.7284,
-      "step": 59910
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.676135666262087e-05,
-      "loss": 0.6218,
-      "step": 59920
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.67340815797728e-05,
-      "loss": 0.7056,
-      "step": 59930
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6706806496924735e-05,
-      "loss": 0.6479,
-      "step": 59940
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6679531414076676e-05,
-      "loss": 0.6495,
-      "step": 59950
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.665225633122861e-05,
-      "loss": 0.6206,
-      "step": 59960
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6624981248380544e-05,
-      "loss": 0.574,
-      "step": 59970
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.659770616553248e-05,
-      "loss": 0.6256,
-      "step": 59980
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.657043108268441e-05,
-      "loss": 0.6599,
-      "step": 59990
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.654315599983635e-05,
-      "loss": 0.6324,
-      "step": 60000
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6515880916988286e-05,
-      "loss": 0.6952,
-      "step": 60010
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.648860583414022e-05,
-      "loss": 0.7172,
-      "step": 60020
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.646133075129216e-05,
-      "loss": 0.62,
-      "step": 60030
-    },
-    {
-      "epoch": 2.45,
-      "learning_rate": 3.6434055668444095e-05,
-      "loss": 0.6056,
-      "step": 60040
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.640678058559603e-05,
-      "loss": 0.7276,
-      "step": 60050
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.637950550274797e-05,
-      "loss": 0.606,
-      "step": 60060
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.63522304198999e-05,
-      "loss": 0.6334,
-      "step": 60070
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.632495533705184e-05,
-      "loss": 0.6102,
-      "step": 60080
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.629768025420378e-05,
-      "loss": 0.6319,
-      "step": 60090
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.627040517135571e-05,
-      "loss": 0.5849,
-      "step": 60100
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.6243130088507646e-05,
-      "loss": 0.7835,
-      "step": 60110
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.621585500565958e-05,
-      "loss": 0.6487,
-      "step": 60120
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.6188579922811514e-05,
-      "loss": 0.6814,
-      "step": 60130
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.6161304839963454e-05,
-      "loss": 0.7005,
-      "step": 60140
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.613402975711539e-05,
-      "loss": 0.7341,
-      "step": 60150
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.610675467426732e-05,
-      "loss": 0.6336,
-      "step": 60160
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.607947959141926e-05,
-      "loss": 0.6749,
-      "step": 60170
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.60522045085712e-05,
-      "loss": 0.6675,
-      "step": 60180
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.602492942572313e-05,
-      "loss": 0.6668,
-      "step": 60190
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.599765434287507e-05,
-      "loss": 0.6356,
-      "step": 60200
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.5970379260027005e-05,
-      "loss": 0.7039,
-      "step": 60210
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.594310417717894e-05,
-      "loss": 0.6467,
-      "step": 60220
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.591582909433088e-05,
-      "loss": 0.624,
-      "step": 60230
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.5888554011482814e-05,
-      "loss": 0.6228,
-      "step": 60240
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.586127892863475e-05,
-      "loss": 0.5832,
-      "step": 60250
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.583400384578668e-05,
-      "loss": 0.6881,
-      "step": 60260
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.5806728762938616e-05,
-      "loss": 0.6829,
-      "step": 60270
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.5779453680090556e-05,
-      "loss": 0.6,
-      "step": 60280
-    },
-    {
-      "epoch": 2.46,
-      "learning_rate": 3.575217859724249e-05,
-      "loss": 0.7688,
-      "step": 60290
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5724903514394424e-05,
-      "loss": 0.6543,
-      "step": 60300
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5697628431546365e-05,
-      "loss": 0.6107,
-      "step": 60310
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.56703533486983e-05,
-      "loss": 0.5983,
-      "step": 60320
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.564307826585023e-05,
-      "loss": 0.5986,
-      "step": 60330
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.561580318300217e-05,
-      "loss": 0.6321,
-      "step": 60340
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.558852810015411e-05,
-      "loss": 0.6446,
-      "step": 60350
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.556125301730604e-05,
-      "loss": 0.7128,
-      "step": 60360
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.553397793445798e-05,
-      "loss": 0.6419,
-      "step": 60370
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5506702851609916e-05,
-      "loss": 0.5791,
-      "step": 60380
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.547942776876185e-05,
-      "loss": 0.6172,
-      "step": 60390
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5452152685913784e-05,
-      "loss": 0.6353,
-      "step": 60400
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.542487760306572e-05,
-      "loss": 0.7069,
-      "step": 60410
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.539760252021765e-05,
-      "loss": 0.5997,
-      "step": 60420
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.537032743736959e-05,
-      "loss": 0.71,
-      "step": 60430
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5343052354521526e-05,
-      "loss": 0.6041,
-      "step": 60440
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.531577727167347e-05,
-      "loss": 0.5945,
-      "step": 60450
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.52885021888254e-05,
-      "loss": 0.6379,
-      "step": 60460
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5261227105977335e-05,
-      "loss": 0.588,
-      "step": 60470
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5233952023129275e-05,
-      "loss": 0.6524,
-      "step": 60480
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.520667694028121e-05,
-      "loss": 0.5177,
-      "step": 60490
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.517940185743314e-05,
-      "loss": 0.6261,
-      "step": 60500
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.5152126774585084e-05,
-      "loss": 0.5807,
-      "step": 60510
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.512485169173702e-05,
-      "loss": 0.6676,
-      "step": 60520
-    },
-    {
-      "epoch": 2.47,
-      "learning_rate": 3.509757660888895e-05,
-      "loss": 0.6536,
-      "step": 60530
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.5070301526040886e-05,
-      "loss": 0.6815,
-      "step": 60540
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.504302644319282e-05,
-      "loss": 0.5918,
-      "step": 60550
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.5015751360344753e-05,
-      "loss": 0.6397,
-      "step": 60560
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4988476277496694e-05,
-      "loss": 0.7202,
-      "step": 60570
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.496120119464863e-05,
-      "loss": 0.6992,
-      "step": 60580
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.493392611180057e-05,
-      "loss": 0.5982,
-      "step": 60590
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.49066510289525e-05,
-      "loss": 0.6259,
-      "step": 60600
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4879375946104437e-05,
-      "loss": 0.6498,
-      "step": 60610
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.485210086325638e-05,
-      "loss": 0.6334,
-      "step": 60620
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.482482578040831e-05,
-      "loss": 0.646,
-      "step": 60630
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4797550697560245e-05,
-      "loss": 0.6453,
-      "step": 60640
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4770275614712186e-05,
-      "loss": 0.6191,
-      "step": 60650
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.474300053186412e-05,
-      "loss": 0.6504,
-      "step": 60660
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4715725449016054e-05,
-      "loss": 0.6615,
-      "step": 60670
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.468845036616799e-05,
-      "loss": 0.6792,
-      "step": 60680
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.466117528331992e-05,
-      "loss": 0.6212,
-      "step": 60690
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4633900200471855e-05,
-      "loss": 0.5941,
-      "step": 60700
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4606625117623796e-05,
-      "loss": 0.5865,
-      "step": 60710
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.457935003477573e-05,
-      "loss": 0.5684,
-      "step": 60720
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4552074951927664e-05,
-      "loss": 0.6999,
-      "step": 60730
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.4524799869079605e-05,
-      "loss": 0.6771,
-      "step": 60740
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.449752478623154e-05,
-      "loss": 0.5947,
-      "step": 60750
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.447024970338348e-05,
-      "loss": 0.6032,
-      "step": 60760
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.444297462053541e-05,
-      "loss": 0.6073,
-      "step": 60770
-    },
-    {
-      "epoch": 2.48,
-      "learning_rate": 3.441569953768735e-05,
-      "loss": 0.6159,
-      "step": 60780
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.438842445483929e-05,
-      "loss": 0.5981,
-      "step": 60790
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.436114937199122e-05,
-      "loss": 0.7013,
-      "step": 60800
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.4333874289143156e-05,
-      "loss": 0.5483,
-      "step": 60810
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.430659920629509e-05,
-      "loss": 0.6208,
-      "step": 60820
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.4279324123447023e-05,
-      "loss": 0.6622,
-      "step": 60830
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.425204904059896e-05,
-      "loss": 0.68,
-      "step": 60840
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.42247739577509e-05,
-      "loss": 0.6352,
-      "step": 60850
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.419749887490283e-05,
-      "loss": 0.6288,
-      "step": 60860
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.4170223792054766e-05,
-      "loss": 0.7499,
-      "step": 60870
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.4142948709206707e-05,
-      "loss": 0.6103,
-      "step": 60880
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.411567362635864e-05,
-      "loss": 0.6194,
-      "step": 60890
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.408839854351058e-05,
-      "loss": 0.6983,
-      "step": 60900
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.4061123460662515e-05,
-      "loss": 0.6141,
-      "step": 60910
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.403384837781445e-05,
-      "loss": 0.7219,
-      "step": 60920
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.400657329496639e-05,
-      "loss": 0.6528,
-      "step": 60930
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.3979298212118324e-05,
-      "loss": 0.7101,
-      "step": 60940
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.395202312927026e-05,
-      "loss": 0.6082,
-      "step": 60950
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.392474804642219e-05,
-      "loss": 0.5783,
-      "step": 60960
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.3897472963574125e-05,
-      "loss": 0.6148,
-      "step": 60970
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.387019788072606e-05,
-      "loss": 0.7922,
-      "step": 60980
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.3842922797878e-05,
-      "loss": 0.6338,
-      "step": 60990
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.3815647715029934e-05,
-      "loss": 0.6708,
-      "step": 61000
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.378837263218187e-05,
-      "loss": 0.7058,
-      "step": 61010
-    },
-    {
-      "epoch": 2.49,
-      "learning_rate": 3.376109754933381e-05,
-      "loss": 0.5708,
-      "step": 61020
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.373382246648574e-05,
-      "loss": 0.7677,
-      "step": 61030
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.370654738363768e-05,
-      "loss": 0.6012,
-      "step": 61040
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.367927230078962e-05,
-      "loss": 0.6622,
-      "step": 61050
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.365199721794155e-05,
-      "loss": 0.672,
-      "step": 61060
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.362472213509349e-05,
-      "loss": 0.6113,
-      "step": 61070
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3597447052245426e-05,
-      "loss": 0.626,
-      "step": 61080
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.357017196939736e-05,
-      "loss": 0.5785,
-      "step": 61090
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3542896886549294e-05,
-      "loss": 0.744,
-      "step": 61100
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.351562180370123e-05,
-      "loss": 0.6891,
-      "step": 61110
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.348834672085316e-05,
-      "loss": 0.6469,
-      "step": 61120
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.34610716380051e-05,
-      "loss": 0.5924,
-      "step": 61130
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3433796555157036e-05,
-      "loss": 0.613,
-      "step": 61140
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.340652147230897e-05,
-      "loss": 0.6234,
-      "step": 61150
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.337924638946091e-05,
-      "loss": 0.6575,
-      "step": 61160
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3351971306612844e-05,
-      "loss": 0.6597,
-      "step": 61170
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.332469622376478e-05,
-      "loss": 0.6511,
-      "step": 61180
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.329742114091672e-05,
-      "loss": 0.6469,
-      "step": 61190
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.327014605806865e-05,
-      "loss": 0.6814,
-      "step": 61200
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3242870975220594e-05,
-      "loss": 0.6079,
-      "step": 61210
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.321559589237253e-05,
-      "loss": 0.6822,
-      "step": 61220
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.318832080952446e-05,
-      "loss": 0.6075,
-      "step": 61230
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3161045726676395e-05,
-      "loss": 0.6654,
-      "step": 61240
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.313377064382833e-05,
-      "loss": 0.5786,
-      "step": 61250
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.310649556098026e-05,
-      "loss": 0.6724,
-      "step": 61260
-    },
-    {
-      "epoch": 2.5,
-      "learning_rate": 3.3079220478132204e-05,
-      "loss": 0.6715,
-      "step": 61270
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.305194539528414e-05,
-      "loss": 0.5855,
-      "step": 61280
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.302467031243607e-05,
-      "loss": 0.5557,
-      "step": 61290
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.299739522958801e-05,
-      "loss": 0.6242,
-      "step": 61300
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2970120146739946e-05,
-      "loss": 0.6365,
-      "step": 61310
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.294284506389188e-05,
-      "loss": 0.6716,
-      "step": 61320
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.291556998104382e-05,
-      "loss": 0.5786,
-      "step": 61330
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2888294898195755e-05,
-      "loss": 0.6264,
-      "step": 61340
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2861019815347696e-05,
-      "loss": 0.6736,
-      "step": 61350
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.283374473249963e-05,
-      "loss": 0.6296,
-      "step": 61360
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2806469649651564e-05,
-      "loss": 0.6248,
-      "step": 61370
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.27791945668035e-05,
-      "loss": 0.6349,
-      "step": 61380
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.275191948395543e-05,
-      "loss": 0.6679,
-      "step": 61390
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2724644401107365e-05,
-      "loss": 0.661,
-      "step": 61400
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2697369318259306e-05,
-      "loss": 0.678,
-      "step": 61410
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.267009423541124e-05,
-      "loss": 0.6284,
-      "step": 61420
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2642819152563174e-05,
-      "loss": 0.5603,
-      "step": 61430
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2615544069715115e-05,
-      "loss": 0.5559,
-      "step": 61440
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.258826898686705e-05,
-      "loss": 0.6443,
-      "step": 61450
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.256099390401898e-05,
-      "loss": 0.6902,
-      "step": 61460
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.253371882117092e-05,
-      "loss": 0.691,
-      "step": 61470
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.250644373832286e-05,
-      "loss": 0.6998,
-      "step": 61480
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.247916865547479e-05,
-      "loss": 0.6734,
-      "step": 61490
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.245189357262673e-05,
-      "loss": 0.6816,
-      "step": 61500
-    },
-    {
-      "epoch": 2.51,
-      "learning_rate": 3.2424618489778666e-05,
-      "loss": 0.5129,
-      "step": 61510
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.23973434069306e-05,
-      "loss": 0.6665,
-      "step": 61520
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.237006832408253e-05,
-      "loss": 0.7507,
-      "step": 61530
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.234279324123447e-05,
-      "loss": 0.6802,
-      "step": 61540
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.231551815838641e-05,
-      "loss": 0.6425,
-      "step": 61550
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.228824307553834e-05,
-      "loss": 0.5969,
-      "step": 61560
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.2260967992690276e-05,
-      "loss": 0.6446,
-      "step": 61570
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.2233692909842217e-05,
-      "loss": 0.6333,
-      "step": 61580
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.220641782699415e-05,
-      "loss": 0.5531,
-      "step": 61590
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.2179142744146084e-05,
-      "loss": 0.5704,
-      "step": 61600
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.2151867661298025e-05,
-      "loss": 0.8224,
-      "step": 61610
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.212459257844996e-05,
-      "loss": 0.744,
-      "step": 61620
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.209731749560189e-05,
-      "loss": 0.5819,
-      "step": 61630
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.2070042412753834e-05,
-      "loss": 0.5544,
-      "step": 61640
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.204276732990577e-05,
-      "loss": 0.6473,
-      "step": 61650
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.20154922470577e-05,
-      "loss": 0.6206,
-      "step": 61660
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.1988217164209635e-05,
-      "loss": 0.6984,
-      "step": 61670
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.196094208136157e-05,
-      "loss": 0.6409,
-      "step": 61680
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.193366699851351e-05,
-      "loss": 0.6546,
-      "step": 61690
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.1906391915665444e-05,
-      "loss": 0.6411,
-      "step": 61700
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.187911683281738e-05,
-      "loss": 0.6332,
-      "step": 61710
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.185184174996932e-05,
-      "loss": 0.644,
-      "step": 61720
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.182456666712125e-05,
-      "loss": 0.6716,
-      "step": 61730
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.1797291584273186e-05,
-      "loss": 0.6706,
-      "step": 61740
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.177001650142513e-05,
-      "loss": 0.6211,
-      "step": 61750
-    },
-    {
-      "epoch": 2.52,
-      "learning_rate": 3.174274141857706e-05,
-      "loss": 0.5843,
-      "step": 61760
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1715466335728995e-05,
-      "loss": 0.591,
-      "step": 61770
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1688191252880936e-05,
-      "loss": 0.6304,
-      "step": 61780
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.166091617003287e-05,
-      "loss": 0.6277,
-      "step": 61790
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1633641087184803e-05,
-      "loss": 0.6314,
-      "step": 61800
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.160636600433674e-05,
-      "loss": 0.6396,
-      "step": 61810
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.157909092148867e-05,
-      "loss": 0.6117,
-      "step": 61820
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.155181583864061e-05,
-      "loss": 0.5813,
-      "step": 61830
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1524540755792546e-05,
-      "loss": 0.6581,
-      "step": 61840
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.149726567294448e-05,
-      "loss": 0.7141,
-      "step": 61850
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.146999059009642e-05,
-      "loss": 0.7166,
-      "step": 61860
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1442715507248354e-05,
-      "loss": 0.6701,
-      "step": 61870
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.141544042440029e-05,
-      "loss": 0.7286,
-      "step": 61880
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.138816534155223e-05,
-      "loss": 0.6503,
-      "step": 61890
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.136089025870416e-05,
-      "loss": 0.5902,
-      "step": 61900
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.13336151758561e-05,
-      "loss": 0.621,
-      "step": 61910
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.130634009300804e-05,
-      "loss": 0.6552,
-      "step": 61920
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.127906501015997e-05,
-      "loss": 0.6868,
-      "step": 61930
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1251789927311905e-05,
-      "loss": 0.6315,
-      "step": 61940
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.122451484446384e-05,
-      "loss": 0.7406,
-      "step": 61950
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.119723976161577e-05,
-      "loss": 0.6376,
-      "step": 61960
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.1169964678767714e-05,
-      "loss": 0.5703,
-      "step": 61970
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.114268959591965e-05,
-      "loss": 0.7018,
-      "step": 61980
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.111541451307158e-05,
-      "loss": 0.6552,
-      "step": 61990
-    },
-    {
-      "epoch": 2.53,
-      "learning_rate": 3.108813943022352e-05,
-      "loss": 0.5977,
-      "step": 62000
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.1060864347375456e-05,
-      "loss": 0.6274,
-      "step": 62010
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.103358926452739e-05,
-      "loss": 0.6676,
-      "step": 62020
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.100631418167933e-05,
-      "loss": 0.6946,
-      "step": 62030
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0979039098831265e-05,
-      "loss": 0.6288,
-      "step": 62040
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.09517640159832e-05,
-      "loss": 0.7342,
-      "step": 62050
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.092448893313514e-05,
-      "loss": 0.6095,
-      "step": 62060
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0897213850287073e-05,
-      "loss": 0.708,
-      "step": 62070
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.086993876743901e-05,
-      "loss": 0.7284,
-      "step": 62080
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.084266368459094e-05,
-      "loss": 0.6896,
-      "step": 62090
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0815388601742875e-05,
-      "loss": 0.6851,
-      "step": 62100
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0788113518894816e-05,
-      "loss": 0.6202,
-      "step": 62110
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.076083843604675e-05,
-      "loss": 0.762,
-      "step": 62120
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0733563353198684e-05,
-      "loss": 0.6321,
-      "step": 62130
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0706288270350624e-05,
-      "loss": 0.62,
-      "step": 62140
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.067901318750256e-05,
-      "loss": 0.5405,
-      "step": 62150
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.065173810465449e-05,
-      "loss": 0.6893,
-      "step": 62160
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.062446302180643e-05,
-      "loss": 0.6425,
-      "step": 62170
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.059718793895837e-05,
-      "loss": 0.758,
-      "step": 62180
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.05699128561103e-05,
-      "loss": 0.7015,
-      "step": 62190
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.054263777326224e-05,
-      "loss": 0.6426,
-      "step": 62200
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0515362690414172e-05,
-      "loss": 0.6506,
-      "step": 62210
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0488087607566106e-05,
-      "loss": 0.6503,
-      "step": 62220
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.0460812524718047e-05,
-      "loss": 0.767,
-      "step": 62230
-    },
-    {
-      "epoch": 2.54,
-      "learning_rate": 3.043353744186998e-05,
-      "loss": 0.6973,
-      "step": 62240
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0406262359021915e-05,
-      "loss": 0.5794,
-      "step": 62250
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0378987276173855e-05,
-      "loss": 0.718,
-      "step": 62260
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.035171219332579e-05,
-      "loss": 0.6196,
-      "step": 62270
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0324437110477726e-05,
-      "loss": 0.6682,
-      "step": 62280
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.029716202762966e-05,
-      "loss": 0.6474,
-      "step": 62290
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0269886944781594e-05,
-      "loss": 0.6502,
-      "step": 62300
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0242611861933535e-05,
-      "loss": 0.6181,
-      "step": 62310
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.021533677908547e-05,
-      "loss": 0.6851,
-      "step": 62320
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0188061696237403e-05,
-      "loss": 0.6116,
-      "step": 62330
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.016078661338934e-05,
-      "loss": 0.5928,
-      "step": 62340
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0133511530541274e-05,
-      "loss": 0.6925,
-      "step": 62350
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0106236447693208e-05,
-      "loss": 0.5567,
-      "step": 62360
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.007896136484515e-05,
-      "loss": 0.6604,
-      "step": 62370
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0051686281997083e-05,
-      "loss": 0.6489,
-      "step": 62380
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 3.0024411199149017e-05,
-      "loss": 0.6422,
-      "step": 62390
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9997136116300957e-05,
-      "loss": 0.6848,
-      "step": 62400
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.996986103345289e-05,
-      "loss": 0.6443,
-      "step": 62410
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.994258595060483e-05,
-      "loss": 0.7382,
-      "step": 62420
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9915310867756762e-05,
-      "loss": 0.7022,
-      "step": 62430
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9888035784908696e-05,
-      "loss": 0.6256,
-      "step": 62440
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9860760702060637e-05,
-      "loss": 0.6554,
-      "step": 62450
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.983348561921257e-05,
-      "loss": 0.7728,
-      "step": 62460
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9806210536364505e-05,
-      "loss": 0.6631,
-      "step": 62470
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9778935453516442e-05,
-      "loss": 0.5879,
-      "step": 62480
-    },
-    {
-      "epoch": 2.55,
-      "learning_rate": 2.9751660370668376e-05,
-      "loss": 0.6665,
-      "step": 62490
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.972438528782031e-05,
-      "loss": 0.6173,
-      "step": 62500
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.969711020497225e-05,
-      "loss": 0.6273,
-      "step": 62510
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9669835122124185e-05,
-      "loss": 0.6566,
-      "step": 62520
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.964256003927612e-05,
-      "loss": 0.6852,
-      "step": 62530
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.961528495642806e-05,
-      "loss": 0.6411,
-      "step": 62540
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9588009873579993e-05,
-      "loss": 0.69,
-      "step": 62550
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.956073479073193e-05,
-      "loss": 0.6152,
-      "step": 62560
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9533459707883864e-05,
-      "loss": 0.6193,
-      "step": 62570
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9506184625035798e-05,
-      "loss": 0.5721,
-      "step": 62580
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.947890954218774e-05,
-      "loss": 0.6829,
-      "step": 62590
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9451634459339673e-05,
-      "loss": 0.6746,
-      "step": 62600
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9424359376491607e-05,
-      "loss": 0.5696,
-      "step": 62610
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9397084293643544e-05,
-      "loss": 0.6455,
-      "step": 62620
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9369809210795478e-05,
-      "loss": 0.6864,
-      "step": 62630
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9342534127947412e-05,
-      "loss": 0.6861,
-      "step": 62640
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9315259045099353e-05,
-      "loss": 0.5761,
-      "step": 62650
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9287983962251287e-05,
-      "loss": 0.6868,
-      "step": 62660
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.926070887940322e-05,
-      "loss": 0.6428,
-      "step": 62670
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.923343379655516e-05,
-      "loss": 0.659,
-      "step": 62680
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9206158713707095e-05,
-      "loss": 0.5968,
-      "step": 62690
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.917888363085903e-05,
-      "loss": 0.7935,
-      "step": 62700
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.9151608548010966e-05,
-      "loss": 0.5848,
-      "step": 62710
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.91243334651629e-05,
-      "loss": 0.626,
-      "step": 62720
-    },
-    {
-      "epoch": 2.56,
-      "learning_rate": 2.909705838231484e-05,
-      "loss": 0.7381,
-      "step": 62730
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.9069783299466775e-05,
-      "loss": 0.5981,
-      "step": 62740
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.904250821661871e-05,
-      "loss": 0.6788,
-      "step": 62750
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.9015233133770646e-05,
-      "loss": 0.6952,
-      "step": 62760
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.898795805092258e-05,
-      "loss": 0.6949,
-      "step": 62770
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8960682968074514e-05,
-      "loss": 0.6729,
-      "step": 62780
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8933407885226455e-05,
-      "loss": 0.6422,
-      "step": 62790
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.890613280237839e-05,
-      "loss": 0.6041,
-      "step": 62800
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8878857719530322e-05,
-      "loss": 0.6945,
-      "step": 62810
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8851582636682263e-05,
-      "loss": 0.6846,
-      "step": 62820
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8824307553834197e-05,
-      "loss": 0.5377,
-      "step": 62830
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.879703247098613e-05,
-      "loss": 0.6549,
-      "step": 62840
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.876975738813807e-05,
-      "loss": 0.6519,
-      "step": 62850
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8742482305290002e-05,
-      "loss": 0.6691,
-      "step": 62860
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8715207222441943e-05,
-      "loss": 0.5696,
-      "step": 62870
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8687932139593877e-05,
-      "loss": 0.6302,
-      "step": 62880
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.866065705674581e-05,
-      "loss": 0.6629,
-      "step": 62890
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8633381973897748e-05,
-      "loss": 0.695,
-      "step": 62900
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8606106891049682e-05,
-      "loss": 0.6403,
-      "step": 62910
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8578831808201616e-05,
-      "loss": 0.6282,
-      "step": 62920
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8551556725353557e-05,
-      "loss": 0.616,
-      "step": 62930
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.852428164250549e-05,
-      "loss": 0.5603,
-      "step": 62940
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8497006559657424e-05,
-      "loss": 0.6184,
-      "step": 62950
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8469731476809365e-05,
-      "loss": 0.6383,
-      "step": 62960
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.84424563939613e-05,
-      "loss": 0.6394,
-      "step": 62970
-    },
-    {
-      "epoch": 2.57,
-      "learning_rate": 2.8415181311113233e-05,
-      "loss": 0.6727,
-      "step": 62980
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.838790622826517e-05,
-      "loss": 0.5868,
-      "step": 62990
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8360631145417104e-05,
-      "loss": 0.6584,
-      "step": 63000
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8333356062569045e-05,
-      "loss": 0.6304,
-      "step": 63010
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.830608097972098e-05,
-      "loss": 0.631,
-      "step": 63020
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8278805896872913e-05,
-      "loss": 0.7883,
-      "step": 63030
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.825153081402485e-05,
-      "loss": 0.6591,
-      "step": 63040
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8224255731176784e-05,
-      "loss": 0.5659,
-      "step": 63050
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8196980648328718e-05,
-      "loss": 0.483,
-      "step": 63060
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.816970556548066e-05,
-      "loss": 0.6095,
-      "step": 63070
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8142430482632593e-05,
-      "loss": 0.6361,
-      "step": 63080
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8115155399784526e-05,
-      "loss": 0.655,
-      "step": 63090
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.809060782522127e-05,
-      "loss": 0.665,
-      "step": 63100
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8063332742373204e-05,
-      "loss": 0.6991,
-      "step": 63110
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.8036057659525145e-05,
-      "loss": 0.7429,
-      "step": 63120
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.800878257667708e-05,
-      "loss": 0.7176,
-      "step": 63130
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7981507493829013e-05,
-      "loss": 0.6548,
-      "step": 63140
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.795423241098095e-05,
-      "loss": 0.6157,
-      "step": 63150
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7926957328132884e-05,
-      "loss": 0.6475,
-      "step": 63160
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7899682245284818e-05,
-      "loss": 0.7068,
-      "step": 63170
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.787240716243676e-05,
-      "loss": 0.7378,
-      "step": 63180
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7845132079588693e-05,
-      "loss": 0.5985,
-      "step": 63190
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7817856996740626e-05,
-      "loss": 0.5458,
-      "step": 63200
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.7790581913892567e-05,
-      "loss": 0.6133,
-      "step": 63210
-    },
-    {
-      "epoch": 2.58,
-      "learning_rate": 2.77633068310445e-05,
-      "loss": 0.6048,
-      "step": 63220
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7736031748196435e-05,
-      "loss": 0.6441,
-      "step": 63230
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7708756665348372e-05,
-      "loss": 0.7379,
-      "step": 63240
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7681481582500306e-05,
-      "loss": 0.6132,
-      "step": 63250
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7654206499652247e-05,
-      "loss": 0.6704,
-      "step": 63260
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.762693141680418e-05,
-      "loss": 0.734,
-      "step": 63270
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7599656333956115e-05,
-      "loss": 0.5789,
-      "step": 63280
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7572381251108052e-05,
-      "loss": 0.6802,
-      "step": 63290
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7545106168259986e-05,
-      "loss": 0.6374,
-      "step": 63300
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.751783108541192e-05,
-      "loss": 0.6326,
-      "step": 63310
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.749055600256386e-05,
-      "loss": 0.6126,
-      "step": 63320
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7463280919715795e-05,
-      "loss": 0.6033,
-      "step": 63330
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.743600583686773e-05,
-      "loss": 0.6925,
-      "step": 63340
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.740873075401967e-05,
-      "loss": 0.6714,
-      "step": 63350
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7381455671171603e-05,
-      "loss": 0.6222,
-      "step": 63360
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7354180588323537e-05,
-      "loss": 0.6067,
-      "step": 63370
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7326905505475474e-05,
-      "loss": 0.6671,
-      "step": 63380
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7299630422627408e-05,
-      "loss": 0.5737,
-      "step": 63390
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.727235533977935e-05,
-      "loss": 0.6979,
-      "step": 63400
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7245080256931283e-05,
-      "loss": 0.6385,
-      "step": 63410
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7217805174083217e-05,
-      "loss": 0.6787,
-      "step": 63420
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7190530091235154e-05,
-      "loss": 0.6817,
-      "step": 63430
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7163255008387088e-05,
-      "loss": 0.6264,
-      "step": 63440
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7135979925539022e-05,
-      "loss": 0.6247,
-      "step": 63450
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7108704842690963e-05,
-      "loss": 0.6295,
-      "step": 63460
-    },
-    {
-      "epoch": 2.59,
-      "learning_rate": 2.7081429759842897e-05,
-      "loss": 0.6109,
-      "step": 63470
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.705415467699483e-05,
-      "loss": 0.759,
-      "step": 63480
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.702687959414677e-05,
-      "loss": 0.725,
-      "step": 63490
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6999604511298705e-05,
-      "loss": 0.6684,
-      "step": 63500
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.697232942845064e-05,
-      "loss": 0.6214,
-      "step": 63510
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6945054345602576e-05,
-      "loss": 0.7128,
-      "step": 63520
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.691777926275451e-05,
-      "loss": 0.6205,
-      "step": 63530
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6890504179906444e-05,
-      "loss": 0.7001,
-      "step": 63540
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6863229097058385e-05,
-      "loss": 0.6916,
-      "step": 63550
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.683595401421032e-05,
-      "loss": 0.557,
-      "step": 63560
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6808678931362256e-05,
-      "loss": 0.5582,
-      "step": 63570
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.678140384851419e-05,
-      "loss": 0.7263,
-      "step": 63580
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6754128765666124e-05,
-      "loss": 0.6178,
-      "step": 63590
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6726853682818065e-05,
-      "loss": 0.5764,
-      "step": 63600
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.669957859997e-05,
-      "loss": 0.7157,
-      "step": 63610
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6672303517121932e-05,
-      "loss": 0.6819,
-      "step": 63620
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6645028434273873e-05,
-      "loss": 0.7395,
-      "step": 63630
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6617753351425807e-05,
-      "loss": 0.6053,
-      "step": 63640
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.659047826857774e-05,
-      "loss": 0.6744,
-      "step": 63650
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6563203185729678e-05,
-      "loss": 0.6024,
-      "step": 63660
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6535928102881612e-05,
-      "loss": 0.6539,
-      "step": 63670
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6508653020033546e-05,
-      "loss": 0.6273,
-      "step": 63680
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6481377937185487e-05,
-      "loss": 0.6506,
-      "step": 63690
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.645410285433742e-05,
-      "loss": 0.6643,
-      "step": 63700
-    },
-    {
-      "epoch": 2.6,
-      "learning_rate": 2.6426827771489358e-05,
-      "loss": 0.5504,
-      "step": 63710
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6399552688641292e-05,
-      "loss": 0.6526,
-      "step": 63720
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6372277605793226e-05,
-      "loss": 0.6067,
-      "step": 63730
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6345002522945167e-05,
-      "loss": 0.5249,
-      "step": 63740
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.63177274400971e-05,
-      "loss": 0.5609,
-      "step": 63750
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6290452357249034e-05,
-      "loss": 0.6115,
-      "step": 63760
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6263177274400975e-05,
-      "loss": 0.7113,
-      "step": 63770
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.623590219155291e-05,
-      "loss": 0.6719,
-      "step": 63780
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6208627108704843e-05,
-      "loss": 0.6938,
-      "step": 63790
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.618135202585678e-05,
-      "loss": 0.6361,
-      "step": 63800
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6154076943008714e-05,
-      "loss": 0.7201,
-      "step": 63810
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6126801860160648e-05,
-      "loss": 0.6509,
-      "step": 63820
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.609952677731259e-05,
-      "loss": 0.6381,
-      "step": 63830
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6072251694464523e-05,
-      "loss": 0.7091,
-      "step": 63840
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6044976611616463e-05,
-      "loss": 0.5744,
-      "step": 63850
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.6017701528768394e-05,
-      "loss": 0.6552,
-      "step": 63860
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5990426445920328e-05,
-      "loss": 0.6884,
-      "step": 63870
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.596315136307227e-05,
-      "loss": 0.7169,
-      "step": 63880
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5935876280224202e-05,
-      "loss": 0.7787,
-      "step": 63890
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5908601197376136e-05,
-      "loss": 0.5644,
-      "step": 63900
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5881326114528077e-05,
-      "loss": 0.6191,
-      "step": 63910
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.585405103168001e-05,
-      "loss": 0.6169,
-      "step": 63920
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5826775948831945e-05,
-      "loss": 0.6483,
-      "step": 63930
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5799500865983882e-05,
-      "loss": 0.6709,
-      "step": 63940
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.5772225783135816e-05,
-      "loss": 0.6275,
-      "step": 63950
-    },
-    {
-      "epoch": 2.61,
-      "learning_rate": 2.574495070028775e-05,
-      "loss": 0.5236,
-      "step": 63960
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.571767561743969e-05,
-      "loss": 0.6016,
-      "step": 63970
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5690400534591625e-05,
-      "loss": 0.6943,
-      "step": 63980
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.566312545174356e-05,
-      "loss": 0.5846,
-      "step": 63990
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5635850368895496e-05,
-      "loss": 0.7384,
-      "step": 64000
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.560857528604743e-05,
-      "loss": 0.6054,
-      "step": 64010
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.558130020319937e-05,
-      "loss": 0.7127,
-      "step": 64020
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5554025120351304e-05,
-      "loss": 0.6645,
-      "step": 64030
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.552675003750324e-05,
-      "loss": 0.5924,
-      "step": 64040
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.549947495465518e-05,
-      "loss": 0.6737,
-      "step": 64050
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5472199871807113e-05,
-      "loss": 0.5864,
-      "step": 64060
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5444924788959047e-05,
-      "loss": 0.6005,
-      "step": 64070
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5417649706110984e-05,
-      "loss": 0.558,
-      "step": 64080
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5390374623262918e-05,
-      "loss": 0.6597,
-      "step": 64090
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5363099540414852e-05,
-      "loss": 0.6374,
-      "step": 64100
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5335824457566793e-05,
-      "loss": 0.6012,
-      "step": 64110
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5308549374718727e-05,
-      "loss": 0.6703,
-      "step": 64120
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.528127429187066e-05,
-      "loss": 0.6294,
-      "step": 64130
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.52539992090226e-05,
-      "loss": 0.7244,
-      "step": 64140
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5226724126174532e-05,
-      "loss": 0.7287,
-      "step": 64150
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5199449043326473e-05,
-      "loss": 0.661,
-      "step": 64160
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5172173960478406e-05,
-      "loss": 0.5854,
-      "step": 64170
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.514489887763034e-05,
-      "loss": 0.6057,
-      "step": 64180
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.511762379478228e-05,
-      "loss": 0.7809,
-      "step": 64190
-    },
-    {
-      "epoch": 2.62,
-      "learning_rate": 2.5090348711934215e-05,
-      "loss": 0.6843,
-      "step": 64200
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.506307362908615e-05,
-      "loss": 0.593,
-      "step": 64210
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.5035798546238086e-05,
-      "loss": 0.7148,
-      "step": 64220
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.500852346339002e-05,
-      "loss": 0.6505,
-      "step": 64230
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4981248380541957e-05,
-      "loss": 0.6719,
-      "step": 64240
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.495397329769389e-05,
-      "loss": 0.7492,
-      "step": 64250
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.492669821484583e-05,
-      "loss": 0.6526,
-      "step": 64260
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4899423131997766e-05,
-      "loss": 0.5833,
-      "step": 64270
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4872148049149703e-05,
-      "loss": 0.6562,
-      "step": 64280
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4844872966301634e-05,
-      "loss": 0.6509,
-      "step": 64290
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.481759788345357e-05,
-      "loss": 0.7536,
-      "step": 64300
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.479032280060551e-05,
-      "loss": 0.6472,
-      "step": 64310
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4763047717757442e-05,
-      "loss": 0.7837,
-      "step": 64320
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.473577263490938e-05,
-      "loss": 0.6339,
-      "step": 64330
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4708497552061317e-05,
-      "loss": 0.7182,
-      "step": 64340
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4681222469213254e-05,
-      "loss": 0.6438,
-      "step": 64350
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4653947386365188e-05,
-      "loss": 0.6634,
-      "step": 64360
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4626672303517122e-05,
-      "loss": 0.6827,
-      "step": 64370
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.459939722066906e-05,
-      "loss": 0.6903,
-      "step": 64380
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4572122137820993e-05,
-      "loss": 0.5964,
-      "step": 64390
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.454484705497293e-05,
-      "loss": 0.6491,
-      "step": 64400
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4517571972124868e-05,
-      "loss": 0.6334,
-      "step": 64410
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4490296889276805e-05,
-      "loss": 0.6184,
-      "step": 64420
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.446302180642874e-05,
-      "loss": 0.6212,
-      "step": 64430
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.4435746723580673e-05,
-      "loss": 0.5722,
-      "step": 64440
-    },
-    {
-      "epoch": 2.63,
-      "learning_rate": 2.440847164073261e-05,
-      "loss": 0.6366,
-      "step": 64450
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4381196557884544e-05,
-      "loss": 0.6615,
-      "step": 64460
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.435392147503648e-05,
-      "loss": 0.6372,
-      "step": 64470
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.432664639218842e-05,
-      "loss": 0.6308,
-      "step": 64480
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4299371309340356e-05,
-      "loss": 0.6716,
-      "step": 64490
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.427209622649229e-05,
-      "loss": 0.6987,
-      "step": 64500
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4244821143644224e-05,
-      "loss": 0.6893,
-      "step": 64510
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.421754606079616e-05,
-      "loss": 0.6332,
-      "step": 64520
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4190270977948095e-05,
-      "loss": 0.6508,
-      "step": 64530
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4162995895100033e-05,
-      "loss": 0.5558,
-      "step": 64540
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.413572081225197e-05,
-      "loss": 0.6802,
-      "step": 64550
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4108445729403904e-05,
-      "loss": 0.6485,
-      "step": 64560
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.408117064655584e-05,
-      "loss": 0.616,
-      "step": 64570
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4053895563707775e-05,
-      "loss": 0.635,
-      "step": 64580
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.4026620480859712e-05,
-      "loss": 0.5778,
-      "step": 64590
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3999345398011646e-05,
-      "loss": 0.5534,
-      "step": 64600
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3972070315163584e-05,
-      "loss": 0.5858,
-      "step": 64610
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.394479523231552e-05,
-      "loss": 0.6449,
-      "step": 64620
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3917520149467455e-05,
-      "loss": 0.6269,
-      "step": 64630
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3890245066619392e-05,
-      "loss": 0.6935,
-      "step": 64640
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3862969983771326e-05,
-      "loss": 0.5934,
-      "step": 64650
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3835694900923263e-05,
-      "loss": 0.647,
-      "step": 64660
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3808419818075197e-05,
-      "loss": 0.6086,
-      "step": 64670
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3781144735227135e-05,
-      "loss": 0.6409,
-      "step": 64680
-    },
-    {
-      "epoch": 2.64,
-      "learning_rate": 2.3753869652379072e-05,
-      "loss": 0.7106,
-      "step": 64690
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3726594569531006e-05,
-      "loss": 0.6917,
-      "step": 64700
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3699319486682943e-05,
-      "loss": 0.5898,
-      "step": 64710
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3672044403834877e-05,
-      "loss": 0.6975,
-      "step": 64720
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3644769320986814e-05,
-      "loss": 0.6628,
-      "step": 64730
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.361749423813875e-05,
-      "loss": 0.6238,
-      "step": 64740
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3590219155290686e-05,
-      "loss": 0.5735,
-      "step": 64750
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3562944072442623e-05,
-      "loss": 0.6494,
-      "step": 64760
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3535668989594557e-05,
-      "loss": 0.5996,
-      "step": 64770
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3508393906746494e-05,
-      "loss": 0.6715,
-      "step": 64780
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3481118823898428e-05,
-      "loss": 0.654,
-      "step": 64790
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3453843741050365e-05,
-      "loss": 0.6855,
-      "step": 64800
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.34265686582023e-05,
-      "loss": 0.6853,
-      "step": 64810
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3399293575354237e-05,
-      "loss": 0.7087,
-      "step": 64820
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3372018492506174e-05,
-      "loss": 0.5922,
-      "step": 64830
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3344743409658108e-05,
-      "loss": 0.7063,
-      "step": 64840
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3317468326810045e-05,
-      "loss": 0.6436,
-      "step": 64850
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.329019324396198e-05,
-      "loss": 0.7259,
-      "step": 64860
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3262918161113916e-05,
-      "loss": 0.6769,
-      "step": 64870
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.323564307826585e-05,
-      "loss": 0.7556,
-      "step": 64880
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3208367995417788e-05,
-      "loss": 0.695,
-      "step": 64890
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3181092912569725e-05,
-      "loss": 0.6787,
-      "step": 64900
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.315381782972166e-05,
-      "loss": 0.7181,
-      "step": 64910
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.3126542746873596e-05,
-      "loss": 0.6178,
-      "step": 64920
-    },
-    {
-      "epoch": 2.65,
-      "learning_rate": 2.309926766402553e-05,
-      "loss": 0.7012,
-      "step": 64930
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.3071992581177464e-05,
-      "loss": 0.6628,
-      "step": 64940
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.30447174983294e-05,
-      "loss": 0.7184,
-      "step": 64950
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.301744241548134e-05,
-      "loss": 0.7273,
-      "step": 64960
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2990167332633276e-05,
-      "loss": 0.6538,
-      "step": 64970
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.296289224978521e-05,
-      "loss": 0.6036,
-      "step": 64980
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2935617166937147e-05,
-      "loss": 0.6139,
-      "step": 64990
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.290834208408908e-05,
-      "loss": 0.7057,
-      "step": 65000
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2881067001241015e-05,
-      "loss": 0.6739,
-      "step": 65010
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2853791918392952e-05,
-      "loss": 0.6884,
-      "step": 65020
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.282651683554489e-05,
-      "loss": 0.7124,
-      "step": 65030
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2799241752696827e-05,
-      "loss": 0.6073,
-      "step": 65040
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.277196666984876e-05,
-      "loss": 0.7043,
-      "step": 65050
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2744691587000698e-05,
-      "loss": 0.7815,
-      "step": 65060
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2717416504152632e-05,
-      "loss": 0.598,
-      "step": 65070
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2690141421304566e-05,
-      "loss": 0.6718,
-      "step": 65080
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2662866338456503e-05,
-      "loss": 0.7787,
-      "step": 65090
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.263559125560844e-05,
-      "loss": 0.7819,
-      "step": 65100
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2608316172760378e-05,
-      "loss": 0.6598,
-      "step": 65110
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2581041089912312e-05,
-      "loss": 0.6502,
-      "step": 65120
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2556493515349052e-05,
-      "loss": 0.7526,
-      "step": 65130
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.252921843250099e-05,
-      "loss": 0.695,
-      "step": 65140
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2501943349652927e-05,
-      "loss": 0.5757,
-      "step": 65150
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.247466826680486e-05,
-      "loss": 0.6591,
-      "step": 65160
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2447393183956798e-05,
-      "loss": 0.7446,
-      "step": 65170
-    },
-    {
-      "epoch": 2.66,
-      "learning_rate": 2.2420118101108732e-05,
-      "loss": 0.6185,
-      "step": 65180
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.239284301826067e-05,
-      "loss": 0.6028,
-      "step": 65190
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2365567935412603e-05,
-      "loss": 0.7065,
-      "step": 65200
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.233829285256454e-05,
-      "loss": 0.7577,
-      "step": 65210
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2311017769716478e-05,
-      "loss": 0.6843,
-      "step": 65220
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2283742686868412e-05,
-      "loss": 0.665,
-      "step": 65230
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.225646760402035e-05,
-      "loss": 0.6755,
-      "step": 65240
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2229192521172283e-05,
-      "loss": 0.6185,
-      "step": 65250
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.220191743832422e-05,
-      "loss": 0.6318,
-      "step": 65260
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2174642355476154e-05,
-      "loss": 0.6562,
-      "step": 65270
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.214736727262809e-05,
-      "loss": 0.6814,
-      "step": 65280
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.212009218978003e-05,
-      "loss": 0.6933,
-      "step": 65290
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2092817106931963e-05,
-      "loss": 0.5802,
-      "step": 65300
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.20655420240839e-05,
-      "loss": 0.7229,
-      "step": 65310
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.2038266941235834e-05,
-      "loss": 0.698,
-      "step": 65320
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.201099185838777e-05,
-      "loss": 0.5972,
-      "step": 65330
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1983716775539705e-05,
-      "loss": 0.7086,
-      "step": 65340
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1956441692691643e-05,
-      "loss": 0.6606,
-      "step": 65350
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.192916660984358e-05,
-      "loss": 0.6368,
-      "step": 65360
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1901891526995514e-05,
-      "loss": 0.6266,
-      "step": 65370
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.187461644414745e-05,
-      "loss": 0.6176,
-      "step": 65380
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1847341361299385e-05,
-      "loss": 0.6117,
-      "step": 65390
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.182006627845132e-05,
-      "loss": 0.7045,
-      "step": 65400
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1792791195603256e-05,
-      "loss": 0.6177,
-      "step": 65410
-    },
-    {
-      "epoch": 2.67,
-      "learning_rate": 2.1765516112755194e-05,
-      "loss": 0.6978,
-      "step": 65420
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.173824102990713e-05,
-      "loss": 0.6536,
-      "step": 65430
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1710965947059065e-05,
-      "loss": 0.6354,
-      "step": 65440
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1683690864211002e-05,
-      "loss": 0.6301,
-      "step": 65450
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1656415781362936e-05,
-      "loss": 0.6422,
-      "step": 65460
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.162914069851487e-05,
-      "loss": 0.5423,
-      "step": 65470
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1601865615666807e-05,
-      "loss": 0.5633,
-      "step": 65480
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1574590532818745e-05,
-      "loss": 0.6249,
-      "step": 65490
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1547315449970682e-05,
-      "loss": 0.717,
-      "step": 65500
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1520040367122616e-05,
-      "loss": 0.6027,
-      "step": 65510
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1492765284274553e-05,
-      "loss": 0.6876,
-      "step": 65520
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1465490201426487e-05,
-      "loss": 0.586,
-      "step": 65530
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.143821511857842e-05,
-      "loss": 0.6032,
-      "step": 65540
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.141094003573036e-05,
-      "loss": 0.6133,
-      "step": 65550
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1383664952882296e-05,
-      "loss": 0.7003,
-      "step": 65560
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1356389870034233e-05,
-      "loss": 0.7325,
-      "step": 65570
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1329114787186167e-05,
-      "loss": 0.7049,
-      "step": 65580
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1301839704338104e-05,
-      "loss": 0.6342,
-      "step": 65590
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1274564621490038e-05,
-      "loss": 0.616,
-      "step": 65600
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1247289538641972e-05,
-      "loss": 0.6445,
-      "step": 65610
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.122001445579391e-05,
-      "loss": 0.6786,
-      "step": 65620
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1192739372945847e-05,
-      "loss": 0.6844,
-      "step": 65630
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1165464290097784e-05,
-      "loss": 0.7201,
-      "step": 65640
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1138189207249718e-05,
-      "loss": 0.5982,
-      "step": 65650
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.1110914124401655e-05,
-      "loss": 0.6382,
-      "step": 65660
-    },
-    {
-      "epoch": 2.68,
-      "learning_rate": 2.108363904155359e-05,
-      "loss": 0.5221,
-      "step": 65670
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.1056363958705523e-05,
-      "loss": 0.5746,
-      "step": 65680
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.102908887585746e-05,
-      "loss": 0.704,
-      "step": 65690
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.1001813793009398e-05,
-      "loss": 0.6847,
-      "step": 65700
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0974538710161335e-05,
-      "loss": 0.6628,
-      "step": 65710
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.094726362731327e-05,
-      "loss": 0.6301,
-      "step": 65720
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0919988544465206e-05,
-      "loss": 0.6969,
-      "step": 65730
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.089271346161714e-05,
-      "loss": 0.6736,
-      "step": 65740
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0865438378769074e-05,
-      "loss": 0.7122,
-      "step": 65750
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.083816329592101e-05,
-      "loss": 0.7101,
-      "step": 65760
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.081088821307295e-05,
-      "loss": 0.7252,
-      "step": 65770
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0783613130224886e-05,
-      "loss": 0.5992,
-      "step": 65780
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.075633804737682e-05,
-      "loss": 0.7607,
-      "step": 65790
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0729062964528757e-05,
-      "loss": 0.5399,
-      "step": 65800
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.070178788168069e-05,
-      "loss": 0.705,
-      "step": 65810
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0674512798832625e-05,
-      "loss": 0.6564,
-      "step": 65820
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0647237715984562e-05,
-      "loss": 0.7398,
-      "step": 65830
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.06199626331365e-05,
-      "loss": 0.711,
-      "step": 65840
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0592687550288434e-05,
-      "loss": 0.6753,
-      "step": 65850
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.056541246744037e-05,
-      "loss": 0.7823,
-      "step": 65860
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0538137384592308e-05,
-      "loss": 0.587,
-      "step": 65870
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0510862301744242e-05,
-      "loss": 0.7409,
-      "step": 65880
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0483587218896176e-05,
-      "loss": 0.5935,
-      "step": 65890
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.0456312136048113e-05,
-      "loss": 0.608,
-      "step": 65900
-    },
-    {
-      "epoch": 2.69,
-      "learning_rate": 2.042903705320005e-05,
-      "loss": 0.7212,
-      "step": 65910
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0401761970351985e-05,
-      "loss": 0.6159,
-      "step": 65920
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0374486887503922e-05,
-      "loss": 0.6696,
-      "step": 65930
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.034721180465586e-05,
-      "loss": 0.6767,
-      "step": 65940
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0319936721807793e-05,
-      "loss": 0.6365,
-      "step": 65950
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0292661638959727e-05,
-      "loss": 0.6144,
-      "step": 65960
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0265386556111664e-05,
-      "loss": 0.6192,
-      "step": 65970
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.02381114732636e-05,
-      "loss": 0.6301,
-      "step": 65980
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0210836390415535e-05,
-      "loss": 0.5803,
-      "step": 65990
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0183561307567473e-05,
-      "loss": 0.6198,
-      "step": 66000
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.015628622471941e-05,
-      "loss": 0.6549,
-      "step": 66010
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0129011141871344e-05,
-      "loss": 0.6381,
-      "step": 66020
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0101736059023278e-05,
-      "loss": 0.7099,
-      "step": 66030
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0074460976175215e-05,
-      "loss": 0.6335,
-      "step": 66040
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0047185893327153e-05,
-      "loss": 0.6624,
-      "step": 66050
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 2.0019910810479086e-05,
-      "loss": 0.6098,
-      "step": 66060
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9992635727631024e-05,
-      "loss": 0.5413,
-      "step": 66070
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.996536064478296e-05,
-      "loss": 0.5942,
-      "step": 66080
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9938085561934895e-05,
-      "loss": 0.6348,
-      "step": 66090
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.991081047908683e-05,
-      "loss": 0.6446,
-      "step": 66100
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9883535396238766e-05,
-      "loss": 0.7074,
-      "step": 66110
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9856260313390704e-05,
-      "loss": 0.7549,
-      "step": 66120
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9828985230542637e-05,
-      "loss": 0.7227,
-      "step": 66130
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9801710147694575e-05,
-      "loss": 0.6658,
-      "step": 66140
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9774435064846512e-05,
-      "loss": 0.6914,
-      "step": 66150
-    },
-    {
-      "epoch": 2.7,
-      "learning_rate": 1.9747159981998446e-05,
-      "loss": 0.6339,
-      "step": 66160
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.971988489915038e-05,
-      "loss": 0.6563,
-      "step": 66170
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9692609816302317e-05,
-      "loss": 0.6658,
-      "step": 66180
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9665334733454255e-05,
-      "loss": 0.7614,
-      "step": 66190
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.963805965060619e-05,
-      "loss": 0.6529,
-      "step": 66200
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9610784567758126e-05,
-      "loss": 0.7124,
-      "step": 66210
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9583509484910063e-05,
-      "loss": 0.6256,
-      "step": 66220
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9556234402061997e-05,
-      "loss": 0.5906,
-      "step": 66230
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.952895931921393e-05,
-      "loss": 0.6598,
-      "step": 66240
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9501684236365868e-05,
-      "loss": 0.6629,
-      "step": 66250
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9474409153517806e-05,
-      "loss": 0.7082,
-      "step": 66260
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.944713407066974e-05,
-      "loss": 0.6023,
-      "step": 66270
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9419858987821677e-05,
-      "loss": 0.6696,
-      "step": 66280
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9392583904973614e-05,
-      "loss": 0.7036,
-      "step": 66290
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9365308822125548e-05,
-      "loss": 0.6826,
-      "step": 66300
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9338033739277482e-05,
-      "loss": 0.583,
-      "step": 66310
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.931075865642942e-05,
-      "loss": 0.7126,
-      "step": 66320
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9283483573581357e-05,
-      "loss": 0.7455,
-      "step": 66330
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.925620849073329e-05,
-      "loss": 0.6191,
-      "step": 66340
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9228933407885228e-05,
-      "loss": 0.7333,
-      "step": 66350
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9201658325037165e-05,
-      "loss": 0.6313,
-      "step": 66360
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.91743832421891e-05,
-      "loss": 0.5786,
-      "step": 66370
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9147108159341033e-05,
-      "loss": 0.5793,
-      "step": 66380
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.911983307649297e-05,
-      "loss": 0.6551,
-      "step": 66390
-    },
-    {
-      "epoch": 2.71,
-      "learning_rate": 1.9092557993644908e-05,
-      "loss": 0.5899,
-      "step": 66400
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.906528291079684e-05,
-      "loss": 0.6857,
-      "step": 66410
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.903800782794878e-05,
-      "loss": 0.6664,
-      "step": 66420
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.9010732745100716e-05,
-      "loss": 0.6477,
-      "step": 66430
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.898345766225265e-05,
-      "loss": 0.7086,
-      "step": 66440
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8956182579404584e-05,
-      "loss": 0.5973,
-      "step": 66450
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.892890749655652e-05,
-      "loss": 0.6327,
-      "step": 66460
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.890163241370846e-05,
-      "loss": 0.6705,
-      "step": 66470
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8874357330860392e-05,
-      "loss": 0.5202,
-      "step": 66480
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.884708224801233e-05,
-      "loss": 0.5695,
-      "step": 66490
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8819807165164267e-05,
-      "loss": 0.6616,
-      "step": 66500
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.87925320823162e-05,
-      "loss": 0.6996,
-      "step": 66510
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8765256999468135e-05,
-      "loss": 0.61,
-      "step": 66520
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8737981916620072e-05,
-      "loss": 0.5539,
-      "step": 66530
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.871070683377201e-05,
-      "loss": 0.7529,
-      "step": 66540
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8683431750923943e-05,
-      "loss": 0.6055,
-      "step": 66550
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.865615666807588e-05,
-      "loss": 0.6381,
-      "step": 66560
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8628881585227818e-05,
-      "loss": 0.6149,
-      "step": 66570
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8601606502379752e-05,
-      "loss": 0.7378,
-      "step": 66580
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8574331419531686e-05,
-      "loss": 0.6623,
-      "step": 66590
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8547056336683623e-05,
-      "loss": 0.6273,
-      "step": 66600
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8519781253835557e-05,
-      "loss": 0.6865,
-      "step": 66610
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8492506170987494e-05,
-      "loss": 0.628,
-      "step": 66620
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.8465231088139432e-05,
-      "loss": 0.6089,
-      "step": 66630
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.843795600529137e-05,
-      "loss": 0.5772,
-      "step": 66640
-    },
-    {
-      "epoch": 2.72,
-      "learning_rate": 1.841340843072811e-05,
-      "loss": 0.6433,
-      "step": 66650
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8386133347880043e-05,
-      "loss": 0.6232,
-      "step": 66660
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.835885826503198e-05,
-      "loss": 0.8059,
-      "step": 66670
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8331583182183918e-05,
-      "loss": 0.6825,
-      "step": 66680
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8304308099335852e-05,
-      "loss": 0.621,
-      "step": 66690
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.827703301648779e-05,
-      "loss": 0.6735,
-      "step": 66700
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8249757933639723e-05,
-      "loss": 0.5667,
-      "step": 66710
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.822248285079166e-05,
-      "loss": 0.6043,
-      "step": 66720
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8195207767943594e-05,
-      "loss": 0.6487,
-      "step": 66730
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8167932685095532e-05,
-      "loss": 0.6017,
-      "step": 66740
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.814065760224747e-05,
-      "loss": 0.6063,
-      "step": 66750
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8113382519399403e-05,
-      "loss": 0.563,
-      "step": 66760
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.808610743655134e-05,
-      "loss": 0.6023,
-      "step": 66770
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8058832353703274e-05,
-      "loss": 0.6555,
-      "step": 66780
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.803155727085521e-05,
-      "loss": 0.7085,
-      "step": 66790
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.8004282188007145e-05,
-      "loss": 0.7341,
-      "step": 66800
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7977007105159083e-05,
-      "loss": 0.6152,
-      "step": 66810
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.794973202231102e-05,
-      "loss": 0.5997,
-      "step": 66820
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7922456939462954e-05,
-      "loss": 0.6463,
-      "step": 66830
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.789518185661489e-05,
-      "loss": 0.6864,
-      "step": 66840
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7867906773766825e-05,
-      "loss": 0.6049,
-      "step": 66850
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7840631690918763e-05,
-      "loss": 0.5488,
-      "step": 66860
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7813356608070696e-05,
-      "loss": 0.6719,
-      "step": 66870
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.7786081525222634e-05,
-      "loss": 0.7077,
-      "step": 66880
-    },
-    {
-      "epoch": 2.73,
-      "learning_rate": 1.775880644237457e-05,
-      "loss": 0.6452,
-      "step": 66890
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7731531359526505e-05,
-      "loss": 0.6746,
-      "step": 66900
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7704256276678442e-05,
-      "loss": 0.6808,
-      "step": 66910
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7676981193830376e-05,
-      "loss": 0.6778,
-      "step": 66920
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7649706110982314e-05,
-      "loss": 0.5567,
-      "step": 66930
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7622431028134247e-05,
-      "loss": 0.636,
-      "step": 66940
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7595155945286185e-05,
-      "loss": 0.7407,
-      "step": 66950
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7567880862438122e-05,
-      "loss": 0.6855,
-      "step": 66960
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7540605779590056e-05,
-      "loss": 0.5894,
-      "step": 66970
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7513330696741993e-05,
-      "loss": 0.6678,
-      "step": 66980
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7486055613893927e-05,
-      "loss": 0.7399,
-      "step": 66990
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7458780531045865e-05,
-      "loss": 0.6877,
-      "step": 67000
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.74315054481978e-05,
-      "loss": 0.8008,
-      "step": 67010
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7404230365349736e-05,
-      "loss": 0.5979,
-      "step": 67020
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7376955282501673e-05,
-      "loss": 0.6815,
-      "step": 67030
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7349680199653607e-05,
-      "loss": 0.6363,
-      "step": 67040
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7322405116805544e-05,
-      "loss": 0.6418,
-      "step": 67050
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7295130033957478e-05,
-      "loss": 0.6894,
-      "step": 67060
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7267854951109416e-05,
-      "loss": 0.624,
-      "step": 67070
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.724057986826135e-05,
-      "loss": 0.6422,
-      "step": 67080
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7213304785413287e-05,
-      "loss": 0.582,
-      "step": 67090
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7186029702565224e-05,
-      "loss": 0.7333,
-      "step": 67100
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7158754619717158e-05,
-      "loss": 0.6316,
-      "step": 67110
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7131479536869095e-05,
-      "loss": 0.6504,
-      "step": 67120
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.710420445402103e-05,
-      "loss": 0.6041,
-      "step": 67130
-    },
-    {
-      "epoch": 2.74,
-      "learning_rate": 1.7076929371172963e-05,
-      "loss": 0.6064,
-      "step": 67140
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.70496542883249e-05,
-      "loss": 0.6967,
-      "step": 67150
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.7022379205476838e-05,
-      "loss": 0.6647,
-      "step": 67160
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6995104122628775e-05,
-      "loss": 0.7978,
-      "step": 67170
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.696782903978071e-05,
-      "loss": 0.7398,
-      "step": 67180
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6940553956932646e-05,
-      "loss": 0.6253,
-      "step": 67190
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.691327887408458e-05,
-      "loss": 0.6991,
-      "step": 67200
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6886003791236514e-05,
-      "loss": 0.5878,
-      "step": 67210
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.685872870838845e-05,
-      "loss": 0.7645,
-      "step": 67220
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.683145362554039e-05,
-      "loss": 0.6693,
-      "step": 67230
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6804178542692326e-05,
-      "loss": 0.6951,
-      "step": 67240
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.677690345984426e-05,
-      "loss": 0.5131,
-      "step": 67250
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6749628376996197e-05,
-      "loss": 0.6671,
-      "step": 67260
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.672235329414813e-05,
-      "loss": 0.5953,
-      "step": 67270
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6695078211300065e-05,
-      "loss": 0.6053,
-      "step": 67280
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6667803128452002e-05,
-      "loss": 0.6338,
-      "step": 67290
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.664052804560394e-05,
-      "loss": 0.6232,
-      "step": 67300
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6613252962755877e-05,
-      "loss": 0.6492,
-      "step": 67310
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.658597787990781e-05,
-      "loss": 0.6834,
-      "step": 67320
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6558702797059748e-05,
-      "loss": 0.575,
-      "step": 67330
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6531427714211682e-05,
-      "loss": 0.661,
-      "step": 67340
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6504152631363616e-05,
-      "loss": 0.6732,
-      "step": 67350
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6476877548515553e-05,
-      "loss": 0.6675,
-      "step": 67360
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.644960246566749e-05,
-      "loss": 0.6531,
-      "step": 67370
-    },
-    {
-      "epoch": 2.75,
-      "learning_rate": 1.6422327382819428e-05,
-      "loss": 0.6997,
-      "step": 67380
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6395052299971362e-05,
-      "loss": 0.6042,
-      "step": 67390
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.63677772171233e-05,
-      "loss": 0.6035,
-      "step": 67400
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6340502134275233e-05,
-      "loss": 0.7115,
-      "step": 67410
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6313227051427167e-05,
-      "loss": 0.6855,
-      "step": 67420
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6285951968579104e-05,
-      "loss": 0.6244,
-      "step": 67430
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6258676885731042e-05,
-      "loss": 0.5893,
-      "step": 67440
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.623140180288298e-05,
-      "loss": 0.7064,
-      "step": 67450
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6204126720034913e-05,
-      "loss": 0.5194,
-      "step": 67460
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.617685163718685e-05,
-      "loss": 0.756,
-      "step": 67470
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6149576554338784e-05,
-      "loss": 0.6312,
-      "step": 67480
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6122301471490718e-05,
-      "loss": 0.5878,
-      "step": 67490
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6095026388642655e-05,
-      "loss": 0.6255,
-      "step": 67500
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6067751305794593e-05,
-      "loss": 0.6377,
-      "step": 67510
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6040476222946527e-05,
-      "loss": 0.6328,
-      "step": 67520
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.6013201140098464e-05,
-      "loss": 0.651,
-      "step": 67530
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.59859260572504e-05,
-      "loss": 0.5662,
-      "step": 67540
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5958650974402335e-05,
-      "loss": 0.6412,
-      "step": 67550
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.593137589155427e-05,
-      "loss": 0.6164,
-      "step": 67560
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5904100808706206e-05,
-      "loss": 0.714,
-      "step": 67570
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5876825725858144e-05,
-      "loss": 0.6094,
-      "step": 67580
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5849550643010078e-05,
-      "loss": 0.6287,
-      "step": 67590
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5822275560162015e-05,
-      "loss": 0.5547,
-      "step": 67600
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5795000477313952e-05,
-      "loss": 0.677,
-      "step": 67610
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.5767725394465886e-05,
-      "loss": 0.782,
-      "step": 67620
-    },
-    {
-      "epoch": 2.76,
-      "learning_rate": 1.574045031161782e-05,
-      "loss": 0.6068,
-      "step": 67630
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5713175228769757e-05,
-      "loss": 0.6366,
-      "step": 67640
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5685900145921695e-05,
-      "loss": 0.7421,
-      "step": 67650
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.565862506307363e-05,
-      "loss": 0.6467,
-      "step": 67660
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5631349980225566e-05,
-      "loss": 0.6515,
-      "step": 67670
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5604074897377503e-05,
-      "loss": 0.6959,
-      "step": 67680
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5576799814529437e-05,
-      "loss": 0.6387,
-      "step": 67690
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.554952473168137e-05,
-      "loss": 0.6804,
-      "step": 67700
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.552224964883331e-05,
-      "loss": 0.8019,
-      "step": 67710
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5494974565985246e-05,
-      "loss": 0.5819,
-      "step": 67720
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.546769948313718e-05,
-      "loss": 0.7152,
-      "step": 67730
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5440424400289117e-05,
-      "loss": 0.7191,
-      "step": 67740
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5413149317441054e-05,
-      "loss": 0.7442,
-      "step": 67750
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5385874234592988e-05,
-      "loss": 0.7252,
-      "step": 67760
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5358599151744922e-05,
-      "loss": 0.6558,
-      "step": 67770
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.533132406889686e-05,
-      "loss": 0.5973,
-      "step": 67780
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5304048986048797e-05,
-      "loss": 0.6532,
-      "step": 67790
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.527677390320073e-05,
-      "loss": 0.6523,
-      "step": 67800
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5249498820352668e-05,
-      "loss": 0.6311,
-      "step": 67810
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5222223737504604e-05,
-      "loss": 0.6755,
-      "step": 67820
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.519494865465654e-05,
-      "loss": 0.5879,
-      "step": 67830
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5167673571808475e-05,
-      "loss": 0.6063,
-      "step": 67840
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.514039848896041e-05,
-      "loss": 0.6106,
-      "step": 67850
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5113123406112348e-05,
-      "loss": 0.6254,
-      "step": 67860
-    },
-    {
-      "epoch": 2.77,
-      "learning_rate": 1.5085848323264282e-05,
-      "loss": 0.735,
-      "step": 67870
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.5058573240416219e-05,
-      "loss": 0.5757,
-      "step": 67880
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.5031298157568155e-05,
-      "loss": 0.6407,
-      "step": 67890
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.5004023074720088e-05,
-      "loss": 0.6204,
-      "step": 67900
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4976747991872026e-05,
-      "loss": 0.6078,
-      "step": 67910
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4949472909023961e-05,
-      "loss": 0.6992,
-      "step": 67920
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4922197826175899e-05,
-      "loss": 0.6782,
-      "step": 67930
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4894922743327833e-05,
-      "loss": 0.6066,
-      "step": 67940
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.486764766047977e-05,
-      "loss": 0.5824,
-      "step": 67950
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4840372577631706e-05,
-      "loss": 0.6018,
-      "step": 67960
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.481309749478364e-05,
-      "loss": 0.6822,
-      "step": 67970
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4785822411935577e-05,
-      "loss": 0.7388,
-      "step": 67980
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4758547329087512e-05,
-      "loss": 0.7127,
-      "step": 67990
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.473127224623945e-05,
-      "loss": 0.6318,
-      "step": 68000
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4703997163391384e-05,
-      "loss": 0.6173,
-      "step": 68010
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4676722080543321e-05,
-      "loss": 0.6816,
-      "step": 68020
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4649446997695256e-05,
-      "loss": 0.6616,
-      "step": 68030
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.462217191484719e-05,
-      "loss": 0.6784,
-      "step": 68040
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4594896831999128e-05,
-      "loss": 0.7507,
-      "step": 68050
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4567621749151063e-05,
-      "loss": 0.6531,
-      "step": 68060
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4540346666303e-05,
-      "loss": 0.5902,
-      "step": 68070
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4513071583454935e-05,
-      "loss": 0.7806,
-      "step": 68080
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4485796500606872e-05,
-      "loss": 0.6088,
-      "step": 68090
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4458521417758807e-05,
-      "loss": 0.6757,
-      "step": 68100
-    },
-    {
-      "epoch": 2.78,
-      "learning_rate": 1.4431246334910741e-05,
-      "loss": 0.77,
-      "step": 68110
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4403971252062679e-05,
-      "loss": 0.6497,
-      "step": 68120
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4376696169214614e-05,
-      "loss": 0.6922,
-      "step": 68130
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4349421086366552e-05,
-      "loss": 0.6617,
-      "step": 68140
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4322146003518486e-05,
-      "loss": 0.6169,
-      "step": 68150
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4294870920670423e-05,
-      "loss": 0.5758,
-      "step": 68160
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4267595837822358e-05,
-      "loss": 0.5484,
-      "step": 68170
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4240320754974292e-05,
-      "loss": 0.6898,
-      "step": 68180
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.421304567212623e-05,
-      "loss": 0.6475,
-      "step": 68190
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4185770589278165e-05,
-      "loss": 0.5966,
-      "step": 68200
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4158495506430103e-05,
-      "loss": 0.5935,
-      "step": 68210
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4131220423582037e-05,
-      "loss": 0.5765,
-      "step": 68220
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4103945340733974e-05,
-      "loss": 0.6705,
-      "step": 68230
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.407667025788591e-05,
-      "loss": 0.5584,
-      "step": 68240
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.4049395175037843e-05,
-      "loss": 0.6302,
-      "step": 68250
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.402212009218978e-05,
-      "loss": 0.6995,
-      "step": 68260
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3994845009341716e-05,
-      "loss": 0.7162,
-      "step": 68270
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.396756992649365e-05,
-      "loss": 0.685,
-      "step": 68280
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3940294843645588e-05,
-      "loss": 0.6441,
-      "step": 68290
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3913019760797525e-05,
-      "loss": 0.6162,
-      "step": 68300
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.388574467794946e-05,
-      "loss": 0.6664,
-      "step": 68310
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3858469595101394e-05,
-      "loss": 0.6741,
-      "step": 68320
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3831194512253332e-05,
-      "loss": 0.7145,
-      "step": 68330
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3803919429405269e-05,
-      "loss": 0.6141,
-      "step": 68340
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3776644346557201e-05,
-      "loss": 0.6721,
-      "step": 68350
-    },
-    {
-      "epoch": 2.79,
-      "learning_rate": 1.3749369263709139e-05,
-      "loss": 0.6287,
-      "step": 68360
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3722094180861076e-05,
-      "loss": 0.7014,
-      "step": 68370
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3694819098013011e-05,
-      "loss": 0.6415,
-      "step": 68380
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3667544015164945e-05,
-      "loss": 0.577,
-      "step": 68390
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3640268932316883e-05,
-      "loss": 0.6911,
-      "step": 68400
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.361299384946882e-05,
-      "loss": 0.6176,
-      "step": 68410
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3585718766620752e-05,
-      "loss": 0.6877,
-      "step": 68420
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.355844368377269e-05,
-      "loss": 0.5729,
-      "step": 68430
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3531168600924627e-05,
-      "loss": 0.5831,
-      "step": 68440
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3503893518076562e-05,
-      "loss": 0.6889,
-      "step": 68450
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3479345943513305e-05,
-      "loss": 0.6621,
-      "step": 68460
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3452070860665239e-05,
-      "loss": 0.6342,
-      "step": 68470
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3424795777817176e-05,
-      "loss": 0.6546,
-      "step": 68480
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3397520694969112e-05,
-      "loss": 0.6146,
-      "step": 68490
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3370245612121045e-05,
-      "loss": 0.5468,
-      "step": 68500
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3342970529272983e-05,
-      "loss": 0.6353,
-      "step": 68510
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.331569544642492e-05,
-      "loss": 0.6822,
-      "step": 68520
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3288420363576856e-05,
-      "loss": 0.6826,
-      "step": 68530
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.326114528072879e-05,
-      "loss": 0.58,
-      "step": 68540
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3233870197880727e-05,
-      "loss": 0.5848,
-      "step": 68550
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3206595115032663e-05,
-      "loss": 0.6857,
-      "step": 68560
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3179320032184596e-05,
-      "loss": 0.5652,
-      "step": 68570
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3152044949336534e-05,
-      "loss": 0.6547,
-      "step": 68580
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3124769866488471e-05,
-      "loss": 0.6568,
-      "step": 68590
-    },
-    {
-      "epoch": 2.8,
-      "learning_rate": 1.3097494783640407e-05,
-      "loss": 0.6635,
-      "step": 68600
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.307021970079234e-05,
-      "loss": 0.7007,
-      "step": 68610
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.3042944617944278e-05,
-      "loss": 0.6897,
-      "step": 68620
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.3015669535096214e-05,
-      "loss": 0.6646,
-      "step": 68630
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2988394452248147e-05,
-      "loss": 0.7238,
-      "step": 68640
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2961119369400085e-05,
-      "loss": 0.6469,
-      "step": 68650
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2933844286552022e-05,
-      "loss": 0.5794,
-      "step": 68660
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2906569203703958e-05,
-      "loss": 0.6226,
-      "step": 68670
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2879294120855892e-05,
-      "loss": 0.5273,
-      "step": 68680
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2852019038007829e-05,
-      "loss": 0.6467,
-      "step": 68690
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2824743955159766e-05,
-      "loss": 0.714,
-      "step": 68700
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2797468872311698e-05,
-      "loss": 0.6404,
-      "step": 68710
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2770193789463636e-05,
-      "loss": 0.6422,
-      "step": 68720
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2742918706615573e-05,
-      "loss": 0.6728,
-      "step": 68730
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2715643623767509e-05,
-      "loss": 0.6553,
-      "step": 68740
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2688368540919443e-05,
-      "loss": 0.686,
-      "step": 68750
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.266109345807138e-05,
-      "loss": 0.5272,
-      "step": 68760
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2633818375223317e-05,
-      "loss": 0.7093,
-      "step": 68770
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.260654329237525e-05,
-      "loss": 0.6927,
-      "step": 68780
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2579268209527187e-05,
-      "loss": 0.6562,
-      "step": 68790
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2551993126679124e-05,
-      "loss": 0.6419,
-      "step": 68800
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2524718043831058e-05,
-      "loss": 0.6011,
-      "step": 68810
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2497442960982994e-05,
-      "loss": 0.6342,
-      "step": 68820
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2470167878134931e-05,
-      "loss": 0.6497,
-      "step": 68830
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2442892795286866e-05,
-      "loss": 0.6748,
-      "step": 68840
-    },
-    {
-      "epoch": 2.81,
-      "learning_rate": 1.2415617712438802e-05,
-      "loss": 0.5436,
-      "step": 68850
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2388342629590738e-05,
-      "loss": 0.6625,
-      "step": 68860
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2361067546742673e-05,
-      "loss": 0.5852,
-      "step": 68870
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.233379246389461e-05,
-      "loss": 0.598,
-      "step": 68880
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2306517381046545e-05,
-      "loss": 0.6525,
-      "step": 68890
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2279242298198482e-05,
-      "loss": 0.6921,
-      "step": 68900
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2251967215350417e-05,
-      "loss": 0.6973,
-      "step": 68910
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2224692132502353e-05,
-      "loss": 0.6272,
-      "step": 68920
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2197417049654289e-05,
-      "loss": 0.6372,
-      "step": 68930
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2170141966806224e-05,
-      "loss": 0.8297,
-      "step": 68940
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2142866883958162e-05,
-      "loss": 0.6357,
-      "step": 68950
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2115591801110096e-05,
-      "loss": 0.7119,
-      "step": 68960
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2088316718262033e-05,
-      "loss": 0.5963,
-      "step": 68970
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2061041635413968e-05,
-      "loss": 0.6523,
-      "step": 68980
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.2033766552565904e-05,
-      "loss": 0.6649,
-      "step": 68990
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.200649146971784e-05,
-      "loss": 0.6188,
-      "step": 69000
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1979216386869775e-05,
-      "loss": 0.6657,
-      "step": 69010
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1951941304021713e-05,
-      "loss": 0.7317,
-      "step": 69020
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1924666221173647e-05,
-      "loss": 0.6325,
-      "step": 69030
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1897391138325584e-05,
-      "loss": 0.5591,
-      "step": 69040
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.187011605547752e-05,
-      "loss": 0.6799,
-      "step": 69050
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1842840972629455e-05,
-      "loss": 0.7487,
-      "step": 69060
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.181556588978139e-05,
-      "loss": 0.6637,
-      "step": 69070
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1788290806933326e-05,
-      "loss": 0.6275,
-      "step": 69080
-    },
-    {
-      "epoch": 2.82,
-      "learning_rate": 1.1761015724085264e-05,
-      "loss": 0.7117,
-      "step": 69090
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1733740641237198e-05,
-      "loss": 0.6209,
-      "step": 69100
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1706465558389135e-05,
-      "loss": 0.6345,
-      "step": 69110
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.167919047554107e-05,
-      "loss": 0.703,
-      "step": 69120
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1651915392693006e-05,
-      "loss": 0.5893,
-      "step": 69130
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1624640309844942e-05,
-      "loss": 0.6599,
-      "step": 69140
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1597365226996877e-05,
-      "loss": 0.6574,
-      "step": 69150
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1570090144148815e-05,
-      "loss": 0.6903,
-      "step": 69160
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1542815061300749e-05,
-      "loss": 0.6216,
-      "step": 69170
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1515539978452686e-05,
-      "loss": 0.7004,
-      "step": 69180
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1488264895604621e-05,
-      "loss": 0.6809,
-      "step": 69190
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1460989812756557e-05,
-      "loss": 0.6153,
-      "step": 69200
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1433714729908493e-05,
-      "loss": 0.5558,
-      "step": 69210
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1406439647060428e-05,
-      "loss": 0.6562,
-      "step": 69220
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1379164564212366e-05,
-      "loss": 0.6197,
-      "step": 69230
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.13518894813643e-05,
-      "loss": 0.6304,
-      "step": 69240
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1324614398516235e-05,
-      "loss": 0.622,
-      "step": 69250
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1297339315668172e-05,
-      "loss": 0.5848,
-      "step": 69260
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1270064232820108e-05,
-      "loss": 0.6642,
-      "step": 69270
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1242789149972044e-05,
-      "loss": 0.751,
-      "step": 69280
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.121551406712398e-05,
-      "loss": 0.7107,
-      "step": 69290
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1188238984275917e-05,
-      "loss": 0.6756,
-      "step": 69300
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.116096390142785e-05,
-      "loss": 0.5923,
-      "step": 69310
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1133688818579786e-05,
-      "loss": 0.6881,
-      "step": 69320
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1106413735731723e-05,
-      "loss": 0.6236,
-      "step": 69330
-    },
-    {
-      "epoch": 2.83,
-      "learning_rate": 1.1079138652883659e-05,
-      "loss": 0.699,
-      "step": 69340
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.1051863570035595e-05,
-      "loss": 0.6431,
-      "step": 69350
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.102458848718753e-05,
-      "loss": 0.6359,
-      "step": 69360
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0997313404339468e-05,
-      "loss": 0.6106,
-      "step": 69370
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0970038321491401e-05,
-      "loss": 0.5472,
-      "step": 69380
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0942763238643337e-05,
-      "loss": 0.6091,
-      "step": 69390
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0915488155795274e-05,
-      "loss": 0.6094,
-      "step": 69400
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.088821307294721e-05,
-      "loss": 0.6833,
-      "step": 69410
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0860937990099146e-05,
-      "loss": 0.6713,
-      "step": 69420
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0833662907251081e-05,
-      "loss": 0.683,
-      "step": 69430
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0806387824403019e-05,
-      "loss": 0.6816,
-      "step": 69440
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0779112741554952e-05,
-      "loss": 0.6602,
-      "step": 69450
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0751837658706888e-05,
-      "loss": 0.64,
-      "step": 69460
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0724562575858825e-05,
-      "loss": 0.7801,
-      "step": 69470
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0697287493010761e-05,
-      "loss": 0.6277,
-      "step": 69480
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0670012410162697e-05,
-      "loss": 0.6146,
-      "step": 69490
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0642737327314632e-05,
-      "loss": 0.5924,
-      "step": 69500
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0615462244466568e-05,
-      "loss": 0.6843,
-      "step": 69510
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0588187161618503e-05,
-      "loss": 0.6279,
-      "step": 69520
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0560912078770439e-05,
-      "loss": 0.5418,
-      "step": 69530
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0533636995922376e-05,
-      "loss": 0.693,
-      "step": 69540
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0506361913074312e-05,
-      "loss": 0.5616,
-      "step": 69550
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0479086830226248e-05,
-      "loss": 0.6073,
-      "step": 69560
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0451811747378183e-05,
-      "loss": 0.6604,
-      "step": 69570
-    },
-    {
-      "epoch": 2.84,
-      "learning_rate": 1.0424536664530119e-05,
-      "loss": 0.7153,
-      "step": 69580
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0397261581682054e-05,
-      "loss": 0.6312,
-      "step": 69590
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.036998649883399e-05,
-      "loss": 0.6371,
-      "step": 69600
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0342711415985927e-05,
-      "loss": 0.7145,
-      "step": 69610
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0315436333137863e-05,
-      "loss": 0.6521,
-      "step": 69620
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0288161250289799e-05,
-      "loss": 0.5759,
-      "step": 69630
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0260886167441734e-05,
-      "loss": 0.5547,
-      "step": 69640
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.023361108459367e-05,
-      "loss": 0.6414,
-      "step": 69650
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0206336001745605e-05,
-      "loss": 0.5358,
-      "step": 69660
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0179060918897541e-05,
-      "loss": 0.5992,
-      "step": 69670
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0151785836049478e-05,
-      "loss": 0.6416,
-      "step": 69680
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0124510753201414e-05,
-      "loss": 0.6175,
-      "step": 69690
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0097235670353348e-05,
-      "loss": 0.5627,
-      "step": 69700
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0069960587505285e-05,
-      "loss": 0.5577,
-      "step": 69710
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.004268550465722e-05,
-      "loss": 0.624,
-      "step": 69720
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 1.0015410421809156e-05,
-      "loss": 0.6703,
-      "step": 69730
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.988135338961092e-06,
-      "loss": 0.7185,
-      "step": 69740
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.96086025611303e-06,
-      "loss": 0.698,
-      "step": 69750
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.933585173264965e-06,
-      "loss": 0.7096,
-      "step": 69760
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.906310090416899e-06,
-      "loss": 0.5724,
-      "step": 69770
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.879035007568836e-06,
-      "loss": 0.6674,
-      "step": 69780
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.851759924720772e-06,
-      "loss": 0.6441,
-      "step": 69790
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.824484841872707e-06,
-      "loss": 0.6394,
-      "step": 69800
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.797209759024643e-06,
-      "loss": 0.5904,
-      "step": 69810
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.76993467617658e-06,
-      "loss": 0.6386,
-      "step": 69820
-    },
-    {
-      "epoch": 2.85,
-      "learning_rate": 9.742659593328516e-06,
-      "loss": 0.6178,
-      "step": 69830
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.71538451048045e-06,
-      "loss": 0.6603,
-      "step": 69840
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.688109427632387e-06,
-      "loss": 0.6939,
-      "step": 69850
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.660834344784323e-06,
-      "loss": 0.6008,
-      "step": 69860
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.633559261936258e-06,
-      "loss": 0.6297,
-      "step": 69870
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.606284179088194e-06,
-      "loss": 0.5372,
-      "step": 69880
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.57900909624013e-06,
-      "loss": 0.7241,
-      "step": 69890
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.551734013392067e-06,
-      "loss": 0.6254,
-      "step": 69900
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.524458930544001e-06,
-      "loss": 0.7147,
-      "step": 69910
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.497183847695938e-06,
-      "loss": 0.7139,
-      "step": 69920
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.469908764847874e-06,
-      "loss": 0.7372,
-      "step": 69930
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.44263368199981e-06,
-      "loss": 0.6129,
-      "step": 69940
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.415358599151745e-06,
-      "loss": 0.6462,
-      "step": 69950
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.38808351630368e-06,
-      "loss": 0.7048,
-      "step": 69960
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.360808433455618e-06,
-      "loss": 0.642,
-      "step": 69970
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.333533350607552e-06,
-      "loss": 0.6724,
-      "step": 69980
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.30625826775949e-06,
-      "loss": 0.6126,
-      "step": 69990
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.278983184911425e-06,
-      "loss": 0.6516,
-      "step": 70000
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.25170810206336e-06,
-      "loss": 0.7044,
-      "step": 70010
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.224433019215296e-06,
-      "loss": 0.6739,
-      "step": 70020
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.197157936367232e-06,
-      "loss": 0.7217,
-      "step": 70030
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.169882853519169e-06,
-      "loss": 0.663,
-      "step": 70040
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.142607770671103e-06,
-      "loss": 0.591,
-      "step": 70050
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.11533268782304e-06,
-      "loss": 0.7095,
-      "step": 70060
-    },
-    {
-      "epoch": 2.86,
-      "learning_rate": 9.088057604974976e-06,
-      "loss": 0.579,
-      "step": 70070
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 9.060782522126911e-06,
-      "loss": 0.6087,
-      "step": 70080
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 9.033507439278847e-06,
-      "loss": 0.6584,
-      "step": 70090
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 9.006232356430783e-06,
-      "loss": 0.5697,
-      "step": 70100
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.97895727358272e-06,
-      "loss": 0.7841,
-      "step": 70110
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.951682190734654e-06,
-      "loss": 0.5409,
-      "step": 70120
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.924407107886591e-06,
-      "loss": 0.7259,
-      "step": 70130
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.897132025038527e-06,
-      "loss": 0.7123,
-      "step": 70140
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.869856942190462e-06,
-      "loss": 0.6599,
-      "step": 70150
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.842581859342398e-06,
-      "loss": 0.6329,
-      "step": 70160
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.815306776494334e-06,
-      "loss": 0.6721,
-      "step": 70170
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.788031693646271e-06,
-      "loss": 0.7217,
-      "step": 70180
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.760756610798205e-06,
-      "loss": 0.6341,
-      "step": 70190
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.733481527950142e-06,
-      "loss": 0.6998,
-      "step": 70200
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.706206445102078e-06,
-      "loss": 0.7338,
-      "step": 70210
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.678931362254013e-06,
-      "loss": 0.6408,
-      "step": 70220
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.651656279405949e-06,
-      "loss": 0.6517,
-      "step": 70230
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.624381196557885e-06,
-      "loss": 0.7163,
-      "step": 70240
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.597106113709822e-06,
-      "loss": 0.7438,
-      "step": 70250
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.569831030861756e-06,
-      "loss": 0.5321,
-      "step": 70260
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.542555948013691e-06,
-      "loss": 0.678,
-      "step": 70270
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.515280865165629e-06,
-      "loss": 0.7042,
-      "step": 70280
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.488005782317564e-06,
-      "loss": 0.5493,
-      "step": 70290
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.4607306994695e-06,
-      "loss": 0.5957,
-      "step": 70300
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.433455616621436e-06,
-      "loss": 0.6211,
-      "step": 70310
-    },
-    {
-      "epoch": 2.87,
-      "learning_rate": 8.406180533773373e-06,
-      "loss": 0.6113,
-      "step": 70320
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.378905450925307e-06,
-      "loss": 0.6655,
-      "step": 70330
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.351630368077242e-06,
-      "loss": 0.6841,
-      "step": 70340
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.32435528522918e-06,
-      "loss": 0.6829,
-      "step": 70350
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.297080202381115e-06,
-      "loss": 0.7013,
-      "step": 70360
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.269805119533051e-06,
-      "loss": 0.5767,
-      "step": 70370
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.242530036684987e-06,
-      "loss": 0.7044,
-      "step": 70380
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.215254953836924e-06,
-      "loss": 0.5959,
-      "step": 70390
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.187979870988858e-06,
-      "loss": 0.717,
-      "step": 70400
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.160704788140793e-06,
-      "loss": 0.5735,
-      "step": 70410
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.13342970529273e-06,
-      "loss": 0.6863,
-      "step": 70420
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.106154622444666e-06,
-      "loss": 0.7578,
-      "step": 70430
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.078879539596602e-06,
-      "loss": 0.61,
-      "step": 70440
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.051604456748538e-06,
-      "loss": 0.6754,
-      "step": 70450
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 8.024329373900473e-06,
-      "loss": 0.6099,
-      "step": 70460
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.997054291052409e-06,
-      "loss": 0.6685,
-      "step": 70470
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.969779208204344e-06,
-      "loss": 0.7252,
-      "step": 70480
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.942504125356282e-06,
-      "loss": 0.6579,
-      "step": 70490
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.915229042508217e-06,
-      "loss": 0.6545,
-      "step": 70500
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.887953959660153e-06,
-      "loss": 0.636,
-      "step": 70510
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.860678876812089e-06,
-      "loss": 0.6488,
-      "step": 70520
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.833403793964024e-06,
-      "loss": 0.6817,
-      "step": 70530
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.80612871111596e-06,
-      "loss": 0.632,
-      "step": 70540
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.778853628267895e-06,
-      "loss": 0.7153,
-      "step": 70550
-    },
-    {
-      "epoch": 2.88,
-      "learning_rate": 7.751578545419833e-06,
-      "loss": 0.6343,
-      "step": 70560
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.724303462571768e-06,
-      "loss": 0.685,
-      "step": 70570
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.697028379723704e-06,
-      "loss": 0.656,
-      "step": 70580
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.66975329687564e-06,
-      "loss": 0.6831,
-      "step": 70590
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.642478214027575e-06,
-      "loss": 0.5392,
-      "step": 70600
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.615203131179512e-06,
-      "loss": 0.6376,
-      "step": 70610
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.587928048331447e-06,
-      "loss": 0.5481,
-      "step": 70620
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.560652965483384e-06,
-      "loss": 0.7426,
-      "step": 70630
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.5333778826353185e-06,
-      "loss": 0.6268,
-      "step": 70640
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.506102799787254e-06,
-      "loss": 0.734,
-      "step": 70650
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.478827716939191e-06,
-      "loss": 0.5435,
-      "step": 70660
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.451552634091126e-06,
-      "loss": 0.6924,
-      "step": 70670
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.424277551243063e-06,
-      "loss": 0.6657,
-      "step": 70680
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.397002468394998e-06,
-      "loss": 0.5606,
-      "step": 70690
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.369727385546935e-06,
-      "loss": 0.5781,
-      "step": 70700
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.3424523026988695e-06,
-      "loss": 0.6706,
-      "step": 70710
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.315177219850805e-06,
-      "loss": 0.6753,
-      "step": 70720
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.287902137002742e-06,
-      "loss": 0.6326,
-      "step": 70730
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.260627054154677e-06,
-      "loss": 0.5872,
-      "step": 70740
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.233351971306614e-06,
-      "loss": 0.7026,
-      "step": 70750
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.206076888458549e-06,
-      "loss": 0.6518,
-      "step": 70760
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.178801805610486e-06,
-      "loss": 0.6835,
-      "step": 70770
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.1515267227624205e-06,
-      "loss": 0.6136,
-      "step": 70780
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.124251639914356e-06,
-      "loss": 0.5614,
-      "step": 70790
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.0969765570662926e-06,
-      "loss": 0.7129,
-      "step": 70800
-    },
-    {
-      "epoch": 2.89,
-      "learning_rate": 7.069701474218228e-06,
-      "loss": 0.6793,
-      "step": 70810
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 7.042426391370165e-06,
-      "loss": 0.6117,
-      "step": 70820
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 7.0151513085221e-06,
-      "loss": 0.6055,
-      "step": 70830
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.987876225674035e-06,
-      "loss": 0.69,
-      "step": 70840
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.9606011428259715e-06,
-      "loss": 0.6286,
-      "step": 70850
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.933326059977907e-06,
-      "loss": 0.6441,
-      "step": 70860
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.9060509771298436e-06,
-      "loss": 0.6356,
-      "step": 70870
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.878775894281779e-06,
-      "loss": 0.5369,
-      "step": 70880
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.851500811433716e-06,
-      "loss": 0.6157,
-      "step": 70890
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.824225728585651e-06,
-      "loss": 0.7262,
-      "step": 70900
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.796950645737586e-06,
-      "loss": 0.6474,
-      "step": 70910
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.7696755628895225e-06,
-      "loss": 0.6158,
-      "step": 70920
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.742400480041458e-06,
-      "loss": 0.6321,
-      "step": 70930
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.7151253971933946e-06,
-      "loss": 0.6324,
-      "step": 70940
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.68785031434533e-06,
-      "loss": 0.7346,
-      "step": 70950
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.660575231497267e-06,
-      "loss": 0.6873,
-      "step": 70960
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.633300148649202e-06,
-      "loss": 0.6539,
-      "step": 70970
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.606025065801137e-06,
-      "loss": 0.6365,
-      "step": 70980
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.5787499829530735e-06,
-      "loss": 0.6059,
-      "step": 70990
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.551474900105009e-06,
-      "loss": 0.7049,
-      "step": 71000
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.5241998172569456e-06,
-      "loss": 0.5958,
-      "step": 71010
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.496924734408881e-06,
-      "loss": 0.7632,
-      "step": 71020
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.469649651560816e-06,
-      "loss": 0.5918,
-      "step": 71030
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.442374568712753e-06,
-      "loss": 0.5858,
-      "step": 71040
-    },
-    {
-      "epoch": 2.9,
-      "learning_rate": 6.415099485864688e-06,
-      "loss": 0.6495,
-      "step": 71050
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.3878244030166245e-06,
-      "loss": 0.6554,
-      "step": 71060
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.36054932016856e-06,
-      "loss": 0.6277,
-      "step": 71070
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.3332742373204965e-06,
-      "loss": 0.6177,
-      "step": 71080
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.305999154472432e-06,
-      "loss": 0.6398,
-      "step": 71090
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.278724071624367e-06,
-      "loss": 0.6377,
-      "step": 71100
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.251448988776304e-06,
-      "loss": 0.6454,
-      "step": 71110
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.22417390592824e-06,
-      "loss": 0.7009,
-      "step": 71120
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.1968988230801755e-06,
-      "loss": 0.5754,
-      "step": 71130
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.169623740232111e-06,
-      "loss": 0.7184,
-      "step": 71140
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.142348657384047e-06,
-      "loss": 0.6647,
-      "step": 71150
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.115073574535983e-06,
-      "loss": 0.5991,
-      "step": 71160
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.087798491687919e-06,
-      "loss": 0.7572,
-      "step": 71170
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.060523408839855e-06,
-      "loss": 0.7151,
-      "step": 71180
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.03324832599179e-06,
-      "loss": 0.5979,
-      "step": 71190
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 6.0059732431437265e-06,
-      "loss": 0.7243,
-      "step": 71200
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.978698160295662e-06,
-      "loss": 0.6855,
-      "step": 71210
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.951423077447598e-06,
-      "loss": 0.6951,
-      "step": 71220
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.924147994599534e-06,
-      "loss": 0.6941,
-      "step": 71230
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.89687291175147e-06,
-      "loss": 0.7077,
-      "step": 71240
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.869597828903405e-06,
-      "loss": 0.6382,
-      "step": 71250
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.842322746055341e-06,
-      "loss": 0.6597,
-      "step": 71260
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.8150476632072774e-06,
-      "loss": 0.6353,
-      "step": 71270
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.787772580359213e-06,
-      "loss": 0.6483,
-      "step": 71280
-    },
-    {
-      "epoch": 2.91,
-      "learning_rate": 5.760497497511149e-06,
-      "loss": 0.6214,
-      "step": 71290
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.733222414663085e-06,
-      "loss": 0.5866,
-      "step": 71300
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.705947331815021e-06,
-      "loss": 0.7109,
-      "step": 71310
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.678672248966956e-06,
-      "loss": 0.6004,
-      "step": 71320
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.651397166118892e-06,
-      "loss": 0.6533,
-      "step": 71330
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.6241220832708284e-06,
-      "loss": 0.7581,
-      "step": 71340
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.596847000422764e-06,
-      "loss": 0.6764,
-      "step": 71350
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.5695719175747e-06,
-      "loss": 0.6461,
-      "step": 71360
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.542296834726636e-06,
-      "loss": 0.522,
-      "step": 71370
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.515021751878571e-06,
-      "loss": 0.5853,
-      "step": 71380
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.487746669030507e-06,
-      "loss": 0.6317,
-      "step": 71390
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.460471586182443e-06,
-      "loss": 0.5991,
-      "step": 71400
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.4331965033343794e-06,
-      "loss": 0.6546,
-      "step": 71410
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.405921420486315e-06,
-      "loss": 0.6351,
-      "step": 71420
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.378646337638251e-06,
-      "loss": 0.5447,
-      "step": 71430
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.351371254790187e-06,
-      "loss": 0.5877,
-      "step": 71440
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.324096171942122e-06,
-      "loss": 0.6314,
-      "step": 71450
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.296821089094058e-06,
-      "loss": 0.6328,
-      "step": 71460
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.269546006245994e-06,
-      "loss": 0.5739,
-      "step": 71470
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.2422709233979304e-06,
-      "loss": 0.706,
-      "step": 71480
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.214995840549866e-06,
-      "loss": 0.6237,
-      "step": 71490
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.187720757701802e-06,
-      "loss": 0.7001,
-      "step": 71500
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.160445674853737e-06,
-      "loss": 0.6438,
-      "step": 71510
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.133170592005673e-06,
-      "loss": 0.7352,
-      "step": 71520
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.105895509157609e-06,
-      "loss": 0.6356,
-      "step": 71530
-    },
-    {
-      "epoch": 2.92,
-      "learning_rate": 5.078620426309545e-06,
-      "loss": 0.7648,
-      "step": 71540
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 5.051345343461481e-06,
-      "loss": 0.6565,
-      "step": 71550
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 5.024070260613417e-06,
-      "loss": 0.6002,
-      "step": 71560
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.996795177765353e-06,
-      "loss": 0.5773,
-      "step": 71570
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.969520094917288e-06,
-      "loss": 0.5606,
-      "step": 71580
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.942245012069225e-06,
-      "loss": 0.6328,
-      "step": 71590
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.91496992922116e-06,
-      "loss": 0.4784,
-      "step": 71600
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.887694846373096e-06,
-      "loss": 0.7301,
-      "step": 71610
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.860419763525032e-06,
-      "loss": 0.6109,
-      "step": 71620
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.833144680676968e-06,
-      "loss": 0.5946,
-      "step": 71630
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.805869597828904e-06,
-      "loss": 0.6304,
-      "step": 71640
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.778594514980839e-06,
-      "loss": 0.6493,
-      "step": 71650
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.751319432132776e-06,
-      "loss": 0.7321,
-      "step": 71660
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.724044349284711e-06,
-      "loss": 0.6757,
-      "step": 71670
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.696769266436647e-06,
-      "loss": 0.6735,
-      "step": 71680
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.669494183588583e-06,
-      "loss": 0.5575,
-      "step": 71690
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.642219100740518e-06,
-      "loss": 0.64,
-      "step": 71700
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.614944017892455e-06,
-      "loss": 0.6667,
-      "step": 71710
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.58766893504439e-06,
-      "loss": 0.6401,
-      "step": 71720
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.560393852196327e-06,
-      "loss": 0.6497,
-      "step": 71730
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.533118769348262e-06,
-      "loss": 0.613,
-      "step": 71740
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.505843686500198e-06,
-      "loss": 0.6711,
-      "step": 71750
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.4785686036521335e-06,
-      "loss": 0.7102,
-      "step": 71760
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.451293520804069e-06,
-      "loss": 0.7634,
-      "step": 71770
-    },
-    {
-      "epoch": 2.93,
-      "learning_rate": 4.424018437956006e-06,
-      "loss": 0.6615,
-      "step": 71780
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.396743355107941e-06,
-      "loss": 0.615,
-      "step": 71790
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.369468272259878e-06,
-      "loss": 0.6131,
-      "step": 71800
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.342193189411813e-06,
-      "loss": 0.6141,
-      "step": 71810
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.314918106563749e-06,
-      "loss": 0.6134,
-      "step": 71820
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.2876430237156845e-06,
-      "loss": 0.7031,
-      "step": 71830
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.26036794086762e-06,
-      "loss": 0.6672,
-      "step": 71840
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.233092858019557e-06,
-      "loss": 0.574,
-      "step": 71850
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.205817775171492e-06,
-      "loss": 0.7384,
-      "step": 71860
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.178542692323429e-06,
-      "loss": 0.6087,
-      "step": 71870
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.151267609475364e-06,
-      "loss": 0.6105,
-      "step": 71880
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.1239925266273e-06,
-      "loss": 0.713,
-      "step": 71890
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.0967174437792355e-06,
-      "loss": 0.6008,
-      "step": 71900
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.069442360931171e-06,
-      "loss": 0.6759,
-      "step": 71910
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.042167278083108e-06,
-      "loss": 0.5884,
-      "step": 71920
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 4.014892195235043e-06,
-      "loss": 0.6816,
-      "step": 71930
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.98761711238698e-06,
-      "loss": 0.6637,
-      "step": 71940
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.9603420295389144e-06,
-      "loss": 0.5973,
-      "step": 71950
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.933066946690851e-06,
-      "loss": 0.6007,
-      "step": 71960
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.9057918638427865e-06,
-      "loss": 0.7395,
-      "step": 71970
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.878516780994722e-06,
-      "loss": 0.5884,
-      "step": 71980
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.851241698146659e-06,
-      "loss": 0.59,
-      "step": 71990
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.823966615298594e-06,
-      "loss": 0.6695,
-      "step": 72000
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.7966915324505303e-06,
-      "loss": 0.6312,
-      "step": 72010
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.7694164496024654e-06,
-      "loss": 0.6807,
-      "step": 72020
-    },
-    {
-      "epoch": 2.94,
-      "learning_rate": 3.7421413667544015e-06,
-      "loss": 0.8175,
-      "step": 72030
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.7148662839063375e-06,
-      "loss": 0.7201,
-      "step": 72040
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.6875912010582736e-06,
-      "loss": 0.6818,
-      "step": 72050
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.6603161182102096e-06,
-      "loss": 0.563,
-      "step": 72060
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.633041035362145e-06,
-      "loss": 0.6215,
-      "step": 72070
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.605765952514081e-06,
-      "loss": 0.7763,
-      "step": 72080
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.5784908696660164e-06,
-      "loss": 0.6892,
-      "step": 72090
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.5512157868179525e-06,
-      "loss": 0.5626,
-      "step": 72100
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.5239407039698885e-06,
-      "loss": 0.7254,
-      "step": 72110
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.4966656211218245e-06,
-      "loss": 0.6794,
-      "step": 72120
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.4693905382737606e-06,
-      "loss": 0.6034,
-      "step": 72130
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.4421154554256958e-06,
-      "loss": 0.6736,
-      "step": 72140
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.414840372577632e-06,
-      "loss": 0.6613,
-      "step": 72150
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.3875652897295674e-06,
-      "loss": 0.7008,
-      "step": 72160
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.3602902068815035e-06,
-      "loss": 0.647,
-      "step": 72170
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.3330151240334395e-06,
-      "loss": 0.6602,
-      "step": 72180
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.3057400411853755e-06,
-      "loss": 0.6576,
-      "step": 72190
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.2784649583373116e-06,
-      "loss": 0.6847,
-      "step": 72200
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.2511898754892468e-06,
-      "loss": 0.6508,
-      "step": 72210
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.223914792641183e-06,
-      "loss": 0.6639,
-      "step": 72220
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.1966397097931184e-06,
-      "loss": 0.6307,
-      "step": 72230
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.1693646269450545e-06,
-      "loss": 0.6107,
-      "step": 72240
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.1420895440969905e-06,
-      "loss": 0.5685,
-      "step": 72250
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.114814461248926e-06,
-      "loss": 0.6735,
-      "step": 72260
-    },
-    {
-      "epoch": 2.95,
-      "learning_rate": 3.087539378400862e-06,
-      "loss": 0.5944,
-      "step": 72270
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 3.060264295552798e-06,
-      "loss": 0.6488,
-      "step": 72280
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 3.032989212704734e-06,
-      "loss": 0.664,
-      "step": 72290
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 3.0057141298566694e-06,
-      "loss": 0.6471,
-      "step": 72300
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.9784390470086054e-06,
-      "loss": 0.7396,
-      "step": 72310
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.951163964160541e-06,
-      "loss": 0.6222,
-      "step": 72320
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.923888881312477e-06,
-      "loss": 0.6002,
-      "step": 72330
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.896613798464413e-06,
-      "loss": 0.69,
-      "step": 72340
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.8693387156163488e-06,
-      "loss": 0.5995,
-      "step": 72350
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.842063632768285e-06,
-      "loss": 0.5916,
-      "step": 72360
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.814788549920221e-06,
-      "loss": 0.5983,
-      "step": 72370
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.787513467072156e-06,
-      "loss": 0.6313,
-      "step": 72380
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.760238384224092e-06,
-      "loss": 0.5884,
-      "step": 72390
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.732963301376028e-06,
-      "loss": 0.6425,
-      "step": 72400
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.7056882185279637e-06,
-      "loss": 0.6268,
-      "step": 72410
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.6784131356798997e-06,
-      "loss": 0.5931,
-      "step": 72420
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.6511380528318358e-06,
-      "loss": 0.6465,
-      "step": 72430
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.623862969983772e-06,
-      "loss": 0.6452,
-      "step": 72440
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.5965878871357074e-06,
-      "loss": 0.6712,
-      "step": 72450
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.569312804287643e-06,
-      "loss": 0.6303,
-      "step": 72460
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.542037721439579e-06,
-      "loss": 0.6088,
-      "step": 72470
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.5147626385915147e-06,
-      "loss": 0.6296,
-      "step": 72480
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.4874875557434507e-06,
-      "loss": 0.6764,
-      "step": 72490
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.4602124728953868e-06,
-      "loss": 0.4918,
-      "step": 72500
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.4329373900473224e-06,
-      "loss": 0.6309,
-      "step": 72510
-    },
-    {
-      "epoch": 2.96,
-      "learning_rate": 2.4056623071992584e-06,
-      "loss": 0.7001,
-      "step": 72520
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.378387224351194e-06,
-      "loss": 0.6501,
-      "step": 72530
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.3511121415031297e-06,
-      "loss": 0.5761,
-      "step": 72540
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.3265645669398723e-06,
-      "loss": 0.6905,
-      "step": 72550
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.299289484091808e-06,
-      "loss": 0.6861,
-      "step": 72560
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.272014401243744e-06,
-      "loss": 0.6428,
-      "step": 72570
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.24473931839568e-06,
-      "loss": 0.6587,
-      "step": 72580
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.2174642355476156e-06,
-      "loss": 0.6556,
-      "step": 72590
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.1901891526995512e-06,
-      "loss": 0.6057,
-      "step": 72600
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.1629140698514873e-06,
-      "loss": 0.7332,
-      "step": 72610
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.135638987003423e-06,
-      "loss": 0.5368,
-      "step": 72620
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.108363904155359e-06,
-      "loss": 0.6745,
-      "step": 72630
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.081088821307295e-06,
-      "loss": 0.6694,
-      "step": 72640
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.053813738459231e-06,
-      "loss": 0.7214,
-      "step": 72650
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 2.0265386556111666e-06,
-      "loss": 0.6048,
-      "step": 72660
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.999263572763102e-06,
-      "loss": 0.6793,
-      "step": 72670
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.9719884899150382e-06,
-      "loss": 0.5259,
-      "step": 72680
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.944713407066974e-06,
-      "loss": 0.6048,
-      "step": 72690
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.91743832421891e-06,
-      "loss": 0.6653,
-      "step": 72700
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.890163241370846e-06,
-      "loss": 0.6321,
-      "step": 72710
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.8628881585227816e-06,
-      "loss": 0.6127,
-      "step": 72720
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.8356130756747174e-06,
-      "loss": 0.7727,
-      "step": 72730
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.8083379928266534e-06,
-      "loss": 0.6043,
-      "step": 72740
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.781062909978589e-06,
-      "loss": 0.6851,
-      "step": 72750
-    },
-    {
-      "epoch": 2.97,
-      "learning_rate": 1.753787827130525e-06,
-      "loss": 0.5777,
-      "step": 72760
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.7265127442824609e-06,
-      "loss": 0.6297,
-      "step": 72770
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.6992376614343965e-06,
-      "loss": 0.6103,
-      "step": 72780
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.6719625785863325e-06,
-      "loss": 0.6214,
-      "step": 72790
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.6446874957382684e-06,
-      "loss": 0.6742,
-      "step": 72800
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.617412412890204e-06,
-      "loss": 0.7117,
-      "step": 72810
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.59013733004214e-06,
-      "loss": 0.5799,
-      "step": 72820
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.562862247194076e-06,
-      "loss": 0.7465,
-      "step": 72830
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.5355871643460117e-06,
-      "loss": 0.6187,
-      "step": 72840
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.5083120814979475e-06,
-      "loss": 0.6351,
-      "step": 72850
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.4810369986498835e-06,
-      "loss": 0.6841,
-      "step": 72860
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.4537619158018194e-06,
-      "loss": 0.5603,
-      "step": 72870
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.4264868329537552e-06,
-      "loss": 0.6664,
-      "step": 72880
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.399211750105691e-06,
-      "loss": 0.6791,
-      "step": 72890
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.3719366672576268e-06,
-      "loss": 0.574,
-      "step": 72900
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.3446615844095627e-06,
-      "loss": 0.5769,
-      "step": 72910
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.3173865015614985e-06,
-      "loss": 0.6231,
-      "step": 72920
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.2901114187134343e-06,
-      "loss": 0.5911,
-      "step": 72930
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.2628363358653704e-06,
-      "loss": 0.6365,
-      "step": 72940
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.2355612530173062e-06,
-      "loss": 0.6607,
-      "step": 72950
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.208286170169242e-06,
-      "loss": 0.6219,
-      "step": 72960
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.1810110873211778e-06,
-      "loss": 0.6662,
-      "step": 72970
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.1537360044731137e-06,
-      "loss": 0.6317,
-      "step": 72980
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.1264609216250495e-06,
-      "loss": 0.5823,
-      "step": 72990
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.0991858387769853e-06,
-      "loss": 0.6241,
-      "step": 73000
-    },
-    {
-      "epoch": 2.98,
-      "learning_rate": 1.0719107559289211e-06,
-      "loss": 0.6455,
-      "step": 73010
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 1.0446356730808572e-06,
-      "loss": 0.6394,
-      "step": 73020
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 1.0173605902327928e-06,
-      "loss": 0.6429,
-      "step": 73030
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 9.900855073847286e-07,
-      "loss": 0.6191,
-      "step": 73040
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 9.628104245366646e-07,
-      "loss": 0.6163,
-      "step": 73050
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 9.355353416886004e-07,
-      "loss": 0.6741,
-      "step": 73060
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 9.082602588405363e-07,
-      "loss": 0.7418,
-      "step": 73070
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 8.809851759924721e-07,
-      "loss": 0.6473,
-      "step": 73080
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 8.53710093144408e-07,
-      "loss": 0.6583,
-      "step": 73090
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 8.264350102963439e-07,
-      "loss": 0.6082,
-      "step": 73100
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 7.991599274482796e-07,
-      "loss": 0.6763,
-      "step": 73110
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 7.718848446002155e-07,
-      "loss": 0.7389,
-      "step": 73120
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 7.446097617521514e-07,
-      "loss": 0.7023,
-      "step": 73130
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 7.173346789040872e-07,
-      "loss": 0.6559,
-      "step": 73140
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 6.90059596056023e-07,
-      "loss": 0.6342,
-      "step": 73150
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 6.627845132079589e-07,
-      "loss": 0.6762,
-      "step": 73160
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 6.355094303598948e-07,
-      "loss": 0.6468,
-      "step": 73170
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 6.082343475118306e-07,
-      "loss": 0.643,
-      "step": 73180
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 5.809592646637664e-07,
-      "loss": 0.7619,
-      "step": 73190
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 5.536841818157024e-07,
-      "loss": 0.6122,
-      "step": 73200
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 5.264090989676382e-07,
-      "loss": 0.6877,
-      "step": 73210
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 4.99134016119574e-07,
-      "loss": 0.6508,
-      "step": 73220
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 4.7185893327150983e-07,
-      "loss": 0.586,
-      "step": 73230
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 4.445838504234457e-07,
-      "loss": 0.6176,
-      "step": 73240
-    },
-    {
-      "epoch": 2.99,
-      "learning_rate": 4.173087675753815e-07,
-      "loss": 0.6197,
-      "step": 73250
-    }
-  ],
-  "max_steps": 73377,
-  "num_train_epochs": 3,
-  "total_flos": 3.1772664426725376e+16,
-  "trial_name": null,
-  "trial_params": null
-}