diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,79036 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9999544439888843,
+  "global_step": 65850,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 9.4062,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 8.9437,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 8.8313,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 8.4375,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-05,
+      "loss": 8.2531,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 7.9813,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 7.9875,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 7.825,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000135,
+      "loss": 7.6281,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015,
+      "loss": 7.4781,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000165,
+      "loss": 7.4969,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 7.8219,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000195,
+      "loss": 7.9813,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 7.5187,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000225,
+      "loss": 7.5094,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 7.6813,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 7.5812,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00027,
+      "loss": 7.6594,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000285,
+      "loss": 7.4219,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0003,
+      "loss": 7.5594,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999999957197428,
+      "loss": 7.6719,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999998287897126,
+      "loss": 7.5687,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999999614776863,
+      "loss": 7.5875,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299999931515889,
+      "loss": 8.2438,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999989299358117,
+      "loss": 8.0594,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299999845910765,
+      "loss": 7.9562,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999979026744313,
+      "loss": 7.6375,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999997260636187,
+      "loss": 7.5156,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999996532992954,
+      "loss": 7.3563,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999995719744775,
+      "loss": 7.2219,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999948208916954,
+      "loss": 7.15,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999938364337663,
+      "loss": 7.2219,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999992766371044,
+      "loss": 7.1,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999916107035894,
+      "loss": 6.9844,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999903694314696,
+      "loss": 6.95,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999989042554754,
+      "loss": 6.8125,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999987630073519,
+      "loss": 6.775,
+      "step": 185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999861319878455,
+      "loss": 6.7438,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999845482978185,
+      "loss": 6.7313,
+      "step": 195
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999828790035286,
+      "loss": 6.6094,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999981124105071,
+      "loss": 6.6094,
+      "step": 205
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999979283602546,
+      "loss": 6.575,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999977357496058,
+      "loss": 6.6,
+      "step": 215
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999753457857177,
+      "loss": 6.5812,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299997324847164,
+      "loss": 6.65,
+      "step": 225
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999710655539437,
+      "loss": 6.6344,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999687970327536,
+      "loss": 6.575,
+      "step": 235
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999664429082,
+      "loss": 6.5625,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999640031804165,
+      "loss": 6.5312,
+      "step": 245
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999614778495427,
+      "loss": 6.5094,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999958866915722,
+      "loss": 6.5031,
+      "step": 255
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999956170379105,
+      "loss": 6.4906,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999953388239844,
+      "loss": 6.3875,
+      "step": 265
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999950520498099,
+      "loss": 6.3781,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999947567154032,
+      "loss": 6.2781,
+      "step": 275
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999944528207813,
+      "loss": 6.4125,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999414036596146,
+      "loss": 6.3125,
+      "step": 285
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999938193509616,
+      "loss": 6.4719,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999348977579997,
+      "loss": 6.4469,
+      "step": 295
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999931516404954,
+      "loss": 6.3281,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999928049450672,
+      "loss": 6.3438,
+      "step": 305
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999244968953513,
+      "loss": 6.3563,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999920858739194,
+      "loss": 6.2812,
+      "step": 315
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299991713498241,
+      "loss": 6.2781,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999913325625209,
+      "loss": 6.2406,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999094306678106,
+      "loss": 6.2281,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999905450110436,
+      "loss": 6.1969,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999013839533125,
+      "loss": 6.2844,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999897232196672,
+      "loss": 6.3219,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999892994840752,
+      "loss": 6.2781,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999888671885794,
+      "loss": 6.3344,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999884263332044,
+      "loss": 6.2375,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998797691797554,
+      "loss": 6.1594,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999875189429183,
+      "loss": 6.1594,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999870524080589,
+      "loss": 6.1625,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999865773134239,
+      "loss": 6.1063,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999860936590405,
+      "loss": 6.0469,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998560144493624,
+      "loss": 6.0094,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998510067113924,
+      "loss": 5.9656,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998459133767804,
+      "loss": 6.0312,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998407344458174,
+      "loss": 5.9688,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999835469918799,
+      "loss": 5.9469,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998301197960256,
+      "loss": 5.9813,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998246840778024,
+      "loss": 5.9094,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998191627644393,
+      "loss": 5.9313,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998135558562523,
+      "loss": 5.975,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999807863353561,
+      "loss": 5.9281,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299980208525669,
+      "loss": 6.0031,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999796221565969,
+      "loss": 5.9938,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999790272281733,
+      "loss": 5.9406,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999784237404321,
+      "loss": 5.9188,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997781169340785,
+      "loss": 5.9156,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997719108713536,
+      "loss": 5.9062,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999765619216501,
+      "loss": 5.9344,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299975924196988,
+      "loss": 5.9656,
+      "step": 475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999752779131854,
+      "loss": 5.8312,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997462307027923,
+      "loss": 5.8031,
+      "step": 485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999739596683068,
+      "loss": 5.8594,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999732877073061,
+      "loss": 5.85,
+      "step": 495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997260718731537,
+      "loss": 5.8469,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997191810837346,
+      "loss": 5.85,
+      "step": 505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997122047051967,
+      "loss": 5.9437,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997051427379384,
+      "loss": 5.8438,
+      "step": 515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996979951823637,
+      "loss": 5.8563,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999690762038879,
+      "loss": 5.8187,
+      "step": 525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999683443307898,
+      "loss": 5.7875,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996760389898377,
+      "loss": 5.8656,
+      "step": 535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999668549085121,
+      "loss": 5.8,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996609735941765,
+      "loss": 5.85,
+      "step": 545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999653312517435,
+      "loss": 5.8375,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996455658553336,
+      "loss": 5.8094,
+      "step": 555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999637733608316,
+      "loss": 5.7438,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999629815776827,
+      "loss": 5.7656,
+      "step": 565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996218123613203,
+      "loss": 5.8156,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996137233622523,
+      "loss": 5.7219,
+      "step": 575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999605548780084,
+      "loss": 5.8,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999597288615282,
+      "loss": 5.7125,
+      "step": 585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995889428683183,
+      "loss": 5.7875,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999580511539669,
+      "loss": 5.7531,
+      "step": 595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999571994629815,
+      "loss": 5.8156,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995633921392424,
+      "loss": 5.8063,
+      "step": 605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999554704068442,
+      "loss": 5.875,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995459304179103,
+      "loss": 5.9,
+      "step": 615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999537071188148,
+      "loss": 5.7625,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999528126379659,
+      "loss": 5.7594,
+      "step": 625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999519095992956,
+      "loss": 5.8656,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999509980028553,
+      "loss": 5.7781,
+      "step": 635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995007784869716,
+      "loss": 5.7781,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994914913687357,
+      "loss": 5.8688,
+      "step": 645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999482118674375,
+      "loss": 6.0156,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994726604044256,
+      "loss": 6.325,
+      "step": 655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994631165594263,
+      "loss": 6.1094,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999453487139923,
+      "loss": 5.9406,
+      "step": 665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999443772146463,
+      "loss": 5.9094,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999433971579603,
+      "loss": 5.825,
+      "step": 675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999424085439902,
+      "loss": 5.8281,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994141137279235,
+      "loss": 5.8375,
+      "step": 685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994040564442365,
+      "loss": 5.8187,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999393913589415,
+      "loss": 5.7625,
+      "step": 695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993836851640383,
+      "loss": 5.7938,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993733711686905,
+      "loss": 5.8812,
+      "step": 705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999362971603959,
+      "loss": 5.9031,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993524864704385,
+      "loss": 5.7938,
+      "step": 715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993419157687266,
+      "loss": 5.9969,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999331259499427,
+      "loss": 5.8969,
+      "step": 725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999320517663147,
+      "loss": 5.8719,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993096902605006,
+      "loss": 5.8125,
+      "step": 735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999298777292106,
+      "loss": 5.7656,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999287778758585,
+      "loss": 5.8594,
+      "step": 745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999276694660566,
+      "loss": 5.8375,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999265524998681,
+      "loss": 5.8344,
+      "step": 755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992542697735683,
+      "loss": 5.7906,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999242928985869,
+      "loss": 5.8063,
+      "step": 765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992315026362317,
+      "loss": 5.6844,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999219990725307,
+      "loss": 5.7906,
+      "step": 775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992083932537533,
+      "loss": 5.75,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991967102222314,
+      "loss": 5.7281,
+      "step": 785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999184941631409,
+      "loss": 5.7719,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991730874819576,
+      "loss": 5.7344,
+      "step": 795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999161147774553,
+      "loss": 5.7031,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991491225098767,
+      "loss": 5.7156,
+      "step": 805
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991370116886154,
+      "loss": 5.6406,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299912481531146,
+      "loss": 5.7812,
+      "step": 815
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999112533379107,
+      "loss": 5.7188,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991001658922573,
+      "loss": 5.7375,
+      "step": 825
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999087712851616,
+      "loss": 5.6875,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990751742578936,
+      "loss": 5.75,
+      "step": 835
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990625501118075,
+      "loss": 5.7469,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999049840414076,
+      "loss": 5.675,
+      "step": 845
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990370451654256,
+      "loss": 5.7375,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999024164366587,
+      "loss": 5.7,
+      "step": 855
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029990111980182936,
+      "loss": 5.6937,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989981461212874,
+      "loss": 5.7156,
+      "step": 865
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989850086763116,
+      "loss": 5.7062,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989717856841166,
+      "loss": 5.7031,
+      "step": 875
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998958477145458,
+      "loss": 5.6875,
+      "step": 880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998945083061094,
+      "loss": 5.6375,
+      "step": 885
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989316034317886,
+      "loss": 5.6281,
+      "step": 890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998918038258313,
+      "loss": 5.6312,
+      "step": 895
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029989043875414394,
+      "loss": 5.5906,
+      "step": 900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998890651281948,
+      "loss": 5.6188,
+      "step": 905
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998876829480623,
+      "loss": 5.75,
+      "step": 910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998862922138252,
+      "loss": 5.6375,
+      "step": 915
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988489292556296,
+      "loss": 5.6562,
+      "step": 920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998834850833554,
+      "loss": 5.7531,
+      "step": 925
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988206868728287,
+      "loss": 5.6656,
+      "step": 930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029988064373742625,
+      "loss": 5.5875,
+      "step": 935
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998792102338668,
+      "loss": 5.6813,
+      "step": 940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029987776817668637,
+      "loss": 5.5687,
+      "step": 945
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998763175659672,
+      "loss": 5.55,
+      "step": 950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029987485840179213,
+      "loss": 5.6688,
+      "step": 955
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998733906842445,
+      "loss": 5.5594,
+      "step": 960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998719144134079,
+      "loss": 5.625,
+      "step": 965
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998704295893667,
+      "loss": 5.6469,
+      "step": 970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029986893621220564,
+      "loss": 5.6344,
+      "step": 975
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998674342820099,
+      "loss": 5.5438,
+      "step": 980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002998659237988652,
+      "loss": 5.5875,
+      "step": 985
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986440476285783,
+      "loss": 5.6344,
+      "step": 990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998628771740744,
+      "loss": 5.5156,
+      "step": 995
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986134103260207,
+      "loss": 5.6125,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998597963385285,
+      "loss": 5.5812,
+      "step": 1005
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985824309194194,
+      "loss": 5.5375,
+      "step": 1010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299856681292931,
+      "loss": 5.6781,
+      "step": 1015
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985511094158477,
+      "loss": 5.6813,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985353203799286,
+      "loss": 5.6844,
+      "step": 1025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998519445822455,
+      "loss": 5.6594,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998503485744331,
+      "loss": 5.6719,
+      "step": 1035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998487440146469,
+      "loss": 5.6219,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998471309029784,
+      "loss": 5.6188,
+      "step": 1045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984550923951967,
+      "loss": 5.55,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998438790243632,
+      "loss": 5.5531,
+      "step": 1055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998422402576022,
+      "loss": 5.5187,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984059293933,
+      "loss": 5.5563,
+      "step": 1065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998389370696408,
+      "loss": 5.6031,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998372726486289,
+      "loss": 5.5781,
+      "step": 1075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998355996763894,
+      "loss": 5.5875,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998339181530178,
+      "loss": 5.5531,
+      "step": 1085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983222807861,
+      "loss": 5.4719,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998305294532625,
+      "loss": 5.5219,
+      "step": 1095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998288222770722,
+      "loss": 5.5906,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982710655013657,
+      "loss": 5.4969,
+      "step": 1105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998253822725535,
+      "loss": 5.5469,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982364944442137,
+      "loss": 5.6156,
+      "step": 1115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982190806583913,
+      "loss": 5.6031,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982015813690614,
+      "loss": 5.7125,
+      "step": 1125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998183996577222,
+      "loss": 5.5625,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029981663262838783,
+      "loss": 5.6063,
+      "step": 1135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029981485704900373,
+      "loss": 5.5438,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998130729196713,
+      "loss": 5.475,
+      "step": 1145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998112802404923,
+      "loss": 5.4906,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998094790115691,
+      "loss": 5.425,
+      "step": 1155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980766923300443,
+      "loss": 5.4688,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980585090490165,
+      "loss": 5.4781,
+      "step": 1165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998040240273645,
+      "loss": 5.3875,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980218860049725,
+      "loss": 5.4156,
+      "step": 1175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980034462440456,
+      "loss": 5.4188,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997984920991918,
+      "loss": 5.4656,
+      "step": 1185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979663102496466,
+      "loss": 5.5125,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979476140182926,
+      "loss": 5.5125,
+      "step": 1195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979288322989244,
+      "loss": 5.55,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979099650926126,
+      "loss": 5.4813,
+      "step": 1205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997891012400435,
+      "loss": 5.5469,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997871974223472,
+      "loss": 5.4875,
+      "step": 1215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978528505628115,
+      "loss": 5.4531,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997833641419544,
+      "loss": 5.4125,
+      "step": 1225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997814346794766,
+      "loss": 5.4,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997794966689579,
+      "loss": 5.4625,
+      "step": 1235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997775501105088,
+      "loss": 5.5844,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997755950042405,
+      "loss": 5.5281,
+      "step": 1245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029977363135026447,
+      "loss": 5.4625,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997716591486929,
+      "loss": 5.5375,
+      "step": 1255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997696783996383,
+      "loss": 5.4469,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029976768910321366,
+      "loss": 5.3906,
+      "step": 1265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997656912595326,
+      "loss": 5.4219,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029976368486870904,
+      "loss": 5.3906,
+      "step": 1275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997616699308575,
+      "loss": 5.3781,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299759646446093,
+      "loss": 5.4,
+      "step": 1285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997576144145311,
+      "loss": 5.3844,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975557383628764,
+      "loss": 5.4125,
+      "step": 1295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975352471147906,
+      "loss": 5.5187,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997514670402225,
+      "loss": 5.4781,
+      "step": 1305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029974940082263515,
+      "loss": 5.5375,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997473260588351,
+      "loss": 5.3875,
+      "step": 1315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997452427489407,
+      "loss": 5.4469,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997431508930708,
+      "loss": 5.4313,
+      "step": 1325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029974105049134484,
+      "loss": 5.4,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997389415438827,
+      "loss": 5.4437,
+      "step": 1335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997368240508047,
+      "loss": 5.4406,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973469801223163,
+      "loss": 5.4281,
+      "step": 1345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973256342828497,
+      "loss": 5.3844,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973042029908645,
+      "loss": 5.3656,
+      "step": 1355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029972826862475837,
+      "loss": 5.3656,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997261084054235,
+      "loss": 5.3375,
+      "step": 1365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997239396412053,
+      "loss": 5.3937,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997217623322273,
+      "loss": 5.3781,
+      "step": 1375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997195764786139,
+      "loss": 5.3844,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997173820804899,
+      "loss": 5.4125,
+      "step": 1385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971517913798037,
+      "loss": 5.3594,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997129676512111,
+      "loss": 5.3,
+      "step": 1395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971074762030834,
+      "loss": 5.35,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997085190453987,
+      "loss": 5.3156,
+      "step": 1405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997062819266095,
+      "loss": 5.2938,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029970403626406833,
+      "loss": 5.3594,
+      "step": 1415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997017820579034,
+      "loss": 5.3031,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996995193082432,
+      "loss": 5.2719,
+      "step": 1425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969724801521706,
+      "loss": 5.3375,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996949681789545,
+      "loss": 5.2406,
+      "step": 1435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996926797995857,
+      "loss": 5.3,
+      "step": 1440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969038287724117,
+      "loss": 5.325,
+      "step": 1445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996880774120521,
+      "loss": 5.2875,
+      "step": 1450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029968576340414995,
+      "loss": 5.3312,
+      "step": 1455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029968344085366686,
+      "loss": 5.2625,
+      "step": 1460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996811097607353,
+      "loss": 5.35,
+      "step": 1465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996787701254884,
+      "loss": 5.35,
+      "step": 1470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029967642194805963,
+      "loss": 5.2656,
+      "step": 1475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299674065228583,
+      "loss": 5.2125,
+      "step": 1480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029967169996719304,
+      "loss": 5.25,
+      "step": 1485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996693261640247,
+      "loss": 5.2375,
+      "step": 1490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029966694381921346,
+      "loss": 5.3187,
+      "step": 1495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996645529328953,
+      "loss": 5.2375,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996621535052066,
+      "loss": 5.2625,
+      "step": 1505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996597455362844,
+      "loss": 5.2719,
+      "step": 1510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029965732902626613,
+      "loss": 5.2188,
+      "step": 1515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029965490397528964,
+      "loss": 5.2406,
+      "step": 1520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996524703834933,
+      "loss": 5.15,
+      "step": 1525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299650028251016,
+      "loss": 5.2,
+      "step": 1530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029964757757799715,
+      "loss": 5.2281,
+      "step": 1535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996451183645767,
+      "loss": 5.1937,
+      "step": 1540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996426506108948,
+      "loss": 5.2406,
+      "step": 1545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996401743170924,
+      "loss": 5.2313,
+      "step": 1550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996376894833108,
+      "loss": 5.2062,
+      "step": 1555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996351961096918,
+      "loss": 5.2031,
+      "step": 1560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996326941963777,
+      "loss": 5.2938,
+      "step": 1565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029963018374351135,
+      "loss": 5.2906,
+      "step": 1570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299627664751236,
+      "loss": 5.2344,
+      "step": 1575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029962513721969533,
+      "loss": 5.2313,
+      "step": 1580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029962260114903364,
+      "loss": 5.3125,
+      "step": 1585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996200565393956,
+      "loss": 5.2562,
+      "step": 1590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029961750339092656,
+      "loss": 5.2156,
+      "step": 1595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029961494170377217,
+      "loss": 5.2313,
+      "step": 1600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996123714780786,
+      "loss": 5.2438,
+      "step": 1605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029960979271399254,
+      "loss": 5.2313,
+      "step": 1610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029960720541166117,
+      "loss": 5.3187,
+      "step": 1615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996046095712321,
+      "loss": 5.2719,
+      "step": 1620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996020051928536,
+      "loss": 5.2562,
+      "step": 1625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029959939227667415,
+      "loss": 5.3438,
+      "step": 1630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299596770822843,
+      "loss": 5.3812,
+      "step": 1635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002995941408315097,
+      "loss": 5.2438,
+      "step": 1640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002995915023028243,
+      "loss": 5.2219,
+      "step": 1645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995888552369374,
+      "loss": 5.2094,
+      "step": 1650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958619963400015,
+      "loss": 5.2406,
+      "step": 1655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958353549416406,
+      "loss": 5.1719,
+      "step": 1660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958086281758114,
+      "loss": 5.2094,
+      "step": 1665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029957818160440393,
+      "loss": 5.3031,
+      "step": 1670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995754918547854,
+      "loss": 5.2438,
+      "step": 1675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995727935688792,
+      "loss": 5.2844,
+      "step": 1680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995700867468392,
+      "loss": 5.2,
+      "step": 1685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995673713888199,
+      "loss": 5.2,
+      "step": 1690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995646474949763,
+      "loss": 5.2094,
+      "step": 1695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995619150654638,
+      "loss": 5.2438,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995591741004384,
+      "loss": 5.1875,
+      "step": 1705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995564246000565,
+      "loss": 5.1562,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299553666564475,
+      "loss": 5.1594,
+      "step": 1715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955089999385124,
+      "loss": 5.1156,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995481248883433,
+      "loss": 5.1719,
+      "step": 1725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995453412481094,
+      "loss": 5.1625,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995425490733084,
+      "loss": 5.1688,
+      "step": 1735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953974836409973,
+      "loss": 5.0719,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953693912064316,
+      "loss": 5.1719,
+      "step": 1745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953412134309904,
+      "loss": 5.2062,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953129503162816,
+      "loss": 5.2125,
+      "step": 1755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995284601863919,
+      "loss": 5.2,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952561680755195,
+      "loss": 5.15,
+      "step": 1765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995227648952706,
+      "loss": 5.1219,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995199044497106,
+      "loss": 5.1312,
+      "step": 1775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995170354710353,
+      "loss": 5.1531,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995141579594083,
+      "loss": 5.1063,
+      "step": 1785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995112719149939,
+      "loss": 5.1719,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950837733795676,
+      "loss": 5.1594,
+      "step": 1795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995054742284621,
+      "loss": 5.1312,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950256258667553,
+      "loss": 5.1719,
+      "step": 1805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994996424127634,
+      "loss": 5.1125,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949671370689215,
+      "loss": 5.1437,
+      "step": 1815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949377646922904,
+      "loss": 5.0938,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949083069994174,
+      "loss": 5.0906,
+      "step": 1825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029948787639919825,
+      "loss": 5.1437,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029948491356716715,
+      "loss": 5.1469,
+      "step": 1835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994819422040177,
+      "loss": 5.1,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029947896230991935,
+      "loss": 5.1656,
+      "step": 1845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994759738850422,
+      "loss": 5.1312,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029947297692955684,
+      "loss": 5.0781,
+      "step": 1855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029946997144363415,
+      "loss": 5.1531,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994669574274459,
+      "loss": 5.0938,
+      "step": 1865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994639348811638,
+      "loss": 5.1469,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994609038049606,
+      "loss": 5.1281,
+      "step": 1875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029945786419900916,
+      "loss": 5.1656,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029945481606348306,
+      "loss": 5.2156,
+      "step": 1885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994517593985561,
+      "loss": 5.0844,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994486942044028,
+      "loss": 5.0938,
+      "step": 1895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994456204811982,
+      "loss": 5.1281,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029944253822911744,
+      "loss": 5.0969,
+      "step": 1905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029943944744833677,
+      "loss": 5.1469,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994363481390323,
+      "loss": 5.1344,
+      "step": 1915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029943324030138105,
+      "loss": 5.1281,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994301239355603,
+      "loss": 5.125,
+      "step": 1925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299426999041748,
+      "loss": 5.1,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029942386562012245,
+      "loss": 5.1094,
+      "step": 1935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994207236708625,
+      "loss": 5.1281,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994175731941474,
+      "loss": 5.1562,
+      "step": 1945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994144141901569,
+      "loss": 5.1406,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994112466590714,
+      "loss": 5.1781,
+      "step": 1955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994080706010717,
+      "loss": 5.1813,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994048860163389,
+      "loss": 5.0906,
+      "step": 1965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029940169290505487,
+      "loss": 5.05,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993984912674018,
+      "loss": 5.0781,
+      "step": 1975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029939528110356245,
+      "loss": 5.0906,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993920624137199,
+      "loss": 5.1781,
+      "step": 1985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299388835198058,
+      "loss": 5.1406,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993855994567608,
+      "loss": 5.1406,
+      "step": 1995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029938235519001304,
+      "loss": 5.0969,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029937910239799986,
+      "loss": 5.1406,
+      "step": 2005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993758410809069,
+      "loss": 5.0812,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029937257123892027,
+      "loss": 5.0687,
+      "step": 2015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029936929287222654,
+      "loss": 5.2219,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029936600598101285,
+      "loss": 5.1469,
+      "step": 2025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029936271056546683,
+      "loss": 5.1531,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029935940662577647,
+      "loss": 5.1,
+      "step": 2035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993560941621303,
+      "loss": 5.1125,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993527731747175,
+      "loss": 5.0938,
+      "step": 2045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993494436637275,
+      "loss": 5.0687,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029934610562935035,
+      "loss": 5.0781,
+      "step": 2055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993427590717765,
+      "loss": 5.0406,
+      "step": 2060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299339403991197,
+      "loss": 5.0344,
+      "step": 2065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029933604038780327,
+      "loss": 5.1312,
+      "step": 2070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993326682617873,
+      "loss": 5.025,
+      "step": 2075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993292876133416,
+      "loss": 5.0187,
+      "step": 2080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029932589844265895,
+      "loss": 5.0812,
+      "step": 2085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993225007499329,
+      "loss": 5.0062,
+      "step": 2090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029931909453535734,
+      "loss": 5.05,
+      "step": 2095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993156797991266,
+      "loss": 5.025,
+      "step": 2100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029931225654143565,
+      "loss": 5.0469,
+      "step": 2105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029930882476247975,
+      "loss": 5.0656,
+      "step": 2110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993053844624548,
+      "loss": 5.1188,
+      "step": 2115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993019356415572,
+      "loss": 5.1156,
+      "step": 2120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929847829998375,
+      "loss": 5.0187,
+      "step": 2125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929501243793164,
+      "loss": 5.0094,
+      "step": 2130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992915380555988,
+      "loss": 4.9562,
+      "step": 2135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029928805515318344,
+      "loss": 5.0406,
+      "step": 2140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029928456373088443,
+      "loss": 5.0938,
+      "step": 2145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992810637889009,
+      "loss": 5.0719,
+      "step": 2150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029927755532743267,
+      "loss": 4.9875,
+      "step": 2155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029927403834668,
+      "loss": 5.0312,
+      "step": 2160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029927051284684347,
+      "loss": 5.0406,
+      "step": 2165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029926697882812437,
+      "loss": 5.0062,
+      "step": 2170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029926343629072444,
+      "loss": 4.9594,
+      "step": 2175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029925988523484573,
+      "loss": 5.0187,
+      "step": 2180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299256325660691,
+      "loss": 5.0312,
+      "step": 2185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992527575684633,
+      "loss": 4.9875,
+      "step": 2190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029924918095836637,
+      "loss": 5.0844,
+      "step": 2195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029924559583060427,
+      "loss": 5.0375,
+      "step": 2200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992420021853816,
+      "loss": 4.975,
+      "step": 2205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029923840002290343,
+      "loss": 5.0094,
+      "step": 2210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029923478934337544,
+      "loss": 5.0312,
+      "step": 2215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029923117014700354,
+      "loss": 4.95,
+      "step": 2220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992275424339944,
+      "loss": 4.8937,
+      "step": 2225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299223906204555,
+      "loss": 5.0125,
+      "step": 2230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029922026145889284,
+      "loss": 4.9406,
+      "step": 2235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029921660819721605,
+      "loss": 4.9531,
+      "step": 2240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029921294641973293,
+      "loss": 5.025,
+      "step": 2245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029920927612665256,
+      "loss": 5.0219,
+      "step": 2250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992055973181844,
+      "loss": 4.9531,
+      "step": 2255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029920190999453846,
+      "loss": 4.9594,
+      "step": 2260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029919821415592507,
+      "loss": 5.0125,
+      "step": 2265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002991945098025552,
+      "loss": 4.9688,
+      "step": 2270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002991907969346402,
+      "loss": 5.0094,
+      "step": 2275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029918707555239205,
+      "loss": 5.0,
+      "step": 2280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029918334565602313,
+      "loss": 4.9813,
+      "step": 2285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029917960724574626,
+      "loss": 4.9813,
+      "step": 2290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002991758603217748,
+      "loss": 4.9813,
+      "step": 2295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002991721048843226,
+      "loss": 5.0438,
+      "step": 2300
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029916834093360393,
+      "loss": 5.025,
+      "step": 2305
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991645684698337,
+      "loss": 5.0406,
+      "step": 2310
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991607874932271,
+      "loss": 5.0031,
+      "step": 2315
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000299156998004,
+      "loss": 4.9813,
+      "step": 2320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029915320000236864,
+      "loss": 5.0094,
+      "step": 2325
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991493934885497,
+      "loss": 4.9906,
+      "step": 2330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991455784627605,
+      "loss": 4.9,
+      "step": 2335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029914175492521875,
+      "loss": 5.0094,
+      "step": 2340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913792287614263,
+      "loss": 4.9656,
+      "step": 2345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991340823157509,
+      "loss": 5.0438,
+      "step": 2350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913023324426265,
+      "loss": 4.9938,
+      "step": 2355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991263756618976,
+      "loss": 5.0187,
+      "step": 2360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991225095688759,
+      "loss": 4.9344,
+      "step": 2365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029911863496541815,
+      "loss": 5.0406,
+      "step": 2370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991147518517456,
+      "loss": 4.9813,
+      "step": 2375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991108602280797,
+      "loss": 4.9469,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991069600946426,
+      "loss": 4.9469,
+      "step": 2385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029910305145165695,
+      "loss": 4.8969,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990991342993457,
+      "loss": 5.0531,
+      "step": 2395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990952086379325,
+      "loss": 4.9781,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029909127446764135,
+      "loss": 4.9469,
+      "step": 2405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990873317886968,
+      "loss": 4.9437,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029908338060132374,
+      "loss": 4.9594,
+      "step": 2415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907942090574786,
+      "loss": 4.9719,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907545270219495,
+      "loss": 4.9219,
+      "step": 2425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907147599089165,
+      "loss": 4.9437,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990674907720648,
+      "loss": 4.9562,
+      "step": 2435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029906349704594186,
+      "loss": 4.9562,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029905949481275073,
+      "loss": 4.9625,
+      "step": 2445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029905548407271984,
+      "loss": 4.9437,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029905146482607807,
+      "loss": 4.9594,
+      "step": 2455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029904743707305487,
+      "loss": 4.9469,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029904340081388,
+      "loss": 4.9437,
+      "step": 2465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029903935604878385,
+      "loss": 4.9562,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990353027779973,
+      "loss": 4.9062,
+      "step": 2475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029903124100175165,
+      "loss": 4.9969,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990271707202787,
+      "loss": 4.9969,
+      "step": 2485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029902309193381065,
+      "loss": 4.9656,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990190046425804,
+      "loss": 5.0219,
+      "step": 2495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901490884682116,
+      "loss": 4.9062,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901080454676677,
+      "loss": 5.0062,
+      "step": 2505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029900669174265127,
+      "loss": 4.9656,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029900257043470954,
+      "loss": 4.9,
+      "step": 2515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989984406231767,
+      "loss": 4.9531,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989943023082885,
+      "loss": 5.0375,
+      "step": 2525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989901554902811,
+      "loss": 4.8937,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989860001693911,
+      "loss": 4.9656,
+      "step": 2535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989818363458557,
+      "loss": 4.9719,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029897766401991254,
+      "loss": 5.0031,
+      "step": 2545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029897348319179973,
+      "loss": 5.0187,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989692938617558,
+      "loss": 4.95,
+      "step": 2555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989650960300199,
+      "loss": 4.8844,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029896088969683164,
+      "loss": 4.9938,
+      "step": 2565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298956674862431,
+      "loss": 4.8656,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029895245152705853,
+      "loss": 4.9656,
+      "step": 2575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029894821969095526,
+      "loss": 4.9656,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029894397935436277,
+      "loss": 4.9562,
+      "step": 2585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893973051752296,
+      "loss": 4.9156,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893547318067835,
+      "loss": 4.9469,
+      "step": 2595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893120734407187,
+      "loss": 4.9281,
+      "step": 2600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989269330079471,
+      "loss": 4.9344,
+      "step": 2605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989226501725478,
+      "loss": 4.925,
+      "step": 2610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989183588381186,
+      "loss": 4.875,
+      "step": 2615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891405900490416,
+      "loss": 4.9094,
+      "step": 2620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029890975067315003,
+      "loss": 5.0156,
+      "step": 2625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989054338431021,
+      "loss": 4.9813,
+      "step": 2630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989011085150066,
+      "loss": 4.9469,
+      "step": 2635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029889677468911054,
+      "loss": 4.8469,
+      "step": 2640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029889243236566116,
+      "loss": 4.8812,
+      "step": 2645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029888808154490633,
+      "loss": 4.9,
+      "step": 2650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029888372222709427,
+      "loss": 4.9562,
+      "step": 2655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029887935441247376,
+      "loss": 4.9406,
+      "step": 2660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029887497810129416,
+      "loss": 4.9813,
+      "step": 2665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988705932938052,
+      "loss": 4.9156,
+      "step": 2670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988661999902571,
+      "loss": 4.8625,
+      "step": 2675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988617981909006,
+      "loss": 4.9,
+      "step": 2680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029885738789598695,
+      "loss": 4.85,
+      "step": 2685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029885296910576775,
+      "loss": 4.8219,
+      "step": 2690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988485418204953,
+      "loss": 4.8688,
+      "step": 2695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988441060404221,
+      "loss": 4.9156,
+      "step": 2700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029883966176580146,
+      "loss": 4.8625,
+      "step": 2705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029883520899688697,
+      "loss": 4.9094,
+      "step": 2710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988307477339327,
+      "loss": 4.8625,
+      "step": 2715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988262779771933,
+      "loss": 4.9594,
+      "step": 2720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029882179972692385,
+      "loss": 4.8344,
+      "step": 2725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029881731298337993,
+      "loss": 4.9188,
+      "step": 2730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988128177468176,
+      "loss": 4.85,
+      "step": 2735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029880831401749334,
+      "loss": 4.7812,
+      "step": 2740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988038017956643,
+      "loss": 4.8875,
+      "step": 2745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987992810815879,
+      "loss": 4.8812,
+      "step": 2750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987947518755222,
+      "loss": 4.8531,
+      "step": 2755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987902141777256,
+      "loss": 4.8094,
+      "step": 2760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029878566798845717,
+      "loss": 4.8781,
+      "step": 2765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029878111330797625,
+      "loss": 4.9094,
+      "step": 2770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987765501365428,
+      "loss": 4.8844,
+      "step": 2775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029877197847441735,
+      "loss": 4.8469,
+      "step": 2780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987673983218607,
+      "loss": 4.8563,
+      "step": 2785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987628096791343,
+      "loss": 4.8469,
+      "step": 2790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987582125464999,
+      "loss": 4.8656,
+      "step": 2795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029875360692422,
+      "loss": 4.8438,
+      "step": 2800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987489928125574,
+      "loss": 4.875,
+      "step": 2805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029874437021177545,
+      "loss": 4.9375,
+      "step": 2810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987397391221379,
+      "loss": 4.8125,
+      "step": 2815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987350995439091,
+      "loss": 4.8406,
+      "step": 2820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987304514773538,
+      "loss": 4.8187,
+      "step": 2825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029872579492273724,
+      "loss": 4.7906,
+      "step": 2830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987211298803253,
+      "loss": 4.8187,
+      "step": 2835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029871645635038405,
+      "loss": 4.825,
+      "step": 2840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029871177433318026,
+      "loss": 4.8688,
+      "step": 2845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987070838289812,
+      "loss": 4.8656,
+      "step": 2850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029870238483805446,
+      "loss": 4.7906,
+      "step": 2855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986976773606683,
+      "loss": 4.8781,
+      "step": 2860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986929613970913,
+      "loss": 4.8719,
+      "step": 2865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986882369475927,
+      "loss": 4.8219,
+      "step": 2870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986835040124421,
+      "loss": 4.85,
+      "step": 2875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986787625919095,
+      "loss": 4.8844,
+      "step": 2880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986740126862656,
+      "loss": 4.8125,
+      "step": 2885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986692542957814,
+      "loss": 4.8531,
+      "step": 2890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029866448742072853,
+      "loss": 4.8375,
+      "step": 2895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298659712061379,
+      "loss": 4.8469,
+      "step": 2900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986549282180054,
+      "loss": 4.8187,
+      "step": 2905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029865013589088066,
+      "loss": 4.8125,
+      "step": 2910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986453350802783,
+      "loss": 4.8219,
+      "step": 2915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029864052578647233,
+      "loss": 4.8219,
+      "step": 2920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986357080097372,
+      "loss": 4.7844,
+      "step": 2925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986308817503479,
+      "loss": 4.8094,
+      "step": 2930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986260470085798,
+      "loss": 4.7094,
+      "step": 2935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029862120378470883,
+      "loss": 4.8469,
+      "step": 2940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986163520790115,
+      "loss": 4.8125,
+      "step": 2945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986114918917645,
+      "loss": 4.7938,
+      "step": 2950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002986066232232454,
+      "loss": 4.8094,
+      "step": 2955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029860174607373194,
+      "loss": 4.8312,
+      "step": 2960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029859686044350246,
+      "loss": 4.85,
+      "step": 2965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985919663328358,
+      "loss": 4.8031,
+      "step": 2970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985870637420113,
+      "loss": 4.8,
+      "step": 2975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985821526713088,
+      "loss": 4.8125,
+      "step": 2980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985772331210084,
+      "loss": 4.8031,
+      "step": 2985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029857230509139105,
+      "loss": 4.7656,
+      "step": 2990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029856736858273786,
+      "loss": 4.8,
+      "step": 2995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029856242359533057,
+      "loss": 4.8187,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029855747012945145,
+      "loss": 4.7562,
+      "step": 3005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985525081853832,
+      "loss": 4.7625,
+      "step": 3010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985475377634089,
+      "loss": 4.7188,
+      "step": 3015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985425588638123,
+      "loss": 4.7656,
+      "step": 3020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985375714868776,
+      "loss": 4.7875,
+      "step": 3025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985325756328893,
+      "loss": 4.7719,
+      "step": 3030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985275713021326,
+      "loss": 4.7844,
+      "step": 3035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029852255849489303,
+      "loss": 4.8563,
+      "step": 3040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985175372114567,
+      "loss": 4.7687,
+      "step": 3045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029851250745211024,
+      "loss": 4.8594,
+      "step": 3050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985074692171406,
+      "loss": 4.6906,
+      "step": 3055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029850242250683537,
+      "loss": 4.6531,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029849736732148253,
+      "loss": 4.7594,
+      "step": 3065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029849230366137065,
+      "loss": 4.7125,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984872315267887,
+      "loss": 4.7531,
+      "step": 3075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298482150918026,
+      "loss": 4.7031,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029847706183537266,
+      "loss": 4.7562,
+      "step": 3085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984719642791191,
+      "loss": 4.7781,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029846685824955617,
+      "loss": 4.6656,
+      "step": 3095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029846174374697533,
+      "loss": 4.7219,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984566207716684,
+      "loss": 4.7812,
+      "step": 3105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029845148932392787,
+      "loss": 4.7,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029844634940404645,
+      "loss": 4.6781,
+      "step": 3115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029844120101231756,
+      "loss": 4.7687,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029843604414903503,
+      "loss": 4.7531,
+      "step": 3125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029843087881449305,
+      "loss": 4.6594,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984257050089866,
+      "loss": 4.7687,
+      "step": 3135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029842052273281074,
+      "loss": 4.7281,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029841533198626135,
+      "loss": 4.6844,
+      "step": 3145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984101327696346,
+      "loss": 4.7125,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984049250832273,
+      "loss": 4.7031,
+      "step": 3155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983997089273366,
+      "loss": 4.7031,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983944843022602,
+      "loss": 4.775,
+      "step": 3165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983892512082962,
+      "loss": 4.6437,
+      "step": 3170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029838400964574336,
+      "loss": 4.7031,
+      "step": 3175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029837875961490073,
+      "loss": 4.7625,
+      "step": 3180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298373501116068,
+      "loss": 4.6937,
+      "step": 3185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983682341495452,
+      "loss": 4.7531,
+      "step": 3190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029836295871563296,
+      "loss": 4.75,
+      "step": 3195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029835767481463236,
+      "loss": 4.7188,
+      "step": 3200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029835238244684496,
+      "loss": 4.6813,
+      "step": 3205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983470816125728,
+      "loss": 4.7531,
+      "step": 3210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029834177231211827,
+      "loss": 4.6937,
+      "step": 3215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029833645454578454,
+      "loss": 4.6875,
+      "step": 3220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029833112831387503,
+      "loss": 4.7531,
+      "step": 3225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983257936166937,
+      "loss": 4.7125,
+      "step": 3230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029832045045454495,
+      "loss": 4.7375,
+      "step": 3235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029831509882773387,
+      "loss": 4.7594,
+      "step": 3240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029830973873656575,
+      "loss": 4.65,
+      "step": 3245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983043701813465,
+      "loss": 4.7438,
+      "step": 3250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029829899316238253,
+      "loss": 4.7062,
+      "step": 3255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982936076799807,
+      "loss": 4.6656,
+      "step": 3260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029828821373444845,
+      "loss": 4.7469,
+      "step": 3265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982828113260934,
+      "loss": 4.7125,
+      "step": 3270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029827740045522405,
+      "loss": 4.7531,
+      "step": 3275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029827198112214915,
+      "loss": 4.65,
+      "step": 3280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029826655332717796,
+      "loss": 4.6688,
+      "step": 3285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982611170706203,
+      "loss": 4.5812,
+      "step": 3290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029825567235278633,
+      "loss": 4.6625,
+      "step": 3295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029825021917398686,
+      "loss": 4.7781,
+      "step": 3300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029824475753453307,
+      "loss": 4.7313,
+      "step": 3305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029823928743473666,
+      "loss": 4.6562,
+      "step": 3310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982338088749098,
+      "loss": 4.6406,
+      "step": 3315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982283218553651,
+      "loss": 4.675,
+      "step": 3320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982228263764158,
+      "loss": 4.675,
+      "step": 3325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982173224383755,
+      "loss": 4.5563,
+      "step": 3330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029821181004155835,
+      "loss": 4.7125,
+      "step": 3335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982062891862788,
+      "loss": 4.6281,
+      "step": 3340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029820075987285207,
+      "loss": 4.6031,
+      "step": 3345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029819522210159364,
+      "loss": 4.6281,
+      "step": 3350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029818967587281965,
+      "loss": 4.6031,
+      "step": 3355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981841211868465,
+      "loss": 4.6937,
+      "step": 3360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029817855804399124,
+      "loss": 4.6844,
+      "step": 3365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981729864445714,
+      "loss": 4.6562,
+      "step": 3370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981674063889049,
+      "loss": 4.7313,
+      "step": 3375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029816181787731017,
+      "loss": 4.6719,
+      "step": 3380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981562209101063,
+      "loss": 4.7219,
+      "step": 3385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029815061548761255,
+      "loss": 4.625,
+      "step": 3390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029814500161014885,
+      "loss": 4.675,
+      "step": 3395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981393792780356,
+      "loss": 4.6719,
+      "step": 3400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981337484915937,
+      "loss": 4.5906,
+      "step": 3405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981281092511445,
+      "loss": 4.7281,
+      "step": 3410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029812246155700974,
+      "loss": 4.6937,
+      "step": 3415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029811680540951183,
+      "loss": 4.6531,
+      "step": 3420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981111408089735,
+      "loss": 4.6094,
+      "step": 3425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029810546775571817,
+      "loss": 4.6937,
+      "step": 3430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980997862500694,
+      "loss": 4.6781,
+      "step": 3435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980940962923516,
+      "loss": 4.6469,
+      "step": 3440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980883978828894,
+      "loss": 4.6281,
+      "step": 3445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029808269102200803,
+      "loss": 4.8156,
+      "step": 3450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980769757100332,
+      "loss": 4.6688,
+      "step": 3455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980712519472911,
+      "loss": 4.6344,
+      "step": 3460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980655197341083,
+      "loss": 4.6156,
+      "step": 3465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298059779070812,
+      "loss": 4.6719,
+      "step": 3470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029805402995772984,
+      "loss": 4.6406,
+      "step": 3475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029804827239518993,
+      "loss": 4.6375,
+      "step": 3480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029804250638352077,
+      "loss": 4.6156,
+      "step": 3485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980367319230515,
+      "loss": 4.6281,
+      "step": 3490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029803094901411167,
+      "loss": 4.6969,
+      "step": 3495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029802515765703125,
+      "loss": 4.6844,
+      "step": 3500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029801935785214083,
+      "loss": 4.6063,
+      "step": 3505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980135495997713,
+      "loss": 4.6375,
+      "step": 3510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980077329002543,
+      "loss": 4.5969,
+      "step": 3515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980019077539216,
+      "loss": 4.6969,
+      "step": 3520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029799607416110583,
+      "loss": 4.6063,
+      "step": 3525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029799023212213974,
+      "loss": 4.6406,
+      "step": 3530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002979843816373569,
+      "loss": 4.6281,
+      "step": 3535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029797852270709106,
+      "loss": 4.6344,
+      "step": 3540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029797265533167665,
+      "loss": 4.6437,
+      "step": 3545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029796677951144854,
+      "loss": 4.6156,
+      "step": 3550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029796089524674203,
+      "loss": 4.5812,
+      "step": 3555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029795500253789294,
+      "loss": 4.6188,
+      "step": 3560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029794910138523754,
+      "loss": 4.5438,
+      "step": 3565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002979431917891127,
+      "loss": 4.5875,
+      "step": 3570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002979372737498556,
+      "loss": 4.7125,
+      "step": 3575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000297931347267804,
+      "loss": 4.7094,
+      "step": 3580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029792541234329616,
+      "loss": 4.5719,
+      "step": 3585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002979194689766707,
+      "loss": 4.5906,
+      "step": 3590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002979135171682669,
+      "loss": 4.6031,
+      "step": 3595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029790755691842444,
+      "loss": 4.6188,
+      "step": 3600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029790158822748337,
+      "loss": 4.65,
+      "step": 3605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002978956110957844,
+      "loss": 4.6188,
+      "step": 3610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002978896255236686,
+      "loss": 4.5938,
+      "step": 3615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029788363151147764,
+      "loss": 4.5844,
+      "step": 3620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978776290595535,
+      "loss": 4.5531,
+      "step": 3625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978716181682389,
+      "loss": 4.6469,
+      "step": 3630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029786559883787666,
+      "loss": 4.675,
+      "step": 3635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978595710688105,
+      "loss": 4.6281,
+      "step": 3640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978535348613843,
+      "loss": 4.6375,
+      "step": 3645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978474902159426,
+      "loss": 4.6219,
+      "step": 3650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029784143713283034,
+      "loss": 4.6312,
+      "step": 3655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029783537561239306,
+      "loss": 4.55,
+      "step": 3660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978293056549766,
+      "loss": 4.6281,
+      "step": 3665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978232272609274,
+      "loss": 4.5687,
+      "step": 3670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029781714043059226,
+      "loss": 4.55,
+      "step": 3675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978110451643187,
+      "loss": 4.5969,
+      "step": 3680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978049414624546,
+      "loss": 4.5812,
+      "step": 3685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029779882932534814,
+      "loss": 4.5438,
+      "step": 3690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029779270875334823,
+      "loss": 4.5187,
+      "step": 3695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029778657974680425,
+      "loss": 4.5687,
+      "step": 3700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977804423060658,
+      "loss": 4.55,
+      "step": 3705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029777429643148325,
+      "loss": 4.5844,
+      "step": 3710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977681421234074,
+      "loss": 4.5656,
+      "step": 3715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029776197938218935,
+      "loss": 4.5812,
+      "step": 3720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977558082081809,
+      "loss": 4.5438,
+      "step": 3725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029774962860173423,
+      "loss": 4.5594,
+      "step": 3730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977434405632019,
+      "loss": 4.5438,
+      "step": 3735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977372440929373,
+      "loss": 4.5687,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977310391912938,
+      "loss": 4.4938,
+      "step": 3745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029772482585862565,
+      "loss": 4.5687,
+      "step": 3750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029771860409528745,
+      "loss": 4.5719,
+      "step": 3755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977123739016343,
+      "loss": 4.4531,
+      "step": 3760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977061352780217,
+      "loss": 4.525,
+      "step": 3765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976998882248057,
+      "loss": 4.5219,
+      "step": 3770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029769363274234276,
+      "loss": 4.5156,
+      "step": 3775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029768736883099,
+      "loss": 4.5156,
+      "step": 3780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976810964911048,
+      "loss": 4.4719,
+      "step": 3785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029767481572304515,
+      "loss": 4.5656,
+      "step": 3790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976685265271696,
+      "loss": 4.5469,
+      "step": 3795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029766222890383694,
+      "loss": 4.6156,
+      "step": 3800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976559228534066,
+      "loss": 4.4781,
+      "step": 3805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029764960837623853,
+      "loss": 4.5156,
+      "step": 3810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297643285472693,
+      "loss": 4.5781,
+      "step": 3815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297636954143131,
+      "loss": 4.4938,
+      "step": 3820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976306143879137,
+      "loss": 4.5094,
+      "step": 3825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029762426620740304,
+      "loss": 4.5,
+      "step": 3830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976179096019612,
+      "loss": 4.4688,
+      "step": 3835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029761154457195104,
+      "loss": 4.4781,
+      "step": 3840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976051711177358,
+      "loss": 4.4188,
+      "step": 3845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975987892396792,
+      "loss": 4.5094,
+      "step": 3850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975923989381454,
+      "loss": 4.5125,
+      "step": 3855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029758600021349916,
+      "loss": 4.525,
+      "step": 3860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975795930661057,
+      "loss": 4.4688,
+      "step": 3865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029757317749633055,
+      "loss": 4.525,
+      "step": 3870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029756675350453994,
+      "loss": 4.4906,
+      "step": 3875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975603210911005,
+      "loss": 4.5,
+      "step": 3880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029755388025637917,
+      "loss": 4.4875,
+      "step": 3885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975474310007438,
+      "loss": 4.4938,
+      "step": 3890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029754097332456217,
+      "loss": 4.4469,
+      "step": 3895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029753450722820303,
+      "loss": 4.5187,
+      "step": 3900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975280327120353,
+      "loss": 4.5187,
+      "step": 3905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975215497764285,
+      "loss": 4.4469,
+      "step": 3910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029751505842175267,
+      "loss": 4.3812,
+      "step": 3915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029750855864837814,
+      "loss": 4.4344,
+      "step": 3920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029750205045667595,
+      "loss": 4.5281,
+      "step": 3925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974955338470175,
+      "loss": 4.4031,
+      "step": 3930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974890088197747,
+      "loss": 4.475,
+      "step": 3935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974824753753199,
+      "loss": 4.4875,
+      "step": 3940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297475933514026,
+      "loss": 4.4094,
+      "step": 3945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974693832362664,
+      "loss": 4.525,
+      "step": 3950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029746282454241483,
+      "loss": 4.5156,
+      "step": 3955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974562574328456,
+      "loss": 4.4813,
+      "step": 3960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974496819079336,
+      "loss": 4.4344,
+      "step": 3965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029744309796805396,
+      "loss": 4.4375,
+      "step": 3970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029743650561358246,
+      "loss": 4.5031,
+      "step": 3975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974299048448954,
+      "loss": 4.4969,
+      "step": 3980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029742329566236937,
+      "loss": 4.4688,
+      "step": 3985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029741667806638174,
+      "loss": 4.4594,
+      "step": 3990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029741005205731,
+      "loss": 4.4813,
+      "step": 3995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974034176355324,
+      "loss": 4.4469,
+      "step": 4000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029739677480142745,
+      "loss": 4.3969,
+      "step": 4005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029739012355537434,
+      "loss": 4.4281,
+      "step": 4010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973834638977527,
+      "loss": 4.4313,
+      "step": 4015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029737679582894253,
+      "loss": 4.4344,
+      "step": 4020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973701193493244,
+      "loss": 4.4531,
+      "step": 4025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973634344592793,
+      "loss": 4.5125,
+      "step": 4030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029735674115918885,
+      "loss": 4.4562,
+      "step": 4035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029735003944943487,
+      "loss": 4.4781,
+      "step": 4040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029734332933039996,
+      "loss": 4.3812,
+      "step": 4045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297336610802467,
+      "loss": 4.3625,
+      "step": 4050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973298838660195,
+      "loss": 4.425,
+      "step": 4055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029732314852144124,
+      "loss": 4.4656,
+      "step": 4060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029731640476911666,
+      "loss": 4.5281,
+      "step": 4065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973096526094307,
+      "loss": 4.4469,
+      "step": 4070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973028920427686,
+      "loss": 4.4375,
+      "step": 4075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029729612306951623,
+      "loss": 4.3875,
+      "step": 4080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972893456900599,
+      "loss": 4.5094,
+      "step": 4085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029728255990478644,
+      "loss": 4.4781,
+      "step": 4090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029727576571408307,
+      "loss": 4.4844,
+      "step": 4095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029726896311833747,
+      "loss": 4.4437,
+      "step": 4100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029726215211793794,
+      "loss": 4.4969,
+      "step": 4105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029725533271327323,
+      "loss": 4.5,
+      "step": 4110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972485049047324,
+      "loss": 4.3937,
+      "step": 4115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972416686927052,
+      "loss": 4.4938,
+      "step": 4120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972348240775818,
+      "loss": 4.4531,
+      "step": 4125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029722797105975274,
+      "loss": 4.3937,
+      "step": 4130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029722110963960915,
+      "loss": 4.4031,
+      "step": 4135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029721423981754263,
+      "loss": 4.4531,
+      "step": 4140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029720736159394523,
+      "loss": 4.3844,
+      "step": 4145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029720047496920954,
+      "loss": 4.5125,
+      "step": 4150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029719357994372847,
+      "loss": 4.4562,
+      "step": 4155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971866765178956,
+      "loss": 4.4656,
+      "step": 4160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971797646921049,
+      "loss": 4.4313,
+      "step": 4165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971728444667508,
+      "loss": 4.4125,
+      "step": 4170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971659158422283,
+      "loss": 4.45,
+      "step": 4175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029715897881893276,
+      "loss": 4.4437,
+      "step": 4180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029715203339726007,
+      "loss": 4.4094,
+      "step": 4185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029714507957760667,
+      "loss": 4.4375,
+      "step": 4190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971381173603694,
+      "loss": 4.4406,
+      "step": 4195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002971311467459455,
+      "loss": 4.4469,
+      "step": 4200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029712416773473285,
+      "loss": 4.4375,
+      "step": 4205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029711718032712976,
+      "loss": 4.4531,
+      "step": 4210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297110184523535,
+      "loss": 4.3469,
+      "step": 4215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029710318032434785,
+      "loss": 4.3844,
+      "step": 4220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029709616772996794,
+      "loss": 4.3969,
+      "step": 4225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002970891467407955,
+      "loss": 4.5031,
+      "step": 4230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029708211735723135,
+      "loss": 4.3063,
+      "step": 4235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002970750795796765,
+      "loss": 4.4375,
+      "step": 4240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002970680334085327,
+      "loss": 4.4156,
+      "step": 4245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297060978844202,
+      "loss": 4.3594,
+      "step": 4250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029705391588708707,
+      "loss": 4.4281,
+      "step": 4255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297046844537591,
+      "loss": 4.425,
+      "step": 4260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029703976479611725,
+      "loss": 4.3781,
+      "step": 4265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029703267666306995,
+      "loss": 4.4375,
+      "step": 4270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029702558013885365,
+      "loss": 4.4125,
+      "step": 4275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002970184752238732,
+      "loss": 4.3656,
+      "step": 4280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970113619185343,
+      "loss": 4.3594,
+      "step": 4285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970042402232427,
+      "loss": 4.3937,
+      "step": 4290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029699711013840487,
+      "loss": 4.4625,
+      "step": 4295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969899716644279,
+      "loss": 4.3656,
+      "step": 4300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296982824801719,
+      "loss": 4.3844,
+      "step": 4305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029697566955068605,
+      "loss": 4.3594,
+      "step": 4310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969685059117375,
+      "loss": 4.3891,
+      "step": 4315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029696133388528215,
+      "loss": 4.475,
+      "step": 4320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029695415347172924,
+      "loss": 4.3656,
+      "step": 4325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029694696467148866,
+      "loss": 4.4281,
+      "step": 4330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029693976748497057,
+      "loss": 4.4656,
+      "step": 4335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029693256191258577,
+      "loss": 4.3469,
+      "step": 4340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029692534795474555,
+      "loss": 4.4344,
+      "step": 4345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969181256118615,
+      "loss": 4.3531,
+      "step": 4350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969108948843458,
+      "loss": 4.3344,
+      "step": 4355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969036557726111,
+      "loss": 4.45,
+      "step": 4360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029689640827707073,
+      "loss": 4.3125,
+      "step": 4365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968891523981381,
+      "loss": 4.325,
+      "step": 4370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029688188813622735,
+      "loss": 4.3219,
+      "step": 4375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029687461549175303,
+      "loss": 4.3453,
+      "step": 4380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968673344651303,
+      "loss": 4.4531,
+      "step": 4385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968600450567746,
+      "loss": 4.3656,
+      "step": 4390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968527472671019,
+      "loss": 4.3937,
+      "step": 4395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968454410965288,
+      "loss": 4.3594,
+      "step": 4400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029683812654547213,
+      "loss": 4.3781,
+      "step": 4405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968308036143495,
+      "loss": 4.3656,
+      "step": 4410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968234723035786,
+      "loss": 4.3688,
+      "step": 4415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029681613261357804,
+      "loss": 4.3656,
+      "step": 4420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029680878454476663,
+      "loss": 4.3594,
+      "step": 4425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968014280975637,
+      "loss": 4.3156,
+      "step": 4430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967940632723891,
+      "loss": 4.3344,
+      "step": 4435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967866900696631,
+      "loss": 4.4219,
+      "step": 4440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967793084898066,
+      "loss": 4.3812,
+      "step": 4445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967719185332407,
+      "loss": 4.3187,
+      "step": 4450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029676452020038736,
+      "loss": 4.3563,
+      "step": 4455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967571134916686,
+      "loss": 4.3625,
+      "step": 4460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967496984075073,
+      "loss": 4.3656,
+      "step": 4465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029674227494832643,
+      "loss": 4.375,
+      "step": 4470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967348431145498,
+      "loss": 4.2969,
+      "step": 4475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967274029066015,
+      "loss": 4.4062,
+      "step": 4480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029671995432490624,
+      "loss": 4.3312,
+      "step": 4485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029671249736988895,
+      "loss": 4.3,
+      "step": 4490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967050320419752,
+      "loss": 4.3187,
+      "step": 4495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966975583415913,
+      "loss": 4.2594,
+      "step": 4500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966900762691634,
+      "loss": 4.3844,
+      "step": 4505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029668258582511876,
+      "loss": 4.4656,
+      "step": 4510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966750870098848,
+      "loss": 4.3219,
+      "step": 4515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966675798238895,
+      "loss": 4.3031,
+      "step": 4520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029666006426756124,
+      "loss": 4.3281,
+      "step": 4525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966525403413289,
+      "loss": 4.3906,
+      "step": 4530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029664500804562195,
+      "loss": 4.2969,
+      "step": 4535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966374673808702,
+      "loss": 4.3063,
+      "step": 4540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029662991834750413,
+      "loss": 4.3156,
+      "step": 4545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966223609459544,
+      "loss": 4.3531,
+      "step": 4550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029661479517665234,
+      "loss": 4.3531,
+      "step": 4555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966072210400298,
+      "loss": 4.3781,
+      "step": 4560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029659963853651905,
+      "loss": 4.4,
+      "step": 4565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029659204766655276,
+      "loss": 4.4156,
+      "step": 4570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029658444843056413,
+      "loss": 4.2969,
+      "step": 4575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965768408289869,
+      "loss": 4.3688,
+      "step": 4580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029656922486225517,
+      "loss": 4.3656,
+      "step": 4585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965616005308037,
+      "loss": 4.2969,
+      "step": 4590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965539678350675,
+      "loss": 4.3063,
+      "step": 4595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965463267754822,
+      "loss": 4.3406,
+      "step": 4600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965386773524839,
+      "loss": 4.3812,
+      "step": 4605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965310195665091,
+      "loss": 4.3125,
+      "step": 4610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029652335341799494,
+      "loss": 4.2641,
+      "step": 4615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965156789073788,
+      "loss": 4.3406,
+      "step": 4620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029650799603509874,
+      "loss": 4.3625,
+      "step": 4625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965003048015932,
+      "loss": 4.3312,
+      "step": 4630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029649260520730114,
+      "loss": 4.35,
+      "step": 4635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029648489725266194,
+      "loss": 4.3375,
+      "step": 4640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964771809381155,
+      "loss": 4.3375,
+      "step": 4645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029646945626410223,
+      "loss": 4.3156,
+      "step": 4650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029646172323106296,
+      "loss": 4.2719,
+      "step": 4655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296453981839439,
+      "loss": 4.3156,
+      "step": 4660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029644623208967215,
+      "loss": 4.3031,
+      "step": 4665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964384739822047,
+      "loss": 4.3156,
+      "step": 4670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964307075174794,
+      "loss": 4.2969,
+      "step": 4675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964229326959395,
+      "loss": 4.3312,
+      "step": 4680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029641514951802867,
+      "loss": 4.2969,
+      "step": 4685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029640735798419117,
+      "loss": 4.3,
+      "step": 4690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029639955809487157,
+      "loss": 4.3625,
+      "step": 4695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963917498505151,
+      "loss": 4.3406,
+      "step": 4700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963839332515673,
+      "loss": 4.3063,
+      "step": 4705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029637610829847424,
+      "loss": 4.35,
+      "step": 4710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029636827499168267,
+      "loss": 4.3,
+      "step": 4715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029636043333163947,
+      "loss": 4.2812,
+      "step": 4720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963525833187922,
+      "loss": 4.3375,
+      "step": 4725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963447249535889,
+      "loss": 4.3937,
+      "step": 4730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296336858236478,
+      "loss": 4.2844,
+      "step": 4735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029632898316790847,
+      "loss": 4.2938,
+      "step": 4740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963210997483297,
+      "loss": 4.4156,
+      "step": 4745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963132079781917,
+      "loss": 4.275,
+      "step": 4750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029630530785794483,
+      "loss": 4.2469,
+      "step": 4755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962973993880399,
+      "loss": 4.3438,
+      "step": 4760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962894825689282,
+      "loss": 4.2375,
+      "step": 4765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029628155740106163,
+      "loss": 4.3625,
+      "step": 4770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962736238848925,
+      "loss": 4.275,
+      "step": 4775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962656820208735,
+      "loss": 4.3812,
+      "step": 4780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029625773180945794,
+      "loss": 4.2719,
+      "step": 4785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029624977325109944,
+      "loss": 4.2719,
+      "step": 4790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962418063462523,
+      "loss": 4.3375,
+      "step": 4795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962338310953712,
+      "loss": 4.3969,
+      "step": 4800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962258474989112,
+      "loss": 4.2969,
+      "step": 4805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029621785555732793,
+      "loss": 4.1937,
+      "step": 4810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962098552710776,
+      "loss": 4.2703,
+      "step": 4815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029620184664061667,
+      "loss": 4.275,
+      "step": 4820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002961938296664023,
+      "loss": 4.2906,
+      "step": 4825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029618580434889193,
+      "loss": 4.3031,
+      "step": 4830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029617777068854354,
+      "loss": 4.2828,
+      "step": 4835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002961697286858157,
+      "loss": 4.3625,
+      "step": 4840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002961616783411674,
+      "loss": 4.3406,
+      "step": 4845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002961536196550579,
+      "loss": 4.3375,
+      "step": 4850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029614555262794725,
+      "loss": 4.1719,
+      "step": 4855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002961374772602959,
+      "loss": 4.275,
+      "step": 4860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029612939355256446,
+      "loss": 4.3438,
+      "step": 4865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029612130150521456,
+      "loss": 4.3094,
+      "step": 4870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029611320111870776,
+      "loss": 4.2188,
+      "step": 4875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029610509239350654,
+      "loss": 4.3625,
+      "step": 4880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960969753300736,
+      "loss": 4.3094,
+      "step": 4885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029608884992887214,
+      "loss": 4.3781,
+      "step": 4890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960807161903659,
+      "loss": 4.3438,
+      "step": 4895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960725741150191,
+      "loss": 4.2656,
+      "step": 4900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960644237032964,
+      "loss": 4.2172,
+      "step": 4905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296056264955663,
+      "loss": 4.275,
+      "step": 4910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960480978725844,
+      "loss": 4.325,
+      "step": 4915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002960399224545268,
+      "loss": 4.2391,
+      "step": 4920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029603173870195663,
+      "loss": 4.2562,
+      "step": 4925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029602354661534107,
+      "loss": 4.2969,
+      "step": 4930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029601534619514764,
+      "loss": 4.275,
+      "step": 4935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029600713744184435,
+      "loss": 4.2594,
+      "step": 4940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029599892035589957,
+      "loss": 4.2781,
+      "step": 4945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029599069493778233,
+      "loss": 4.2375,
+      "step": 4950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029598246118796203,
+      "loss": 4.1625,
+      "step": 4955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029597421910690857,
+      "loss": 4.2812,
+      "step": 4960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959659686950924,
+      "loss": 4.2922,
+      "step": 4965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959577099529842,
+      "loss": 4.2875,
+      "step": 4970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029594944288105545,
+      "loss": 4.3406,
+      "step": 4975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029594116747977796,
+      "loss": 4.2812,
+      "step": 4980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959328837496238,
+      "loss": 4.2719,
+      "step": 4985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959245916910661,
+      "loss": 4.2812,
+      "step": 4990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959162913045777,
+      "loss": 4.2812,
+      "step": 4995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029590798259063254,
+      "loss": 4.2266,
+      "step": 5000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958996655497047,
+      "loss": 4.2812,
+      "step": 5005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029589134018226886,
+      "loss": 4.325,
+      "step": 5010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029588300648880015,
+      "loss": 4.3406,
+      "step": 5015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958746644697742,
+      "loss": 4.2094,
+      "step": 5020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029586631412566704,
+      "loss": 4.2781,
+      "step": 5025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958579554569553,
+      "loss": 4.2562,
+      "step": 5030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958495884641159,
+      "loss": 4.2375,
+      "step": 5035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958412131476265,
+      "loss": 4.2938,
+      "step": 5040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958328295079649,
+      "loss": 4.2406,
+      "step": 5045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958244375456097,
+      "loss": 4.2281,
+      "step": 5050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958160372610397,
+      "loss": 4.2531,
+      "step": 5055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958076286547344,
+      "loss": 4.3094,
+      "step": 5060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957992117271737,
+      "loss": 4.3031,
+      "step": 5065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957907864788379,
+      "loss": 4.2297,
+      "step": 5070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957823529102078,
+      "loss": 4.3125,
+      "step": 5075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957739110217648,
+      "loss": 4.2219,
+      "step": 5080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029576546081399056,
+      "loss": 4.2687,
+      "step": 5085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029575700228736746,
+      "loss": 4.1891,
+      "step": 5090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957485354423782,
+      "loss": 4.25,
+      "step": 5095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957400602795059,
+      "loss": 4.2516,
+      "step": 5100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957315767992343,
+      "loss": 4.25,
+      "step": 5105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957230850020475,
+      "loss": 4.2375,
+      "step": 5110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957145848884302,
+      "loss": 4.3063,
+      "step": 5115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957060764588675,
+      "loss": 4.2422,
+      "step": 5120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029569755971384497,
+      "loss": 4.2719,
+      "step": 5125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956890346538486,
+      "loss": 4.2313,
+      "step": 5130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029568050127936496,
+      "loss": 4.2875,
+      "step": 5135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956719595908811,
+      "loss": 4.3063,
+      "step": 5140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956634095888844,
+      "loss": 4.3125,
+      "step": 5145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029565485127386283,
+      "loss": 4.2562,
+      "step": 5150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956462846463049,
+      "loss": 4.2313,
+      "step": 5155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956377097066994,
+      "loss": 4.3,
+      "step": 5160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029562912645553577,
+      "loss": 4.2656,
+      "step": 5165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956205348933038,
+      "loss": 4.275,
+      "step": 5170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956119350204939,
+      "loss": 4.2078,
+      "step": 5175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956033268375968,
+      "loss": 4.2844,
+      "step": 5180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029559471034510375,
+      "loss": 4.1594,
+      "step": 5185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955860855435065,
+      "loss": 4.225,
+      "step": 5190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955774524332973,
+      "loss": 4.3312,
+      "step": 5195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029556881101496895,
+      "loss": 4.3281,
+      "step": 5200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955601612890144,
+      "loss": 4.2875,
+      "step": 5205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029555150325592745,
+      "loss": 4.2281,
+      "step": 5210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955428369162021,
+      "loss": 4.2328,
+      "step": 5215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029553416227033304,
+      "loss": 4.1813,
+      "step": 5220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029552547931881526,
+      "loss": 4.2375,
+      "step": 5225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029551678806214436,
+      "loss": 4.3,
+      "step": 5230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955080885008163,
+      "loss": 4.2516,
+      "step": 5235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954993806353276,
+      "loss": 4.2609,
+      "step": 5240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954906644661752,
+      "loss": 4.3094,
+      "step": 5245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954819399938565,
+      "loss": 4.2141,
+      "step": 5250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954732072188695,
+      "loss": 4.2812,
+      "step": 5255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029546446614171247,
+      "loss": 4.2313,
+      "step": 5260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954557167628843,
+      "loss": 4.2922,
+      "step": 5265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954469590828844,
+      "loss": 4.2766,
+      "step": 5270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029543819310221247,
+      "loss": 4.1937,
+      "step": 5275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954294188213688,
+      "loss": 4.2422,
+      "step": 5280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954206362408542,
+      "loss": 4.1688,
+      "step": 5285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954118453611698,
+      "loss": 4.2406,
+      "step": 5290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954030461828174,
+      "loss": 4.3063,
+      "step": 5295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953942387062991,
+      "loss": 4.2687,
+      "step": 5300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029538542293211756,
+      "loss": 4.2406,
+      "step": 5305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953765988607759,
+      "loss": 4.1453,
+      "step": 5310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953677664927777,
+      "loss": 4.2562,
+      "step": 5315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029535892582862705,
+      "loss": 4.1906,
+      "step": 5320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029535007686882846,
+      "loss": 4.2906,
+      "step": 5325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295341219613887,
+      "loss": 4.2094,
+      "step": 5330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295332354064308,
+      "loss": 4.2406,
+      "step": 5335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953234802205976,
+      "loss": 4.2875,
+      "step": 5340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029531459808326214,
+      "loss": 4.3438,
+      "step": 5345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953057076528085,
+      "loss": 4.2172,
+      "step": 5350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029529680892974417,
+      "loss": 4.2938,
+      "step": 5355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952879019145769,
+      "loss": 4.275,
+      "step": 5360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952789866078151,
+      "loss": 4.2156,
+      "step": 5365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952700630099674,
+      "loss": 4.1937,
+      "step": 5370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029526113112154326,
+      "loss": 4.2812,
+      "step": 5375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952521909430523,
+      "loss": 4.2906,
+      "step": 5380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952432424750048,
+      "loss": 4.2422,
+      "step": 5385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952342857179114,
+      "loss": 4.1469,
+      "step": 5390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029522532067228336,
+      "loss": 4.1875,
+      "step": 5395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952163473386322,
+      "loss": 4.175,
+      "step": 5400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952073657174701,
+      "loss": 4.1531,
+      "step": 5405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951983758093096,
+      "loss": 4.2578,
+      "step": 5410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951893776146638,
+      "loss": 4.2141,
+      "step": 5415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951803711340462,
+      "loss": 4.1594,
+      "step": 5420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029517135636797083,
+      "loss": 4.25,
+      "step": 5425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951623333169521,
+      "loss": 4.1672,
+      "step": 5430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295153301981505,
+      "loss": 4.2094,
+      "step": 5435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951442623621449,
+      "loss": 4.3031,
+      "step": 5440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029513521445938783,
+      "loss": 4.2875,
+      "step": 5445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029512615827375,
+      "loss": 4.275,
+      "step": 5450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029511709380574834,
+      "loss": 4.2531,
+      "step": 5455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951080210559001,
+      "loss": 4.2406,
+      "step": 5460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950989400247231,
+      "loss": 4.2188,
+      "step": 5465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950898507127356,
+      "loss": 4.2094,
+      "step": 5470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950807531204563,
+      "loss": 4.2469,
+      "step": 5475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950716472484044,
+      "loss": 4.2531,
+      "step": 5480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950625330970996,
+      "loss": 4.225,
+      "step": 5485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295053410667062,
+      "loss": 4.2406,
+      "step": 5490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950442799588123,
+      "loss": 4.2469,
+      "step": 5495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950351409728716,
+      "loss": 4.2,
+      "step": 5500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950259937097613,
+      "loss": 4.2297,
+      "step": 5505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029501683817000356,
+      "loss": 4.2234,
+      "step": 5510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002950076743541209,
+      "loss": 4.2516,
+      "step": 5515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029499850226263625,
+      "loss": 4.2531,
+      "step": 5520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029498932189607306,
+      "loss": 4.2438,
+      "step": 5525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029498013325495527,
+      "loss": 4.1813,
+      "step": 5530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029497093633980737,
+      "loss": 4.2156,
+      "step": 5535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029496173115115405,
+      "loss": 4.1766,
+      "step": 5540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002949525176895207,
+      "loss": 4.225,
+      "step": 5545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029494329595543327,
+      "loss": 4.2375,
+      "step": 5550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029493406594941786,
+      "loss": 4.2422,
+      "step": 5555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002949248276720014,
+      "loss": 4.2313,
+      "step": 5560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000294915581123711,
+      "loss": 4.2938,
+      "step": 5565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002949063263050744,
+      "loss": 4.2438,
+      "step": 5570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002948970632166198,
+      "loss": 4.2875,
+      "step": 5575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029488779185887575,
+      "loss": 4.3063,
+      "step": 5580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002948785122323714,
+      "loss": 4.2812,
+      "step": 5585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002948692243376365,
+      "loss": 4.1188,
+      "step": 5590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002948599281752009,
+      "loss": 4.2812,
+      "step": 5595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029485062374559525,
+      "loss": 4.2344,
+      "step": 5600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948413110493505,
+      "loss": 4.2375,
+      "step": 5605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029483199008699813,
+      "loss": 4.2656,
+      "step": 5610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029482266085907014,
+      "loss": 4.2141,
+      "step": 5615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948133233660989,
+      "loss": 4.2328,
+      "step": 5620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948039776086173,
+      "loss": 4.1969,
+      "step": 5625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029479462358715877,
+      "loss": 4.2375,
+      "step": 5630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029478526130225706,
+      "loss": 4.2203,
+      "step": 5635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029477589075444654,
+      "loss": 4.2031,
+      "step": 5640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947665119442619,
+      "loss": 4.1937,
+      "step": 5645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029475712487223846,
+      "loss": 4.3,
+      "step": 5650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029474772953891196,
+      "loss": 4.3156,
+      "step": 5655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029473832594481856,
+      "loss": 4.2641,
+      "step": 5660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947289140904949,
+      "loss": 4.1984,
+      "step": 5665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947194939764782,
+      "loss": 4.2234,
+      "step": 5670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029471006560330594,
+      "loss": 4.1734,
+      "step": 5675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947006289715163,
+      "loss": 4.1547,
+      "step": 5680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946911840816478,
+      "loss": 4.1875,
+      "step": 5685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029468173093423944,
+      "loss": 4.1641,
+      "step": 5690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029467226952983075,
+      "loss": 4.2313,
+      "step": 5695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029466279986896165,
+      "loss": 4.1203,
+      "step": 5700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029465332195217263,
+      "loss": 4.2188,
+      "step": 5705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029464383578000454,
+      "loss": 4.2531,
+      "step": 5710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029463434135299876,
+      "loss": 4.1813,
+      "step": 5715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029462483867169714,
+      "loss": 4.2875,
+      "step": 5720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946153277366421,
+      "loss": 4.1766,
+      "step": 5725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946058085483763,
+      "loss": 4.1531,
+      "step": 5730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029459628110744304,
+      "loss": 4.2297,
+      "step": 5735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029458674541438603,
+      "loss": 4.2281,
+      "step": 5740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029457720146974954,
+      "loss": 4.1531,
+      "step": 5745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945676492740782,
+      "loss": 4.2016,
+      "step": 5750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945580888279172,
+      "loss": 4.0984,
+      "step": 5755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294548520131812,
+      "loss": 4.1406,
+      "step": 5760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029453894318630895,
+      "loss": 4.1937,
+      "step": 5765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945293579919544,
+      "loss": 4.1922,
+      "step": 5770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029451976454929543,
+      "loss": 4.1719,
+      "step": 5775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029451016285887954,
+      "loss": 4.2969,
+      "step": 5780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945005529212547,
+      "loss": 4.2938,
+      "step": 5785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944909347369694,
+      "loss": 4.1625,
+      "step": 5790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029448130830657247,
+      "loss": 4.2406,
+      "step": 5795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029447167363061335,
+      "loss": 4.2031,
+      "step": 5800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944620307096418,
+      "loss": 4.2281,
+      "step": 5805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944523795442083,
+      "loss": 4.2625,
+      "step": 5810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944427201348635,
+      "loss": 4.2344,
+      "step": 5815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029443305248215873,
+      "loss": 4.225,
+      "step": 5820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029442337658664575,
+      "loss": 4.1937,
+      "step": 5825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944136924488767,
+      "loss": 4.1641,
+      "step": 5830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944040000694043,
+      "loss": 4.2922,
+      "step": 5835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029439429944878165,
+      "loss": 4.1984,
+      "step": 5840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943845905875624,
+      "loss": 4.2203,
+      "step": 5845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029437487348630057,
+      "loss": 4.1609,
+      "step": 5850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029436514814555083,
+      "loss": 4.2625,
+      "step": 5855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943554145658682,
+      "loss": 4.1406,
+      "step": 5860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029434567274780804,
+      "loss": 4.2313,
+      "step": 5865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029433592269192645,
+      "loss": 4.1844,
+      "step": 5870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029432616439877983,
+      "loss": 4.1094,
+      "step": 5875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029431639786892507,
+      "loss": 4.2062,
+      "step": 5880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943066231029195,
+      "loss": 4.2969,
+      "step": 5885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029429684010132107,
+      "loss": 4.2188,
+      "step": 5890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294287048864688,
+      "loss": 4.2313,
+      "step": 5895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029427724939357915,
+      "loss": 4.1719,
+      "step": 5900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942674416885538,
+      "loss": 4.2844,
+      "step": 5905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029425762575017154,
+      "loss": 4.2531,
+      "step": 5910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942478015789927,
+      "loss": 4.2156,
+      "step": 5915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942379691755779,
+      "loss": 4.2016,
+      "step": 5920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029422812854048826,
+      "loss": 4.2812,
+      "step": 5925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942182796742854,
+      "loss": 4.2125,
+      "step": 5930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942084225775314,
+      "loss": 4.1875,
+      "step": 5935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941985572507888,
+      "loss": 4.1531,
+      "step": 5940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029418868369462064,
+      "loss": 4.2453,
+      "step": 5945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029417880190959035,
+      "loss": 4.2,
+      "step": 5950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294168911896262,
+      "loss": 4.2938,
+      "step": 5955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029415901365519986,
+      "loss": 4.2172,
+      "step": 5960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029414910718696883,
+      "loss": 4.2062,
+      "step": 5965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029413919249213444,
+      "loss": 4.2391,
+      "step": 5970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941292695712624,
+      "loss": 4.2188,
+      "step": 5975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294119338424919,
+      "loss": 4.3406,
+      "step": 5980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941093990536711,
+      "loss": 4.3031,
+      "step": 5985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029409945145808584,
+      "loss": 4.2281,
+      "step": 5990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294089495638731,
+      "loss": 4.2625,
+      "step": 5995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940795315961747,
+      "loss": 4.1547,
+      "step": 6000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940695593309857,
+      "loss": 4.1484,
+      "step": 6005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294059578843733,
+      "loss": 4.2094,
+      "step": 6010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940495901349862,
+      "loss": 4.2375,
+      "step": 6015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029403959320531534,
+      "loss": 4.1406,
+      "step": 6020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940295880552911,
+      "loss": 4.1813,
+      "step": 6025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940195746854843,
+      "loss": 4.2125,
+      "step": 6030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940095530964665,
+      "loss": 4.1734,
+      "step": 6035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029399952328880956,
+      "loss": 4.1594,
+      "step": 6040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000293989485263086,
+      "loss": 4.2219,
+      "step": 6045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029397943901986857,
+      "loss": 4.1547,
+      "step": 6050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939693845597306,
+      "loss": 4.1719,
+      "step": 6055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029395932188324607,
+      "loss": 4.1875,
+      "step": 6060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939492509909891,
+      "loss": 4.2422,
+      "step": 6065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029393917188353443,
+      "loss": 4.15,
+      "step": 6070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939290845614574,
+      "loss": 4.1281,
+      "step": 6075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939189890253336,
+      "loss": 4.2,
+      "step": 6080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029390888527573917,
+      "loss": 4.1625,
+      "step": 6085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938987733132508,
+      "loss": 4.275,
+      "step": 6090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938886531384455,
+      "loss": 4.1594,
+      "step": 6095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938785247519009,
+      "loss": 4.2031,
+      "step": 6100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029386838815419506,
+      "loss": 4.2219,
+      "step": 6105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029385824334590634,
+      "loss": 4.1141,
+      "step": 6110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938480903276138,
+      "loss": 4.2141,
+      "step": 6115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938379290998969,
+      "loss": 4.2188,
+      "step": 6120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938277596633355,
+      "loss": 4.1594,
+      "step": 6125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029381758201851,
+      "loss": 4.2969,
+      "step": 6130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002938073961660011,
+      "loss": 4.225,
+      "step": 6135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002937972021063903,
+      "loss": 4.2375,
+      "step": 6140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002937869998402593,
+      "loss": 4.2016,
+      "step": 6145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029377678936819027,
+      "loss": 4.2531,
+      "step": 6150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029376657069076607,
+      "loss": 4.2,
+      "step": 6155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002937563438085698,
+      "loss": 4.1875,
+      "step": 6160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000293746108722185,
+      "loss": 4.2656,
+      "step": 6165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000293735865432196,
+      "loss": 4.2297,
+      "step": 6170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029372561393918725,
+      "loss": 4.1359,
+      "step": 6175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002937153542437439,
+      "loss": 4.1625,
+      "step": 6180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029370508634645137,
+      "loss": 4.15,
+      "step": 6185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936948102478957,
+      "loss": 4.1781,
+      "step": 6190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029368452594866326,
+      "loss": 4.1703,
+      "step": 6195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029367423344934114,
+      "loss": 4.1875,
+      "step": 6200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029366393275051663,
+      "loss": 4.1016,
+      "step": 6205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936536238527776,
+      "loss": 4.0656,
+      "step": 6210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936433067567124,
+      "loss": 4.1688,
+      "step": 6215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936329814629098,
+      "loss": 4.0656,
+      "step": 6220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936226479719591,
+      "loss": 4.2016,
+      "step": 6225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029361230628445,
+      "loss": 4.2078,
+      "step": 6230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002936019564009727,
+      "loss": 4.1406,
+      "step": 6235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002935915983221179,
+      "loss": 4.1703,
+      "step": 6240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002935812320484768,
+      "loss": 4.1891,
+      "step": 6245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029357085758064083,
+      "loss": 4.1953,
+      "step": 6250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029356047491920216,
+      "loss": 4.2078,
+      "step": 6255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935500840647534,
+      "loss": 4.0938,
+      "step": 6260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935396850178874,
+      "loss": 4.1453,
+      "step": 6265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935292777791978,
+      "loss": 4.1922,
+      "step": 6270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935188623492784,
+      "loss": 4.1625,
+      "step": 6275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935084387287237,
+      "loss": 4.1437,
+      "step": 6280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029349800691812856,
+      "loss": 4.2375,
+      "step": 6285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934875669180883,
+      "loss": 4.2484,
+      "step": 6290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934771187291987,
+      "loss": 4.2344,
+      "step": 6295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934666623520562,
+      "loss": 4.1828,
+      "step": 6300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934561977872573,
+      "loss": 4.1797,
+      "step": 6305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029344572503539943,
+      "loss": 4.2125,
+      "step": 6310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029343524409708015,
+      "loss": 4.1781,
+      "step": 6315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029342475497289765,
+      "loss": 4.2281,
+      "step": 6320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029341425766345056,
+      "loss": 4.2266,
+      "step": 6325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029340375216933796,
+      "loss": 4.1266,
+      "step": 6330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933932384911593,
+      "loss": 4.2406,
+      "step": 6335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933827166295147,
+      "loss": 4.1297,
+      "step": 6340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933721865850047,
+      "loss": 4.1969,
+      "step": 6345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029336164835823007,
+      "loss": 4.2344,
+      "step": 6350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933511019497924,
+      "loss": 4.2281,
+      "step": 6355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933405473602935,
+      "loss": 4.1969,
+      "step": 6360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933299845903357,
+      "loss": 4.2344,
+      "step": 6365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933194136405218,
+      "loss": 4.1469,
+      "step": 6370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029330883451145516,
+      "loss": 4.1766,
+      "step": 6375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029329824720373957,
+      "loss": 4.1969,
+      "step": 6380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932876517179791,
+      "loss": 4.1437,
+      "step": 6385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932770480547785,
+      "loss": 4.2109,
+      "step": 6390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293266436214743,
+      "loss": 4.1484,
+      "step": 6395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029325581619847813,
+      "loss": 4.2719,
+      "step": 6400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029324518800659,
+      "loss": 4.2266,
+      "step": 6405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932345516396851,
+      "loss": 4.2484,
+      "step": 6410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932239070983706,
+      "loss": 4.2438,
+      "step": 6415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932132543832539,
+      "loss": 4.1656,
+      "step": 6420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932025934949429,
+      "loss": 4.1922,
+      "step": 6425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931919244340461,
+      "loss": 4.1984,
+      "step": 6430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029318124720117234,
+      "loss": 4.1672,
+      "step": 6435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029317056179693096,
+      "loss": 4.25,
+      "step": 6440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931598682219318,
+      "loss": 4.1734,
+      "step": 6445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931491664767852,
+      "loss": 4.2219,
+      "step": 6450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931384565621018,
+      "loss": 4.1906,
+      "step": 6455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931277384784929,
+      "loss": 4.1906,
+      "step": 6460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931170122265701,
+      "loss": 4.1172,
+      "step": 6465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029310627780694564,
+      "loss": 4.2344,
+      "step": 6470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293095535220232,
+      "loss": 4.1906,
+      "step": 6475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930847844670425,
+      "loss": 4.2594,
+      "step": 6480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930740255479904,
+      "loss": 4.275,
+      "step": 6485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029306325846368994,
+      "loss": 4.2156,
+      "step": 6490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029305248321475545,
+      "loss": 4.2609,
+      "step": 6495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029304169980180197,
+      "loss": 4.2156,
+      "step": 6500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029303090822544483,
+      "loss": 4.1078,
+      "step": 6505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029302010848630003,
+      "loss": 4.1,
+      "step": 6510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029300930058498377,
+      "loss": 4.1813,
+      "step": 6515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029299848452211296,
+      "loss": 4.1813,
+      "step": 6520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029298766029830484,
+      "loss": 4.1609,
+      "step": 6525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029297682791417707,
+      "loss": 4.2031,
+      "step": 6530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292965987370348,
+      "loss": 4.2469,
+      "step": 6535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029295513866743617,
+      "loss": 4.3531,
+      "step": 6540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029294428180606086,
+      "loss": 4.2094,
+      "step": 6545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929334167868415,
+      "loss": 4.2578,
+      "step": 6550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929225436103983,
+      "loss": 4.2188,
+      "step": 6555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029291166227735175,
+      "loss": 4.1531,
+      "step": 6560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929007727883228,
+      "loss": 4.1844,
+      "step": 6565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292889875143933,
+      "loss": 4.2313,
+      "step": 6570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928789693448042,
+      "loss": 4.2406,
+      "step": 6575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928680553915588,
+      "loss": 4.25,
+      "step": 6580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928571332848197,
+      "loss": 4.1688,
+      "step": 6585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928462030252103,
+      "loss": 4.1359,
+      "step": 6590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928352646133542,
+      "loss": 4.1359,
+      "step": 6595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928243180498758,
+      "loss": 4.2391,
+      "step": 6600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928133633353998,
+      "loss": 4.0781,
+      "step": 6605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928024004705513,
+      "loss": 4.1437,
+      "step": 6610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029279142945595613,
+      "loss": 4.1422,
+      "step": 6615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927804502922402,
+      "loss": 4.1281,
+      "step": 6620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927694629800302,
+      "loss": 4.1437,
+      "step": 6625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927584675199532,
+      "loss": 4.1391,
+      "step": 6630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029274746391263665,
+      "loss": 4.1844,
+      "step": 6635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029273645215870857,
+      "loss": 4.1937,
+      "step": 6640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927254322587974,
+      "loss": 4.1094,
+      "step": 6645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927144042135321,
+      "loss": 4.2219,
+      "step": 6650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927033680235418,
+      "loss": 4.1891,
+      "step": 6655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029269232368945664,
+      "loss": 4.2031,
+      "step": 6660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029268127121190676,
+      "loss": 4.0875,
+      "step": 6665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926702105915229,
+      "loss": 4.2281,
+      "step": 6670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926591418289364,
+      "loss": 4.1609,
+      "step": 6675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292648064924779,
+      "loss": 4.1391,
+      "step": 6680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029263697987968264,
+      "loss": 4.15,
+      "step": 6685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926258866942801,
+      "loss": 4.2266,
+      "step": 6690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029261478536920447,
+      "loss": 4.1828,
+      "step": 6695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926036759050892,
+      "loss": 4.1391,
+      "step": 6700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029259255830256845,
+      "loss": 4.1266,
+      "step": 6705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925814325622766,
+      "loss": 4.1266,
+      "step": 6710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029257029868484864,
+      "loss": 4.2297,
+      "step": 6715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029255915667092,
+      "loss": 4.1453,
+      "step": 6720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029254800652112647,
+      "loss": 4.1063,
+      "step": 6725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925368482361045,
+      "loss": 4.2188,
+      "step": 6730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925256818164908,
+      "loss": 4.1484,
+      "step": 6735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925145072629227,
+      "loss": 4.2375,
+      "step": 6740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925033245760379,
+      "loss": 4.1656,
+      "step": 6745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029249213375647464,
+      "loss": 4.2734,
+      "step": 6750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029248093480487155,
+      "loss": 4.1797,
+      "step": 6755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029246972772186775,
+      "loss": 4.2062,
+      "step": 6760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924585125081028,
+      "loss": 4.1078,
+      "step": 6765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924472891642168,
+      "loss": 4.2203,
+      "step": 6770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029243605769085034,
+      "loss": 4.175,
+      "step": 6775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924248180886442,
+      "loss": 4.1688,
+      "step": 6780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029241357035824,
+      "loss": 4.2594,
+      "step": 6785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924023145002796,
+      "loss": 4.275,
+      "step": 6790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029239105051540536,
+      "loss": 4.225,
+      "step": 6795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002923797784042601,
+      "loss": 4.1672,
+      "step": 6800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002923684981674872,
+      "loss": 4.1281,
+      "step": 6805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002923572098057303,
+      "loss": 4.1312,
+      "step": 6810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002923459133196337,
+      "loss": 4.25,
+      "step": 6815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029233460870984213,
+      "loss": 4.1234,
+      "step": 6820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029232329597700065,
+      "loss": 4.1891,
+      "step": 6825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029231197512175495,
+      "loss": 4.1609,
+      "step": 6830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029230064614475115,
+      "loss": 4.1312,
+      "step": 6835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029228930904663566,
+      "loss": 4.2625,
+      "step": 6840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002922779638280556,
+      "loss": 4.3063,
+      "step": 6845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029226661048965835,
+      "loss": 4.1719,
+      "step": 6850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292255249032092,
+      "loss": 4.1063,
+      "step": 6855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002922438794560048,
+      "loss": 4.1391,
+      "step": 6860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029223250176204567,
+      "loss": 4.2031,
+      "step": 6865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029222111595086393,
+      "loss": 4.1813,
+      "step": 6870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002922097220231094,
+      "loss": 4.1781,
+      "step": 6875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002921983199794323,
+      "loss": 4.1891,
+      "step": 6880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002921869098204834,
+      "loss": 4.1078,
+      "step": 6885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029217549154691373,
+      "loss": 4.2781,
+      "step": 6890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002921640651593751,
+      "loss": 4.1594,
+      "step": 6895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029215263065851955,
+      "loss": 4.2141,
+      "step": 6900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002921411880449996,
+      "loss": 4.1656,
+      "step": 6905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029212973731946836,
+      "loss": 4.1859,
+      "step": 6910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921182784825793,
+      "loss": 4.1875,
+      "step": 6915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029210681153498633,
+      "loss": 4.0969,
+      "step": 6920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029209533647734395,
+      "loss": 4.1109,
+      "step": 6925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029208385331030696,
+      "loss": 4.2687,
+      "step": 6930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029207236203453076,
+      "loss": 4.2125,
+      "step": 6935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029206086265067113,
+      "loss": 4.2281,
+      "step": 6940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920493551593844,
+      "loss": 4.0453,
+      "step": 6945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029203783956132723,
+      "loss": 4.1141,
+      "step": 6950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029202631585715686,
+      "loss": 4.1266,
+      "step": 6955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920147840475309,
+      "loss": 4.1594,
+      "step": 6960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920032441331076,
+      "loss": 4.2375,
+      "step": 6965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029199169611454536,
+      "loss": 4.1328,
+      "step": 6970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029198013999250333,
+      "loss": 4.0984,
+      "step": 6975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291968575767641,
+      "loss": 4.2531,
+      "step": 6980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029195700344061837,
+      "loss": 4.1828,
+      "step": 6985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029194542301209584,
+      "loss": 4.1125,
+      "step": 6990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919338344827343,
+      "loss": 4.2656,
+      "step": 6995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029192223785319516,
+      "loss": 4.1906,
+      "step": 7000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919106331241402,
+      "loss": 4.2062,
+      "step": 7005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029189902029623163,
+      "loss": 4.1719,
+      "step": 7010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029188739937013235,
+      "loss": 4.1719,
+      "step": 7015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918757703465055,
+      "loss": 4.1844,
+      "step": 7020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918641332260147,
+      "loss": 4.225,
+      "step": 7025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918524880093242,
+      "loss": 4.2281,
+      "step": 7030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918408346970984,
+      "loss": 4.1156,
+      "step": 7035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029182917329000253,
+      "loss": 4.2031,
+      "step": 7040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918175037887021,
+      "loss": 4.1531,
+      "step": 7045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029180582619386296,
+      "loss": 4.1484,
+      "step": 7050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029179414050615164,
+      "loss": 4.1297,
+      "step": 7055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917824467262351,
+      "loss": 4.0969,
+      "step": 7060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029177074485478056,
+      "loss": 4.225,
+      "step": 7065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029175903489245595,
+      "loss": 4.2719,
+      "step": 7070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029174731683992953,
+      "loss": 4.2344,
+      "step": 7075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029173559069787006,
+      "loss": 4.2641,
+      "step": 7080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917238564669468,
+      "loss": 4.2078,
+      "step": 7085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917121141478293,
+      "loss": 4.175,
+      "step": 7090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029170036374118777,
+      "loss": 4.1688,
+      "step": 7095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029168860524769286,
+      "loss": 4.2047,
+      "step": 7100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916768386680155,
+      "loss": 4.0859,
+      "step": 7105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916650640028273,
+      "loss": 4.2328,
+      "step": 7110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916532812528003,
+      "loss": 4.2047,
+      "step": 7115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029164149041860675,
+      "loss": 4.1031,
+      "step": 7120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029162969150091974,
+      "loss": 4.1375,
+      "step": 7125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029161788450041257,
+      "loss": 4.1813,
+      "step": 7130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291606069417759,
+      "loss": 4.1609,
+      "step": 7135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915942462536334,
+      "loss": 4.2281,
+      "step": 7140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029158241500871053,
+      "loss": 4.2781,
+      "step": 7145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029157057568366556,
+      "loss": 4.1891,
+      "step": 7150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915587282791742,
+      "loss": 4.2656,
+      "step": 7155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915468727959125,
+      "loss": 4.0859,
+      "step": 7160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029153500923455707,
+      "loss": 4.2109,
+      "step": 7165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915231375957851,
+      "loss": 4.1766,
+      "step": 7170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029151125788027393,
+      "loss": 4.0625,
+      "step": 7175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914993700887017,
+      "loss": 4.1828,
+      "step": 7180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914874742217466,
+      "loss": 4.1266,
+      "step": 7185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914755702800878,
+      "loss": 4.1109,
+      "step": 7190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029146365826440457,
+      "loss": 4.1922,
+      "step": 7195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914517381753767,
+      "loss": 4.1594,
+      "step": 7200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914398100136844,
+      "loss": 4.1578,
+      "step": 7205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029142787378000855,
+      "loss": 4.1531,
+      "step": 7210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914159294750303,
+      "loss": 4.1625,
+      "step": 7215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029140397709943123,
+      "loss": 4.1547,
+      "step": 7220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913920166538936,
+      "loss": 4.1141,
+      "step": 7225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029138004813909996,
+      "loss": 4.1906,
+      "step": 7230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029136807155573326,
+      "loss": 4.2,
+      "step": 7235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913560869044771,
+      "loss": 4.1344,
+      "step": 7240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913440941860154,
+      "loss": 4.2047,
+      "step": 7245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913320934010326,
+      "loss": 4.1516,
+      "step": 7250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029132008455021364,
+      "loss": 4.1719,
+      "step": 7255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029130806763424374,
+      "loss": 4.1109,
+      "step": 7260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912960426538088,
+      "loss": 4.1984,
+      "step": 7265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912840096095951,
+      "loss": 4.2062,
+      "step": 7270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029127196850228934,
+      "loss": 4.1094,
+      "step": 7275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029125991933257864,
+      "loss": 4.1703,
+      "step": 7280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029124786210115073,
+      "loss": 4.1984,
+      "step": 7285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029123579680869366,
+      "loss": 4.3125,
+      "step": 7290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029122372345589606,
+      "loss": 4.1875,
+      "step": 7295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291211642043447,
+      "loss": 4.2031,
+      "step": 7300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029119955257203583,
+      "loss": 4.1781,
+      "step": 7305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911874550423526,
+      "loss": 4.2531,
+      "step": 7310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029117534945508766,
+      "loss": 4.2125,
+      "step": 7315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029116323581093186,
+      "loss": 4.1734,
+      "step": 7320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029115111411057664,
+      "loss": 4.2578,
+      "step": 7325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029113898435471366,
+      "loss": 4.1813,
+      "step": 7330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911268465440352,
+      "loss": 4.1422,
+      "step": 7335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291114700679234,
+      "loss": 4.1953,
+      "step": 7340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029110254676100323,
+      "loss": 4.1188,
+      "step": 7345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910903847900365,
+      "loss": 4.1703,
+      "step": 7350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910782147670279,
+      "loss": 4.2297,
+      "step": 7355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291066036692672,
+      "loss": 4.2469,
+      "step": 7360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029105385056766366,
+      "loss": 4.2,
+      "step": 7365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910416563926985,
+      "loss": 4.2188,
+      "step": 7370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910294541684724,
+      "loss": 4.1641,
+      "step": 7375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910172438956818,
+      "loss": 4.2,
+      "step": 7380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910050255750234,
+      "loss": 4.1641,
+      "step": 7385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909927992071946,
+      "loss": 4.1891,
+      "step": 7390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909805647928931,
+      "loss": 4.1359,
+      "step": 7395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909683223328172,
+      "loss": 4.1937,
+      "step": 7400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029095607182766554,
+      "loss": 4.2516,
+      "step": 7405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909438132781372,
+      "loss": 4.2078,
+      "step": 7410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909315466849319,
+      "loss": 4.2344,
+      "step": 7415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029091927204874963,
+      "loss": 4.1578,
+      "step": 7420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029090698937029084,
+      "loss": 4.1516,
+      "step": 7425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908946986502566,
+      "loss": 4.2062,
+      "step": 7430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908823998893483,
+      "loss": 4.1547,
+      "step": 7435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908700930882679,
+      "loss": 4.1406,
+      "step": 7440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908577782477176,
+      "loss": 4.1906,
+      "step": 7445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029084545536840035,
+      "loss": 4.0656,
+      "step": 7450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908331244510193,
+      "loss": 4.1516,
+      "step": 7455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029082078549627833,
+      "loss": 4.1797,
+      "step": 7460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002908084385048815,
+      "loss": 4.1406,
+      "step": 7465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029079608347753344,
+      "loss": 4.1797,
+      "step": 7470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029078372041493937,
+      "loss": 4.1375,
+      "step": 7475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029077134931780477,
+      "loss": 4.175,
+      "step": 7480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029075897018683566,
+      "loss": 4.1453,
+      "step": 7485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002907465830227386,
+      "loss": 4.1969,
+      "step": 7490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002907341878262204,
+      "loss": 4.1328,
+      "step": 7495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002907217845979885,
+      "loss": 4.2156,
+      "step": 7500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002907093733387508,
+      "loss": 4.175,
+      "step": 7505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002906969540492156,
+      "loss": 4.2188,
+      "step": 7510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002906845267300916,
+      "loss": 4.225,
+      "step": 7515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002906720913820881,
+      "loss": 4.15,
+      "step": 7520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002906596480059148,
+      "loss": 4.1953,
+      "step": 7525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029064719660228176,
+      "loss": 4.1328,
+      "step": 7530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029063473717189965,
+      "loss": 4.1734,
+      "step": 7535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002906222697154795,
+      "loss": 4.2062,
+      "step": 7540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029060979423373287,
+      "loss": 4.1484,
+      "step": 7545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002905973107273717,
+      "loss": 4.0891,
+      "step": 7550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029058481919710846,
+      "loss": 4.2453,
+      "step": 7555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029057231964365595,
+      "loss": 4.1984,
+      "step": 7560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029055981206772765,
+      "loss": 4.0578,
+      "step": 7565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002905472964700373,
+      "loss": 4.1688,
+      "step": 7570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029053477285129913,
+      "loss": 4.1453,
+      "step": 7575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029052224121222796,
+      "loss": 4.1906,
+      "step": 7580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905097015535389,
+      "loss": 4.1063,
+      "step": 7585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904971538759476,
+      "loss": 4.1656,
+      "step": 7590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904845981801702,
+      "loss": 4.1578,
+      "step": 7595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029047203446692314,
+      "loss": 4.1875,
+      "step": 7600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904594627369236,
+      "loss": 4.2313,
+      "step": 7605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029044688299088893,
+      "loss": 4.2219,
+      "step": 7610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029043429522953706,
+      "loss": 4.1641,
+      "step": 7615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029042169945358643,
+      "loss": 4.1719,
+      "step": 7620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904090956637559,
+      "loss": 4.0922,
+      "step": 7625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029039648386076466,
+      "loss": 4.1094,
+      "step": 7630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903838640453326,
+      "loss": 4.1391,
+      "step": 7635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029037123621817977,
+      "loss": 4.1344,
+      "step": 7640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029035860038002705,
+      "loss": 4.1609,
+      "step": 7645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903459565315954,
+      "loss": 4.1312,
+      "step": 7650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903333046736065,
+      "loss": 4.1641,
+      "step": 7655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903206448067823,
+      "loss": 4.1344,
+      "step": 7660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903079769318453,
+      "loss": 4.1656,
+      "step": 7665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029029530104951863,
+      "loss": 4.0922,
+      "step": 7670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029028261716052553,
+      "loss": 4.2297,
+      "step": 7675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029026992526559,
+      "loss": 4.2094,
+      "step": 7680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902572253654362,
+      "loss": 4.2062,
+      "step": 7685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290244517460789,
+      "loss": 4.1312,
+      "step": 7690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029023180155237374,
+      "loss": 4.1672,
+      "step": 7695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029021907764091603,
+      "loss": 4.2062,
+      "step": 7700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029020634572714195,
+      "loss": 4.2078,
+      "step": 7705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029019360581177824,
+      "loss": 4.2344,
+      "step": 7710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029018085789555187,
+      "loss": 4.1312,
+      "step": 7715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901681019791904,
+      "loss": 4.2219,
+      "step": 7720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901553380634218,
+      "loss": 4.0969,
+      "step": 7725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029014256614897467,
+      "loss": 4.1531,
+      "step": 7730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901297862365776,
+      "loss": 4.1578,
+      "step": 7735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901169983269602,
+      "loss": 4.1828,
+      "step": 7740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901042024208521,
+      "loss": 4.0875,
+      "step": 7745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029009139851898373,
+      "loss": 4.1813,
+      "step": 7750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900785866220857,
+      "loss": 4.1406,
+      "step": 7755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029006576673088917,
+      "loss": 4.1906,
+      "step": 7760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029005293884612585,
+      "loss": 4.2297,
+      "step": 7765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900401029685278,
+      "loss": 4.2156,
+      "step": 7770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029002725909882746,
+      "loss": 4.1906,
+      "step": 7775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290014407237758,
+      "loss": 4.1516,
+      "step": 7780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029000154738605284,
+      "loss": 4.1609,
+      "step": 7785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028998867954444575,
+      "loss": 4.1813,
+      "step": 7790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028997580371367134,
+      "loss": 4.1469,
+      "step": 7795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899629198944642,
+      "loss": 4.1453,
+      "step": 7800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028995002808755976,
+      "loss": 4.175,
+      "step": 7805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899371282936937,
+      "loss": 4.1344,
+      "step": 7810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028992422051360217,
+      "loss": 4.25,
+      "step": 7815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899113047480219,
+      "loss": 4.1312,
+      "step": 7820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028989838099769,
+      "loss": 4.2406,
+      "step": 7825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028988544926334395,
+      "loss": 4.2062,
+      "step": 7830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898725095457218,
+      "loss": 4.1437,
+      "step": 7835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028985956184556205,
+      "loss": 4.1922,
+      "step": 7840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898466061636036,
+      "loss": 4.2062,
+      "step": 7845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028983364250058586,
+      "loss": 4.1234,
+      "step": 7850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898206708572486,
+      "loss": 4.2188,
+      "step": 7855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898076912343322,
+      "loss": 4.2125,
+      "step": 7860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028979470363257734,
+      "loss": 4.2094,
+      "step": 7865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028978170805272525,
+      "loss": 4.1781,
+      "step": 7870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897687044955176,
+      "loss": 4.1156,
+      "step": 7875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897556929616965,
+      "loss": 4.1453,
+      "step": 7880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028974267345200445,
+      "loss": 4.2047,
+      "step": 7885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897296459671846,
+      "loss": 4.1937,
+      "step": 7890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028971661050798035,
+      "loss": 4.1437,
+      "step": 7895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028970356707513564,
+      "loss": 4.2609,
+      "step": 7900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896905156693949,
+      "loss": 4.1516,
+      "step": 7905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028967745629150297,
+      "loss": 4.2188,
+      "step": 7910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028966438894220507,
+      "loss": 4.1609,
+      "step": 7915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028965131362224697,
+      "loss": 4.2812,
+      "step": 7920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000289638230332375,
+      "loss": 4.1594,
+      "step": 7925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028962513907333575,
+      "loss": 4.1172,
+      "step": 7930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896120398458763,
+      "loss": 4.0875,
+      "step": 7935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028959893265074425,
+      "loss": 4.1609,
+      "step": 7940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895858174886876,
+      "loss": 4.2156,
+      "step": 7945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895726943604549,
+      "loss": 4.1094,
+      "step": 7950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028955956326679513,
+      "loss": 4.2125,
+      "step": 7955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895464242084576,
+      "loss": 4.1547,
+      "step": 7960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895332771861921,
+      "loss": 4.0875,
+      "step": 7965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028952012220074906,
+      "loss": 4.1609,
+      "step": 7970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028950695925287914,
+      "loss": 4.1047,
+      "step": 7975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028949378834333363,
+      "loss": 4.1734,
+      "step": 7980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894806094728641,
+      "loss": 4.0969,
+      "step": 7985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894674226422227,
+      "loss": 4.1328,
+      "step": 7990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028945422785216213,
+      "loss": 4.2516,
+      "step": 7995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028944102510343526,
+      "loss": 4.1813,
+      "step": 8000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028942781439679564,
+      "loss": 4.1172,
+      "step": 8005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028941459573299716,
+      "loss": 4.1578,
+      "step": 8010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028940136911279426,
+      "loss": 4.1266,
+      "step": 8015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028938813453694177,
+      "loss": 4.1688,
+      "step": 8020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000289374892006195,
+      "loss": 4.2297,
+      "step": 8025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893616415213097,
+      "loss": 4.1906,
+      "step": 8030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028934838308304205,
+      "loss": 4.1984,
+      "step": 8035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893351166921487,
+      "loss": 4.2391,
+      "step": 8040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893218423493868,
+      "loss": 4.1016,
+      "step": 8045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028930856005551396,
+      "loss": 4.1312,
+      "step": 8050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002892952698112881,
+      "loss": 4.0906,
+      "step": 8055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028928197161746775,
+      "loss": 4.1125,
+      "step": 8060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028926866547481184,
+      "loss": 4.2281,
+      "step": 8065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028925535138407975,
+      "loss": 4.1562,
+      "step": 8070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002892420293460313,
+      "loss": 4.0734,
+      "step": 8075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028922869936142683,
+      "loss": 4.2062,
+      "step": 8080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000289215361431027,
+      "loss": 4.1844,
+      "step": 8085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028920201555559307,
+      "loss": 4.2031,
+      "step": 8090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002891886617358867,
+      "loss": 4.1266,
+      "step": 8095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028917529997266994,
+      "loss": 4.1625,
+      "step": 8100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002891619302667054,
+      "loss": 4.1125,
+      "step": 8105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028914855261875607,
+      "loss": 4.1625,
+      "step": 8110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028913516702958536,
+      "loss": 4.2016,
+      "step": 8115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028912177349995727,
+      "loss": 4.1437,
+      "step": 8120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002891083720306361,
+      "loss": 4.1688,
+      "step": 8125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002890949626223868,
+      "loss": 4.2141,
+      "step": 8130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002890815452759745,
+      "loss": 4.1391,
+      "step": 8135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000289068119992165,
+      "loss": 4.1344,
+      "step": 8140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002890546867717244,
+      "loss": 4.1781,
+      "step": 8145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028904124561541947,
+      "loss": 4.2125,
+      "step": 8150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028902779652401723,
+      "loss": 4.1531,
+      "step": 8155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002890143394982852,
+      "loss": 4.1391,
+      "step": 8160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002890008745389914,
+      "loss": 4.0828,
+      "step": 8165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889874016469043,
+      "loss": 4.2391,
+      "step": 8170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889739208227928,
+      "loss": 4.1406,
+      "step": 8175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889604320674261,
+      "loss": 4.2422,
+      "step": 8180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889469353815742,
+      "loss": 4.1531,
+      "step": 8185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889334307660073,
+      "loss": 4.1469,
+      "step": 8190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889199182214961,
+      "loss": 4.05,
+      "step": 8195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002889063977488117,
+      "loss": 4.0938,
+      "step": 8200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028889286934872583,
+      "loss": 4.2562,
+      "step": 8205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028887933302201044,
+      "loss": 4.1484,
+      "step": 8210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002888657887694382,
+      "loss": 4.175,
+      "step": 8215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002888522365917819,
+      "loss": 4.175,
+      "step": 8220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002888386764898151,
+      "loss": 4.125,
+      "step": 8225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028882510846431163,
+      "loss": 4.1703,
+      "step": 8230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888115325160458,
+      "loss": 4.1688,
+      "step": 8235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028879794864579243,
+      "loss": 4.1422,
+      "step": 8240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028878435685432673,
+      "loss": 4.1391,
+      "step": 8245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887707571424244,
+      "loss": 4.2281,
+      "step": 8250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887571495108615,
+      "loss": 4.125,
+      "step": 8255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887435339604147,
+      "loss": 4.2188,
+      "step": 8260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028872991049186104,
+      "loss": 4.1484,
+      "step": 8265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887162791059781,
+      "loss": 4.1828,
+      "step": 8270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887026398035436,
+      "loss": 4.1688,
+      "step": 8275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886889925853361,
+      "loss": 4.0578,
+      "step": 8280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028867533745213433,
+      "loss": 4.1297,
+      "step": 8285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886616744047177,
+      "loss": 4.2094,
+      "step": 8290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288648003443866,
+      "loss": 4.1203,
+      "step": 8295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028863432457035927,
+      "loss": 4.1297,
+      "step": 8300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886206377849783,
+      "loss": 4.1672,
+      "step": 8305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028860694308850416,
+      "loss": 4.1234,
+      "step": 8310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885932404817184,
+      "loss": 4.1312,
+      "step": 8315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288579529965403,
+      "loss": 4.2141,
+      "step": 8320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885658115403404,
+      "loss": 4.1594,
+      "step": 8325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028855208520731367,
+      "loss": 4.1813,
+      "step": 8330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288538350967106,
+      "loss": 4.1969,
+      "step": 8335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885246088205013,
+      "loss": 4.1188,
+      "step": 8340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885108587682837,
+      "loss": 4.1844,
+      "step": 8345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028849710081123813,
+      "loss": 4.0125,
+      "step": 8350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884833349501496,
+      "loss": 4.1797,
+      "step": 8355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884695611858038,
+      "loss": 4.1156,
+      "step": 8360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028845577951898676,
+      "loss": 4.1266,
+      "step": 8365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028844198995048504,
+      "loss": 4.2031,
+      "step": 8370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028842819248108556,
+      "loss": 4.1094,
+      "step": 8375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884143871115758,
+      "loss": 4.1281,
+      "step": 8380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028840057384274363,
+      "loss": 4.1578,
+      "step": 8385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028838675267537733,
+      "loss": 4.1188,
+      "step": 8390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028837292361026564,
+      "loss": 4.1547,
+      "step": 8395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028835908664819786,
+      "loss": 4.1125,
+      "step": 8400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028834524178996367,
+      "loss": 4.1688,
+      "step": 8405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883313890363532,
+      "loss": 4.1344,
+      "step": 8410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288317528388157,
+      "loss": 4.275,
+      "step": 8415
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028830365984616605,
+      "loss": 4.0891,
+      "step": 8420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028828978341117193,
+      "loss": 4.0875,
+      "step": 8425
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882758990839665,
+      "loss": 4.2,
+      "step": 8430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882620068653422,
+      "loss": 4.1625,
+      "step": 8435
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028824810675609177,
+      "loss": 4.0891,
+      "step": 8440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882341987570086,
+      "loss": 4.1484,
+      "step": 8445
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882202828688863,
+      "loss": 4.2281,
+      "step": 8450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882063590925191,
+      "loss": 4.1312,
+      "step": 8455
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028819242742870175,
+      "loss": 4.1031,
+      "step": 8460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881784878782291,
+      "loss": 4.1641,
+      "step": 8465
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881645404418969,
+      "loss": 4.1281,
+      "step": 8470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028815058512050097,
+      "loss": 4.2047,
+      "step": 8475
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028813662191483786,
+      "loss": 4.1859,
+      "step": 8480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028812265082570445,
+      "loss": 4.1531,
+      "step": 8485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028810867185389796,
+      "loss": 4.2078,
+      "step": 8490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028809468500021624,
+      "loss": 4.1781,
+      "step": 8495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880806902654575,
+      "loss": 4.1766,
+      "step": 8500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028806668765042045,
+      "loss": 4.2313,
+      "step": 8505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880526771559043,
+      "loss": 4.2516,
+      "step": 8510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880386587827084,
+      "loss": 4.1797,
+      "step": 8515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288024632531633,
+      "loss": 4.2344,
+      "step": 8520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880105984034784,
+      "loss": 4.1844,
+      "step": 8525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028799655639904576,
+      "loss": 4.1437,
+      "step": 8530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028798250651913625,
+      "loss": 4.1891,
+      "step": 8535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028796844876455177,
+      "loss": 4.1937,
+      "step": 8540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879543831360946,
+      "loss": 4.1766,
+      "step": 8545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879403096345675,
+      "loss": 4.1875,
+      "step": 8550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879262282607736,
+      "loss": 4.1922,
+      "step": 8555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879121390155166,
+      "loss": 4.1625,
+      "step": 8560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878980418996004,
+      "loss": 4.1578,
+      "step": 8565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878839369138297,
+      "loss": 4.1906,
+      "step": 8570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028786982405900935,
+      "loss": 4.2125,
+      "step": 8575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878557033359449,
+      "loss": 4.0938,
+      "step": 8580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878415747454421,
+      "loss": 4.0641,
+      "step": 8585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878274382883074,
+      "loss": 4.2594,
+      "step": 8590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028781329396534744,
+      "loss": 4.2094,
+      "step": 8595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877991417773695,
+      "loss": 4.2078,
+      "step": 8600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877849817251812,
+      "loss": 4.2062,
+      "step": 8605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028777081380959075,
+      "loss": 4.1891,
+      "step": 8610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028775663803140664,
+      "loss": 4.1516,
+      "step": 8615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877424543914379,
+      "loss": 4.1891,
+      "step": 8620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028772826289049394,
+      "loss": 4.1281,
+      "step": 8625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877140635293848,
+      "loss": 4.1109,
+      "step": 8630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028769985630892073,
+      "loss": 4.1844,
+      "step": 8635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876856412299126,
+      "loss": 4.1719,
+      "step": 8640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028767141829317156,
+      "loss": 4.0844,
+      "step": 8645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028765718749950944,
+      "loss": 4.0797,
+      "step": 8650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876429488497383,
+      "loss": 4.1219,
+      "step": 8655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876287023446708,
+      "loss": 4.1063,
+      "step": 8660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028761444798512003,
+      "loss": 4.1297,
+      "step": 8665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876001857718994,
+      "loss": 4.0812,
+      "step": 8670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028758591570582283,
+      "loss": 4.1688,
+      "step": 8675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875716377877048,
+      "loss": 4.1125,
+      "step": 8680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875573520183601,
+      "loss": 4.0969,
+      "step": 8685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875430583986041,
+      "loss": 4.1219,
+      "step": 8690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028752875692925246,
+      "loss": 4.15,
+      "step": 8695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028751444761112134,
+      "loss": 4.1625,
+      "step": 8700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028750013044502744,
+      "loss": 4.1328,
+      "step": 8705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028748580543178786,
+      "loss": 4.2375,
+      "step": 8710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874714725722201,
+      "loss": 4.1516,
+      "step": 8715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028745713186714207,
+      "loss": 4.1719,
+      "step": 8720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874427833173723,
+      "loss": 4.2141,
+      "step": 8725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874284269237296,
+      "loss": 4.1516,
+      "step": 8730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874140626870333,
+      "loss": 4.1156,
+      "step": 8735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002873996906081032,
+      "loss": 4.1937,
+      "step": 8740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002873853106877595,
+      "loss": 4.1578,
+      "step": 8745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028737092292682286,
+      "loss": 4.1469,
+      "step": 8750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028735652732611437,
+      "loss": 4.2656,
+      "step": 8755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002873421238864556,
+      "loss": 4.1188,
+      "step": 8760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028732771260866856,
+      "loss": 4.1594,
+      "step": 8765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002873132934935757,
+      "loss": 4.1359,
+      "step": 8770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028729886654199995,
+      "loss": 4.1844,
+      "step": 8775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002872844317547646,
+      "loss": 4.1781,
+      "step": 8780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028726998913269355,
+      "loss": 4.1813,
+      "step": 8785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002872555386766109,
+      "loss": 4.2687,
+      "step": 8790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028724108038734146,
+      "loss": 4.1844,
+      "step": 8795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028722661426571025,
+      "loss": 4.0891,
+      "step": 8800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002872121403125429,
+      "loss": 4.2687,
+      "step": 8805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002871976585286655,
+      "loss": 4.0859,
+      "step": 8810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002871831689149045,
+      "loss": 4.1344,
+      "step": 8815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028716867147208683,
+      "loss": 4.1359,
+      "step": 8820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002871541662010398,
+      "loss": 4.1734,
+      "step": 8825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028713965310259127,
+      "loss": 4.0734,
+      "step": 8830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002871251321775695,
+      "loss": 4.2562,
+      "step": 8835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002871106034268032,
+      "loss": 4.1219,
+      "step": 8840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028709606685112157,
+      "loss": 4.175,
+      "step": 8845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028708152245135405,
+      "loss": 4.1047,
+      "step": 8850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028706697022833094,
+      "loss": 4.1734,
+      "step": 8855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028705241018288255,
+      "loss": 4.1672,
+      "step": 8860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002870378423158399,
+      "loss": 4.1719,
+      "step": 8865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002870232666280343,
+      "loss": 4.0984,
+      "step": 8870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002870086831202977,
+      "loss": 4.1188,
+      "step": 8875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002869940917934623,
+      "loss": 4.1328,
+      "step": 8880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002869794926483609,
+      "loss": 4.1297,
+      "step": 8885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028696488568582656,
+      "loss": 4.1359,
+      "step": 8890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869502709066931,
+      "loss": 4.1719,
+      "step": 8895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869356483117944,
+      "loss": 4.0531,
+      "step": 8900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028692101790196494,
+      "loss": 4.0969,
+      "step": 8905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028690637967803984,
+      "loss": 4.1937,
+      "step": 8910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028689173364085445,
+      "loss": 4.0812,
+      "step": 8915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028687707979124463,
+      "loss": 4.1969,
+      "step": 8920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868624181300466,
+      "loss": 4.1453,
+      "step": 8925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028684774865809717,
+      "loss": 4.15,
+      "step": 8930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028683307137623355,
+      "loss": 4.1422,
+      "step": 8935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868183862852933,
+      "loss": 4.0703,
+      "step": 8940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028680369338611456,
+      "loss": 4.2188,
+      "step": 8945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867889926795358,
+      "loss": 4.0766,
+      "step": 8950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028677428416639604,
+      "loss": 4.0828,
+      "step": 8955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867595678475347,
+      "loss": 4.1609,
+      "step": 8960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867448437237916,
+      "loss": 4.1984,
+      "step": 8965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028673011179600704,
+      "loss": 4.1344,
+      "step": 8970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028671537206502185,
+      "loss": 4.1328,
+      "step": 8975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028670062453167716,
+      "loss": 4.1562,
+      "step": 8980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028668586919681463,
+      "loss": 4.1,
+      "step": 8985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028667110606127636,
+      "loss": 4.1047,
+      "step": 8990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866563351259048,
+      "loss": 4.2062,
+      "step": 8995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028664155639154316,
+      "loss": 4.1672,
+      "step": 9000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866267698590346,
+      "loss": 4.1969,
+      "step": 9005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028661197552922314,
+      "loss": 4.0812,
+      "step": 9010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286597173402953,
+      "loss": 4.1219,
+      "step": 9015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286582363481069,
+      "loss": 4.1891,
+      "step": 9020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028656754576441635,
+      "loss": 4.1469,
+      "step": 9025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028655272025384065,
+      "loss": 4.1922,
+      "step": 9030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028653788695018806,
+      "loss": 4.1078,
+      "step": 9035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286523045854305,
+      "loss": 4.2219,
+      "step": 9040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028650819696703864,
+      "loss": 4.175,
+      "step": 9045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028649334028923627,
+      "loss": 4.1859,
+      "step": 9050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864784758217458,
+      "loss": 4.0359,
+      "step": 9055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864636035654155,
+      "loss": 4.1625,
+      "step": 9060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028644872352109424,
+      "loss": 4.1188,
+      "step": 9065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864338356896311,
+      "loss": 4.2094,
+      "step": 9070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864189400718758,
+      "loss": 4.0969,
+      "step": 9075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028640403666867843,
+      "loss": 4.1516,
+      "step": 9080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028638912548088947,
+      "loss": 4.15,
+      "step": 9085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028637420650936006,
+      "loss": 4.0625,
+      "step": 9090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863592797549414,
+      "loss": 4.1437,
+      "step": 9095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863443452184856,
+      "loss": 4.1672,
+      "step": 9100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863294029008448,
+      "loss": 4.0938,
+      "step": 9105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863144528028718,
+      "loss": 4.125,
+      "step": 9110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028629949492541977,
+      "loss": 4.1625,
+      "step": 9115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862845292693425,
+      "loss": 4.2344,
+      "step": 9120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028626955583549397,
+      "loss": 4.1125,
+      "step": 9125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862545746247287,
+      "loss": 4.1922,
+      "step": 9130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862395856379017,
+      "loss": 4.1031,
+      "step": 9135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028622458887586844,
+      "loss": 4.1688,
+      "step": 9140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028620958433948473,
+      "loss": 4.1391,
+      "step": 9145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861945720296069,
+      "loss": 4.2,
+      "step": 9150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028617955194709167,
+      "loss": 4.1141,
+      "step": 9155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861645240927963,
+      "loss": 4.0672,
+      "step": 9160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028614948846757837,
+      "loss": 4.1437,
+      "step": 9165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286134445072296,
+      "loss": 4.1484,
+      "step": 9170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861193939078077,
+      "loss": 4.1891,
+      "step": 9175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861043349749724,
+      "loss": 4.1312,
+      "step": 9180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860892682746497,
+      "loss": 4.1328,
+      "step": 9185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860741938076992,
+      "loss": 4.1891,
+      "step": 9190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028605911157498137,
+      "loss": 4.1844,
+      "step": 9195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860440215773569,
+      "loss": 4.1391,
+      "step": 9200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286028923815687,
+      "loss": 4.1391,
+      "step": 9205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028601381829083326,
+      "loss": 4.2062,
+      "step": 9210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859987050036578,
+      "loss": 4.1328,
+      "step": 9215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028598358395502313,
+      "loss": 4.1656,
+      "step": 9220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028596845514579215,
+      "loss": 4.1391,
+      "step": 9225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028595331857682836,
+      "loss": 4.1625,
+      "step": 9230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028593817424899556,
+      "loss": 4.1016,
+      "step": 9235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028592302216315797,
+      "loss": 4.1797,
+      "step": 9240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028590786232018044,
+      "loss": 4.1984,
+      "step": 9245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858926947209281,
+      "loss": 4.1797,
+      "step": 9250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858775193662665,
+      "loss": 4.1094,
+      "step": 9255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028586233625706175,
+      "loss": 4.1547,
+      "step": 9260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858471453941804,
+      "loss": 4.1172,
+      "step": 9265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028583194677848933,
+      "loss": 4.1391,
+      "step": 9270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028581674041085593,
+      "loss": 4.1531,
+      "step": 9275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028580152629214807,
+      "loss": 4.1547,
+      "step": 9280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000285786304423234,
+      "loss": 4.1312,
+      "step": 9285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028577107480498245,
+      "loss": 3.9937,
+      "step": 9290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028575583743826246,
+      "loss": 4.15,
+      "step": 9295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857405923239438,
+      "loss": 4.1578,
+      "step": 9300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028572533946289645,
+      "loss": 4.1703,
+      "step": 9305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857100788559908,
+      "loss": 4.1625,
+      "step": 9310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856948105040979,
+      "loss": 4.1969,
+      "step": 9315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028567953440808907,
+      "loss": 4.0734,
+      "step": 9320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856642505688361,
+      "loss": 4.0891,
+      "step": 9325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028564895898721126,
+      "loss": 4.1813,
+      "step": 9330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856336596640872,
+      "loss": 4.0328,
+      "step": 9335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856183526003371,
+      "loss": 4.2062,
+      "step": 9340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028560303779683456,
+      "loss": 4.1594,
+      "step": 9345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028558771525445355,
+      "loss": 4.1594,
+      "step": 9350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855723849740685,
+      "loss": 4.1781,
+      "step": 9355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028555704695655445,
+      "loss": 4.1,
+      "step": 9360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855417012027866,
+      "loss": 4.1156,
+      "step": 9365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028552634771364073,
+      "loss": 4.225,
+      "step": 9370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855109864899932,
+      "loss": 4.1609,
+      "step": 9375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854956175327205,
+      "loss": 4.15,
+      "step": 9380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854802408426999,
+      "loss": 4.2219,
+      "step": 9385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854648564208088,
+      "loss": 4.1859,
+      "step": 9390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028544946426792533,
+      "loss": 4.15,
+      "step": 9395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854340643849279,
+      "loss": 4.125,
+      "step": 9400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854186567726952,
+      "loss": 4.1344,
+      "step": 9405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028540324143210683,
+      "loss": 4.125,
+      "step": 9410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002853878183640423,
+      "loss": 4.2125,
+      "step": 9415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028537238756938196,
+      "loss": 4.1469,
+      "step": 9420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002853569490490064,
+      "loss": 4.2188,
+      "step": 9425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028534150280379664,
+      "loss": 4.1531,
+      "step": 9430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002853260488346343,
+      "loss": 4.1781,
+      "step": 9435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028531058714240126,
+      "loss": 4.2188,
+      "step": 9440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028529511772797995,
+      "loss": 4.1828,
+      "step": 9445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852796405922532,
+      "loss": 4.2328,
+      "step": 9450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852641557361044,
+      "loss": 4.1688,
+      "step": 9455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852486631604171,
+      "loss": 4.0625,
+      "step": 9460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852331628660755,
+      "loss": 4.0641,
+      "step": 9465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852176548539643,
+      "loss": 4.1547,
+      "step": 9470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002852021391249684,
+      "loss": 4.1766,
+      "step": 9475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028518661567997346,
+      "loss": 4.1469,
+      "step": 9480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002851710845198653,
+      "loss": 3.9688,
+      "step": 9485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028515554564553027,
+      "loss": 4.1422,
+      "step": 9490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002851399990578552,
+      "loss": 4.1906,
+      "step": 9495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002851244447577274,
+      "loss": 3.9891,
+      "step": 9500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002851088827460344,
+      "loss": 4.175,
+      "step": 9505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002850933130236645,
+      "loss": 4.0344,
+      "step": 9510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028507773559150617,
+      "loss": 4.1781,
+      "step": 9515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002850621504504484,
+      "loss": 4.1797,
+      "step": 9520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028504655760138066,
+      "loss": 4.1672,
+      "step": 9525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002850309570451929,
+      "loss": 4.1297,
+      "step": 9530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028501534878277535,
+      "loss": 4.1531,
+      "step": 9535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028499973281501876,
+      "loss": 4.1141,
+      "step": 9540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002849841091428145,
+      "loss": 4.1141,
+      "step": 9545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028496847776705403,
+      "loss": 4.1344,
+      "step": 9550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002849528386886295,
+      "loss": 4.1969,
+      "step": 9555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028493719190843345,
+      "loss": 4.2062,
+      "step": 9560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028492153742735885,
+      "loss": 4.0828,
+      "step": 9565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028490587524629906,
+      "loss": 4.1359,
+      "step": 9570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848902053661479,
+      "loss": 4.1359,
+      "step": 9575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848745277877998,
+      "loss": 4.0875,
+      "step": 9580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848588425121493,
+      "loss": 4.1625,
+      "step": 9585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848431495400917,
+      "loss": 4.1266,
+      "step": 9590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848274488725225,
+      "loss": 4.1031,
+      "step": 9595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028481174051033783,
+      "loss": 4.1188,
+      "step": 9600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028479602445443406,
+      "loss": 4.175,
+      "step": 9605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847803007057082,
+      "loss": 4.1047,
+      "step": 9610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847645692650575,
+      "loss": 4.1641,
+      "step": 9615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847488301333799,
+      "loss": 4.1375,
+      "step": 9620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028473308331157353,
+      "loss": 4.1047,
+      "step": 9625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028471732880053707,
+      "loss": 4.1766,
+      "step": 9630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847015666011696,
+      "loss": 4.2469,
+      "step": 9635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846857967143708,
+      "loss": 4.1188,
+      "step": 9640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028467001914104064,
+      "loss": 4.1109,
+      "step": 9645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846542338820794,
+      "loss": 4.1813,
+      "step": 9650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028463844093838805,
+      "loss": 4.0578,
+      "step": 9655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028462264031086786,
+      "loss": 4.0766,
+      "step": 9660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846068320004206,
+      "loss": 4.1406,
+      "step": 9665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028459101600794844,
+      "loss": 4.1875,
+      "step": 9670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028457519233435403,
+      "loss": 4.1734,
+      "step": 9675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845593609805404,
+      "loss": 4.1578,
+      "step": 9680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845435219474111,
+      "loss": 4.0766,
+      "step": 9685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845276752358699,
+      "loss": 4.1359,
+      "step": 9690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028451182084682133,
+      "loss": 4.1344,
+      "step": 9695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844959587811702,
+      "loss": 4.1484,
+      "step": 9700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028448008903982167,
+      "loss": 4.1078,
+      "step": 9705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844642116236815,
+      "loss": 4.1125,
+      "step": 9710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028444832653365577,
+      "loss": 4.1437,
+      "step": 9715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844324337706511,
+      "loss": 4.2562,
+      "step": 9720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028441653333557444,
+      "loss": 4.0344,
+      "step": 9725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844006252293333,
+      "loss": 4.1047,
+      "step": 9730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028438470945283545,
+      "loss": 4.1531,
+      "step": 9735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028436878600698927,
+      "loss": 4.1391,
+      "step": 9740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843528548927035,
+      "loss": 4.1875,
+      "step": 9745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028433691611088734,
+      "loss": 4.2359,
+      "step": 9750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843209696624504,
+      "loss": 4.1797,
+      "step": 9755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843050155483027,
+      "loss": 4.1906,
+      "step": 9760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028428905376935487,
+      "loss": 4.15,
+      "step": 9765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028427308432651775,
+      "loss": 4.2062,
+      "step": 9770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028425710722070275,
+      "loss": 4.1156,
+      "step": 9775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842411224528217,
+      "loss": 4.1109,
+      "step": 9780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842251300237868,
+      "loss": 4.1484,
+      "step": 9785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028420912993451085,
+      "loss": 4.1703,
+      "step": 9790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841931221859068,
+      "loss": 4.1078,
+      "step": 9795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028417710677888834,
+      "loss": 4.2375,
+      "step": 9800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841610837143695,
+      "loss": 4.1688,
+      "step": 9805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841450529932646,
+      "loss": 4.1156,
+      "step": 9810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841290146164886,
+      "loss": 4.1203,
+      "step": 9815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841129685849568,
+      "loss": 4.0359,
+      "step": 9820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840969148995849,
+      "loss": 4.1281,
+      "step": 9825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028408085356128917,
+      "loss": 4.1609,
+      "step": 9830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840647845709862,
+      "loss": 4.2047,
+      "step": 9835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028404870792959295,
+      "loss": 4.1063,
+      "step": 9840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028403262363802705,
+      "loss": 4.1234,
+      "step": 9845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840165316972064,
+      "loss": 4.0625,
+      "step": 9850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840004321080493,
+      "loss": 4.1562,
+      "step": 9855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028398432487147465,
+      "loss": 4.1875,
+      "step": 9860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839682099884016,
+      "loss": 4.0859,
+      "step": 9865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028395208745974986,
+      "loss": 4.1516,
+      "step": 9870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028393595728643965,
+      "loss": 4.0687,
+      "step": 9875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028391981946939137,
+      "loss": 4.0531,
+      "step": 9880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028390367400952604,
+      "loss": 4.2031,
+      "step": 9885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028388752090776514,
+      "loss": 4.1344,
+      "step": 9890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838713601650305,
+      "loss": 4.2094,
+      "step": 9895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838551917822444,
+      "loss": 4.1703,
+      "step": 9900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028383901576032954,
+      "loss": 4.0812,
+      "step": 9905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838228321002092,
+      "loss": 4.1219,
+      "step": 9910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028380664080280687,
+      "loss": 4.1453,
+      "step": 9915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837904418690467,
+      "loss": 4.1406,
+      "step": 9920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028377423529985304,
+      "loss": 4.1937,
+      "step": 9925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028375802109615087,
+      "loss": 4.1219,
+      "step": 9930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028374179925886553,
+      "loss": 4.1953,
+      "step": 9935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837255697889228,
+      "loss": 4.1578,
+      "step": 9940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028370933268724883,
+      "loss": 4.1031,
+      "step": 9945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028369308795477043,
+      "loss": 4.1297,
+      "step": 9950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028367683559241456,
+      "loss": 4.1531,
+      "step": 9955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028366057560110875,
+      "loss": 4.1828,
+      "step": 9960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000283644307981781,
+      "loss": 4.1547,
+      "step": 9965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028362803273535974,
+      "loss": 4.1953,
+      "step": 9970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028361174986277373,
+      "loss": 4.1172,
+      "step": 9975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835954593649522,
+      "loss": 4.1484,
+      "step": 9980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028357916124282503,
+      "loss": 4.1969,
+      "step": 9985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028356285549732213,
+      "loss": 4.0984,
+      "step": 9990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835465421293742,
+      "loss": 4.1328,
+      "step": 9995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835302211399122,
+      "loss": 4.0781,
+      "step": 10000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835138925298676,
+      "loss": 4.0406,
+      "step": 10005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834975563001723,
+      "loss": 4.1328,
+      "step": 10010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028348121245175853,
+      "loss": 4.2156,
+      "step": 10015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834648609855591,
+      "loss": 4.1359,
+      "step": 10020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834485019025072,
+      "loss": 4.1719,
+      "step": 10025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834321352035364,
+      "loss": 4.0984,
+      "step": 10030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834157608895807,
+      "loss": 4.1531,
+      "step": 10035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028339937896157473,
+      "loss": 4.1859,
+      "step": 10040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833829894204533,
+      "loss": 4.1719,
+      "step": 10045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028336659226715173,
+      "loss": 4.0797,
+      "step": 10050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833501875026059,
+      "loss": 4.1844,
+      "step": 10055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000283333775127752,
+      "loss": 4.0891,
+      "step": 10060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833173551435267,
+      "loss": 4.1109,
+      "step": 10065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028330092755086706,
+      "loss": 4.15,
+      "step": 10070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002832844923507106,
+      "loss": 4.1344,
+      "step": 10075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028326804954399536,
+      "loss": 4.175,
+      "step": 10080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028325159913165964,
+      "loss": 4.1437,
+      "step": 10085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002832351411146423,
+      "loss": 4.1172,
+      "step": 10090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002832186754938826,
+      "loss": 4.0766,
+      "step": 10095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028320220227032015,
+      "loss": 4.1656,
+      "step": 10100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028318572144489523,
+      "loss": 4.2125,
+      "step": 10105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002831692330185483,
+      "loss": 4.0891,
+      "step": 10110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002831527369922204,
+      "loss": 4.1688,
+      "step": 10115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028313623336685294,
+      "loss": 4.1906,
+      "step": 10120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002831197221433878,
+      "loss": 4.1,
+      "step": 10125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002831032033227673,
+      "loss": 4.2156,
+      "step": 10130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002830866769059341,
+      "loss": 4.25,
+      "step": 10135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002830701428938314,
+      "loss": 4.1578,
+      "step": 10140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002830536012874028,
+      "loss": 4.1094,
+      "step": 10145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028303705208759236,
+      "loss": 4.1359,
+      "step": 10150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028302049529534444,
+      "loss": 4.175,
+      "step": 10155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002830039309116041,
+      "loss": 4.2469,
+      "step": 10160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002829873589373165,
+      "loss": 4.0938,
+      "step": 10165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028297077937342757,
+      "loss": 4.1156,
+      "step": 10170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028295419222088334,
+      "loss": 4.2156,
+      "step": 10175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002829375974806306,
+      "loss": 4.1219,
+      "step": 10180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002829209951536163,
+      "loss": 4.15,
+      "step": 10185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000282904385240788,
+      "loss": 4.1437,
+      "step": 10190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002828877677430936,
+      "loss": 4.1578,
+      "step": 10195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028287114266148145,
+      "loss": 4.1844,
+      "step": 10200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028285450999690037,
+      "loss": 4.1359,
+      "step": 10205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828378697502996,
+      "loss": 4.1719,
+      "step": 10210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028282122192262876,
+      "loss": 4.1375,
+      "step": 10215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828045665148379,
+      "loss": 4.15,
+      "step": 10220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028278790352787773,
+      "loss": 4.1984,
+      "step": 10225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282771232962699,
+      "loss": 4.1734,
+      "step": 10230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028275455482025324,
+      "loss": 4.1734,
+      "step": 10235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028273786910149224,
+      "loss": 4.0891,
+      "step": 10240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028272117580736817,
+      "loss": 4.1328,
+      "step": 10245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028270447493883386,
+      "loss": 4.1594,
+      "step": 10250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826877664968423,
+      "loss": 4.1766,
+      "step": 10255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028267105048234705,
+      "loss": 4.1266,
+      "step": 10260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826543268963022,
+      "loss": 4.1359,
+      "step": 10265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028263759573966216,
+      "loss": 4.1484,
+      "step": 10270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826208570133817,
+      "loss": 4.1781,
+      "step": 10275
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028260411071841604,
+      "loss": 4.125,
+      "step": 10280
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282587356855721,
+      "loss": 4.0641,
+      "step": 10285
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825705954262528,
+      "loss": 4.1688,
+      "step": 10290
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825538264309679,
+      "loss": 4.1344,
+      "step": 10295
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028253704987082326,
+      "loss": 4.1328,
+      "step": 10300
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028252026574677644,
+      "loss": 4.2156,
+      "step": 10305
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028250347405978527,
+      "loss": 4.0359,
+      "step": 10310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028248667481080807,
+      "loss": 4.1312,
+      "step": 10315
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824698680008035,
+      "loss": 4.0687,
+      "step": 10320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824530536307307,
+      "loss": 4.0656,
+      "step": 10325
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028243623170154944,
+      "loss": 4.1813,
+      "step": 10330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824194022142197,
+      "loss": 4.0828,
+      "step": 10335
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028240256516970174,
+      "loss": 4.125,
+      "step": 10340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028238572056895673,
+      "loss": 4.125,
+      "step": 10345
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028236886841294585,
+      "loss": 4.0625,
+      "step": 10350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028235200870263083,
+      "loss": 4.1016,
+      "step": 10355
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028233514143897396,
+      "loss": 4.0891,
+      "step": 10360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028231826662293775,
+      "loss": 4.1719,
+      "step": 10365
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823013842554853,
+      "loss": 4.1625,
+      "step": 10370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822844943375801,
+      "loss": 4.0969,
+      "step": 10375
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282267596870186,
+      "loss": 4.1891,
+      "step": 10380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822506918542674,
+      "loss": 4.1719,
+      "step": 10385
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282233779290789,
+      "loss": 4.1203,
+      "step": 10390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822168591807161,
+      "loss": 4.0938,
+      "step": 10395
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821999315250143,
+      "loss": 4.1188,
+      "step": 10400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821829963246496,
+      "loss": 4.0891,
+      "step": 10405
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028216605358058857,
+      "loss": 4.1594,
+      "step": 10410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028214910329379814,
+      "loss": 4.1406,
+      "step": 10415
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821321454652455,
+      "loss": 4.1203,
+      "step": 10420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821151800958987,
+      "loss": 4.0766,
+      "step": 10425
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820982071867258,
+      "loss": 4.1016,
+      "step": 10430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820812267386954,
+      "loss": 4.0844,
+      "step": 10435
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028206423875277667,
+      "loss": 4.1531,
+      "step": 10440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820472432299391,
+      "loss": 4.1375,
+      "step": 10445
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820302401711526,
+      "loss": 4.2,
+      "step": 10450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028201322957738755,
+      "loss": 4.1437,
+      "step": 10455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819962114496148,
+      "loss": 4.0781,
+      "step": 10460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819791857888054,
+      "loss": 4.1219,
+      "step": 10465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819621525959312,
+      "loss": 4.1141,
+      "step": 10470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819451118719642,
+      "loss": 4.1266,
+      "step": 10475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028192806361787695,
+      "loss": 4.1234,
+      "step": 10480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028191100783464236,
+      "loss": 4.1578,
+      "step": 10485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028189394452323384,
+      "loss": 4.1609,
+      "step": 10490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818768736846251,
+      "loss": 4.1937,
+      "step": 10495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818597953197905,
+      "loss": 4.1297,
+      "step": 10500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028184270942970463,
+      "loss": 4.1422,
+      "step": 10505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028182561601534266,
+      "loss": 4.1406,
+      "step": 10510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028180851507768003,
+      "loss": 4.1547,
+      "step": 10515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817914066176927,
+      "loss": 4.1094,
+      "step": 10520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817742906363571,
+      "loss": 4.0797,
+      "step": 10525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028175716713465,
+      "loss": 4.1437,
+      "step": 10530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028174003611354867,
+      "loss": 4.175,
+      "step": 10535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028172289757403076,
+      "loss": 4.1719,
+      "step": 10540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817057515170743,
+      "loss": 4.1359,
+      "step": 10545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028168859794365796,
+      "loss": 4.125,
+      "step": 10550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028167143685476064,
+      "loss": 4.1,
+      "step": 10555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028165426825136166,
+      "loss": 4.1719,
+      "step": 10560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028163709213444087,
+      "loss": 4.0938,
+      "step": 10565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816199085049786,
+      "loss": 4.1625,
+      "step": 10570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816027173639554,
+      "loss": 4.1766,
+      "step": 10575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028158551871235236,
+      "loss": 4.1813,
+      "step": 10580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028156831255115114,
+      "loss": 4.1422,
+      "step": 10585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815510988813336,
+      "loss": 4.1437,
+      "step": 10590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815338777038821,
+      "loss": 4.0766,
+      "step": 10595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815166490197796,
+      "loss": 4.1469,
+      "step": 10600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028149941283000916,
+      "loss": 4.1266,
+      "step": 10605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028148216913555463,
+      "loss": 4.0844,
+      "step": 10610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028146491793739996,
+      "loss": 4.0797,
+      "step": 10615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814476592365297,
+      "loss": 4.0359,
+      "step": 10620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814303930339289,
+      "loss": 4.1328,
+      "step": 10625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028141311933058285,
+      "loss": 4.1594,
+      "step": 10630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813958381274774,
+      "loss": 4.1594,
+      "step": 10635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028137854942559876,
+      "loss": 4.1422,
+      "step": 10640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028136125322593365,
+      "loss": 4.1047,
+      "step": 10645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028134394952946914,
+      "loss": 4.1375,
+      "step": 10650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813266383371927,
+      "loss": 4.1453,
+      "step": 10655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813093196500924,
+      "loss": 4.2313,
+      "step": 10660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812919934691565,
+      "loss": 4.2469,
+      "step": 10665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812746597953739,
+      "loss": 4.1781,
+      "step": 10670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812573186297338,
+      "loss": 4.1531,
+      "step": 10675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028123996997322584,
+      "loss": 4.1156,
+      "step": 10680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028122261382684013,
+      "loss": 4.0625,
+      "step": 10685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028120525019156715,
+      "loss": 4.1688,
+      "step": 10690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811878790683979,
+      "loss": 4.0578,
+      "step": 10695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811705004583237,
+      "loss": 4.1406,
+      "step": 10700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811531143623364,
+      "loss": 4.0906,
+      "step": 10705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811357207814282,
+      "loss": 4.0922,
+      "step": 10710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028111831971659176,
+      "loss": 4.1484,
+      "step": 10715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811009111688202,
+      "loss": 4.1344,
+      "step": 10720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002810834951391069,
+      "loss": 4.1188,
+      "step": 10725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028106607162844597,
+      "loss": 4.1969,
+      "step": 10730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002810486406378316,
+      "loss": 4.1688,
+      "step": 10735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002810312021682587,
+      "loss": 4.1734,
+      "step": 10740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002810137562207224,
+      "loss": 4.1312,
+      "step": 10745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028099630279621843,
+      "loss": 4.2156,
+      "step": 10750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002809788418957428,
+      "loss": 4.1594,
+      "step": 10755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028096137352029204,
+      "loss": 4.0797,
+      "step": 10760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000280943897670863,
+      "loss": 4.0594,
+      "step": 10765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028092641434845314,
+      "loss": 4.2125,
+      "step": 10770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002809089235540602,
+      "loss": 4.1625,
+      "step": 10775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808914252886823,
+      "loss": 4.1469,
+      "step": 10780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808739195533181,
+      "loss": 4.1203,
+      "step": 10785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808564063489667,
+      "loss": 4.0625,
+      "step": 10790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808388856766276,
+      "loss": 4.1547,
+      "step": 10795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808213575373006,
+      "loss": 4.1219,
+      "step": 10800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002808038219319861,
+      "loss": 4.0766,
+      "step": 10805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002807862788616849,
+      "loss": 4.1891,
+      "step": 10810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002807687283273981,
+      "loss": 4.1047,
+      "step": 10815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028075117033012737,
+      "loss": 4.1109,
+      "step": 10820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028073360487087477,
+      "loss": 4.2,
+      "step": 10825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002807160319506427,
+      "loss": 4.1937,
+      "step": 10830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000280698451570434,
+      "loss": 4.1656,
+      "step": 10835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002806808637312521,
+      "loss": 4.1469,
+      "step": 10840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002806632684341007,
+      "loss": 4.1594,
+      "step": 10845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028064566567998395,
+      "loss": 4.0812,
+      "step": 10850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028062805546990644,
+      "loss": 4.2062,
+      "step": 10855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002806104378048732,
+      "loss": 4.1266,
+      "step": 10860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028059281268588964,
+      "loss": 4.25,
+      "step": 10865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805751801139617,
+      "loss": 4.1,
+      "step": 10870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805575400900956,
+      "loss": 4.1188,
+      "step": 10875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805398926152981,
+      "loss": 4.1141,
+      "step": 10880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028052223769057627,
+      "loss": 4.0656,
+      "step": 10885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028050457531693786,
+      "loss": 4.1859,
+      "step": 10890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804869054953906,
+      "loss": 4.1656,
+      "step": 10895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028046922822694314,
+      "loss": 4.0641,
+      "step": 10900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804515435126042,
+      "loss": 4.0844,
+      "step": 10905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804338513533831,
+      "loss": 4.1734,
+      "step": 10910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804161517502895,
+      "loss": 4.1766,
+      "step": 10915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028039844470433353,
+      "loss": 4.0906,
+      "step": 10920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803807302165257,
+      "loss": 4.2719,
+      "step": 10925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280363008287877,
+      "loss": 4.1297,
+      "step": 10930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803452789193989,
+      "loss": 4.1562,
+      "step": 10935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803275421121031,
+      "loss": 4.1234,
+      "step": 10940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803097978670019,
+      "loss": 4.1781,
+      "step": 10945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280292046185108,
+      "loss": 4.1453,
+      "step": 10950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802742870674344,
+      "loss": 4.1125,
+      "step": 10955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802565205149947,
+      "loss": 4.0625,
+      "step": 10960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802387465288028,
+      "loss": 4.1609,
+      "step": 10965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280220965109873,
+      "loss": 4.0891,
+      "step": 10970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802031762592202,
+      "loss": 4.1063,
+      "step": 10975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801853799778596,
+      "loss": 4.0719,
+      "step": 10980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801675762668067,
+      "loss": 4.2219,
+      "step": 10985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028014976512707773,
+      "loss": 4.0953,
+      "step": 10990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028013194655968906,
+      "loss": 4.0594,
+      "step": 10995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028011412056565765,
+      "loss": 4.1297,
+      "step": 11000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800962871460008,
+      "loss": 4.1141,
+      "step": 11005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800784463017363,
+      "loss": 4.1109,
+      "step": 11010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028006059803388234,
+      "loss": 4.1844,
+      "step": 11015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028004274234345747,
+      "loss": 4.1641,
+      "step": 11020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800248792314808,
+      "loss": 4.1437,
+      "step": 11025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028000700869897164,
+      "loss": 4.1797,
+      "step": 11030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027998913074694996,
+      "loss": 4.0906,
+      "step": 11035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799712453764361,
+      "loss": 4.1391,
+      "step": 11040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799533525884506,
+      "loss": 4.0469,
+      "step": 11045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027993545238401483,
+      "loss": 4.0828,
+      "step": 11050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799175447641502,
+      "loss": 4.1266,
+      "step": 11055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027989962972987877,
+      "loss": 4.1219,
+      "step": 11060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002798817072822229,
+      "loss": 4.1578,
+      "step": 11065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027986377742220554,
+      "loss": 4.1094,
+      "step": 11070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002798458401508498,
+      "loss": 4.0812,
+      "step": 11075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027982789546917945,
+      "loss": 4.1781,
+      "step": 11080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027980994337821854,
+      "loss": 4.1688,
+      "step": 11085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027979198387899167,
+      "loss": 4.0531,
+      "step": 11090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027977401697252375,
+      "loss": 4.1516,
+      "step": 11095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797560426598401,
+      "loss": 4.0719,
+      "step": 11100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797380609419666,
+      "loss": 4.1469,
+      "step": 11105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027972007181992946,
+      "loss": 4.1047,
+      "step": 11110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797020752947553,
+      "loss": 4.1781,
+      "step": 11115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027968407136747116,
+      "loss": 4.0109,
+      "step": 11120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796660600391045,
+      "loss": 4.0781,
+      "step": 11125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796480413106834,
+      "loss": 4.1781,
+      "step": 11130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027963001518323597,
+      "loss": 4.0641,
+      "step": 11135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796119816577911,
+      "loss": 4.0563,
+      "step": 11140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027959394073537796,
+      "loss": 4.0922,
+      "step": 11145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279575892417026,
+      "loss": 4.1516,
+      "step": 11150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027955783670376546,
+      "loss": 4.1266,
+      "step": 11155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795397735966266,
+      "loss": 4.1688,
+      "step": 11160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795217030966404,
+      "loss": 4.1047,
+      "step": 11165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795036252048381,
+      "loss": 4.15,
+      "step": 11170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027948553992225144,
+      "loss": 4.0641,
+      "step": 11175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027946744724991244,
+      "loss": 4.15,
+      "step": 11180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027944934718885376,
+      "loss": 4.1422,
+      "step": 11185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027943123974010837,
+      "loss": 4.0766,
+      "step": 11190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027941312490470957,
+      "loss": 4.2281,
+      "step": 11195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027939500268369126,
+      "loss": 4.0797,
+      "step": 11200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793768730780877,
+      "loss": 4.0891,
+      "step": 11205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793587360889335,
+      "loss": 4.1437,
+      "step": 11210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793405917172637,
+      "loss": 4.125,
+      "step": 11215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793224399641139,
+      "loss": 4.1766,
+      "step": 11220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027930428083051994,
+      "loss": 4.1875,
+      "step": 11225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027928611431751813,
+      "loss": 4.0703,
+      "step": 11230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792679404261454,
+      "loss": 4.15,
+      "step": 11235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027924975915743876,
+      "loss": 4.1531,
+      "step": 11240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792315705124359,
+      "loss": 4.0609,
+      "step": 11245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027921337449217484,
+      "loss": 4.1312,
+      "step": 11250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027919517109769405,
+      "loss": 4.1188,
+      "step": 11255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027917696033003236,
+      "loss": 4.0016,
+      "step": 11260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791587421902291,
+      "loss": 4.1344,
+      "step": 11265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791405166793239,
+      "loss": 4.1031,
+      "step": 11270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027912228379835697,
+      "loss": 4.1156,
+      "step": 11275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791040435483688,
+      "loss": 4.1016,
+      "step": 11280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027908579593040046,
+      "loss": 4.0422,
+      "step": 11285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027906754094549324,
+      "loss": 4.125,
+      "step": 11290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027904927859468906,
+      "loss": 4.1047,
+      "step": 11295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027903100887903006,
+      "loss": 4.0781,
+      "step": 11300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790127317995589,
+      "loss": 4.1094,
+      "step": 11305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789944473573188,
+      "loss": 4.1422,
+      "step": 11310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278976155553353,
+      "loss": 4.0828,
+      "step": 11315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027895785638870564,
+      "loss": 4.1469,
+      "step": 11320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789395498644209,
+      "loss": 4.0406,
+      "step": 11325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789212359815436,
+      "loss": 4.1141,
+      "step": 11330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278902914741119,
+      "loss": 4.1375,
+      "step": 11335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788845861441925,
+      "loss": 4.0578,
+      "step": 11340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788662501918103,
+      "loss": 4.0625,
+      "step": 11345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788479068850188,
+      "loss": 4.1453,
+      "step": 11350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027882955622486475,
+      "loss": 4.0797,
+      "step": 11355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027881119821239546,
+      "loss": 4.1828,
+      "step": 11360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787928328486587,
+      "loss": 4.2594,
+      "step": 11365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027877446013470255,
+      "loss": 4.1219,
+      "step": 11370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787560800715755,
+      "loss": 4.1047,
+      "step": 11375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027873769266032654,
+      "loss": 4.0703,
+      "step": 11380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027871929790200507,
+      "loss": 4.1172,
+      "step": 11385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787008957976608,
+      "loss": 4.0625,
+      "step": 11390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278682486348344,
+      "loss": 4.1156,
+      "step": 11395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002786640695551053,
+      "loss": 4.0062,
+      "step": 11400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002786456454189957,
+      "loss": 4.1266,
+      "step": 11405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002786272139410667,
+      "loss": 4.0672,
+      "step": 11410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002786087751223702,
+      "loss": 4.15,
+      "step": 11415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027859032896395845,
+      "loss": 4.0859,
+      "step": 11420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002785718754668843,
+      "loss": 4.0641,
+      "step": 11425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002785534146322007,
+      "loss": 4.1,
+      "step": 11430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027853494646096135,
+      "loss": 4.1906,
+      "step": 11435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027851647095422026,
+      "loss": 4.1422,
+      "step": 11440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002784979881130317,
+      "loss": 4.1094,
+      "step": 11445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002784794979384506,
+      "loss": 4.1656,
+      "step": 11450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002784610004315321,
+      "loss": 4.1656,
+      "step": 11455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027844249559333195,
+      "loss": 4.0812,
+      "step": 11460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027842398342490615,
+      "loss": 4.2375,
+      "step": 11465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027840546392731124,
+      "loss": 4.0938,
+      "step": 11470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027838693710160404,
+      "loss": 4.0703,
+      "step": 11475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278368402948842,
+      "loss": 4.1437,
+      "step": 11480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002783498614700828,
+      "loss": 4.0484,
+      "step": 11485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002783313126663847,
+      "loss": 4.0344,
+      "step": 11490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002783127565388061,
+      "loss": 3.9562,
+      "step": 11495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002782941930884061,
+      "loss": 4.1141,
+      "step": 11500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027827562231624414,
+      "loss": 4.1312,
+      "step": 11505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027825704422338,
+      "loss": 4.0234,
+      "step": 11510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027823845881087405,
+      "loss": 4.075,
+      "step": 11515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027821986607978684,
+      "loss": 4.1156,
+      "step": 11520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027820126603117945,
+      "loss": 4.1203,
+      "step": 11525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781826586661135,
+      "loss": 4.0516,
+      "step": 11530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781640439856508,
+      "loss": 4.0781,
+      "step": 11535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027814542199085376,
+      "loss": 4.1562,
+      "step": 11540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781267926827851,
+      "loss": 4.2094,
+      "step": 11545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027810815606250797,
+      "loss": 4.1625,
+      "step": 11550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027808951213108605,
+      "loss": 4.1125,
+      "step": 11555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027807086088958336,
+      "loss": 4.05,
+      "step": 11560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780522023390642,
+      "loss": 4.1344,
+      "step": 11565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780335364805935,
+      "loss": 4.0969,
+      "step": 11570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027801486331523656,
+      "loss": 4.0812,
+      "step": 11575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277996182844059,
+      "loss": 4.1141,
+      "step": 11580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779774950681269,
+      "loss": 4.0938,
+      "step": 11585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027795879998850683,
+      "loss": 4.075,
+      "step": 11590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779400976062657,
+      "loss": 4.1,
+      "step": 11595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027792138792247077,
+      "loss": 4.1328,
+      "step": 11600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027790267093818994,
+      "loss": 4.1141,
+      "step": 11605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778839466544914,
+      "loss": 4.1141,
+      "step": 11610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027786521507244355,
+      "loss": 4.1203,
+      "step": 11615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027784647619311564,
+      "loss": 4.1281,
+      "step": 11620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778277300175769,
+      "loss": 4.1188,
+      "step": 11625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778089765468974,
+      "loss": 4.1188,
+      "step": 11630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777902157821472,
+      "loss": 4.0469,
+      "step": 11635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277771447724397,
+      "loss": 4.1609,
+      "step": 11640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027775267237471795,
+      "loss": 4.1531,
+      "step": 11645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777338897341816,
+      "loss": 4.1906,
+      "step": 11650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027771509980385983,
+      "loss": 4.1469,
+      "step": 11655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277696302584825,
+      "loss": 4.0734,
+      "step": 11660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776774980781498,
+      "loss": 4.1328,
+      "step": 11665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776586862849075,
+      "loss": 4.1016,
+      "step": 11670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027763986720617166,
+      "loss": 4.0797,
+      "step": 11675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027762104084301623,
+      "loss": 4.1906,
+      "step": 11680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027760220719651575,
+      "loss": 4.1063,
+      "step": 11685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027758336626774495,
+      "loss": 4.1531,
+      "step": 11690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002775645180577791,
+      "loss": 4.1422,
+      "step": 11695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027754566256769393,
+      "loss": 4.1312,
+      "step": 11700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027752679979856543,
+      "loss": 4.0469,
+      "step": 11705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002775079297514702,
+      "loss": 4.1047,
+      "step": 11710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774890524274851,
+      "loss": 4.0469,
+      "step": 11715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774701678276875,
+      "loss": 4.0906,
+      "step": 11720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774512759531551,
+      "loss": 4.0938,
+      "step": 11725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027743237680496606,
+      "loss": 4.0828,
+      "step": 11730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027741347038419894,
+      "loss": 4.075,
+      "step": 11735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027739455669193277,
+      "loss": 4.025,
+      "step": 11740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027737563572924703,
+      "loss": 4.1,
+      "step": 11745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027735670749722143,
+      "loss": 4.1344,
+      "step": 11750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773377719969362,
+      "loss": 4.1031,
+      "step": 11755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027731882922947207,
+      "loss": 4.05,
+      "step": 11760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772998791959101,
+      "loss": 4.0734,
+      "step": 11765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027728092189733167,
+      "loss": 4.0938,
+      "step": 11770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772619573348188,
+      "loss": 4.0797,
+      "step": 11775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772429855094537,
+      "loss": 4.1547,
+      "step": 11780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027722400642231907,
+      "loss": 4.0797,
+      "step": 11785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772050200744982,
+      "loss": 4.1531,
+      "step": 11790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027718602646707453,
+      "loss": 4.1766,
+      "step": 11795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002771670256011321,
+      "loss": 4.1531,
+      "step": 11800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027714801747775515,
+      "loss": 4.1063,
+      "step": 11805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027712900209802866,
+      "loss": 4.1328,
+      "step": 11810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002771099794630377,
+      "loss": 4.1562,
+      "step": 11815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027709094957386796,
+      "loss": 4.2062,
+      "step": 11820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002770719124316054,
+      "loss": 4.0625,
+      "step": 11825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002770528680373366,
+      "loss": 4.1109,
+      "step": 11830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027703381639214837,
+      "loss": 4.1266,
+      "step": 11835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002770147574971279,
+      "loss": 4.1437,
+      "step": 11840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276995691353363,
+      "loss": 4.0984,
+      "step": 11845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769766179619417,
+      "loss": 4.1375,
+      "step": 11850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769575373239526,
+      "loss": 4.0,
+      "step": 11855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769384494404846,
+      "loss": 4.0687,
+      "step": 11860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276919354312627,
+      "loss": 4.2156,
+      "step": 11865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027690025194146965,
+      "loss": 4.0812,
+      "step": 11870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768811423281026,
+      "loss": 4.0812,
+      "step": 11875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027686202547361655,
+      "loss": 4.125,
+      "step": 11880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768429013791025,
+      "loss": 4.0984,
+      "step": 11885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027682377004565174,
+      "loss": 4.1703,
+      "step": 11890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768046314743562,
+      "loss": 4.0891,
+      "step": 11895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027678548566630815,
+      "loss": 4.0531,
+      "step": 11900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027676633262260016,
+      "loss": 4.1047,
+      "step": 11905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767471723443254,
+      "loss": 4.1328,
+      "step": 11910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767280048325772,
+      "loss": 4.1234,
+      "step": 11915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767088300884495,
+      "loss": 4.0531,
+      "step": 11920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766896481130367,
+      "loss": 4.0672,
+      "step": 11925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027667045890743345,
+      "loss": 4.0812,
+      "step": 11930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766512624727349,
+      "loss": 4.1047,
+      "step": 11935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766320588100365,
+      "loss": 4.15,
+      "step": 11940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766128479204343,
+      "loss": 4.1437,
+      "step": 11945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765936298050247,
+      "loss": 4.1203,
+      "step": 11950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765744044649044,
+      "loss": 4.1016,
+      "step": 11955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027655517190117065,
+      "loss": 4.1234,
+      "step": 11960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027653593211492096,
+      "loss": 4.075,
+      "step": 11965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765166851072534,
+      "loss": 4.0375,
+      "step": 11970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027649743087926646,
+      "loss": 4.0266,
+      "step": 11975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027647816943205896,
+      "loss": 4.1109,
+      "step": 11980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027645890076673,
+      "loss": 4.125,
+      "step": 11985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002764396248843795,
+      "loss": 3.9594,
+      "step": 11990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002764203417861073,
+      "loss": 4.1406,
+      "step": 11995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027640105147301407,
+      "loss": 4.0781,
+      "step": 12000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763817539462006,
+      "loss": 4.0641,
+      "step": 12005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763624492067682,
+      "loss": 4.0984,
+      "step": 12010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027634313725581866,
+      "loss": 4.1719,
+      "step": 12015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763238180944541,
+      "loss": 4.0609,
+      "step": 12020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276304491723777,
+      "loss": 4.1281,
+      "step": 12025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002762851581448904,
+      "loss": 4.1437,
+      "step": 12030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027626581735889756,
+      "loss": 4.2031,
+      "step": 12035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027624646936690245,
+      "loss": 4.1437,
+      "step": 12040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002762271141700091,
+      "loss": 4.0969,
+      "step": 12045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027620775176932213,
+      "loss": 4.1375,
+      "step": 12050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002761883821659466,
+      "loss": 4.0984,
+      "step": 12055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002761690053609879,
+      "loss": 4.1359,
+      "step": 12060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002761496213555519,
+      "loss": 4.0484,
+      "step": 12065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027613023015074484,
+      "loss": 4.1109,
+      "step": 12070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027611083174767343,
+      "loss": 4.0266,
+      "step": 12075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002760914261474446,
+      "loss": 4.0359,
+      "step": 12080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002760720133511659,
+      "loss": 4.0578,
+      "step": 12085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027605259335994526,
+      "loss": 4.1437,
+      "step": 12090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276033166174891,
+      "loss": 4.0484,
+      "step": 12095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002760137317971117,
+      "loss": 4.0875,
+      "step": 12100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002759942902277166,
+      "loss": 4.1516,
+      "step": 12105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027597484146781514,
+      "loss": 4.1719,
+      "step": 12110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002759553855185174,
+      "loss": 4.1547,
+      "step": 12115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027593592238093356,
+      "loss": 4.1109,
+      "step": 12120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002759164520561745,
+      "loss": 4.075,
+      "step": 12125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002758969745453514,
+      "loss": 4.225,
+      "step": 12130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027587748984957583,
+      "loss": 4.1063,
+      "step": 12135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027585799796995967,
+      "loss": 4.15,
+      "step": 12140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002758384989076155,
+      "loss": 4.0438,
+      "step": 12145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000275818992663656,
+      "loss": 4.0891,
+      "step": 12150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002757994792391945,
+      "loss": 4.1391,
+      "step": 12155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027577995863534453,
+      "loss": 4.1234,
+      "step": 12160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027576043085322024,
+      "loss": 4.0687,
+      "step": 12165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027574089589393593,
+      "loss": 4.1312,
+      "step": 12170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002757213537586066,
+      "loss": 4.0438,
+      "step": 12175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002757018044483475,
+      "loss": 4.0953,
+      "step": 12180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027568224796427424,
+      "loss": 4.1078,
+      "step": 12185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275662684307503,
+      "loss": 4.15,
+      "step": 12190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027564311347915023,
+      "loss": 4.1188,
+      "step": 12195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002756235354803328,
+      "loss": 4.0172,
+      "step": 12200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002756039503121681,
+      "loss": 4.1047,
+      "step": 12205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027558435797577386,
+      "loss": 4.0953,
+      "step": 12210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002755647584722682,
+      "loss": 4.1031,
+      "step": 12215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027554515180276955,
+      "loss": 4.1516,
+      "step": 12220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275525537968397,
+      "loss": 4.1344,
+      "step": 12225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027550591697026995,
+      "loss": 4.1813,
+      "step": 12230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275486288809508,
+      "loss": 4.1375,
+      "step": 12235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027546665348723154,
+      "loss": 4.225,
+      "step": 12240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275447011004561,
+      "loss": 4.0656,
+      "step": 12245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754273613626174,
+      "loss": 4.0594,
+      "step": 12250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027540770456252223,
+      "loss": 4.0469,
+      "step": 12255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002753880406053973,
+      "loss": 4.1266,
+      "step": 12260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002753683694923647,
+      "loss": 4.1937,
+      "step": 12265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027534869122454724,
+      "loss": 4.0703,
+      "step": 12270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027532900580306776,
+      "loss": 4.0391,
+      "step": 12275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002753093132290499,
+      "loss": 4.1531,
+      "step": 12280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027528961350361744,
+      "loss": 4.1328,
+      "step": 12285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002752699066278946,
+      "loss": 4.0391,
+      "step": 12290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027525019260300613,
+      "loss": 4.1813,
+      "step": 12295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027523047143007706,
+      "loss": 4.1656,
+      "step": 12300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002752107431102329,
+      "loss": 4.1172,
+      "step": 12305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751910076445995,
+      "loss": 4.0641,
+      "step": 12310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751712650343033,
+      "loss": 4.0609,
+      "step": 12315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027515151528047086,
+      "loss": 4.0328,
+      "step": 12320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751317583842294,
+      "loss": 4.1109,
+      "step": 12325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027511199434670635,
+      "loss": 4.1141,
+      "step": 12330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750922231690298,
+      "loss": 4.0344,
+      "step": 12335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750724448523279,
+      "loss": 4.1063,
+      "step": 12340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750526593977296,
+      "loss": 4.1375,
+      "step": 12345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750328668063638,
+      "loss": 4.1031,
+      "step": 12350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027501306707936035,
+      "loss": 4.1437,
+      "step": 12355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749932602178491,
+      "loss": 4.1328,
+      "step": 12360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027497344622296035,
+      "loss": 4.1328,
+      "step": 12365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000274953625095825,
+      "loss": 4.125,
+      "step": 12370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749337968375742,
+      "loss": 4.0062,
+      "step": 12375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749139614493396,
+      "loss": 4.0312,
+      "step": 12380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027489411893225307,
+      "loss": 4.1047,
+      "step": 12385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748742692874472,
+      "loss": 4.0875,
+      "step": 12390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748544125160547,
+      "loss": 4.0891,
+      "step": 12395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027483454861920877,
+      "loss": 4.1656,
+      "step": 12400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748146775980432,
+      "loss": 4.0656,
+      "step": 12405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747947994536918,
+      "loss": 4.1969,
+      "step": 12410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747749141872893,
+      "loss": 4.1297,
+      "step": 12415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747550217999703,
+      "loss": 4.1719,
+      "step": 12420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747351222928702,
+      "loss": 4.1031,
+      "step": 12425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747152156671246,
+      "loss": 4.0953,
+      "step": 12430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746953019238696,
+      "loss": 4.0828,
+      "step": 12435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027467538106424173,
+      "loss": 4.1797,
+      "step": 12440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746554530893778,
+      "loss": 4.0406,
+      "step": 12445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027463551800041515,
+      "loss": 4.1437,
+      "step": 12450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027461557579849143,
+      "loss": 4.1016,
+      "step": 12455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027459562648474476,
+      "loss": 4.1063,
+      "step": 12460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745756700603137,
+      "loss": 4.1312,
+      "step": 12465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027455570652633706,
+      "loss": 4.0391,
+      "step": 12470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027453573588395426,
+      "loss": 4.1188,
+      "step": 12475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027451575813430503,
+      "loss": 4.1375,
+      "step": 12480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744957732785294,
+      "loss": 4.0797,
+      "step": 12485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000274475781317768,
+      "loss": 4.1031,
+      "step": 12490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027445578225316177,
+      "loss": 4.1437,
+      "step": 12495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000274435776085852,
+      "loss": 4.1656,
+      "step": 12500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744157628169805,
+      "loss": 4.0969,
+      "step": 12505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743957424476894,
+      "loss": 4.1297,
+      "step": 12510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027437571497912125,
+      "loss": 4.1031,
+      "step": 12515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027435568041241906,
+      "loss": 4.1328,
+      "step": 12520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743356387487262,
+      "loss": 4.1469,
+      "step": 12525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027431558998918643,
+      "loss": 4.125,
+      "step": 12530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002742955341349439,
+      "loss": 4.1422,
+      "step": 12535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027427547118714327,
+      "loss": 4.0984,
+      "step": 12540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027425540114692947,
+      "loss": 4.1203,
+      "step": 12545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027423532401544794,
+      "loss": 4.1594,
+      "step": 12550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002742152397938445,
+      "loss": 4.1109,
+      "step": 12555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741951484832653,
+      "loss": 4.0656,
+      "step": 12560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027417505008485703,
+      "loss": 4.0656,
+      "step": 12565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027415494459976666,
+      "loss": 4.1531,
+      "step": 12570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741348320291416,
+      "loss": 4.0359,
+      "step": 12575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741147123741297,
+      "loss": 4.1406,
+      "step": 12580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740945856358792,
+      "loss": 4.1688,
+      "step": 12585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027407445181553867,
+      "loss": 4.1188,
+      "step": 12590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740543109142572,
+      "loss": 4.0891,
+      "step": 12595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740341629331843,
+      "loss": 4.1156,
+      "step": 12600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027401400787346967,
+      "loss": 4.0828,
+      "step": 12605
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027399384573626367,
+      "loss": 4.1328,
+      "step": 12610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027397367652271697,
+      "loss": 4.0672,
+      "step": 12615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739535002339805,
+      "loss": 4.1047,
+      "step": 12620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739333168712059,
+      "loss": 4.1516,
+      "step": 12625
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027391312643554485,
+      "loss": 4.1234,
+      "step": 12630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027389292892814976,
+      "loss": 4.1156,
+      "step": 12635
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027387272435017325,
+      "loss": 4.1125,
+      "step": 12640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002738525127027684,
+      "loss": 4.25,
+      "step": 12645
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002738322939870887,
+      "loss": 4.0984,
+      "step": 12650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027381206820428804,
+      "loss": 4.1031,
+      "step": 12655
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002737918353555207,
+      "loss": 4.1766,
+      "step": 12660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002737715954419414,
+      "loss": 4.0875,
+      "step": 12665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002737513484647051,
+      "loss": 4.0875,
+      "step": 12670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027373109442496745,
+      "loss": 4.1156,
+      "step": 12675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027371083332388427,
+      "loss": 4.1344,
+      "step": 12680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002736905651626119,
+      "loss": 4.0484,
+      "step": 12685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002736702899423071,
+      "loss": 4.0875,
+      "step": 12690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027365000766412683,
+      "loss": 4.2,
+      "step": 12695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027362971832922873,
+      "loss": 4.1203,
+      "step": 12700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027360942193877067,
+      "loss": 4.125,
+      "step": 12705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027358911849391094,
+      "loss": 4.1031,
+      "step": 12710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027356880799580836,
+      "loss": 4.1625,
+      "step": 12715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002735484904456219,
+      "loss": 4.1172,
+      "step": 12720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002735281658445112,
+      "loss": 4.0984,
+      "step": 12725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027350783419363613,
+      "loss": 4.1125,
+      "step": 12730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027348749549415704,
+      "loss": 4.0594,
+      "step": 12735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002734671497472347,
+      "loss": 4.1094,
+      "step": 12740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027344679695403016,
+      "loss": 4.0516,
+      "step": 12745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000273426437115705,
+      "loss": 4.075,
+      "step": 12750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002734060702334212,
+      "loss": 4.0781,
+      "step": 12755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002733856963083411,
+      "loss": 4.1094,
+      "step": 12760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002733653153416273,
+      "loss": 4.0859,
+      "step": 12765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027334492733444303,
+      "loss": 4.0234,
+      "step": 12770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002733245322879519,
+      "loss": 4.1625,
+      "step": 12775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002733041302033179,
+      "loss": 4.1453,
+      "step": 12780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002732837210817052,
+      "loss": 4.0812,
+      "step": 12785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002732633049242786,
+      "loss": 4.1875,
+      "step": 12790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027324288173220335,
+      "loss": 4.1141,
+      "step": 12795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002732224515066449,
+      "loss": 4.1469,
+      "step": 12800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002732020142487692,
+      "loss": 4.1375,
+      "step": 12805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002731815699597428,
+      "loss": 4.1594,
+      "step": 12810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027316111864073223,
+      "loss": 4.1297,
+      "step": 12815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027314066029290467,
+      "loss": 4.0578,
+      "step": 12820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027312019491742786,
+      "loss": 4.1797,
+      "step": 12825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002730997225154696,
+      "loss": 4.0594,
+      "step": 12830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027307924308819824,
+      "loss": 4.1437,
+      "step": 12835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027305875663678265,
+      "loss": 4.1391,
+      "step": 12840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730382631623919,
+      "loss": 4.1031,
+      "step": 12845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730177626661956,
+      "loss": 4.15,
+      "step": 12850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027299725514936374,
+      "loss": 4.0922,
+      "step": 12855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027297674061306667,
+      "loss": 4.075,
+      "step": 12860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002729562190584751,
+      "loss": 4.1516,
+      "step": 12865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027293569048676025,
+      "loss": 4.1469,
+      "step": 12870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002729151548990937,
+      "loss": 4.0875,
+      "step": 12875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002728946122966474,
+      "loss": 4.0891,
+      "step": 12880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002728740626805936,
+      "loss": 4.1219,
+      "step": 12885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002728535060521053,
+      "loss": 4.0625,
+      "step": 12890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027283294241235554,
+      "loss": 4.0516,
+      "step": 12895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002728123717625179,
+      "loss": 4.0625,
+      "step": 12900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727917941037663,
+      "loss": 4.1188,
+      "step": 12905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727712094372752,
+      "loss": 4.0797,
+      "step": 12910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027275061776421927,
+      "loss": 4.0828,
+      "step": 12915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727300190857738,
+      "loss": 4.05,
+      "step": 12920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027270941340311425,
+      "loss": 4.1141,
+      "step": 12925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726888007174167,
+      "loss": 4.0938,
+      "step": 12930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027266818102985734,
+      "loss": 4.1219,
+      "step": 12935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027264755434161313,
+      "loss": 4.1422,
+      "step": 12940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027262692065386113,
+      "loss": 4.0766,
+      "step": 12945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027260627996777894,
+      "loss": 4.1344,
+      "step": 12950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725856322845445,
+      "loss": 4.1031,
+      "step": 12955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725649776053362,
+      "loss": 4.0422,
+      "step": 12960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725443159313328,
+      "loss": 4.0797,
+      "step": 12965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027252364726371343,
+      "loss": 4.1094,
+      "step": 12970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027250297160365776,
+      "loss": 4.0766,
+      "step": 12975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724822889523456,
+      "loss": 4.0875,
+      "step": 12980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724615993109574,
+      "loss": 4.175,
+      "step": 12985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027244090268067396,
+      "loss": 4.0062,
+      "step": 12990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724201990626763,
+      "loss": 4.1172,
+      "step": 12995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027239948845814616,
+      "loss": 3.9859,
+      "step": 13000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723787708682653,
+      "loss": 4.125,
+      "step": 13005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723580462942162,
+      "loss": 4.1641,
+      "step": 13010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723373147371816,
+      "loss": 4.0781,
+      "step": 13015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027231657619834467,
+      "loss": 4.1562,
+      "step": 13020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722958306788889,
+      "loss": 4.1016,
+      "step": 13025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722750781799983,
+      "loss": 4.1156,
+      "step": 13030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722543187028571,
+      "loss": 4.0922,
+      "step": 13035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722335522486502,
+      "loss": 4.1234,
+      "step": 13040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722127788185627,
+      "loss": 4.1375,
+      "step": 13045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027219199841378,
+      "loss": 4.0531,
+      "step": 13050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027217121103548827,
+      "loss": 4.1469,
+      "step": 13055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721504166848737,
+      "loss": 4.1281,
+      "step": 13060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027212961536312306,
+      "loss": 4.1297,
+      "step": 13065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027210880707142344,
+      "loss": 4.1875,
+      "step": 13070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027208799181096247,
+      "loss": 4.0969,
+      "step": 13075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000272067169582928,
+      "loss": 4.0953,
+      "step": 13080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027204634038850836,
+      "loss": 4.1203,
+      "step": 13085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720255042288923,
+      "loss": 4.1562,
+      "step": 13090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027200466110526895,
+      "loss": 4.0422,
+      "step": 13095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719838110188278,
+      "loss": 4.0516,
+      "step": 13100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719629539707588,
+      "loss": 4.0672,
+      "step": 13105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027194208996225217,
+      "loss": 4.0,
+      "step": 13110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719212189944988,
+      "loss": 4.1406,
+      "step": 13115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719003410686896,
+      "loss": 4.1219,
+      "step": 13120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027187945618601625,
+      "loss": 4.0812,
+      "step": 13125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002718585643476705,
+      "loss": 4.0406,
+      "step": 13130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002718376655548448,
+      "loss": 4.0422,
+      "step": 13135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027181675980873175,
+      "loss": 3.9469,
+      "step": 13140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002717958471105244,
+      "loss": 4.0656,
+      "step": 13145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027177492746141633,
+      "loss": 4.1625,
+      "step": 13150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027175400086260136,
+      "loss": 4.1156,
+      "step": 13155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002717330673152739,
+      "loss": 4.0453,
+      "step": 13160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027171212682062845,
+      "loss": 4.0875,
+      "step": 13165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716911793798602,
+      "loss": 4.0406,
+      "step": 13170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716702249941646,
+      "loss": 4.1219,
+      "step": 13175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027164926366473745,
+      "loss": 4.0391,
+      "step": 13180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716282953927752,
+      "loss": 4.0672,
+      "step": 13185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716073201794743,
+      "loss": 4.1125,
+      "step": 13190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715863380260319,
+      "loss": 4.1641,
+      "step": 13195
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715653489336455,
+      "loss": 4.1234,
+      "step": 13200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027154435290351286,
+      "loss": 4.1375,
+      "step": 13205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715233499368323,
+      "loss": 4.25,
+      "step": 13210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027150234003480247,
+      "loss": 4.0844,
+      "step": 13215
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714813231986223,
+      "loss": 4.0766,
+      "step": 13220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027146029942949125,
+      "loss": 4.0906,
+      "step": 13225
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714392687286093,
+      "loss": 4.1312,
+      "step": 13230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027141823109717644,
+      "loss": 4.1437,
+      "step": 13235
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002713971865363935,
+      "loss": 4.0906,
+      "step": 13240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027137613504746137,
+      "loss": 4.0344,
+      "step": 13245
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027135507663158154,
+      "loss": 4.1031,
+      "step": 13250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002713340112899557,
+      "loss": 4.1859,
+      "step": 13255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027131293902378614,
+      "loss": 4.175,
+      "step": 13260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027129185983427545,
+      "loss": 4.1016,
+      "step": 13265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027127077372262665,
+      "loss": 4.1906,
+      "step": 13270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027124968069004306,
+      "loss": 4.1406,
+      "step": 13275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027122858073772845,
+      "loss": 4.0844,
+      "step": 13280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027120747386688706,
+      "loss": 4.1109,
+      "step": 13285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027118636007872343,
+      "loss": 4.0844,
+      "step": 13290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002711652393744426,
+      "loss": 4.1109,
+      "step": 13295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027114411175524977,
+      "loss": 4.0187,
+      "step": 13300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027112297722235083,
+      "loss": 4.0781,
+      "step": 13305
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027110183577695187,
+      "loss": 4.05,
+      "step": 13310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710806874202595,
+      "loss": 4.1312,
+      "step": 13315
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027105953215348055,
+      "loss": 4.1078,
+      "step": 13320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027103836997782244,
+      "loss": 4.0781,
+      "step": 13325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027101720089449285,
+      "loss": 4.1609,
+      "step": 13330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027099602490469996,
+      "loss": 4.1469,
+      "step": 13335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002709748420096522,
+      "loss": 4.0984,
+      "step": 13340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002709536522105586,
+      "loss": 4.0844,
+      "step": 13345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027093245550862836,
+      "loss": 4.0375,
+      "step": 13350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027091125190507117,
+      "loss": 4.1328,
+      "step": 13355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002708900414010972,
+      "loss": 4.0656,
+      "step": 13360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027086882399791695,
+      "loss": 4.1141,
+      "step": 13365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027084759969674116,
+      "loss": 4.0391,
+      "step": 13370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027082636849878124,
+      "loss": 4.1219,
+      "step": 13375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027080513040524877,
+      "loss": 4.225,
+      "step": 13380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027078388541735583,
+      "loss": 4.1125,
+      "step": 13385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027076263353631496,
+      "loss": 4.1016,
+      "step": 13390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002707413747633389,
+      "loss": 4.1359,
+      "step": 13395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002707201090996409,
+      "loss": 4.0922,
+      "step": 13400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027069883654643465,
+      "loss": 4.0938,
+      "step": 13405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027067755710493416,
+      "loss": 4.0766,
+      "step": 13410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027065627077635385,
+      "loss": 4.15,
+      "step": 13415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027063497756190844,
+      "loss": 4.1594,
+      "step": 13420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002706136774628133,
+      "loss": 4.0969,
+      "step": 13425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002705923704802839,
+      "loss": 4.1375,
+      "step": 13430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002705710566155363,
+      "loss": 4.0703,
+      "step": 13435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027054973586978685,
+      "loss": 4.1422,
+      "step": 13440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027052840824425234,
+      "loss": 4.0438,
+      "step": 13445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002705070737401499,
+      "loss": 4.1141,
+      "step": 13450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002704857323586972,
+      "loss": 4.0797,
+      "step": 13455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000270464384101112,
+      "loss": 4.0187,
+      "step": 13460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002704430289686129,
+      "loss": 4.1219,
+      "step": 13465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027042166696241847,
+      "loss": 4.1516,
+      "step": 13470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027040029808374786,
+      "loss": 4.1078,
+      "step": 13475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027037892233382064,
+      "loss": 4.1375,
+      "step": 13480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002703575397138567,
+      "loss": 4.1156,
+      "step": 13485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027033615022507634,
+      "loss": 4.0594,
+      "step": 13490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002703147538687003,
+      "loss": 4.0938,
+      "step": 13495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027029335064594957,
+      "loss": 4.0875,
+      "step": 13500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027027194055804577,
+      "loss": 4.15,
+      "step": 13505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027025052360621067,
+      "loss": 4.0578,
+      "step": 13510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002702290997916666,
+      "loss": 4.1016,
+      "step": 13515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002702076691156362,
+      "loss": 4.1125,
+      "step": 13520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701862315793425,
+      "loss": 4.0891,
+      "step": 13525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027016478718400904,
+      "loss": 4.1109,
+      "step": 13530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701433359308595,
+      "loss": 3.9969,
+      "step": 13535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701218778211182,
+      "loss": 4.0891,
+      "step": 13540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027010041285600975,
+      "loss": 4.1047,
+      "step": 13545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027007894103675914,
+      "loss": 4.0672,
+      "step": 13550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700574623645918,
+      "loss": 4.075,
+      "step": 13555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700359768407334,
+      "loss": 4.1234,
+      "step": 13560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700144844664103,
+      "loss": 4.1344,
+      "step": 13565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269992985242849,
+      "loss": 4.1172,
+      "step": 13570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026997147917127637,
+      "loss": 4.1297,
+      "step": 13575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026994996625291987,
+      "loss": 4.0578,
+      "step": 13580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026992844648900723,
+      "loss": 4.1094,
+      "step": 13585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699069198807666,
+      "loss": 4.0719,
+      "step": 13590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026988538642942647,
+      "loss": 3.9859,
+      "step": 13595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002698638461362157,
+      "loss": 4.1203,
+      "step": 13600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026984229900236373,
+      "loss": 4.075,
+      "step": 13605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026982074502910014,
+      "loss": 4.0734,
+      "step": 13610
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697991842176551,
+      "loss": 4.1391,
+      "step": 13615
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026977761656925904,
+      "loss": 4.1781,
+      "step": 13620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026975604208514284,
+      "loss": 4.0344,
+      "step": 13625
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697344607665377,
+      "loss": 4.0984,
+      "step": 13630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697128726146755,
+      "loss": 4.0875,
+      "step": 13635
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696912776307879,
+      "loss": 4.0984,
+      "step": 13640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026966967581610766,
+      "loss": 4.15,
+      "step": 13645
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696480671718674,
+      "loss": 4.0062,
+      "step": 13650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026962645169930044,
+      "loss": 4.0984,
+      "step": 13655
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026960482939964033,
+      "loss": 4.0062,
+      "step": 13660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269583200274121,
+      "loss": 4.0187,
+      "step": 13665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026956156432397696,
+      "loss": 4.125,
+      "step": 13670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695399215504429,
+      "loss": 4.0406,
+      "step": 13675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695182719547539,
+      "loss": 4.2047,
+      "step": 13680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026949661553814566,
+      "loss": 4.0141,
+      "step": 13685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026947495230185403,
+      "loss": 4.0953,
+      "step": 13690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694532822471153,
+      "loss": 4.0734,
+      "step": 13695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026943160537516627,
+      "loss": 4.1078,
+      "step": 13700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694099216872439,
+      "loss": 4.1469,
+      "step": 13705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026938823118458587,
+      "loss": 4.1641,
+      "step": 13710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026936653386843,
+      "loss": 4.0359,
+      "step": 13715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693448297400144,
+      "loss": 4.1688,
+      "step": 13720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693231188005779,
+      "loss": 4.0109,
+      "step": 13725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693014010513595,
+      "loss": 4.1359,
+      "step": 13730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692796764935986,
+      "loss": 4.15,
+      "step": 13735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692579451285351,
+      "loss": 4.1937,
+      "step": 13740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026923620695740906,
+      "loss": 4.1359,
+      "step": 13745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026921446198146127,
+      "loss": 4.1219,
+      "step": 13750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691927102019326,
+      "loss": 4.1266,
+      "step": 13755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026917095162006443,
+      "loss": 4.1422,
+      "step": 13760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691491862370986,
+      "loss": 4.0563,
+      "step": 13765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026912741405427717,
+      "loss": 4.1906,
+      "step": 13770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691056350728427,
+      "loss": 4.1547,
+      "step": 13775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690838492940382,
+      "loss": 4.0938,
+      "step": 13780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690620567191069,
+      "loss": 4.0094,
+      "step": 13785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026904025734929244,
+      "loss": 4.0484,
+      "step": 13790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690184511858391,
+      "loss": 4.0406,
+      "step": 13795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689966382299913,
+      "loss": 4.0,
+      "step": 13800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689748184829938,
+      "loss": 4.1,
+      "step": 13805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689529919460919,
+      "loss": 4.1281,
+      "step": 13810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689311586205313,
+      "loss": 4.0984,
+      "step": 13815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268909318507558,
+      "loss": 4.1234,
+      "step": 13820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026888747160841843,
+      "loss": 4.1406,
+      "step": 13825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688656179243594,
+      "loss": 4.1391,
+      "step": 13830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268843757456628,
+      "loss": 4.1141,
+      "step": 13835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026882189020647194,
+      "loss": 4.0609,
+      "step": 13840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026880001617513916,
+      "loss": 4.1063,
+      "step": 13845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026877813536387794,
+      "loss": 4.1344,
+      "step": 13850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687562477739371,
+      "loss": 4.0969,
+      "step": 13855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026873435340656575,
+      "loss": 4.0547,
+      "step": 13860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026871245226301333,
+      "loss": 4.0672,
+      "step": 13865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686905443445298,
+      "loss": 4.1484,
+      "step": 13870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686686296523655,
+      "loss": 4.0625,
+      "step": 13875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026864670818777106,
+      "loss": 4.1,
+      "step": 13880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026862477995199754,
+      "loss": 4.1344,
+      "step": 13885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686028449462963,
+      "loss": 4.1141,
+      "step": 13890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685809031719193,
+      "loss": 4.0344,
+      "step": 13895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685589546301187,
+      "loss": 4.125,
+      "step": 13900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685369993221471,
+      "loss": 4.1094,
+      "step": 13905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685150372492575,
+      "loss": 4.175,
+      "step": 13910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684930684127033,
+      "loss": 4.1578,
+      "step": 13915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684710928137383,
+      "loss": 4.1156,
+      "step": 13920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684491104536165,
+      "loss": 4.1344,
+      "step": 13925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684271213335926,
+      "loss": 4.0547,
+      "step": 13930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026840512545492145,
+      "loss": 4.1469,
+      "step": 13935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683831228188583,
+      "loss": 4.0875,
+      "step": 13940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026836111342665896,
+      "loss": 4.1844,
+      "step": 13945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026833909727957945,
+      "loss": 4.1031,
+      "step": 13950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683170743788762,
+      "loss": 4.0953,
+      "step": 13955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026829504472580617,
+      "loss": 4.0531,
+      "step": 13960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682730083216264,
+      "loss": 4.1,
+      "step": 13965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682509651675947,
+      "loss": 4.0734,
+      "step": 13970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026822891526496903,
+      "loss": 4.0828,
+      "step": 13975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682068586150077,
+      "loss": 4.0594,
+      "step": 13980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026818479521896954,
+      "loss": 4.0812,
+      "step": 13985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681627250781138,
+      "loss": 4.0578,
+      "step": 13990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681406481936998,
+      "loss": 4.0094,
+      "step": 13995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681185645669877,
+      "loss": 4.0594,
+      "step": 14000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026809647419923766,
+      "loss": 4.1344,
+      "step": 14005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026807437709171043,
+      "loss": 4.1766,
+      "step": 14010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026805227324566716,
+      "loss": 4.1047,
+      "step": 14015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680301626623692,
+      "loss": 4.1,
+      "step": 14020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680080453430785,
+      "loss": 4.0797,
+      "step": 14025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026798592128905724,
+      "loss": 4.1203,
+      "step": 14030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002679637905015681,
+      "loss": 4.0547,
+      "step": 14035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026794165298187403,
+      "loss": 4.0422,
+      "step": 14040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026791950873123843,
+      "loss": 4.1219,
+      "step": 14045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002678973577509251,
+      "loss": 4.1547,
+      "step": 14050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026787520004219813,
+      "loss": 4.0906,
+      "step": 14055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026785303560632217,
+      "loss": 4.0719,
+      "step": 14060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002678308644445621,
+      "loss": 4.1094,
+      "step": 14065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026780868655818316,
+      "loss": 4.0219,
+      "step": 14070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002677865019484512,
+      "loss": 4.1312,
+      "step": 14075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002677643106166321,
+      "loss": 4.1047,
+      "step": 14080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002677421125639925,
+      "loss": 4.1281,
+      "step": 14085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002677199077917991,
+      "loss": 4.1094,
+      "step": 14090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026769769630131924,
+      "loss": 4.1234,
+      "step": 14095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002676754780938205,
+      "loss": 4.0359,
+      "step": 14100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026765325317057085,
+      "loss": 4.0219,
+      "step": 14105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002676310215328387,
+      "loss": 4.0563,
+      "step": 14110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002676087831818928,
+      "loss": 4.1172,
+      "step": 14115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026758653811900226,
+      "loss": 4.0906,
+      "step": 14120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026756428634543665,
+      "loss": 4.1156,
+      "step": 14125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002675420278624659,
+      "loss": 4.1016,
+      "step": 14130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026751976267136023,
+      "loss": 4.125,
+      "step": 14135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002674974907733904,
+      "loss": 4.1437,
+      "step": 14140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002674752121698274,
+      "loss": 4.0672,
+      "step": 14145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002674529268619427,
+      "loss": 4.1688,
+      "step": 14150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026743063485100817,
+      "loss": 4.1562,
+      "step": 14155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000267408336138296,
+      "loss": 4.1688,
+      "step": 14160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673860307250786,
+      "loss": 4.0734,
+      "step": 14165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673637186126292,
+      "loss": 4.0203,
+      "step": 14170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026734139980222105,
+      "loss": 4.1688,
+      "step": 14175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673190742951279,
+      "loss": 4.1188,
+      "step": 14180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672967420926238,
+      "loss": 4.05,
+      "step": 14185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026727440319598334,
+      "loss": 4.1422,
+      "step": 14190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672520576064814,
+      "loss": 4.1516,
+      "step": 14195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026722970532539316,
+      "loss": 4.1266,
+      "step": 14200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672073463539943,
+      "loss": 4.0391,
+      "step": 14205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026718498069356093,
+      "loss": 4.1672,
+      "step": 14210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026716260834536935,
+      "loss": 4.1672,
+      "step": 14215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026714022931069643,
+      "loss": 4.1125,
+      "step": 14220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671178435908193,
+      "loss": 4.1781,
+      "step": 14225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670954511870156,
+      "loss": 4.0609,
+      "step": 14230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670730521005631,
+      "loss": 4.0687,
+      "step": 14235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026705064633274027,
+      "loss": 4.0812,
+      "step": 14240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670282338848258,
+      "loss": 4.1375,
+      "step": 14245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670058147580986,
+      "loss": 4.1234,
+      "step": 14250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026698338895383836,
+      "loss": 4.0984,
+      "step": 14255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026696095647332476,
+      "loss": 4.0891,
+      "step": 14260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026693851731783813,
+      "loss": 4.0656,
+      "step": 14265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266916071488659,
+      "loss": 4.1312,
+      "step": 14270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668936189870684,
+      "loss": 4.1688,
+      "step": 14275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026687115981434764,
+      "loss": 4.1156,
+      "step": 14280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668486939717786,
+      "loss": 3.9813,
+      "step": 14285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026682622146064324,
+      "loss": 4.1016,
+      "step": 14290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026680374228222416,
+      "loss": 4.0281,
+      "step": 14295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667812564378042,
+      "loss": 4.1094,
+      "step": 14300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026675876392866666,
+      "loss": 4.1203,
+      "step": 14305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026673626475609525,
+      "loss": 4.0375,
+      "step": 14310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026671375892137395,
+      "loss": 4.125,
+      "step": 14315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666912464257871,
+      "loss": 4.0859,
+      "step": 14320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666687272706196,
+      "loss": 4.1172,
+      "step": 14325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026664620145715656,
+      "loss": 4.1297,
+      "step": 14330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666236689866835,
+      "loss": 4.0563,
+      "step": 14335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026660112986048645,
+      "loss": 4.0594,
+      "step": 14340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026657858407985166,
+      "loss": 4.0984,
+      "step": 14345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665560316460658,
+      "loss": 4.0812,
+      "step": 14350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026653347256041593,
+      "loss": 4.0625,
+      "step": 14355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665109068241896,
+      "loss": 4.0797,
+      "step": 14360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026648833443867456,
+      "loss": 4.0484,
+      "step": 14365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266465755405159,
+      "loss": 4.1375,
+      "step": 14370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026644316972493153,
+      "loss": 4.0953,
+      "step": 14375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664205773992811,
+      "loss": 4.0656,
+      "step": 14380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663979784294972,
+      "loss": 4.0625,
+      "step": 14385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663753728168693,
+      "loss": 4.1469,
+      "step": 14390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026635276056268767,
+      "loss": 4.1031,
+      "step": 14395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026633014166824284,
+      "loss": 4.1094,
+      "step": 14400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663075161348255,
+      "loss": 4.0328,
+      "step": 14405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662848839637271,
+      "loss": 4.1109,
+      "step": 14410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266262245156239,
+      "loss": 3.9734,
+      "step": 14415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026623959971365345,
+      "loss": 4.0875,
+      "step": 14420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662169476372627,
+      "loss": 4.0312,
+      "step": 14425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661942889283595,
+      "loss": 4.1281,
+      "step": 14430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026617162358823703,
+      "loss": 4.0734,
+      "step": 14435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026614895161818883,
+      "loss": 4.0844,
+      "step": 14440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661262730195087,
+      "loss": 4.1203,
+      "step": 14445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266103587793491,
+      "loss": 4.1188,
+      "step": 14450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002660808959414303,
+      "loss": 4.0375,
+      "step": 14455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026605819746462165,
+      "loss": 4.1594,
+      "step": 14460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002660354923643605,
+      "loss": 3.9641,
+      "step": 14465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026601278064194257,
+      "loss": 4.1375,
+      "step": 14470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659900622986641,
+      "loss": 4.1312,
+      "step": 14475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026596733733582153,
+      "loss": 4.1109,
+      "step": 14480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026594460575471185,
+      "loss": 4.1781,
+      "step": 14485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026592186755663235,
+      "loss": 4.0594,
+      "step": 14490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026589912274288063,
+      "loss": 4.1141,
+      "step": 14495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658763713147548,
+      "loss": 4.1453,
+      "step": 14500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658536132735533,
+      "loss": 4.0266,
+      "step": 14505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658308486205749,
+      "loss": 4.0203,
+      "step": 14510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026580807735711873,
+      "loss": 4.1531,
+      "step": 14515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026578529948448446,
+      "loss": 4.0438,
+      "step": 14520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265762515003972,
+      "loss": 4.0438,
+      "step": 14525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026573972391688154,
+      "loss": 4.1141,
+      "step": 14530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026571692622451396,
+      "loss": 4.1484,
+      "step": 14535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026569412192817016,
+      "loss": 4.0641,
+      "step": 14540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656713110291517,
+      "loss": 4.0,
+      "step": 14545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026564849352876034,
+      "loss": 4.15,
+      "step": 14550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026562566942829824,
+      "loss": 4.075,
+      "step": 14555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656028387290681,
+      "loss": 4.1562,
+      "step": 14560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655800014323728,
+      "loss": 4.1281,
+      "step": 14565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655571575395156,
+      "loss": 4.0094,
+      "step": 14570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026553430705180027,
+      "loss": 4.1437,
+      "step": 14575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655114499705309,
+      "loss": 4.075,
+      "step": 14580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026548858629701195,
+      "loss": 4.0687,
+      "step": 14585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654657160325482,
+      "loss": 4.1672,
+      "step": 14590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026544283917844495,
+      "loss": 4.0359,
+      "step": 14595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026541995573600767,
+      "loss": 3.9453,
+      "step": 14600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653970657065424,
+      "loss": 4.0406,
+      "step": 14605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653741690913554,
+      "loss": 4.1625,
+      "step": 14610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653512658917535,
+      "loss": 4.0719,
+      "step": 14615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653283561090436,
+      "loss": 4.1359,
+      "step": 14620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653054397445334,
+      "loss": 4.0391,
+      "step": 14625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026528251679953067,
+      "loss": 4.1375,
+      "step": 14630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652595872753435,
+      "loss": 4.1047,
+      "step": 14635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652366511732805,
+      "loss": 4.1328,
+      "step": 14640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652137084946508,
+      "loss": 4.1,
+      "step": 14645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651907592407636,
+      "loss": 4.2047,
+      "step": 14650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651678034129286,
+      "loss": 4.1703,
+      "step": 14655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265144841012456,
+      "loss": 4.0719,
+      "step": 14660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026512187204065616,
+      "loss": 4.1047,
+      "step": 14665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026509889649884,
+      "loss": 4.1016,
+      "step": 14670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026507591438831873,
+      "loss": 4.0625,
+      "step": 14675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002650529257104039,
+      "loss": 4.1141,
+      "step": 14680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002650299304664075,
+      "loss": 4.1297,
+      "step": 14685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026500692865764186,
+      "loss": 4.0875,
+      "step": 14690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026498392028541964,
+      "loss": 4.0109,
+      "step": 14695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026496090535105407,
+      "loss": 4.0453,
+      "step": 14700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002649378838558585,
+      "loss": 4.0563,
+      "step": 14705
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026491485580114685,
+      "loss": 4.1063,
+      "step": 14710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002648918211882332,
+      "loss": 4.0891,
+      "step": 14715
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002648687800184323,
+      "loss": 4.2,
+      "step": 14720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026484573229305897,
+      "loss": 4.0563,
+      "step": 14725
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026482267801342867,
+      "loss": 3.9859,
+      "step": 14730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000264799617180857,
+      "loss": 4.0781,
+      "step": 14735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002647765497966601,
+      "loss": 4.0891,
+      "step": 14740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002647534758621545,
+      "loss": 4.0687,
+      "step": 14745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026473039537865687,
+      "loss": 4.025,
+      "step": 14750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002647073083474845,
+      "loss": 4.0984,
+      "step": 14755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000264684214769955,
+      "loss": 4.15,
+      "step": 14760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026466111464738633,
+      "loss": 4.1125,
+      "step": 14765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002646380079810968,
+      "loss": 3.9984,
+      "step": 14770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000264614894772405,
+      "loss": 4.0859,
+      "step": 14775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002645917750226301,
+      "loss": 4.0969,
+      "step": 14780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026456864873309157,
+      "loss": 4.0578,
+      "step": 14785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002645455159051092,
+      "loss": 4.1047,
+      "step": 14790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026452237654000317,
+      "loss": 4.0812,
+      "step": 14795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026449923063909405,
+      "loss": 4.0969,
+      "step": 14800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026447607820370277,
+      "loss": 4.0906,
+      "step": 14805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002644529192351507,
+      "loss": 4.0297,
+      "step": 14810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026442975373475945,
+      "loss": 4.0703,
+      "step": 14815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002644065817038511,
+      "loss": 4.05,
+      "step": 14820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002643834031437481,
+      "loss": 4.1391,
+      "step": 14825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026436021805577325,
+      "loss": 4.0719,
+      "step": 14830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026433702644124976,
+      "loss": 4.0812,
+      "step": 14835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000264313828301501,
+      "loss": 4.0859,
+      "step": 14840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026429062363785114,
+      "loss": 4.1234,
+      "step": 14845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026426741245162434,
+      "loss": 4.1031,
+      "step": 14850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026424419474414525,
+      "loss": 4.0453,
+      "step": 14855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000264220970516739,
+      "loss": 4.1344,
+      "step": 14860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641977397707308,
+      "loss": 3.9844,
+      "step": 14865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026417450250744673,
+      "loss": 4.1312,
+      "step": 14870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641512587282127,
+      "loss": 4.0828,
+      "step": 14875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026412800843435535,
+      "loss": 4.0672,
+      "step": 14880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641047516272015,
+      "loss": 4.1328,
+      "step": 14885
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026408148830807843,
+      "loss": 4.1156,
+      "step": 14890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640582184783139,
+      "loss": 4.1031,
+      "step": 14895
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640349421392358,
+      "loss": 3.9953,
+      "step": 14900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640116592921725,
+      "loss": 4.0328,
+      "step": 14905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639883699384528,
+      "loss": 4.0687,
+      "step": 14910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026396507407940587,
+      "loss": 4.0828,
+      "step": 14915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639417717163612,
+      "loss": 4.0687,
+      "step": 14920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639184628506485,
+      "loss": 4.1047,
+      "step": 14925
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638951474835982,
+      "loss": 4.0828,
+      "step": 14930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026387182561654084,
+      "loss": 4.0766,
+      "step": 14935
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026384849725080744,
+      "loss": 4.1297,
+      "step": 14940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638251623877292,
+      "loss": 4.0969,
+      "step": 14945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026380182102863807,
+      "loss": 4.0656,
+      "step": 14950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026377847317486593,
+      "loss": 4.0312,
+      "step": 14955
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002637551188277454,
+      "loss": 4.0578,
+      "step": 14960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026373175798860923,
+      "loss": 4.1,
+      "step": 14965
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026370839065879067,
+      "loss": 4.1328,
+      "step": 14970
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002636850168396232,
+      "loss": 3.9844,
+      "step": 14975
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000263661636532441,
+      "loss": 4.0328,
+      "step": 14980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026363824973857814,
+      "loss": 4.025,
+      "step": 14985
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002636148564593694,
+      "loss": 4.0328,
+      "step": 14990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026359145669614984,
+      "loss": 4.0797,
+      "step": 14995
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026356805045025486,
+      "loss": 4.0625,
+      "step": 15000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026354463772302033,
+      "loss": 4.1156,
+      "step": 15005
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026352121851578236,
+      "loss": 4.0938,
+      "step": 15010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026349779282987745,
+      "loss": 4.1156,
+      "step": 15015
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634743606666426,
+      "loss": 4.0422,
+      "step": 15020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000263450922027415,
+      "loss": 4.0172,
+      "step": 15025
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634274769135323,
+      "loss": 4.0906,
+      "step": 15030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634040253263326,
+      "loss": 4.1125,
+      "step": 15035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002633805672671542,
+      "loss": 4.1219,
+      "step": 15040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026335710273733587,
+      "loss": 4.0531,
+      "step": 15045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026333363173821673,
+      "loss": 4.075,
+      "step": 15050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026331015427113634,
+      "loss": 4.0344,
+      "step": 15055
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002632866703374345,
+      "loss": 4.1312,
+      "step": 15060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026326317993845136,
+      "loss": 3.9656,
+      "step": 15065
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002632396830755277,
+      "loss": 4.0375,
+      "step": 15070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026321617975000436,
+      "loss": 4.0859,
+      "step": 15075
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026319266996322274,
+      "loss": 4.1312,
+      "step": 15080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631691537165245,
+      "loss": 4.1453,
+      "step": 15085
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631456310112517,
+      "loss": 4.0938,
+      "step": 15090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631221018487469,
+      "loss": 4.0875,
+      "step": 15095
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002630985662303528,
+      "loss": 4.1344,
+      "step": 15100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026307502415741255,
+      "loss": 4.0797,
+      "step": 15105
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026305147563126976,
+      "loss": 4.0875,
+      "step": 15110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002630279206532684,
+      "loss": 4.0906,
+      "step": 15115
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026300435922475265,
+      "loss": 4.0734,
+      "step": 15120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026298079134706725,
+      "loss": 4.1078,
+      "step": 15125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026295721702155716,
+      "loss": 4.0906,
+      "step": 15130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026293363624956776,
+      "loss": 4.0641,
+      "step": 15135
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002629100490324449,
+      "loss": 4.0187,
+      "step": 15140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026288645537153463,
+      "loss": 4.1531,
+      "step": 15145
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002628628552681834,
+      "loss": 4.0141,
+      "step": 15150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026283924872373823,
+      "loss": 4.0125,
+      "step": 15155
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026281563573954614,
+      "loss": 4.1469,
+      "step": 15160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026279201631695487,
+      "loss": 3.9797,
+      "step": 15165
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026276839045731235,
+      "loss": 4.1188,
+      "step": 15170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026274475816196683,
+      "loss": 4.1219,
+      "step": 15175
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627211194322672,
+      "loss": 4.0187,
+      "step": 15180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626974742695623,
+      "loss": 4.0953,
+      "step": 15185
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626738226752017,
+      "loss": 4.0703,
+      "step": 15190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626501646505352,
+      "loss": 4.0391,
+      "step": 15195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626265001969129,
+      "loss": 4.0469,
+      "step": 15200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626028293156853,
+      "loss": 4.0875,
+      "step": 15205
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625791520082035,
+      "loss": 4.0719,
+      "step": 15210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026255546827581844,
+      "loss": 4.1328,
+      "step": 15215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625317781198821,
+      "loss": 4.0656,
+      "step": 15220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026250808154174623,
+      "loss": 4.2578,
+      "step": 15225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026248437854276333,
+      "loss": 4.0953,
+      "step": 15230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624606691242861,
+      "loss": 4.1219,
+      "step": 15235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624369532876676,
+      "loss": 4.075,
+      "step": 15240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026241323103426135,
+      "loss": 4.075,
+      "step": 15245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002623895023654211,
+      "loss": 4.1188,
+      "step": 15250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002623657672825012,
+      "loss": 4.0312,
+      "step": 15255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000262342025786856,
+      "loss": 4.1063,
+      "step": 15260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002623182778798406,
+      "loss": 4.0625,
+      "step": 15265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622945235628102,
+      "loss": 4.0656,
+      "step": 15270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026227076283712056,
+      "loss": 4.1781,
+      "step": 15275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622469957041276,
+      "loss": 4.1094,
+      "step": 15280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622232221651878,
+      "loss": 4.0906,
+      "step": 15285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026219944222165785,
+      "loss": 4.1156,
+      "step": 15290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621756558748949,
+      "loss": 4.1156,
+      "step": 15295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621518631262564,
+      "loss": 4.0672,
+      "step": 15300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026212806397710035,
+      "loss": 4.0219,
+      "step": 15305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621042584287848,
+      "loss": 4.0719,
+      "step": 15310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026208044648266837,
+      "loss": 4.0297,
+      "step": 15315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620566281401101,
+      "loss": 4.0297,
+      "step": 15320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620328034024692,
+      "loss": 4.0125,
+      "step": 15325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026200897227110547,
+      "loss": 4.1297,
+      "step": 15330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002619851347473788,
+      "loss": 4.1672,
+      "step": 15335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002619612908326497,
+      "loss": 4.0016,
+      "step": 15340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026193744052827896,
+      "loss": 4.1,
+      "step": 15345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026191358383562764,
+      "loss": 4.15,
+      "step": 15350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002618897207560573,
+      "loss": 4.0953,
+      "step": 15355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026186585129092974,
+      "loss": 4.1203,
+      "step": 15360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026184197544160733,
+      "loss": 3.9641,
+      "step": 15365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002618180932094526,
+      "loss": 4.1109,
+      "step": 15370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026179420459582843,
+      "loss": 4.0687,
+      "step": 15375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002617703096020982,
+      "loss": 4.2,
+      "step": 15380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002617464082296257,
+      "loss": 3.9984,
+      "step": 15385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026172250047977485,
+      "loss": 4.1078,
+      "step": 15390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026169858635391013,
+      "loss": 4.075,
+      "step": 15395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002616746658533962,
+      "loss": 4.0594,
+      "step": 15400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002616507389795984,
+      "loss": 4.0531,
+      "step": 15405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002616268057338822,
+      "loss": 4.0719,
+      "step": 15410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002616028661176134,
+      "loss": 4.1047,
+      "step": 15415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002615789201321582,
+      "loss": 4.1719,
+      "step": 15420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002615549677788833,
+      "loss": 4.0516,
+      "step": 15425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026153100905915557,
+      "loss": 4.0062,
+      "step": 15430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002615070439743424,
+      "loss": 4.0547,
+      "step": 15435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002614830725258115,
+      "loss": 4.0922,
+      "step": 15440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026145909471493085,
+      "loss": 4.0766,
+      "step": 15445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026143511054306897,
+      "loss": 4.0781,
+      "step": 15450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002614111200115945,
+      "loss": 4.1,
+      "step": 15455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002613871231218767,
+      "loss": 4.0812,
+      "step": 15460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000261363119875285,
+      "loss": 4.0156,
+      "step": 15465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002613391102731893,
+      "loss": 4.0906,
+      "step": 15470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026131509431695987,
+      "loss": 4.0797,
+      "step": 15475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612910720079672,
+      "loss": 4.0906,
+      "step": 15480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612670433475824,
+      "loss": 4.0812,
+      "step": 15485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026124300833717663,
+      "loss": 4.1047,
+      "step": 15490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612189669781216,
+      "loss": 4.0391,
+      "step": 15495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026119491927178946,
+      "loss": 4.0953,
+      "step": 15500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026117086521955255,
+      "loss": 4.1531,
+      "step": 15505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611468048227836,
+      "loss": 4.0594,
+      "step": 15510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026112273808285576,
+      "loss": 4.0031,
+      "step": 15515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026109866500114256,
+      "loss": 4.1297,
+      "step": 15520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026107458557901776,
+      "loss": 4.0797,
+      "step": 15525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610504998178557,
+      "loss": 4.0812,
+      "step": 15530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610264077190308,
+      "loss": 4.1781,
+      "step": 15535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610023092839182,
+      "loss": 4.0328,
+      "step": 15540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609782045138931,
+      "loss": 4.0938,
+      "step": 15545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609540934103311,
+      "loss": 4.1719,
+      "step": 15550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026092997597460834,
+      "loss": 4.1656,
+      "step": 15555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609058522081011,
+      "loss": 4.1141,
+      "step": 15560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026088172211218615,
+      "loss": 4.0687,
+      "step": 15565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026085758568824066,
+      "loss": 4.1797,
+      "step": 15570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260833442937642,
+      "loss": 4.1688,
+      "step": 15575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026080929386176807,
+      "loss": 4.1078,
+      "step": 15580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026078513846199705,
+      "loss": 4.0016,
+      "step": 15585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607609767397075,
+      "loss": 4.0766,
+      "step": 15590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607368086962783,
+      "loss": 4.0531,
+      "step": 15595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607126343330888,
+      "loss": 4.1297,
+      "step": 15600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026068845365151845,
+      "loss": 4.1141,
+      "step": 15605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026066426665294747,
+      "loss": 4.1594,
+      "step": 15610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026064007333875605,
+      "loss": 4.0844,
+      "step": 15615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026061587371032497,
+      "loss": 4.1875,
+      "step": 15620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026059166776903534,
+      "loss": 4.0656,
+      "step": 15625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002605674555162685,
+      "loss": 4.0031,
+      "step": 15630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026054323695340635,
+      "loss": 4.1031,
+      "step": 15635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026051901208183094,
+      "loss": 4.0766,
+      "step": 15640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002604947809029249,
+      "loss": 4.0484,
+      "step": 15645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260470543418071,
+      "loss": 4.0531,
+      "step": 15650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026044629962865257,
+      "loss": 4.0984,
+      "step": 15655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026042204953605316,
+      "loss": 4.0906,
+      "step": 15660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603977931416567,
+      "loss": 4.1844,
+      "step": 15665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026037353044684747,
+      "loss": 4.175,
+      "step": 15670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603492614530103,
+      "loss": 4.1156,
+      "step": 15675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603249861615301,
+      "loss": 4.1063,
+      "step": 15680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026030070457379225,
+      "loss": 4.1531,
+      "step": 15685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602764166911826,
+      "loss": 4.1,
+      "step": 15690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602521225150871,
+      "loss": 4.0922,
+      "step": 15695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026022782204689244,
+      "loss": 4.0391,
+      "step": 15700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026020351528798527,
+      "loss": 4.0687,
+      "step": 15705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026017920223975286,
+      "loss": 4.0625,
+      "step": 15710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002601548829035828,
+      "loss": 4.1781,
+      "step": 15715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026013055728086286,
+      "loss": 4.1109,
+      "step": 15720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002601062253729814,
+      "loss": 4.1156,
+      "step": 15725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026008188718132705,
+      "loss": 4.0234,
+      "step": 15730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002600575427072888,
+      "loss": 3.9906,
+      "step": 15735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002600331919522559,
+      "loss": 4.0969,
+      "step": 15740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026000883491761816,
+      "loss": 4.0766,
+      "step": 15745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025998447160476554,
+      "loss": 4.1422,
+      "step": 15750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025996010201508854,
+      "loss": 4.1266,
+      "step": 15755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025993572614997797,
+      "loss": 4.0531,
+      "step": 15760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599113440108248,
+      "loss": 4.0187,
+      "step": 15765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002598869555990207,
+      "loss": 4.2,
+      "step": 15770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025986256091595737,
+      "loss": 4.0484,
+      "step": 15775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025983815996302707,
+      "loss": 4.0703,
+      "step": 15780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002598137527416224,
+      "loss": 4.1266,
+      "step": 15785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025978933925313623,
+      "loss": 4.1625,
+      "step": 15790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025976491949896194,
+      "loss": 4.0781,
+      "step": 15795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025974049348049307,
+      "loss": 4.1234,
+      "step": 15800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002597160611991237,
+      "loss": 4.075,
+      "step": 15805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259691622656248,
+      "loss": 4.1672,
+      "step": 15810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002596671778532609,
+      "loss": 4.0719,
+      "step": 15815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025964272679155736,
+      "loss": 4.1297,
+      "step": 15820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025961826947253276,
+      "loss": 4.0219,
+      "step": 15825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259593805897583,
+      "loss": 4.1109,
+      "step": 15830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002595693360681042,
+      "loss": 4.0687,
+      "step": 15835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002595448599854927,
+      "loss": 4.0969,
+      "step": 15840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025952037765114555,
+      "loss": 3.9656,
+      "step": 15845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594958890664599,
+      "loss": 4.1234,
+      "step": 15850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025947139423283317,
+      "loss": 4.1078,
+      "step": 15855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594468931516635,
+      "loss": 4.0609,
+      "step": 15860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259422385824349,
+      "loss": 4.1875,
+      "step": 15865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002593978722522885,
+      "loss": 4.0766,
+      "step": 15870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025937335243688076,
+      "loss": 4.075,
+      "step": 15875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025934882637952526,
+      "loss": 4.0656,
+      "step": 15880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002593242940816217,
+      "loss": 4.0516,
+      "step": 15885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592997555445701,
+      "loss": 4.1562,
+      "step": 15890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025927521076977094,
+      "loss": 4.0609,
+      "step": 15895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025925065975862494,
+      "loss": 4.1172,
+      "step": 15900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592261025125332,
+      "loss": 4.0516,
+      "step": 15905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025920153903289734,
+      "loss": 4.0578,
+      "step": 15910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025917696932111904,
+      "loss": 4.0438,
+      "step": 15915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025915239337860054,
+      "loss": 4.0781,
+      "step": 15920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025912781120674447,
+      "loss": 4.0172,
+      "step": 15925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591032228069536,
+      "loss": 4.0125,
+      "step": 15930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025907862818063136,
+      "loss": 4.1,
+      "step": 15935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002590540273291812,
+      "loss": 4.1562,
+      "step": 15940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025902942025400724,
+      "loss": 4.0328,
+      "step": 15945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002590048069565137,
+      "loss": 4.0328,
+      "step": 15950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589801874381053,
+      "loss": 4.15,
+      "step": 15955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589555617001872,
+      "loss": 4.1219,
+      "step": 15960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589309297441645,
+      "loss": 4.0531,
+      "step": 15965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025890629157144325,
+      "loss": 3.9469,
+      "step": 15970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588816471834294,
+      "loss": 4.1141,
+      "step": 15975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588569965815294,
+      "loss": 4.0187,
+      "step": 15980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588323397671501,
+      "loss": 4.1453,
+      "step": 15985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588076767416987,
+      "loss": 4.1141,
+      "step": 15990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025878300750658265,
+      "loss": 4.1656,
+      "step": 15995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025875833206320993,
+      "loss": 4.0547,
+      "step": 16000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025873365041298864,
+      "loss": 4.0687,
+      "step": 16005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025870896255732737,
+      "loss": 4.1359,
+      "step": 16010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025868426849763517,
+      "loss": 4.0375,
+      "step": 16015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025865956823532127,
+      "loss": 4.1125,
+      "step": 16020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025863486177179535,
+      "loss": 4.1047,
+      "step": 16025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002586101491084674,
+      "loss": 4.1375,
+      "step": 16030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025858543024674766,
+      "loss": 4.0969,
+      "step": 16035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000258560705188047,
+      "loss": 4.0531,
+      "step": 16040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002585359739337764,
+      "loss": 4.0578,
+      "step": 16045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002585112364853473,
+      "loss": 4.1219,
+      "step": 16050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025848649284417144,
+      "loss": 4.0219,
+      "step": 16055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000258461743011661,
+      "loss": 4.0469,
+      "step": 16060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025843698698922843,
+      "loss": 4.0359,
+      "step": 16065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002584122247782865,
+      "loss": 4.0406,
+      "step": 16070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025838745638024843,
+      "loss": 4.0984,
+      "step": 16075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002583626817965278,
+      "loss": 4.0266,
+      "step": 16080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002583379010285384,
+      "loss": 4.0687,
+      "step": 16085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002583131140776946,
+      "loss": 4.0547,
+      "step": 16090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002582883209454109,
+      "loss": 4.0156,
+      "step": 16095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002582635216331022,
+      "loss": 4.1344,
+      "step": 16100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025823871614218396,
+      "loss": 4.0687,
+      "step": 16105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025821390447407173,
+      "loss": 3.9937,
+      "step": 16110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025818908663018147,
+      "loss": 4.0953,
+      "step": 16115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025816426261192967,
+      "loss": 4.0125,
+      "step": 16120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002581394324207329,
+      "loss": 4.1125,
+      "step": 16125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002581145960580083,
+      "loss": 4.1266,
+      "step": 16130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580897535251733,
+      "loss": 4.0828,
+      "step": 16135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025806490482364555,
+      "loss": 4.0328,
+      "step": 16140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025804004995484334,
+      "loss": 3.9859,
+      "step": 16145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258015188920185,
+      "loss": 4.1281,
+      "step": 16150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025799032172108943,
+      "loss": 4.0578,
+      "step": 16155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025796544835897573,
+      "loss": 4.0703,
+      "step": 16160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002579405688352635,
+      "loss": 4.0375,
+      "step": 16165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002579156831513726,
+      "loss": 4.1234,
+      "step": 16170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025789079130872323,
+      "loss": 4.1109,
+      "step": 16175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257865893308736,
+      "loss": 4.0594,
+      "step": 16180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002578409891528318,
+      "loss": 4.0609,
+      "step": 16185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257816078842432,
+      "loss": 4.1094,
+      "step": 16190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577911623789581,
+      "loss": 4.0297,
+      "step": 16195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577662397638322,
+      "loss": 4.1422,
+      "step": 16200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025774131099847656,
+      "loss": 4.0859,
+      "step": 16205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025771637608431393,
+      "loss": 4.0891,
+      "step": 16210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025769143502276726,
+      "loss": 3.9875,
+      "step": 16215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025766648781526007,
+      "loss": 4.0781,
+      "step": 16220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257641534463216,
+      "loss": 4.1359,
+      "step": 16225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002576165749680592,
+      "loss": 4.1312,
+      "step": 16230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257591609331214,
+      "loss": 3.9906,
+      "step": 16235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025756663755410536,
+      "loss": 4.1437,
+      "step": 16240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025754165963815824,
+      "loss": 4.0672,
+      "step": 16245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002575166755847983,
+      "loss": 4.0609,
+      "step": 16250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025749168539545127,
+      "loss": 4.0312,
+      "step": 16255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002574666890715434,
+      "loss": 4.0375,
+      "step": 16260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002574416866145012,
+      "loss": 4.1437,
+      "step": 16265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025741667802575154,
+      "loss": 4.0953,
+      "step": 16270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573916633067217,
+      "loss": 3.9703,
+      "step": 16275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573666424588393,
+      "loss": 4.0563,
+      "step": 16280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025734161548353227,
+      "loss": 4.05,
+      "step": 16285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573165823822288,
+      "loss": 4.0344,
+      "step": 16290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572915431563577,
+      "loss": 4.1125,
+      "step": 16295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572664978073479,
+      "loss": 4.0891,
+      "step": 16300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025724144633662864,
+      "loss": 4.0734,
+      "step": 16305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572163887456297,
+      "loss": 4.1344,
+      "step": 16310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025719132503578115,
+      "loss": 4.0547,
+      "step": 16315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025716625520851326,
+      "loss": 4.0578,
+      "step": 16320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025714117926525684,
+      "loss": 4.0062,
+      "step": 16325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025711609720744305,
+      "loss": 4.1109,
+      "step": 16330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570910090365032,
+      "loss": 4.1047,
+      "step": 16335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570659147538691,
+      "loss": 4.0344,
+      "step": 16340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025704081436097296,
+      "loss": 4.1,
+      "step": 16345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570157078592472,
+      "loss": 4.1297,
+      "step": 16350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025699059525012465,
+      "loss": 3.9766,
+      "step": 16355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025696547653503854,
+      "loss": 4.1297,
+      "step": 16360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025694035171542235,
+      "loss": 4.0031,
+      "step": 16365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025691522079270994,
+      "loss": 4.0422,
+      "step": 16370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025689008376833555,
+      "loss": 4.0672,
+      "step": 16375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568649406437338,
+      "loss": 4.0875,
+      "step": 16380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568397914203395,
+      "loss": 4.0563,
+      "step": 16385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025681463609958805,
+      "loss": 4.0563,
+      "step": 16390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256789474682915,
+      "loss": 3.9922,
+      "step": 16395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025676430717175633,
+      "loss": 4.0016,
+      "step": 16400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002567391335675484,
+      "loss": 4.0672,
+      "step": 16405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002567139538717277,
+      "loss": 4.0016,
+      "step": 16410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025668876808573134,
+      "loss": 4.0609,
+      "step": 16415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025666357621099674,
+      "loss": 4.0984,
+      "step": 16420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025663837824896156,
+      "loss": 4.1875,
+      "step": 16425
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025661317420106383,
+      "loss": 4.0844,
+      "step": 16430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025658796406874195,
+      "loss": 4.0641,
+      "step": 16435
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025656274785343467,
+      "loss": 4.1422,
+      "step": 16440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025653752555658115,
+      "loss": 4.0516,
+      "step": 16445
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002565122971796207,
+      "loss": 4.0656,
+      "step": 16450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025648706272399317,
+      "loss": 4.0812,
+      "step": 16455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025646182219113867,
+      "loss": 4.0969,
+      "step": 16460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564365755824977,
+      "loss": 4.0375,
+      "step": 16465
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564113228995111,
+      "loss": 4.1188,
+      "step": 16470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002563860641436201,
+      "loss": 4.1094,
+      "step": 16475
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002563607993162661,
+      "loss": 4.0828,
+      "step": 16480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256335528418891,
+      "loss": 4.0609,
+      "step": 16485
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256310251452937,
+      "loss": 4.0484,
+      "step": 16490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002562849684198467,
+      "loss": 4.0453,
+      "step": 16495
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025625967932106305,
+      "loss": 4.0625,
+      "step": 16500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002562343841580292,
+      "loss": 4.0187,
+      "step": 16505
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002562090829321888,
+      "loss": 4.0187,
+      "step": 16510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561837756449858,
+      "loss": 4.0438,
+      "step": 16515
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025615846229786447,
+      "loss": 4.0469,
+      "step": 16520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025613314289226944,
+      "loss": 4.0922,
+      "step": 16525
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561078174296457,
+      "loss": 4.1125,
+      "step": 16530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025608248591143856,
+      "loss": 4.1078,
+      "step": 16535
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002560571483390937,
+      "loss": 4.1078,
+      "step": 16540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002560318047140572,
+      "loss": 4.0141,
+      "step": 16545
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025600645503777534,
+      "loss": 4.1125,
+      "step": 16550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025598109931169484,
+      "loss": 4.1125,
+      "step": 16555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002559557375372628,
+      "loss": 4.0984,
+      "step": 16560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025593036971592657,
+      "loss": 3.9984,
+      "step": 16565
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025590499584913394,
+      "loss": 4.1531,
+      "step": 16570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002558796159383329,
+      "loss": 4.0766,
+      "step": 16575
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025585422998497206,
+      "loss": 4.0828,
+      "step": 16580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002558288379905,
+      "loss": 4.1172,
+      "step": 16585
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000255803439956366,
+      "loss": 4.0875,
+      "step": 16590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025577803588401946,
+      "loss": 4.0609,
+      "step": 16595
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025575262577491023,
+      "loss": 4.1813,
+      "step": 16600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557272096304884,
+      "loss": 4.0812,
+      "step": 16605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557017874522045,
+      "loss": 4.1188,
+      "step": 16610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025567635924150946,
+      "loss": 4.0578,
+      "step": 16615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025565092499985433,
+      "loss": 4.1188,
+      "step": 16620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025562548472869066,
+      "loss": 4.1406,
+      "step": 16625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025560003842947047,
+      "loss": 4.0,
+      "step": 16630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002555745861036459,
+      "loss": 3.9953,
+      "step": 16635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025554912775266943,
+      "loss": 4.1828,
+      "step": 16640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025552366337799415,
+      "loss": 4.1484,
+      "step": 16645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002554981929810731,
+      "loss": 4.1328,
+      "step": 16650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002554727165633601,
+      "loss": 4.1,
+      "step": 16655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025544723412630895,
+      "loss": 4.0734,
+      "step": 16660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025542174567137396,
+      "loss": 4.1922,
+      "step": 16665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025539625120000976,
+      "loss": 4.025,
+      "step": 16670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025537075071367134,
+      "loss": 4.0875,
+      "step": 16675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000255345244213814,
+      "loss": 4.0797,
+      "step": 16680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002553197317018934,
+      "loss": 4.0906,
+      "step": 16685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002552942131793656,
+      "loss": 3.9656,
+      "step": 16690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002552686886476868,
+      "loss": 4.125,
+      "step": 16695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002552431581083138,
+      "loss": 4.0625,
+      "step": 16700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002552176215627036,
+      "loss": 4.0047,
+      "step": 16705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002551920790123136,
+      "loss": 4.0859,
+      "step": 16710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025516653045860155,
+      "loss": 4.15,
+      "step": 16715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002551409759030254,
+      "loss": 4.1156,
+      "step": 16720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025511541534704364,
+      "loss": 4.1203,
+      "step": 16725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002550898487921149,
+      "loss": 4.0609,
+      "step": 16730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025506427623969843,
+      "loss": 4.0578,
+      "step": 16735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025503869769125353,
+      "loss": 4.0672,
+      "step": 16740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025501311314824,
+      "loss": 4.0641,
+      "step": 16745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025498752261211805,
+      "loss": 3.9516,
+      "step": 16750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025496192608434793,
+      "loss": 4.0125,
+      "step": 16755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002549363235663907,
+      "loss": 4.0953,
+      "step": 16760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002549107150597072,
+      "loss": 4.125,
+      "step": 16765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025488510056575917,
+      "loss": 3.9672,
+      "step": 16770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025485948008600824,
+      "loss": 4.1297,
+      "step": 16775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025483385362191674,
+      "loss": 4.0719,
+      "step": 16780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025480822117494704,
+      "loss": 4.0812,
+      "step": 16785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002547825827465621,
+      "loss": 4.1422,
+      "step": 16790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000254756938338225,
+      "loss": 4.1234,
+      "step": 16795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002547312879513993,
+      "loss": 4.0609,
+      "step": 16800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025470563158754895,
+      "loss": 4.1141,
+      "step": 16805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000254679969248138,
+      "loss": 4.0938,
+      "step": 16810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002546543009346312,
+      "loss": 4.1281,
+      "step": 16815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002546286266484933,
+      "loss": 4.1375,
+      "step": 16820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025460294639118963,
+      "loss": 4.1359,
+      "step": 16825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025457726016418563,
+      "loss": 4.1281,
+      "step": 16830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545515679689473,
+      "loss": 4.0563,
+      "step": 16835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025452586980694096,
+      "loss": 4.0672,
+      "step": 16840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545001656796331,
+      "loss": 4.0594,
+      "step": 16845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002544744555884907,
+      "loss": 4.0875,
+      "step": 16850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025444873953498105,
+      "loss": 4.0344,
+      "step": 16855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025442301752057173,
+      "loss": 4.0156,
+      "step": 16860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025439728954673073,
+      "loss": 4.1484,
+      "step": 16865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025437155561492634,
+      "loss": 4.1469,
+      "step": 16870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543458157266272,
+      "loss": 4.1297,
+      "step": 16875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543200698833023,
+      "loss": 4.0547,
+      "step": 16880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542943180864209,
+      "loss": 4.0859,
+      "step": 16885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542685603374527,
+      "loss": 4.1391,
+      "step": 16890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542427966378677,
+      "loss": 4.0016,
+      "step": 16895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542170269891363,
+      "loss": 4.0781,
+      "step": 16900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025419125139272903,
+      "loss": 4.0406,
+      "step": 16905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025416546985011703,
+      "loss": 4.0766,
+      "step": 16910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025413968236277155,
+      "loss": 4.0703,
+      "step": 16915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002541138889321644,
+      "loss": 4.05,
+      "step": 16920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025408808955976756,
+      "loss": 4.1141,
+      "step": 16925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002540622842470534,
+      "loss": 4.0359,
+      "step": 16930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025403647299549457,
+      "loss": 3.9875,
+      "step": 16935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025401065580656426,
+      "loss": 4.0453,
+      "step": 16940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025398483268173575,
+      "loss": 4.0703,
+      "step": 16945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025395900362248285,
+      "loss": 4.1625,
+      "step": 16950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025393316863027947,
+      "loss": 4.0563,
+      "step": 16955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025390732770660024,
+      "loss": 4.0969,
+      "step": 16960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538814808529197,
+      "loss": 4.1516,
+      "step": 16965
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025385562807071305,
+      "loss": 4.0703,
+      "step": 16970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538297693614557,
+      "loss": 4.0625,
+      "step": 16975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025380390472662335,
+      "loss": 4.0547,
+      "step": 16980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537780341676922,
+      "loss": 4.1328,
+      "step": 16985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025375215768613855,
+      "loss": 4.1406,
+      "step": 16990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537262752834393,
+      "loss": 4.1594,
+      "step": 16995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025370038696107146,
+      "loss": 4.0672,
+      "step": 17000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025367449272051257,
+      "loss": 4.0938,
+      "step": 17005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002536485925632404,
+      "loss": 4.0984,
+      "step": 17010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025362268649073297,
+      "loss": 4.075,
+      "step": 17015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025359677450446884,
+      "loss": 4.1156,
+      "step": 17020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535708566059268,
+      "loss": 4.0812,
+      "step": 17025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000253544932796586,
+      "loss": 4.05,
+      "step": 17030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025351900307792584,
+      "loss": 4.075,
+      "step": 17035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025349306745142626,
+      "loss": 3.9859,
+      "step": 17040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025346712591856724,
+      "loss": 4.0297,
+      "step": 17045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534411784808294,
+      "loss": 4.1031,
+      "step": 17050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534152251396935,
+      "loss": 4.1172,
+      "step": 17055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002533892658966408,
+      "loss": 4.0563,
+      "step": 17060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002533633007531526,
+      "loss": 4.0844,
+      "step": 17065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025333732971071085,
+      "loss": 4.1156,
+      "step": 17070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025331135277079775,
+      "loss": 4.1047,
+      "step": 17075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532853699348958,
+      "loss": 4.1203,
+      "step": 17080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532593812044878,
+      "loss": 4.0391,
+      "step": 17085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025323338658105694,
+      "loss": 4.0516,
+      "step": 17090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532073860660868,
+      "loss": 4.0969,
+      "step": 17095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531813796610611,
+      "loss": 3.9844,
+      "step": 17100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025315536736746417,
+      "loss": 4.1391,
+      "step": 17105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531293491867804,
+      "loss": 4.0656,
+      "step": 17110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025310332512049473,
+      "loss": 4.175,
+      "step": 17115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025307729517009235,
+      "loss": 3.9578,
+      "step": 17120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002530512593370588,
+      "loss": 4.0547,
+      "step": 17125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025302521762287993,
+      "loss": 4.0922,
+      "step": 17130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529991700290419,
+      "loss": 4.0812,
+      "step": 17135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025297311655703137,
+      "loss": 4.0672,
+      "step": 17140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529470572083351,
+      "loss": 4.0687,
+      "step": 17145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025292099198444035,
+      "loss": 4.0406,
+      "step": 17150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002528949208868346,
+      "loss": 4.075,
+      "step": 17155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025286884391700584,
+      "loss": 4.1297,
+      "step": 17160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025284276107644224,
+      "loss": 4.0187,
+      "step": 17165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025281667236663226,
+      "loss": 4.0438,
+      "step": 17170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025279057778906494,
+      "loss": 4.0125,
+      "step": 17175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025276447734522936,
+      "loss": 4.0906,
+      "step": 17180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002527383710366152,
+      "loss": 4.0328,
+      "step": 17185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025271225886471226,
+      "loss": 4.1359,
+      "step": 17190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025268614083101077,
+      "loss": 4.0922,
+      "step": 17195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526600169370013,
+      "loss": 4.0234,
+      "step": 17200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526338871841748,
+      "loss": 4.0266,
+      "step": 17205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526077515740225,
+      "loss": 4.1266,
+      "step": 17210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025258161010803575,
+      "loss": 4.0469,
+      "step": 17215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025255546278770677,
+      "loss": 4.0281,
+      "step": 17220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002525293096145276,
+      "loss": 4.0609,
+      "step": 17225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002525031505899908,
+      "loss": 4.0672,
+      "step": 17230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025247698571558924,
+      "loss": 4.0484,
+      "step": 17235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025245081499281627,
+      "loss": 4.2,
+      "step": 17240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524246384231654,
+      "loss": 4.0875,
+      "step": 17245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025239845600813054,
+      "loss": 4.1125,
+      "step": 17250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523722677492059,
+      "loss": 4.0719,
+      "step": 17255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025234607364788605,
+      "loss": 4.125,
+      "step": 17260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523198737056659,
+      "loss": 4.1156,
+      "step": 17265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522936679240407,
+      "loss": 4.1078,
+      "step": 17270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252267456304506,
+      "loss": 4.1813,
+      "step": 17275
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025224123884855763,
+      "loss": 4.1203,
+      "step": 17280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252215015557692,
+      "loss": 4.0984,
+      "step": 17285
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025218878643340546,
+      "loss": 4.0641,
+      "step": 17290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252162551477195,
+      "loss": 4.125,
+      "step": 17295
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252136310690558,
+      "loss": 4.0328,
+      "step": 17300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002521100640749918,
+      "loss": 4.0312,
+      "step": 17305
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025208381163199437,
+      "loss": 4.1063,
+      "step": 17310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252057553363064,
+      "loss": 4.0969,
+      "step": 17315
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002520312892696992,
+      "loss": 4.0234,
+      "step": 17320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025200501935339884,
+      "loss": 4.0156,
+      "step": 17325
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025197874361566224,
+      "loss": 4.0297,
+      "step": 17330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002519524620579888,
+      "loss": 4.05,
+      "step": 17335
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025192617468187864,
+      "loss": 4.1047,
+      "step": 17340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002518998814888317,
+      "loss": 4.1844,
+      "step": 17345
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002518735824803488,
+      "loss": 4.0766,
+      "step": 17350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002518472776579307,
+      "loss": 4.0438,
+      "step": 17355
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002518209670230786,
+      "loss": 4.1109,
+      "step": 17360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025179465057729414,
+      "loss": 4.0359,
+      "step": 17365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002517683283220791,
+      "loss": 4.1375,
+      "step": 17370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002517420002589357,
+      "loss": 4.1297,
+      "step": 17375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025171566638936655,
+      "loss": 4.0797,
+      "step": 17380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025168932671487454,
+      "loss": 4.1047,
+      "step": 17385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002516629812369628,
+      "loss": 4.1328,
+      "step": 17390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002516366299571349,
+      "loss": 4.1109,
+      "step": 17395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002516102728768948,
+      "loss": 4.0312,
+      "step": 17400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025158390999774653,
+      "loss": 4.0734,
+      "step": 17405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025155754132119476,
+      "loss": 4.1188,
+      "step": 17410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002515311668487442,
+      "loss": 4.1516,
+      "step": 17415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002515047865819003,
+      "loss": 3.9922,
+      "step": 17420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025147840052216835,
+      "loss": 4.0406,
+      "step": 17425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025145200867105426,
+      "loss": 4.0656,
+      "step": 17430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025142561103006433,
+      "loss": 3.9844,
+      "step": 17435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002513992076007049,
+      "loss": 3.9969,
+      "step": 17440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025137279838448296,
+      "loss": 4.0953,
+      "step": 17445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002513463833829056,
+      "loss": 4.0953,
+      "step": 17450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025131996259748036,
+      "loss": 4.0328,
+      "step": 17455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025129353602971506,
+      "loss": 4.075,
+      "step": 17460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512671036811179,
+      "loss": 4.0734,
+      "step": 17465
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025124066555319737,
+      "loss": 4.1047,
+      "step": 17470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025121422164746225,
+      "loss": 4.0891,
+      "step": 17475
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511877719654218,
+      "loss": 4.1031,
+      "step": 17480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025116131650858536,
+      "loss": 4.1359,
+      "step": 17485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511348552784628,
+      "loss": 4.0844,
+      "step": 17490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511083882765644,
+      "loss": 4.0672,
+      "step": 17495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510819155044005,
+      "loss": 4.0781,
+      "step": 17500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510554369634818,
+      "loss": 4.1547,
+      "step": 17505
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510289526553197,
+      "loss": 4.0812,
+      "step": 17510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025100246258142543,
+      "loss": 4.0719,
+      "step": 17515
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002509759667433109,
+      "loss": 3.9953,
+      "step": 17520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002509494651424882,
+      "loss": 4.0047,
+      "step": 17525
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002509229577804698,
+      "loss": 4.0938,
+      "step": 17530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025089644465876843,
+      "loss": 4.1016,
+      "step": 17535
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002508699257788972,
+      "loss": 4.0734,
+      "step": 17540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002508434011423696,
+      "loss": 4.0328,
+      "step": 17545
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025081687075069936,
+      "loss": 4.0703,
+      "step": 17550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507903346054006,
+      "loss": 4.0875,
+      "step": 17555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025076379270798767,
+      "loss": 4.0266,
+      "step": 17560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025073724505997536,
+      "loss": 4.0797,
+      "step": 17565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507106916628788,
+      "loss": 4.0781,
+      "step": 17570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002506841325182133,
+      "loss": 4.1281,
+      "step": 17575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025065756762749465,
+      "loss": 4.0266,
+      "step": 17580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025063099699223893,
+      "loss": 4.1406,
+      "step": 17585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025060442061396247,
+      "loss": 4.0891,
+      "step": 17590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250577838494182,
+      "loss": 4.0766,
+      "step": 17595
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025055125063441465,
+      "loss": 4.0219,
+      "step": 17600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505246570361776,
+      "loss": 4.0766,
+      "step": 17605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002504980577009887,
+      "loss": 4.0406,
+      "step": 17610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250471452630366,
+      "loss": 4.0516,
+      "step": 17615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002504448418258278,
+      "loss": 4.0406,
+      "step": 17620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025041822528889277,
+      "loss": 3.9922,
+      "step": 17625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025039160302107994,
+      "loss": 4.0406,
+      "step": 17630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002503649750239086,
+      "loss": 4.0594,
+      "step": 17635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025033834129889845,
+      "loss": 4.0969,
+      "step": 17640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002503117018475695,
+      "loss": 4.075,
+      "step": 17645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025028505667144205,
+      "loss": 4.1078,
+      "step": 17650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002502584057720367,
+      "loss": 4.0953,
+      "step": 17655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002502317491508745,
+      "loss": 4.1125,
+      "step": 17660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025020508680947663,
+      "loss": 4.0906,
+      "step": 17665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002501784187493648,
+      "loss": 4.0141,
+      "step": 17670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250151744972061,
+      "loss": 4.0563,
+      "step": 17675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002501250654790874,
+      "loss": 4.0141,
+      "step": 17680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025009838027196667,
+      "loss": 4.0719,
+      "step": 17685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002500716893522217,
+      "loss": 4.0547,
+      "step": 17690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025004499272137575,
+      "loss": 4.0766,
+      "step": 17695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025001829038095243,
+      "loss": 4.0891,
+      "step": 17700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499915823324756,
+      "loss": 4.0734,
+      "step": 17705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024996486857746953,
+      "loss": 4.1047,
+      "step": 17710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499381491174588,
+      "loss": 4.1281,
+      "step": 17715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499114239539682,
+      "loss": 4.0969,
+      "step": 17720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024988469308852296,
+      "loss": 4.1437,
+      "step": 17725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498579565226487,
+      "loss": 4.0078,
+      "step": 17730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498312142578712,
+      "loss": 4.0297,
+      "step": 17735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498044662957167,
+      "loss": 4.1344,
+      "step": 17740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497777126377116,
+      "loss": 4.0359,
+      "step": 17745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497509532853829,
+      "loss": 4.0938,
+      "step": 17750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497241882402576,
+      "loss": 4.1469,
+      "step": 17755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024969741750386325,
+      "loss": 4.0062,
+      "step": 17760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496706410777277,
+      "loss": 4.0156,
+      "step": 17765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000249643858963379,
+      "loss": 4.0891,
+      "step": 17770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496170711623457,
+      "loss": 4.0234,
+      "step": 17775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024959027767615654,
+      "loss": 4.0891,
+      "step": 17780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002495634785063405,
+      "loss": 4.0547,
+      "step": 17785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024953667365442724,
+      "loss": 4.0297,
+      "step": 17790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002495098631219464,
+      "loss": 4.0187,
+      "step": 17795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024948304691042807,
+      "loss": 3.9641,
+      "step": 17800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002494562250214026,
+      "loss": 4.1781,
+      "step": 17805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002494293974564008,
+      "loss": 4.0687,
+      "step": 17810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024940256421695367,
+      "loss": 4.0109,
+      "step": 17815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493757253045926,
+      "loss": 4.1125,
+      "step": 17820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493488807208493,
+      "loss": 4.0812,
+      "step": 17825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024932203046725584,
+      "loss": 4.0203,
+      "step": 17830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492951745453445,
+      "loss": 4.0906,
+      "step": 17835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492683129566479,
+      "loss": 4.1125,
+      "step": 17840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492414457026991,
+      "loss": 3.9688,
+      "step": 17845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024921457278503147,
+      "loss": 4.1203,
+      "step": 17850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491876942051786,
+      "loss": 3.9672,
+      "step": 17855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024916080996467444,
+      "loss": 4.0594,
+      "step": 17860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024913392006505324,
+      "loss": 4.0516,
+      "step": 17865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491070245078497,
+      "loss": 3.9906,
+      "step": 17870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024908012329459865,
+      "loss": 4.0891,
+      "step": 17875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002490532164268354,
+      "loss": 4.0703,
+      "step": 17880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002490263039060956,
+      "loss": 4.1656,
+      "step": 17885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248999385733915,
+      "loss": 4.0969,
+      "step": 17890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024897246191182994,
+      "loss": 4.0344,
+      "step": 17895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489455324413769,
+      "loss": 4.0375,
+      "step": 17900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489185973240928,
+      "loss": 4.0828,
+      "step": 17905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488916565615148,
+      "loss": 4.0953,
+      "step": 17910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488647101551804,
+      "loss": 4.0703,
+      "step": 17915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024883775810662746,
+      "loss": 3.9672,
+      "step": 17920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024881080041739406,
+      "loss": 4.0,
+      "step": 17925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487838370890188,
+      "loss": 4.0547,
+      "step": 17930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024875686812304046,
+      "loss": 4.0922,
+      "step": 17935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024872989352099807,
+      "loss": 4.0172,
+      "step": 17940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487029132844311,
+      "loss": 3.9688,
+      "step": 17945
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024867592741487944,
+      "loss": 4.2062,
+      "step": 17950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248648935913883,
+      "loss": 4.0187,
+      "step": 17955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024862193878298226,
+      "loss": 4.1453,
+      "step": 17960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248594936023718,
+      "loss": 4.0625,
+      "step": 17965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024856792763763115,
+      "loss": 4.0891,
+      "step": 17970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024854091362626324,
+      "loss": 4.0531,
+      "step": 17975
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002485138939911558,
+      "loss": 4.0687,
+      "step": 17980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248486868733851,
+      "loss": 4.0984,
+      "step": 17985
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024845983785589103,
+      "loss": 4.0891,
+      "step": 17990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024843280135881864,
+      "loss": 4.0781,
+      "step": 17995
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024840575924417674,
+      "loss": 3.9953,
+      "step": 18000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024837871151350875,
+      "loss": 4.0938,
+      "step": 18005
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002483516581683581,
+      "loss": 4.1156,
+      "step": 18010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002483245992102688,
+      "loss": 4.1688,
+      "step": 18015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024829753464078523,
+      "loss": 4.0266,
+      "step": 18020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024827046446145186,
+      "loss": 4.0219,
+      "step": 18025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024824338867381355,
+      "loss": 4.0219,
+      "step": 18030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024821630727941556,
+      "loss": 4.0578,
+      "step": 18035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024818922027980346,
+      "loss": 4.0609,
+      "step": 18040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002481621276765231,
+      "loss": 4.0859,
+      "step": 18045
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002481350294711206,
+      "loss": 4.0812,
+      "step": 18050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024810792566514253,
+      "loss": 4.0,
+      "step": 18055
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002480808162601356,
+      "loss": 4.1578,
+      "step": 18060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002480537012576471,
+      "loss": 4.0563,
+      "step": 18065
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024802658065922436,
+      "loss": 4.1063,
+      "step": 18070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002479994544664152,
+      "loss": 4.0891,
+      "step": 18075
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024797232268076774,
+      "loss": 4.1281,
+      "step": 18080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024794518530383034,
+      "loss": 4.0453,
+      "step": 18085
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024791804233715176,
+      "loss": 3.9875,
+      "step": 18090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002478908937822811,
+      "loss": 4.1328,
+      "step": 18095
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024786373964076765,
+      "loss": 3.9703,
+      "step": 18100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002478365799141611,
+      "loss": 4.0609,
+      "step": 18105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024780941460401154,
+      "loss": 4.0797,
+      "step": 18110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024778224371186926,
+      "loss": 4.0266,
+      "step": 18115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024775506723928483,
+      "loss": 4.0703,
+      "step": 18120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024772788518780933,
+      "loss": 3.9734,
+      "step": 18125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002477006975589939,
+      "loss": 4.1078,
+      "step": 18130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476735043543903,
+      "loss": 4.0344,
+      "step": 18135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476463055755504,
+      "loss": 3.9766,
+      "step": 18140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024761910122402637,
+      "loss": 4.0641,
+      "step": 18145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475918913013708,
+      "loss": 4.1781,
+      "step": 18150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475646758091366,
+      "loss": 4.0734,
+      "step": 18155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024753745474887694,
+      "loss": 4.0484,
+      "step": 18160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475102281221453,
+      "loss": 4.05,
+      "step": 18165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474829959304955,
+      "loss": 4.0547,
+      "step": 18170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474557581754818,
+      "loss": 4.05,
+      "step": 18175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024742851485865853,
+      "loss": 3.9484,
+      "step": 18180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024740126598158047,
+      "loss": 4.0656,
+      "step": 18185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002473740115458028,
+      "loss": 4.1,
+      "step": 18190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024734675155288087,
+      "loss": 4.0906,
+      "step": 18195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024731948600437045,
+      "loss": 4.1672,
+      "step": 18200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002472922149018276,
+      "loss": 4.0719,
+      "step": 18205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002472649382468086,
+      "loss": 4.0031,
+      "step": 18210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002472376560408702,
+      "loss": 3.9937,
+      "step": 18215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024721036828556945,
+      "loss": 4.0141,
+      "step": 18220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002471830749824635,
+      "loss": 4.0141,
+      "step": 18225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024715577613311013,
+      "loss": 4.0703,
+      "step": 18230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002471284717390672,
+      "loss": 4.0406,
+      "step": 18235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000247101161801893,
+      "loss": 4.0531,
+      "step": 18240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470738463231462,
+      "loss": 4.0641,
+      "step": 18245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024704652530438555,
+      "loss": 4.0734,
+      "step": 18250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470191987471704,
+      "loss": 3.9766,
+      "step": 18255
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024699186665306014,
+      "loss": 4.0078,
+      "step": 18260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024696452902361475,
+      "loss": 4.0,
+      "step": 18265
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024693718586039427,
+      "loss": 4.0031,
+      "step": 18270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024690983716495933,
+      "loss": 4.0844,
+      "step": 18275
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024688248293887056,
+      "loss": 4.0687,
+      "step": 18280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024685512318368914,
+      "loss": 4.0031,
+      "step": 18285
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002468277579009765,
+      "loss": 4.0797,
+      "step": 18290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024680038709229434,
+      "loss": 4.0344,
+      "step": 18295
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002467730107592047,
+      "loss": 4.1156,
+      "step": 18300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002467456289032701,
+      "loss": 4.1094,
+      "step": 18305
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024671824152605306,
+      "loss": 4.1125,
+      "step": 18310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024669084862911667,
+      "loss": 4.0563,
+      "step": 18315
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466634502140242,
+      "loss": 3.9906,
+      "step": 18320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466360462823393,
+      "loss": 4.0516,
+      "step": 18325
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466086368356259,
+      "loss": 4.0203,
+      "step": 18330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002465812218754483,
+      "loss": 4.0703,
+      "step": 18335
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024655380140337096,
+      "loss": 4.1437,
+      "step": 18340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024652637542095897,
+      "loss": 4.0656,
+      "step": 18345
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024649894392977733,
+      "loss": 4.1328,
+      "step": 18350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024647150693139167,
+      "loss": 4.0438,
+      "step": 18355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002464440644273678,
+      "loss": 4.0531,
+      "step": 18360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002464166164192719,
+      "loss": 4.0719,
+      "step": 18365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024638916290867035,
+      "loss": 4.0656,
+      "step": 18370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024636170389713,
+      "loss": 4.1391,
+      "step": 18375
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463342393862179,
+      "loss": 4.0766,
+      "step": 18380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463067693775014,
+      "loss": 4.0516,
+      "step": 18385
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024627929387254833,
+      "loss": 4.0828,
+      "step": 18390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002462518128729267,
+      "loss": 4.0734,
+      "step": 18395
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024622432638020473,
+      "loss": 4.0359,
+      "step": 18400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002461968343959512,
+      "loss": 4.1156,
+      "step": 18405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002461693369217351,
+      "loss": 4.0625,
+      "step": 18410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024614183395912557,
+      "loss": 4.0187,
+      "step": 18415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024611432550969236,
+      "loss": 4.0547,
+      "step": 18420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460868115750053,
+      "loss": 3.9906,
+      "step": 18425
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024605929215663463,
+      "loss": 3.9984,
+      "step": 18430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024603176725615083,
+      "loss": 4.0641,
+      "step": 18435
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024600423687512485,
+      "loss": 4.0906,
+      "step": 18440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459767010151278,
+      "loss": 3.9906,
+      "step": 18445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459491596777312,
+      "loss": 4.0203,
+      "step": 18450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459216128645067,
+      "loss": 4.0812,
+      "step": 18455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024589406057702655,
+      "loss": 4.0766,
+      "step": 18460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458665028168631,
+      "loss": 4.1063,
+      "step": 18465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458389395855891,
+      "loss": 3.9594,
+      "step": 18470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458113708847775,
+      "loss": 4.0266,
+      "step": 18475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024578379671600183,
+      "loss": 4.0391,
+      "step": 18480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024575621708083554,
+      "loss": 4.0469,
+      "step": 18485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024572863198085276,
+      "loss": 3.9906,
+      "step": 18490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024570104141762765,
+      "loss": 3.9937,
+      "step": 18495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456734453927349,
+      "loss": 4.0984,
+      "step": 18500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456458439077494,
+      "loss": 4.0703,
+      "step": 18505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456182369642463,
+      "loss": 4.0953,
+      "step": 18510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024559062456380124,
+      "loss": 4.0672,
+      "step": 18515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024556300670799,
+      "loss": 4.0297,
+      "step": 18520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024553538339838873,
+      "loss": 4.0312,
+      "step": 18525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024550775463657394,
+      "loss": 4.0578,
+      "step": 18530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024548012042412235,
+      "loss": 4.0328,
+      "step": 18535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024545248076261104,
+      "loss": 4.0625,
+      "step": 18540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002454248356536175,
+      "loss": 4.1031,
+      "step": 18545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024539718509871935,
+      "loss": 4.1375,
+      "step": 18550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024536952909949463,
+      "loss": 4.0812,
+      "step": 18555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002453418676575216,
+      "loss": 4.0453,
+      "step": 18560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002453142007743791,
+      "loss": 4.1063,
+      "step": 18565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024528652845164594,
+      "loss": 4.0594,
+      "step": 18570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002452588506909014,
+      "loss": 4.0484,
+      "step": 18575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000245231167493725,
+      "loss": 4.075,
+      "step": 18580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024520347886169676,
+      "loss": 4.1078,
+      "step": 18585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024517578479639674,
+      "loss": 4.0906,
+      "step": 18590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451480852994055,
+      "loss": 4.0828,
+      "step": 18595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451203803723038,
+      "loss": 4.0422,
+      "step": 18600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024509267001667287,
+      "loss": 4.0031,
+      "step": 18605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024506495423409406,
+      "loss": 4.0391,
+      "step": 18610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450372330261492,
+      "loss": 4.0516,
+      "step": 18615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450095063944202,
+      "loss": 3.9734,
+      "step": 18620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024498177434048953,
+      "loss": 4.1031,
+      "step": 18625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002449540368659398,
+      "loss": 4.0984,
+      "step": 18630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002449262939723541,
+      "loss": 4.0078,
+      "step": 18635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002448985456613155,
+      "loss": 4.0187,
+      "step": 18640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002448707919344078,
+      "loss": 4.0203,
+      "step": 18645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024484303279321493,
+      "loss": 4.0703,
+      "step": 18650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024481526823932095,
+      "loss": 4.0344,
+      "step": 18655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002447874982743105,
+      "loss": 4.0609,
+      "step": 18660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002447597228997683,
+      "loss": 4.1125,
+      "step": 18665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002447319421172796,
+      "loss": 4.1312,
+      "step": 18670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002447041559284299,
+      "loss": 4.0672,
+      "step": 18675
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024467636433480477,
+      "loss": 4.0266,
+      "step": 18680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002446485673379904,
+      "loss": 3.95,
+      "step": 18685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002446207649395732,
+      "loss": 3.9906,
+      "step": 18690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002445929571411398,
+      "loss": 4.0563,
+      "step": 18695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024456514394427715,
+      "loss": 4.0453,
+      "step": 18700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002445373253505727,
+      "loss": 4.0469,
+      "step": 18705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002445095013616139,
+      "loss": 4.0828,
+      "step": 18710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024448167197898874,
+      "loss": 3.9453,
+      "step": 18715
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024445383720428544,
+      "loss": 4.1719,
+      "step": 18720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024442599703909256,
+      "loss": 3.9703,
+      "step": 18725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002443981514849989,
+      "loss": 4.0344,
+      "step": 18730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024437030054359364,
+      "loss": 4.0469,
+      "step": 18735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002443424442164662,
+      "loss": 4.0734,
+      "step": 18740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002443145825052063,
+      "loss": 4.0969,
+      "step": 18745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002442867154114042,
+      "loss": 4.0203,
+      "step": 18750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002442588429366501,
+      "loss": 4.0344,
+      "step": 18755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002442309650825347,
+      "loss": 4.0391,
+      "step": 18760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024420308185064905,
+      "loss": 3.9672,
+      "step": 18765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024417519324258444,
+      "loss": 3.9406,
+      "step": 18770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024414729925993243,
+      "loss": 3.9562,
+      "step": 18775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024411939990428495,
+      "loss": 4.0906,
+      "step": 18780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024409149517723428,
+      "loss": 4.0984,
+      "step": 18785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024406358508037286,
+      "loss": 4.0594,
+      "step": 18790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024403566961529358,
+      "loss": 4.0812,
+      "step": 18795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024400774878358952,
+      "loss": 4.0812,
+      "step": 18800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002439798225868542,
+      "loss": 4.0406,
+      "step": 18805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024395189102668132,
+      "loss": 4.0609,
+      "step": 18810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243923954104665,
+      "loss": 4.1031,
+      "step": 18815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024389601182239946,
+      "loss": 4.0328,
+      "step": 18820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024386806418147955,
+      "loss": 4.0,
+      "step": 18825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024384011118350012,
+      "loss": 4.0734,
+      "step": 18830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024381215283005646,
+      "loss": 4.0125,
+      "step": 18835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002437841891227442,
+      "loss": 4.0719,
+      "step": 18840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002437562200631592,
+      "loss": 4.0375,
+      "step": 18845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024372824565289773,
+      "loss": 4.0453,
+      "step": 18850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024370026589355616,
+      "loss": 4.0297,
+      "step": 18855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002436722807867314,
+      "loss": 4.025,
+      "step": 18860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024364429033402055,
+      "loss": 4.0453,
+      "step": 18865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024361629453702096,
+      "loss": 4.0781,
+      "step": 18870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024358829339733048,
+      "loss": 4.0047,
+      "step": 18875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024356028691654698,
+      "loss": 4.0875,
+      "step": 18880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024353227509626895,
+      "loss": 4.0734,
+      "step": 18885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024350425793809492,
+      "loss": 4.0719,
+      "step": 18890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024347623544362384,
+      "loss": 4.0109,
+      "step": 18895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024344820761445504,
+      "loss": 4.0266,
+      "step": 18900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243420174452188,
+      "loss": 4.0687,
+      "step": 18905
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002433921359584225,
+      "loss": 4.0656,
+      "step": 18910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024336409213475884,
+      "loss": 4.1375,
+      "step": 18915
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024333604298279745,
+      "loss": 4.0,
+      "step": 18920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243307988504139,
+      "loss": 4.0484,
+      "step": 18925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024327992870038474,
+      "loss": 4.0609,
+      "step": 18930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002432518635731359,
+      "loss": 4.0906,
+      "step": 18935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002432237931239942,
+      "loss": 4.05,
+      "step": 18940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024319571735456164,
+      "loss": 4.0703,
+      "step": 18945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002431676362664405,
+      "loss": 4.0438,
+      "step": 18950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002431395498612333,
+      "loss": 4.0219,
+      "step": 18955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002431114581405431,
+      "loss": 4.1188,
+      "step": 18960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024308336110597293,
+      "loss": 4.0422,
+      "step": 18965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024305525875912635,
+      "loss": 4.0453,
+      "step": 18970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024302715110160724,
+      "loss": 3.9906,
+      "step": 18975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024299903813501957,
+      "loss": 4.0969,
+      "step": 18980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024297091986096787,
+      "loss": 4.0547,
+      "step": 18985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002429427962810567,
+      "loss": 4.0281,
+      "step": 18990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024291466739689128,
+      "loss": 4.0109,
+      "step": 18995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024288653321007678,
+      "loss": 3.9906,
+      "step": 19000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024285839372221882,
+      "loss": 4.0875,
+      "step": 19005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024283024893492346,
+      "loss": 4.025,
+      "step": 19010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024280209884979678,
+      "loss": 4.0844,
+      "step": 19015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024277394346844535,
+      "loss": 3.9766,
+      "step": 19020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024274578279247602,
+      "loss": 4.0469,
+      "step": 19025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024271761682349594,
+      "loss": 4.0312,
+      "step": 19030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024268944556311248,
+      "loss": 4.0047,
+      "step": 19035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024266126901293343,
+      "loss": 4.0469,
+      "step": 19040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024263308717456682,
+      "loss": 4.1,
+      "step": 19045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024260490004962096,
+      "loss": 4.0109,
+      "step": 19050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024257670763970453,
+      "loss": 4.1437,
+      "step": 19055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024254850994642646,
+      "loss": 4.0891,
+      "step": 19060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000242520306971396,
+      "loss": 4.0687,
+      "step": 19065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024249209871622266,
+      "loss": 3.9266,
+      "step": 19070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024246388518251637,
+      "loss": 4.0891,
+      "step": 19075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024243566637188716,
+      "loss": 4.0656,
+      "step": 19080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024240744228594562,
+      "loss": 4.0406,
+      "step": 19085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024237921292630238,
+      "loss": 4.0984,
+      "step": 19090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024235097829456855,
+      "loss": 4.1031,
+      "step": 19095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024232273839235548,
+      "loss": 4.0281,
+      "step": 19100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002422944932212748,
+      "loss": 4.0766,
+      "step": 19105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024226624278293854,
+      "loss": 4.0812,
+      "step": 19110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024223798707895882,
+      "loss": 4.0625,
+      "step": 19115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024220972611094832,
+      "loss": 4.0609,
+      "step": 19120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024218145988051982,
+      "loss": 4.0766,
+      "step": 19125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024215318838928654,
+      "loss": 4.0531,
+      "step": 19130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002421249116388619,
+      "loss": 4.1016,
+      "step": 19135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024209662963085962,
+      "loss": 4.0141,
+      "step": 19140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024206834236689386,
+      "loss": 4.1297,
+      "step": 19145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024204004984857883,
+      "loss": 3.95,
+      "step": 19150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024201175207752936,
+      "loss": 4.0734,
+      "step": 19155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024198344905536026,
+      "loss": 4.0391,
+      "step": 19160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024195514078368684,
+      "loss": 4.05,
+      "step": 19165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002419268272641247,
+      "loss": 4.0703,
+      "step": 19170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024189850849828964,
+      "loss": 4.0594,
+      "step": 19175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024187018448779783,
+      "loss": 4.0672,
+      "step": 19180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024184185523426567,
+      "loss": 4.0984,
+      "step": 19185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024181352073931003,
+      "loss": 4.1578,
+      "step": 19190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024178518100454783,
+      "loss": 4.0719,
+      "step": 19195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024175683603159658,
+      "loss": 4.0406,
+      "step": 19200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024172848582207376,
+      "loss": 4.1016,
+      "step": 19205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024170013037759746,
+      "loss": 4.1219,
+      "step": 19210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024167176969978587,
+      "loss": 4.0516,
+      "step": 19215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002416434037902575,
+      "loss": 3.9625,
+      "step": 19220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024161503265063126,
+      "loss": 3.9391,
+      "step": 19225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024158665628252617,
+      "loss": 4.0016,
+      "step": 19230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024155827468756187,
+      "loss": 3.9734,
+      "step": 19235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024152988786735796,
+      "loss": 4.1016,
+      "step": 19240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024150149582353452,
+      "loss": 4.0781,
+      "step": 19245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024147309855771187,
+      "loss": 4.1266,
+      "step": 19250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024144469607151068,
+      "loss": 4.0828,
+      "step": 19255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024141628836655185,
+      "loss": 4.0297,
+      "step": 19260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024138787544445662,
+      "loss": 3.9906,
+      "step": 19265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002413594573068465,
+      "loss": 4.1437,
+      "step": 19270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002413310339553433,
+      "loss": 4.0656,
+      "step": 19275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002413026053915692,
+      "loss": 4.0328,
+      "step": 19280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024127417161714664,
+      "loss": 4.1281,
+      "step": 19285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024124573263369827,
+      "loss": 3.9953,
+      "step": 19290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002412172884428471,
+      "loss": 4.0422,
+      "step": 19295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002411888390462165,
+      "loss": 4.0719,
+      "step": 19300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024116038444543006,
+      "loss": 4.0625,
+      "step": 19305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002411319246421117,
+      "loss": 4.0156,
+      "step": 19310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002411034596378855,
+      "loss": 4.1047,
+      "step": 19315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024107498943437618,
+      "loss": 3.9953,
+      "step": 19320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024104651403320838,
+      "loss": 4.0203,
+      "step": 19325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024101803343600723,
+      "loss": 4.0687,
+      "step": 19330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002409895476443981,
+      "loss": 4.125,
+      "step": 19335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024096105666000675,
+      "loss": 4.0719,
+      "step": 19340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024093256048445913,
+      "loss": 4.0172,
+      "step": 19345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024090405911938144,
+      "loss": 4.1141,
+      "step": 19350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024087555256640036,
+      "loss": 4.1328,
+      "step": 19355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002408470408271427,
+      "loss": 4.0391,
+      "step": 19360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002408185239032357,
+      "loss": 4.0891,
+      "step": 19365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024079000179630673,
+      "loss": 4.0203,
+      "step": 19370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024076147450798359,
+      "loss": 3.9875,
+      "step": 19375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024073294203989435,
+      "loss": 4.0047,
+      "step": 19380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024070440439366735,
+      "loss": 4.0516,
+      "step": 19385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024067586157093128,
+      "loss": 4.0219,
+      "step": 19390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024064731357331494,
+      "loss": 4.0938,
+      "step": 19395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024061876040244777,
+      "loss": 3.9688,
+      "step": 19400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024059020205995912,
+      "loss": 4.1047,
+      "step": 19405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002405616385474789,
+      "loss": 4.0891,
+      "step": 19410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002405330698666373,
+      "loss": 3.9906,
+      "step": 19415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024050449601906458,
+      "loss": 4.0594,
+      "step": 19420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024047591700639156,
+      "loss": 4.0078,
+      "step": 19425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024044733283024923,
+      "loss": 4.1172,
+      "step": 19430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002404187434922689,
+      "loss": 4.0578,
+      "step": 19435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002403901489940821,
+      "loss": 4.0563,
+      "step": 19440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024036154933732083,
+      "loss": 4.0422,
+      "step": 19445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024033294452361714,
+      "loss": 3.9859,
+      "step": 19450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024030433455460366,
+      "loss": 4.0594,
+      "step": 19455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024027571943191308,
+      "loss": 4.0156,
+      "step": 19460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024024709915717845,
+      "loss": 3.9875,
+      "step": 19465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024021847373203314,
+      "loss": 4.0344,
+      "step": 19470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401898431581109,
+      "loss": 4.1125,
+      "step": 19475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401612074370456,
+      "loss": 4.0766,
+      "step": 19480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401325665704714,
+      "loss": 4.1281,
+      "step": 19485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024010392056002301,
+      "loss": 4.0266,
+      "step": 19490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002400752694073351,
+      "loss": 4.1437,
+      "step": 19495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002400466131140429,
+      "loss": 4.0953,
+      "step": 19500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002400179516817818,
+      "loss": 4.1422,
+      "step": 19505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023998928511218754,
+      "loss": 3.9875,
+      "step": 19510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023996061340689608,
+      "loss": 4.1141,
+      "step": 19515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023993193656754374,
+      "loss": 4.0766,
+      "step": 19520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023990325459576705,
+      "loss": 4.0734,
+      "step": 19525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023987456749320298,
+      "loss": 4.1437,
+      "step": 19530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023984587526148867,
+      "loss": 4.0391,
+      "step": 19535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023981717790226155,
+      "loss": 3.9953,
+      "step": 19540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023978847541715945,
+      "loss": 4.0891,
+      "step": 19545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023975976780782037,
+      "loss": 4.1406,
+      "step": 19550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023973105507588268,
+      "loss": 4.0984,
+      "step": 19555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023970233722298505,
+      "loss": 4.1141,
+      "step": 19560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002396736142507663,
+      "loss": 3.9781,
+      "step": 19565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002396448861608658,
+      "loss": 4.0656,
+      "step": 19570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023961615295492293,
+      "loss": 3.9375,
+      "step": 19575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023958741463457763,
+      "loss": 4.0391,
+      "step": 19580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002395586712014699,
+      "loss": 4.0969,
+      "step": 19585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023952992265724012,
+      "loss": 4.0469,
+      "step": 19590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002395011690035291,
+      "loss": 4.1063,
+      "step": 19595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023947241024197766,
+      "loss": 4.0516,
+      "step": 19600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023944364637422714,
+      "loss": 4.0531,
+      "step": 19605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023941487740191907,
+      "loss": 4.0469,
+      "step": 19610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023938610332669536,
+      "loss": 4.0438,
+      "step": 19615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023935732415019805,
+      "loss": 4.0531,
+      "step": 19620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002393285398740697,
+      "loss": 4.0609,
+      "step": 19625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023929975049995287,
+      "loss": 4.0187,
+      "step": 19630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002392709560294907,
+      "loss": 4.0922,
+      "step": 19635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023924215646432646,
+      "loss": 4.1063,
+      "step": 19640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023921335180610373,
+      "loss": 4.1078,
+      "step": 19645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023918454205646642,
+      "loss": 4.0172,
+      "step": 19650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023915572721705864,
+      "loss": 3.9937,
+      "step": 19655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023912690728952494,
+      "loss": 4.0391,
+      "step": 19660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023909808227551,
+      "loss": 4.1469,
+      "step": 19665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023906925217665893,
+      "loss": 3.9875,
+      "step": 19670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023904041699461707,
+      "loss": 4.0656,
+      "step": 19675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023901157673102995,
+      "loss": 4.0375,
+      "step": 19680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023898273138754357,
+      "loss": 4.0016,
+      "step": 19685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023895388096580416,
+      "loss": 3.9875,
+      "step": 19690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023892502546745816,
+      "loss": 4.1063,
+      "step": 19695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023889616489415237,
+      "loss": 4.1266,
+      "step": 19700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023886729924753385,
+      "loss": 4.025,
+      "step": 19705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023883842852925,
+      "loss": 4.0625,
+      "step": 19710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023880955274094845,
+      "loss": 4.0281,
+      "step": 19715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023878067188427715,
+      "loss": 4.0312,
+      "step": 19720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023875178596088434,
+      "loss": 4.0641,
+      "step": 19725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023872289497241854,
+      "loss": 4.1141,
+      "step": 19730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002386939989205286,
+      "loss": 4.0797,
+      "step": 19735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002386650978068635,
+      "loss": 4.0187,
+      "step": 19740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023863619163307277,
+      "loss": 3.9297,
+      "step": 19745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023860728040080604,
+      "loss": 4.0203,
+      "step": 19750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023857836411171317,
+      "loss": 4.0828,
+      "step": 19755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023854944276744463,
+      "loss": 4.0922,
+      "step": 19760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023852051636965077,
+      "loss": 4.0531,
+      "step": 19765
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023849158491998248,
+      "loss": 4.0844,
+      "step": 19770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023846264842009092,
+      "loss": 3.9531,
+      "step": 19775
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023843370687162747,
+      "loss": 4.0625,
+      "step": 19780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002384047602762439,
+      "loss": 4.0703,
+      "step": 19785
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023837580863559203,
+      "loss": 3.9734,
+      "step": 19790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002383468519513243,
+      "loss": 4.0469,
+      "step": 19795
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023831789022509317,
+      "loss": 4.0859,
+      "step": 19800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023828892345855154,
+      "loss": 4.0484,
+      "step": 19805
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023825995165335247,
+      "loss": 4.0141,
+      "step": 19810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023823097481114945,
+      "loss": 4.0062,
+      "step": 19815
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023820199293359624,
+      "loss": 4.0531,
+      "step": 19820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023817300602234674,
+      "loss": 4.0141,
+      "step": 19825
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023814401407905528,
+      "loss": 4.0187,
+      "step": 19830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023811501710537642,
+      "loss": 4.1219,
+      "step": 19835
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023808601510296506,
+      "loss": 4.0281,
+      "step": 19840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023805700807347628,
+      "loss": 4.075,
+      "step": 19845
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023802799601856556,
+      "loss": 4.0453,
+      "step": 19850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023799897893988863,
+      "loss": 4.0406,
+      "step": 19855
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023796995683910143,
+      "loss": 4.0625,
+      "step": 19860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023794092971786033,
+      "loss": 4.0266,
+      "step": 19865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002379118975778219,
+      "loss": 4.0703,
+      "step": 19870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023788286042064296,
+      "loss": 4.0547,
+      "step": 19875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023785381824798068,
+      "loss": 4.0969,
+      "step": 19880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023782477106149257,
+      "loss": 4.0344,
+      "step": 19885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023779571886283626,
+      "loss": 3.9937,
+      "step": 19890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002377666616536698,
+      "loss": 4.0219,
+      "step": 19895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023773759943565147,
+      "loss": 3.9578,
+      "step": 19900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023770853221043988,
+      "loss": 3.9469,
+      "step": 19905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002376794599796939,
+      "loss": 4.1219,
+      "step": 19910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002376503827450727,
+      "loss": 3.9672,
+      "step": 19915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023762130050823563,
+      "loss": 4.0938,
+      "step": 19920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002375922132708425,
+      "loss": 3.9937,
+      "step": 19925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023756312103455335,
+      "loss": 4.0734,
+      "step": 19930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023753402380102844,
+      "loss": 4.1359,
+      "step": 19935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002375049215719283,
+      "loss": 4.0812,
+      "step": 19940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002374758143489138,
+      "loss": 4.1422,
+      "step": 19945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023744670213364622,
+      "loss": 3.9641,
+      "step": 19950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023741758492778686,
+      "loss": 4.1344,
+      "step": 19955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023738846273299745,
+      "loss": 3.9953,
+      "step": 19960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023735933555094012,
+      "loss": 4.0547,
+      "step": 19965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023733020338327705,
+      "loss": 4.0469,
+      "step": 19970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023730106623167082,
+      "loss": 4.0969,
+      "step": 19975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023727192409778432,
+      "loss": 4.0344,
+      "step": 19980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002372427769832807,
+      "loss": 4.0531,
+      "step": 19985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023721362488982335,
+      "loss": 4.0625,
+      "step": 19990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000237184467819076,
+      "loss": 4.0687,
+      "step": 19995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002371553057727027,
+      "loss": 4.0359,
+      "step": 20000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002371261387523677,
+      "loss": 3.8672,
+      "step": 20005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023709696675973548,
+      "loss": 4.0781,
+      "step": 20010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000237067789796471,
+      "loss": 4.05,
+      "step": 20015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023703860786423937,
+      "loss": 3.9547,
+      "step": 20020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023700942096470593,
+      "loss": 4.0391,
+      "step": 20025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002369802290995365,
+      "loss": 4.0703,
+      "step": 20030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023695103227039697,
+      "loss": 4.0172,
+      "step": 20035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002369218304789536,
+      "loss": 4.0734,
+      "step": 20040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023689262372687304,
+      "loss": 4.0125,
+      "step": 20045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000236863412015822,
+      "loss": 4.0219,
+      "step": 20050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002368341953474677,
+      "loss": 4.0766,
+      "step": 20055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023680497372347746,
+      "loss": 4.1406,
+      "step": 20060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000236775747145519,
+      "loss": 4.0859,
+      "step": 20065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023674651561526029,
+      "loss": 3.9891,
+      "step": 20070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002367172791343695,
+      "loss": 4.0406,
+      "step": 20075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023668803770451524,
+      "loss": 3.9422,
+      "step": 20080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023665879132736631,
+      "loss": 3.9453,
+      "step": 20085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023662954000459188,
+      "loss": 3.9312,
+      "step": 20090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023660028373786112,
+      "loss": 4.1172,
+      "step": 20095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023657102252884387,
+      "loss": 4.0625,
+      "step": 20100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023654175637920997,
+      "loss": 4.0266,
+      "step": 20105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002365124852906297,
+      "loss": 3.9422,
+      "step": 20110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002364832092647735,
+      "loss": 3.9438,
+      "step": 20115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023645392830331228,
+      "loss": 4.0344,
+      "step": 20120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023642464240791698,
+      "loss": 4.1172,
+      "step": 20125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236395351580259,
+      "loss": 3.9531,
+      "step": 20130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023636605582200997,
+      "loss": 4.0172,
+      "step": 20135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023633675513484175,
+      "loss": 4.1875,
+      "step": 20140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023630744952042666,
+      "loss": 4.0172,
+      "step": 20145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023627813898043705,
+      "loss": 4.1594,
+      "step": 20150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023624882351654575,
+      "loss": 4.1047,
+      "step": 20155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023621950313042572,
+      "loss": 4.0375,
+      "step": 20160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023619017782375033,
+      "loss": 4.1406,
+      "step": 20165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023616084759819323,
+      "loss": 4.0281,
+      "step": 20170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023613151245542818,
+      "loss": 4.0594,
+      "step": 20175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023610217239712942,
+      "loss": 3.9937,
+      "step": 20180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023607282742497138,
+      "loss": 4.0234,
+      "step": 20185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023604347754062875,
+      "loss": 3.9781,
+      "step": 20190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023601412274577654,
+      "loss": 4.1125,
+      "step": 20195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023598476304209008,
+      "loss": 4.0781,
+      "step": 20200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023595539843124488,
+      "loss": 4.0328,
+      "step": 20205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023592602891491678,
+      "loss": 4.1359,
+      "step": 20210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023589665449478192,
+      "loss": 4.075,
+      "step": 20215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023586727517251667,
+      "loss": 4.0625,
+      "step": 20220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023583789094979777,
+      "loss": 4.0156,
+      "step": 20225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023580850182830214,
+      "loss": 3.9937,
+      "step": 20230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023577910780970708,
+      "loss": 4.0187,
+      "step": 20235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023574970889568997,
+      "loss": 4.0547,
+      "step": 20240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023572030508792874,
+      "loss": 4.0563,
+      "step": 20245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002356908963881014,
+      "loss": 4.0531,
+      "step": 20250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023566148279788638,
+      "loss": 4.0906,
+      "step": 20255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023563206431896223,
+      "loss": 3.9625,
+      "step": 20260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023560264095300788,
+      "loss": 4.0719,
+      "step": 20265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023557321270170262,
+      "loss": 4.05,
+      "step": 20270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023554377956672577,
+      "loss": 3.9953,
+      "step": 20275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023551434154975724,
+      "loss": 4.0344,
+      "step": 20280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023548489865247692,
+      "loss": 3.9766,
+      "step": 20285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002354554508765652,
+      "loss": 4.0844,
+      "step": 20290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023542599822370266,
+      "loss": 4.0969,
+      "step": 20295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023539654069557015,
+      "loss": 4.0891,
+      "step": 20300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023536707829384883,
+      "loss": 4.1312,
+      "step": 20305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023533761102022008,
+      "loss": 4.0234,
+      "step": 20310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023530813887636564,
+      "loss": 4.0797,
+      "step": 20315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002352786618639675,
+      "loss": 3.9547,
+      "step": 20320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023524917998470792,
+      "loss": 3.9594,
+      "step": 20325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023521969324026933,
+      "loss": 4.0969,
+      "step": 20330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351902016323347,
+      "loss": 4.0828,
+      "step": 20335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023516070516258698,
+      "loss": 4.0344,
+      "step": 20340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023513120383270966,
+      "loss": 4.1312,
+      "step": 20345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023510169764438625,
+      "loss": 4.0484,
+      "step": 20350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023507218659930075,
+      "loss": 4.0625,
+      "step": 20355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023504267069913737,
+      "loss": 4.1125,
+      "step": 20360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002350131499455806,
+      "loss": 4.0734,
+      "step": 20365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023498362434031507,
+      "loss": 4.0391,
+      "step": 20370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023495409388502597,
+      "loss": 4.0531,
+      "step": 20375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023492455858139852,
+      "loss": 4.1125,
+      "step": 20380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023489501843111828,
+      "loss": 3.9969,
+      "step": 20385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023486547343587116,
+      "loss": 4.0156,
+      "step": 20390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023483592359734327,
+      "loss": 4.1156,
+      "step": 20395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023480636891722105,
+      "loss": 4.0719,
+      "step": 20400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002347768093971912,
+      "loss": 4.0031,
+      "step": 20405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023474724503894055,
+      "loss": 4.0984,
+      "step": 20410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023471767584415653,
+      "loss": 4.1141,
+      "step": 20415
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023468810181452658,
+      "loss": 3.9875,
+      "step": 20420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023465852295173842,
+      "loss": 3.9891,
+      "step": 20425
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023462893925748024,
+      "loss": 4.1078,
+      "step": 20430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002345993507334403,
+      "loss": 4.0922,
+      "step": 20435
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023456975738130726,
+      "loss": 4.1094,
+      "step": 20440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023454015920276999,
+      "loss": 3.9937,
+      "step": 20445
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023451055619951768,
+      "loss": 4.1,
+      "step": 20450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023448094837323974,
+      "loss": 3.9813,
+      "step": 20455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023445133572562598,
+      "loss": 4.0422,
+      "step": 20460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344217182583663,
+      "loss": 4.0312,
+      "step": 20465
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023439209597315105,
+      "loss": 4.0156,
+      "step": 20470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023436246887167066,
+      "loss": 4.0297,
+      "step": 20475
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343328369556161,
+      "loss": 3.9969,
+      "step": 20480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023430320022667833,
+      "loss": 4.1312,
+      "step": 20485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023427355868654883,
+      "loss": 3.9891,
+      "step": 20490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002342439123369192,
+      "loss": 4.0422,
+      "step": 20495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023421426117948137,
+      "loss": 4.0078,
+      "step": 20500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002341846052159275,
+      "loss": 4.0312,
+      "step": 20505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002341549444479501,
+      "loss": 4.0187,
+      "step": 20510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002341252788772419,
+      "loss": 4.1,
+      "step": 20515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023409560850549586,
+      "loss": 4.1359,
+      "step": 20520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002340659333344054,
+      "loss": 4.0703,
+      "step": 20525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023403625336566397,
+      "loss": 4.0812,
+      "step": 20530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023400656860096545,
+      "loss": 3.9844,
+      "step": 20535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023397687904200395,
+      "loss": 4.0922,
+      "step": 20540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023394718469047388,
+      "loss": 4.1016,
+      "step": 20545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023391748554806987,
+      "loss": 4.0312,
+      "step": 20550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023388778161648684,
+      "loss": 4.0094,
+      "step": 20555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023385807289742002,
+      "loss": 3.9594,
+      "step": 20560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023382835939256486,
+      "loss": 3.9937,
+      "step": 20565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337986411036172,
+      "loss": 4.125,
+      "step": 20570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023376891803227292,
+      "loss": 4.0953,
+      "step": 20575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023373919018022845,
+      "loss": 4.0953,
+      "step": 20580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023370945754918037,
+      "loss": 4.0906,
+      "step": 20585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002336797201408254,
+      "loss": 3.9641,
+      "step": 20590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002336499779568607,
+      "loss": 4.1281,
+      "step": 20595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002336202309989837,
+      "loss": 4.0781,
+      "step": 20600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023359047926889206,
+      "loss": 3.9922,
+      "step": 20605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023356072276828373,
+      "loss": 4.0125,
+      "step": 20610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023353096149885687,
+      "loss": 4.0328,
+      "step": 20615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335011954623099,
+      "loss": 4.0453,
+      "step": 20620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023347142466034174,
+      "loss": 4.0719,
+      "step": 20625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002334416490946513,
+      "loss": 4.0531,
+      "step": 20630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023341186876693786,
+      "loss": 3.9875,
+      "step": 20635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023338208367890104,
+      "loss": 4.0375,
+      "step": 20640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023335229383224061,
+      "loss": 4.0656,
+      "step": 20645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023332249922865679,
+      "loss": 4.0703,
+      "step": 20650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002332926998698499,
+      "loss": 4.0516,
+      "step": 20655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023326289575752052,
+      "loss": 3.9594,
+      "step": 20660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023323308689336965,
+      "loss": 3.9906,
+      "step": 20665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002332032732790985,
+      "loss": 4.1016,
+      "step": 20670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002331734549164085,
+      "loss": 4.0266,
+      "step": 20675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023314363180700147,
+      "loss": 4.0047,
+      "step": 20680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023311380395257928,
+      "loss": 4.0172,
+      "step": 20685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023308397135484427,
+      "loss": 4.1188,
+      "step": 20690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000233054134015499,
+      "loss": 4.025,
+      "step": 20695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002330242919362463,
+      "loss": 4.0469,
+      "step": 20700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023299444511878925,
+      "loss": 4.0234,
+      "step": 20705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002329645935648312,
+      "loss": 4.0219,
+      "step": 20710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023293473727607576,
+      "loss": 4.0844,
+      "step": 20715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023290487625422692,
+      "loss": 3.9719,
+      "step": 20720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023287501050098868,
+      "loss": 3.9266,
+      "step": 20725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002328451400180657,
+      "loss": 4.0984,
+      "step": 20730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023281526480716254,
+      "loss": 4.0328,
+      "step": 20735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023278538486998425,
+      "loss": 4.0453,
+      "step": 20740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232755500208236,
+      "loss": 4.0453,
+      "step": 20745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002327256108236234,
+      "loss": 4.0219,
+      "step": 20750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326957167178522,
+      "loss": 4.1437,
+      "step": 20755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023266581789262847,
+      "loss": 3.9859,
+      "step": 20760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023263591434965849,
+      "loss": 3.9203,
+      "step": 20765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023260600609064896,
+      "loss": 4.1172,
+      "step": 20770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023257609311730666,
+      "loss": 4.0906,
+      "step": 20775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002325461754313388,
+      "loss": 4.0328,
+      "step": 20780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023251625303445265,
+      "loss": 3.9922,
+      "step": 20785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023248632592835604,
+      "loss": 4.0703,
+      "step": 20790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023245639411475683,
+      "loss": 3.9828,
+      "step": 20795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023242645759536323,
+      "loss": 4.0594,
+      "step": 20800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023239651637188376,
+      "loss": 4.0594,
+      "step": 20805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023236657044602716,
+      "loss": 4.1141,
+      "step": 20810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002323366198195024,
+      "loss": 4.1,
+      "step": 20815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023230666449401883,
+      "loss": 4.0359,
+      "step": 20820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023227670447128592,
+      "loss": 4.0828,
+      "step": 20825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002322467397530136,
+      "loss": 4.1234,
+      "step": 20830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002322167703409119,
+      "loss": 4.0516,
+      "step": 20835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023218679623669114,
+      "loss": 4.0375,
+      "step": 20840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232156817442062,
+      "loss": 4.0797,
+      "step": 20845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023212683395873538,
+      "loss": 4.1469,
+      "step": 20850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320968457884224,
+      "loss": 4.0062,
+      "step": 20855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320668529328345,
+      "loss": 3.9922,
+      "step": 20860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320368553936834,
+      "loss": 3.9797,
+      "step": 20865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232006853172681,
+      "loss": 4.0766,
+      "step": 20870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023197684627153966,
+      "loss": 4.0344,
+      "step": 20875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023194683469197173,
+      "loss": 4.0047,
+      "step": 20880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023191681843569008,
+      "loss": 4.0484,
+      "step": 20885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023188679750440764,
+      "loss": 3.9781,
+      "step": 20890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023185677189983773,
+      "loss": 3.9516,
+      "step": 20895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023182674162369401,
+      "loss": 3.9297,
+      "step": 20900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023179670667769023,
+      "loss": 4.1203,
+      "step": 20905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023176666706354056,
+      "loss": 3.9859,
+      "step": 20910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002317366227829592,
+      "loss": 4.0219,
+      "step": 20915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023170657383766094,
+      "loss": 4.0469,
+      "step": 20920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023167652022936062,
+      "loss": 4.0922,
+      "step": 20925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023164646195977342,
+      "loss": 3.9453,
+      "step": 20930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023161639903061473,
+      "loss": 4.0391,
+      "step": 20935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023158633144360028,
+      "loss": 4.0469,
+      "step": 20940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023155625920044605,
+      "loss": 4.0547,
+      "step": 20945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023152618230286816,
+      "loss": 3.9984,
+      "step": 20950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023149610075258326,
+      "loss": 4.075,
+      "step": 20955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023146601455130796,
+      "loss": 4.0297,
+      "step": 20960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023143592370075938,
+      "loss": 4.0719,
+      "step": 20965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023140582820265477,
+      "loss": 3.9891,
+      "step": 20970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002313757280587117,
+      "loss": 4.0969,
+      "step": 20975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023134562327064794,
+      "loss": 4.0391,
+      "step": 20980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002313155138401816,
+      "loss": 4.0859,
+      "step": 20985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023128539976903108,
+      "loss": 4.0422,
+      "step": 20990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023125528105891492,
+      "loss": 4.0406,
+      "step": 20995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023122515771155204,
+      "loss": 3.975,
+      "step": 21000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023119502972866157,
+      "loss": 4.0219,
+      "step": 21005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311648971119629,
+      "loss": 4.0047,
+      "step": 21010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023113475986317572,
+      "loss": 4.0594,
+      "step": 21015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023110461798401998,
+      "loss": 3.9922,
+      "step": 21020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023107447147621585,
+      "loss": 4.0875,
+      "step": 21025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002310443203414838,
+      "loss": 4.0469,
+      "step": 21030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023101416458154458,
+      "loss": 4.05,
+      "step": 21035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023098400419811914,
+      "loss": 4.0922,
+      "step": 21040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002309538391929288,
+      "loss": 4.0422,
+      "step": 21045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000230923669567695,
+      "loss": 4.0922,
+      "step": 21050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023089349532413961,
+      "loss": 4.0578,
+      "step": 21055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023086331646398462,
+      "loss": 3.9797,
+      "step": 21060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023083313298895235,
+      "loss": 4.0172,
+      "step": 21065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023080294490076536,
+      "loss": 4.0687,
+      "step": 21070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307727522011465,
+      "loss": 4.0281,
+      "step": 21075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023074255489181892,
+      "loss": 4.0547,
+      "step": 21080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023071235297450588,
+      "loss": 4.0703,
+      "step": 21085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023068214645093112,
+      "loss": 3.9406,
+      "step": 21090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023065193532281842,
+      "loss": 4.0828,
+      "step": 21095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023062171959189204,
+      "loss": 3.9594,
+      "step": 21100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023059149925987632,
+      "loss": 4.0672,
+      "step": 21105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023056127432849593,
+      "loss": 4.0828,
+      "step": 21110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023053104479947583,
+      "loss": 4.0734,
+      "step": 21115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023050081067454127,
+      "loss": 4.0438,
+      "step": 21120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023047057195541766,
+      "loss": 3.9906,
+      "step": 21125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023044032864383073,
+      "loss": 4.0656,
+      "step": 21130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023041008074150654,
+      "loss": 4.0375,
+      "step": 21135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002303798282501712,
+      "loss": 4.1359,
+      "step": 21140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023034957117155138,
+      "loss": 4.0578,
+      "step": 21145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002303193095073737,
+      "loss": 4.0969,
+      "step": 21150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302890432593653,
+      "loss": 4.0234,
+      "step": 21155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023025877242925348,
+      "loss": 3.9906,
+      "step": 21160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023022849701876574,
+      "loss": 4.0875,
+      "step": 21165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023019821702962992,
+      "loss": 3.9391,
+      "step": 21170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023016793246357414,
+      "loss": 4.0797,
+      "step": 21175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002301376433223267,
+      "loss": 4.0312,
+      "step": 21180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023010734960761622,
+      "loss": 4.0219,
+      "step": 21185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023007705132117157,
+      "loss": 4.0516,
+      "step": 21190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023004674846472187,
+      "loss": 4.0266,
+      "step": 21195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023001644103999653,
+      "loss": 4.0109,
+      "step": 21200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022998612904872513,
+      "loss": 4.0281,
+      "step": 21205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022995581249263763,
+      "loss": 4.0781,
+      "step": 21210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022992549137346426,
+      "loss": 4.0703,
+      "step": 21215
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022989516569293532,
+      "loss": 4.05,
+      "step": 21220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022986483545278156,
+      "loss": 4.0,
+      "step": 21225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022983450065473395,
+      "loss": 4.0016,
+      "step": 21230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022980416130052368,
+      "loss": 3.9609,
+      "step": 21235
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022977381739188222,
+      "loss": 3.9766,
+      "step": 21240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022974346893054131,
+      "loss": 3.95,
+      "step": 21245
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002297131159182329,
+      "loss": 4.0547,
+      "step": 21250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022968275835668928,
+      "loss": 4.0187,
+      "step": 21255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000229652396247643,
+      "loss": 4.0469,
+      "step": 21260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022962202959282672,
+      "loss": 3.9984,
+      "step": 21265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022959165839397355,
+      "loss": 3.9828,
+      "step": 21270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002295612826528167,
+      "loss": 4.0469,
+      "step": 21275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022953090237108986,
+      "loss": 3.9906,
+      "step": 21280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022950051755052665,
+      "loss": 4.0469,
+      "step": 21285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002294701281928613,
+      "loss": 3.9766,
+      "step": 21290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022943973429982804,
+      "loss": 4.0516,
+      "step": 21295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022940933587316145,
+      "loss": 4.0438,
+      "step": 21300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022937893291459646,
+      "loss": 4.0484,
+      "step": 21305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022934852542586803,
+      "loss": 4.075,
+      "step": 21310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022931811340871165,
+      "loss": 4.0359,
+      "step": 21315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022928769686486283,
+      "loss": 4.0297,
+      "step": 21320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022925727579605756,
+      "loss": 3.9094,
+      "step": 21325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022922685020403187,
+      "loss": 4.0797,
+      "step": 21330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002291964200905222,
+      "loss": 4.1188,
+      "step": 21335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022916598545726517,
+      "loss": 4.0516,
+      "step": 21340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022913554630599772,
+      "loss": 4.0062,
+      "step": 21345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022910510263845702,
+      "loss": 3.9844,
+      "step": 21350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022907465445638046,
+      "loss": 4.0125,
+      "step": 21355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022904420176150572,
+      "loss": 4.0406,
+      "step": 21360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002290137445555708,
+      "loss": 4.0594,
+      "step": 21365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002289832828403138,
+      "loss": 4.0344,
+      "step": 21370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022895281661747325,
+      "loss": 4.1531,
+      "step": 21375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022892234588878783,
+      "loss": 4.0406,
+      "step": 21380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022889187065599652,
+      "loss": 4.0578,
+      "step": 21385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022886139092083852,
+      "loss": 4.0656,
+      "step": 21390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022883090668505336,
+      "loss": 4.0297,
+      "step": 21395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022880041795038074,
+      "loss": 4.0625,
+      "step": 21400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022876992471856063,
+      "loss": 4.0453,
+      "step": 21405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022873942699133332,
+      "loss": 4.0328,
+      "step": 21410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022870892477043936,
+      "loss": 4.1,
+      "step": 21415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022867841805761943,
+      "loss": 4.025,
+      "step": 21420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022864790685461455,
+      "loss": 4.0453,
+      "step": 21425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002286173911631661,
+      "loss": 4.0547,
+      "step": 21430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022858687098501547,
+      "loss": 4.0922,
+      "step": 21435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022855634632190457,
+      "loss": 4.0141,
+      "step": 21440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022852581717557543,
+      "loss": 4.025,
+      "step": 21445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022849528354777028,
+      "loss": 4.0797,
+      "step": 21450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022846474544023173,
+      "loss": 3.9813,
+      "step": 21455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002284342028547026,
+      "loss": 4.0328,
+      "step": 21460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022840365579292592,
+      "loss": 4.0922,
+      "step": 21465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022837310425664507,
+      "loss": 4.0656,
+      "step": 21470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022834254824760357,
+      "loss": 4.0578,
+      "step": 21475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022831198776754525,
+      "loss": 4.0484,
+      "step": 21480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022828142281821427,
+      "loss": 4.0375,
+      "step": 21485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022825085340135495,
+      "loss": 4.0359,
+      "step": 21490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022822027951871184,
+      "loss": 3.9391,
+      "step": 21495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022818970117202981,
+      "loss": 4.0141,
+      "step": 21500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022815911836305405,
+      "loss": 3.9516,
+      "step": 21505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022812853109352977,
+      "loss": 4.0797,
+      "step": 21510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022809793936520278,
+      "loss": 4.1063,
+      "step": 21515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022806734317981877,
+      "loss": 4.0891,
+      "step": 21520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022803674253912398,
+      "loss": 4.0187,
+      "step": 21525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002280061374448648,
+      "loss": 4.0187,
+      "step": 21530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022797552789878772,
+      "loss": 4.0359,
+      "step": 21535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022794491390263984,
+      "loss": 4.0109,
+      "step": 21540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022791429545816814,
+      "loss": 4.1,
+      "step": 21545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022788367256712012,
+      "loss": 3.9359,
+      "step": 21550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022785304523124337,
+      "loss": 4.0828,
+      "step": 21555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002278224134522858,
+      "loss": 3.9766,
+      "step": 21560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002277917772319956,
+      "loss": 4.0797,
+      "step": 21565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022776113657212112,
+      "loss": 3.9875,
+      "step": 21570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022773049147441113,
+      "loss": 4.0391,
+      "step": 21575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022769984194061444,
+      "loss": 4.0438,
+      "step": 21580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022766918797248033,
+      "loss": 4.0609,
+      "step": 21585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002276385295717581,
+      "loss": 4.0531,
+      "step": 21590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002276078667401975,
+      "loss": 4.0641,
+      "step": 21595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275771994795485,
+      "loss": 4.0359,
+      "step": 21600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275465277915612,
+      "loss": 4.0687,
+      "step": 21605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022751585167798608,
+      "loss": 4.0484,
+      "step": 21610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274851711405738,
+      "loss": 4.0422,
+      "step": 21615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022745448618107533,
+      "loss": 4.0234,
+      "step": 21620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274237968012419,
+      "loss": 3.975,
+      "step": 21625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022739310300282488,
+      "loss": 4.0547,
+      "step": 21630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022736240478757599,
+      "loss": 3.9891,
+      "step": 21635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022733170215724724,
+      "loss": 4.0219,
+      "step": 21640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022730099511359073,
+      "loss": 3.9781,
+      "step": 21645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000227270283658359,
+      "loss": 4.0031,
+      "step": 21650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002272395677933047,
+      "loss": 4.0969,
+      "step": 21655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022720884752018084,
+      "loss": 4.0734,
+      "step": 21660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022717812284074056,
+      "loss": 4.0484,
+      "step": 21665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002271473937567374,
+      "loss": 4.0266,
+      "step": 21670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022711666026992502,
+      "loss": 4.0266,
+      "step": 21675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002270859223820574,
+      "loss": 4.025,
+      "step": 21680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022705518009488872,
+      "loss": 4.0641,
+      "step": 21685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022702443341017346,
+      "loss": 4.0703,
+      "step": 21690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022699368232966641,
+      "loss": 4.0344,
+      "step": 21695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022696292685512242,
+      "loss": 4.0516,
+      "step": 21700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022693216698829678,
+      "loss": 4.0734,
+      "step": 21705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022690140273094498,
+      "loss": 4.0719,
+      "step": 21710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022687063408482268,
+      "loss": 4.0828,
+      "step": 21715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022683986105168585,
+      "loss": 3.9859,
+      "step": 21720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022680908363329077,
+      "loss": 4.0375,
+      "step": 21725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022677830183139385,
+      "loss": 4.0312,
+      "step": 21730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022674751564775184,
+      "loss": 4.0984,
+      "step": 21735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022671672508412172,
+      "loss": 4.0938,
+      "step": 21740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022668593014226065,
+      "loss": 4.1078,
+      "step": 21745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022665513082392616,
+      "loss": 4.0609,
+      "step": 21750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022662432713087594,
+      "loss": 3.9859,
+      "step": 21755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022659351906486802,
+      "loss": 4.0484,
+      "step": 21760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022656270662766055,
+      "loss": 3.9937,
+      "step": 21765
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022653188982101202,
+      "loss": 4.0406,
+      "step": 21770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022650106864668116,
+      "loss": 4.0359,
+      "step": 21775
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022647024310642694,
+      "loss": 3.9922,
+      "step": 21780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022643941320200857,
+      "loss": 3.9906,
+      "step": 21785
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022640857893518543,
+      "loss": 3.9609,
+      "step": 21790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022637774030771733,
+      "loss": 4.0344,
+      "step": 21795
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022634689732136428,
+      "loss": 4.0391,
+      "step": 21800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022631604997788645,
+      "loss": 4.0453,
+      "step": 21805
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022628519827904423,
+      "loss": 4.0766,
+      "step": 21810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022625434222659836,
+      "loss": 4.0078,
+      "step": 21815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022622348182230984,
+      "loss": 4.0422,
+      "step": 21820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022619261706793988,
+      "loss": 4.0234,
+      "step": 21825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002261617479652499,
+      "loss": 4.0281,
+      "step": 21830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022613087451600156,
+      "loss": 4.1266,
+      "step": 21835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002260999967219569,
+      "loss": 4.0141,
+      "step": 21840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022606911458487812,
+      "loss": 4.0266,
+      "step": 21845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022603822810652758,
+      "loss": 4.0172,
+      "step": 21850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022600733728866802,
+      "loss": 4.0203,
+      "step": 21855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002259764421330624,
+      "loss": 4.0016,
+      "step": 21860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022594554264147388,
+      "loss": 3.9875,
+      "step": 21865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002259146388156659,
+      "loss": 4.1156,
+      "step": 21870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002258837306574022,
+      "loss": 4.0938,
+      "step": 21875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002258528181684466,
+      "loss": 4.125,
+      "step": 21880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022582190135056338,
+      "loss": 3.9969,
+      "step": 21885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022579098020551698,
+      "loss": 3.9922,
+      "step": 21890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022576005473507198,
+      "loss": 4.0109,
+      "step": 21895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022572912494099332,
+      "loss": 4.0938,
+      "step": 21900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002256981908250462,
+      "loss": 4.0062,
+      "step": 21905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022566725238899605,
+      "loss": 4.0375,
+      "step": 21910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002256363096346085,
+      "loss": 4.0766,
+      "step": 21915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022560536256364944,
+      "loss": 4.0766,
+      "step": 21920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000225574411177885,
+      "loss": 4.0156,
+      "step": 21925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002255434554790817,
+      "loss": 4.0172,
+      "step": 21930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022551249546900605,
+      "loss": 4.0031,
+      "step": 21935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000225481531149425,
+      "loss": 4.0391,
+      "step": 21940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022545056252210572,
+      "loss": 4.0359,
+      "step": 21945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002254195895888155,
+      "loss": 3.9922,
+      "step": 21950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022538861235132208,
+      "loss": 4.0875,
+      "step": 21955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022535763081139325,
+      "loss": 4.0766,
+      "step": 21960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022532664497079715,
+      "loss": 4.1156,
+      "step": 21965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022529565483130214,
+      "loss": 4.0406,
+      "step": 21970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022526466039467683,
+      "loss": 4.0969,
+      "step": 21975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002252336616626901,
+      "loss": 4.0203,
+      "step": 21980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022520265863711103,
+      "loss": 4.0062,
+      "step": 21985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000225171651319709,
+      "loss": 3.9594,
+      "step": 21990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022514063971225352,
+      "loss": 4.1219,
+      "step": 21995
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022510962381651456,
+      "loss": 4.0297,
+      "step": 22000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022507860363426207,
+      "loss": 4.0359,
+      "step": 22005
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022504757916726642,
+      "loss": 4.1078,
+      "step": 22010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002250165504172982,
+      "loss": 4.0156,
+      "step": 22015
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022498551738612816,
+      "loss": 4.0438,
+      "step": 22020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022495448007552748,
+      "loss": 4.0312,
+      "step": 22025
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002249234384872673,
+      "loss": 4.0172,
+      "step": 22030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022489239262311935,
+      "loss": 4.0844,
+      "step": 22035
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022486134248485526,
+      "loss": 3.9937,
+      "step": 22040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022483028807424713,
+      "loss": 4.0312,
+      "step": 22045
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022479922939306723,
+      "loss": 4.0141,
+      "step": 22050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022476816644308813,
+      "loss": 3.9359,
+      "step": 22055
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002247370992260825,
+      "loss": 3.9625,
+      "step": 22060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022470602774382343,
+      "loss": 4.0266,
+      "step": 22065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022467495199808416,
+      "loss": 3.9094,
+      "step": 22070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022464387199063813,
+      "loss": 4.0625,
+      "step": 22075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002246127877232592,
+      "loss": 4.0203,
+      "step": 22080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245816991977212,
+      "loss": 4.0812,
+      "step": 22085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022455060641579842,
+      "loss": 4.0219,
+      "step": 22090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245195093792654,
+      "loss": 4.0516,
+      "step": 22095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002244884080898967,
+      "loss": 4.0781,
+      "step": 22100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022445730254946742,
+      "loss": 3.9922,
+      "step": 22105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022442619275975267,
+      "loss": 4.0891,
+      "step": 22110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002243950787225279,
+      "loss": 4.0547,
+      "step": 22115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022436396043956888,
+      "loss": 3.975,
+      "step": 22120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002243328379126514,
+      "loss": 3.9859,
+      "step": 22125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022430171114355167,
+      "loss": 3.9641,
+      "step": 22130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022427058013404615,
+      "loss": 4.0594,
+      "step": 22135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022423944488591144,
+      "loss": 4.0734,
+      "step": 22140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022420830540092446,
+      "loss": 4.05,
+      "step": 22145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022417716168086232,
+      "loss": 3.9641,
+      "step": 22150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002241460137275024,
+      "loss": 3.9906,
+      "step": 22155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022411486154262232,
+      "loss": 4.0938,
+      "step": 22160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022408370512799988,
+      "loss": 4.0281,
+      "step": 22165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022405254448541333,
+      "loss": 3.9828,
+      "step": 22170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022402137961664089,
+      "loss": 4.0766,
+      "step": 22175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022399021052346116,
+      "loss": 4.0547,
+      "step": 22180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022395903720765295,
+      "loss": 4.0656,
+      "step": 22185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002239278596709954,
+      "loss": 4.0734,
+      "step": 22190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002238966779152677,
+      "loss": 3.9547,
+      "step": 22195
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022386549194224948,
+      "loss": 4.1375,
+      "step": 22200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022383430175372052,
+      "loss": 4.0797,
+      "step": 22205
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022380310735146082,
+      "loss": 3.9391,
+      "step": 22210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022377190873725067,
+      "loss": 4.025,
+      "step": 22215
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022374070591287055,
+      "loss": 4.0297,
+      "step": 22220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022370949888010123,
+      "loss": 3.9781,
+      "step": 22225
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002236782876407237,
+      "loss": 4.0672,
+      "step": 22230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002236470721965192,
+      "loss": 3.9953,
+      "step": 22235
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022361585254926914,
+      "loss": 4.0234,
+      "step": 22240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022358462870075528,
+      "loss": 3.9797,
+      "step": 22245
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235534006527596,
+      "loss": 4.0156,
+      "step": 22250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022352216840706418,
+      "loss": 4.0203,
+      "step": 22255
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022349093196545155,
+      "loss": 4.0469,
+      "step": 22260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022345969132970436,
+      "loss": 4.0141,
+      "step": 22265
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022342844650160545,
+      "loss": 3.9578,
+      "step": 22270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022339719748293803,
+      "loss": 3.9906,
+      "step": 22275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022336594427548547,
+      "loss": 4.0672,
+      "step": 22280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022333468688103138,
+      "loss": 4.0187,
+      "step": 22285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022330342530135961,
+      "loss": 4.0375,
+      "step": 22290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022327215953825432,
+      "loss": 3.9547,
+      "step": 22295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022324088959349976,
+      "loss": 4.0859,
+      "step": 22300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022320961546888062,
+      "loss": 4.0547,
+      "step": 22305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022317833716618162,
+      "loss": 4.1031,
+      "step": 22310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022314705468718787,
+      "loss": 3.9984,
+      "step": 22315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022311576803368466,
+      "loss": 4.0,
+      "step": 22320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002230844772074575,
+      "loss": 4.0672,
+      "step": 22325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022305318221029212,
+      "loss": 4.0516,
+      "step": 22330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022302188304397464,
+      "loss": 4.0672,
+      "step": 22335
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022299057971029125,
+      "loss": 4.0563,
+      "step": 22340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022295927221102838,
+      "loss": 3.9906,
+      "step": 22345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022292796054797286,
+      "loss": 4.0656,
+      "step": 22350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022289664472291158,
+      "loss": 4.0469,
+      "step": 22355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002228653247376317,
+      "loss": 4.0656,
+      "step": 22360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022283400059392073,
+      "loss": 4.1188,
+      "step": 22365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022280267229356633,
+      "loss": 4.0656,
+      "step": 22370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002227713398383564,
+      "loss": 4.0312,
+      "step": 22375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022274000323007904,
+      "loss": 4.1531,
+      "step": 22380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022270866247052273,
+      "loss": 3.9484,
+      "step": 22385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022267731756147596,
+      "loss": 4.0422,
+      "step": 22390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022264596850472774,
+      "loss": 4.0891,
+      "step": 22395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022261461530206702,
+      "loss": 3.9984,
+      "step": 22400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022258325795528327,
+      "loss": 3.9297,
+      "step": 22405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022255189646616593,
+      "loss": 4.0938,
+      "step": 22410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022252053083650488,
+      "loss": 4.0094,
+      "step": 22415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022248916106809008,
+      "loss": 3.9953,
+      "step": 22420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002224577871627119,
+      "loss": 4.0422,
+      "step": 22425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022242640912216083,
+      "loss": 4.0281,
+      "step": 22430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022239502694822755,
+      "loss": 4.0078,
+      "step": 22435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022236364064270314,
+      "loss": 4.1422,
+      "step": 22440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022233225020737877,
+      "loss": 3.9766,
+      "step": 22445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002223008556440459,
+      "loss": 4.0453,
+      "step": 22450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002222694569544962,
+      "loss": 4.0109,
+      "step": 22455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002222380541405216,
+      "loss": 4.0156,
+      "step": 22460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002222066472039143,
+      "loss": 4.0328,
+      "step": 22465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022217523614646663,
+      "loss": 4.0141,
+      "step": 22470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022214382096997134,
+      "loss": 4.0516,
+      "step": 22475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002221124016762212,
+      "loss": 4.0328,
+      "step": 22480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022208097826700933,
+      "loss": 4.0359,
+      "step": 22485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022204955074412905,
+      "loss": 4.0859,
+      "step": 22490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000222018119109374,
+      "loss": 4.0078,
+      "step": 22495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002219866833645379,
+      "loss": 4.0031,
+      "step": 22500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022195524351141486,
+      "loss": 4.0578,
+      "step": 22505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022192379955179907,
+      "loss": 4.0875,
+      "step": 22510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022189235148748512,
+      "loss": 3.9984,
+      "step": 22515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022186089932026777,
+      "loss": 4.0172,
+      "step": 22520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022182944305194192,
+      "loss": 4.0672,
+      "step": 22525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022179798268430289,
+      "loss": 4.0984,
+      "step": 22530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022176651821914598,
+      "loss": 4.0406,
+      "step": 22535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022173504965826698,
+      "loss": 4.0609,
+      "step": 22540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022170357700346174,
+      "loss": 3.9828,
+      "step": 22545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022167210025652648,
+      "loss": 4.025,
+      "step": 22550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216406194192575,
+      "loss": 4.1016,
+      "step": 22555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216091344934515,
+      "loss": 4.0734,
+      "step": 22560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022157764548090525,
+      "loss": 4.0672,
+      "step": 22565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022154615238341585,
+      "loss": 4.0891,
+      "step": 22570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002215146552027807,
+      "loss": 4.0766,
+      "step": 22575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022148315394079717,
+      "loss": 3.9844,
+      "step": 22580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002214516485992632,
+      "loss": 4.0609,
+      "step": 22585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022142013917997677,
+      "loss": 4.0938,
+      "step": 22590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022138862568473605,
+      "loss": 4.0031,
+      "step": 22595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002213571081153396,
+      "loss": 4.0687,
+      "step": 22600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002213255864735861,
+      "loss": 3.9547,
+      "step": 22605
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022129406076127448,
+      "loss": 4.0219,
+      "step": 22610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022126253098020393,
+      "loss": 3.9969,
+      "step": 22615
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002212309971321739,
+      "loss": 4.0406,
+      "step": 22620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022119945921898393,
+      "loss": 4.05,
+      "step": 22625
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221167917242434,
+      "loss": 4.0578,
+      "step": 22630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022113637120432412,
+      "loss": 4.0375,
+      "step": 22635
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022110482110645474,
+      "loss": 4.0641,
+      "step": 22640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022107326695062628,
+      "loss": 4.0953,
+      "step": 22645
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002210417087386396,
+      "loss": 3.9406,
+      "step": 22650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002210101464722958,
+      "loss": 4.0203,
+      "step": 22655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022097858015339606,
+      "loss": 4.0141,
+      "step": 22660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022094700978374197,
+      "loss": 4.0359,
+      "step": 22665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022091543536513508,
+      "loss": 4.05,
+      "step": 22670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002208838568993775,
+      "loss": 4.0484,
+      "step": 22675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022085227438827136,
+      "loss": 4.0219,
+      "step": 22680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002208206878336191,
+      "loss": 4.0359,
+      "step": 22685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002207890972372233,
+      "loss": 4.075,
+      "step": 22690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002207575026008869,
+      "loss": 4.075,
+      "step": 22695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000220725903926413,
+      "loss": 4.0219,
+      "step": 22700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022069430121560493,
+      "loss": 4.0281,
+      "step": 22705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002206626944702663,
+      "loss": 4.0828,
+      "step": 22710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022063108369220087,
+      "loss": 4.0734,
+      "step": 22715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022059946888321265,
+      "loss": 3.9281,
+      "step": 22720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022056785004510596,
+      "loss": 4.0125,
+      "step": 22725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022053622717968524,
+      "loss": 4.1078,
+      "step": 22730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002205046002887552,
+      "loss": 4.0281,
+      "step": 22735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002204729693741208,
+      "loss": 4.0062,
+      "step": 22740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002204413344375873,
+      "loss": 4.1109,
+      "step": 22745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022040969548096,
+      "loss": 4.0016,
+      "step": 22750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022037805250604462,
+      "loss": 3.9672,
+      "step": 22755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022034640551464693,
+      "loss": 4.0547,
+      "step": 22760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022031475450857315,
+      "loss": 4.0656,
+      "step": 22765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022028309948962953,
+      "loss": 4.0344,
+      "step": 22770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202514404596226,
+      "loss": 3.9578,
+      "step": 22775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202197774203592,
+      "loss": 3.9453,
+      "step": 22780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022018811037364633,
+      "loss": 4.0594,
+      "step": 22785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022015643932129123,
+      "loss": 4.0766,
+      "step": 22790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022012476426510135,
+      "loss": 4.0187,
+      "step": 22795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022009308520688442,
+      "loss": 4.0203,
+      "step": 22800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022006140214844835,
+      "loss": 3.9266,
+      "step": 22805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022002971509160128,
+      "loss": 4.0344,
+      "step": 22810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021999802403815163,
+      "loss": 4.05,
+      "step": 22815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021996632898990795,
+      "loss": 4.0016,
+      "step": 22820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021993462994867916,
+      "loss": 4.0281,
+      "step": 22825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021990292691627427,
+      "loss": 4.0797,
+      "step": 22830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021987121989450263,
+      "loss": 3.9937,
+      "step": 22835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021983950888517369,
+      "loss": 4.0938,
+      "step": 22840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198077938900972,
+      "loss": 4.0812,
+      "step": 22845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021977607491108325,
+      "loss": 4.0125,
+      "step": 22850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021974435194994192,
+      "loss": 4.0578,
+      "step": 22855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021971262500848372,
+      "loss": 4.0062,
+      "step": 22860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021968089408851924,
+      "loss": 3.9719,
+      "step": 22865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021964915919185947,
+      "loss": 4.0375,
+      "step": 22870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021961742032031544,
+      "loss": 4.0687,
+      "step": 22875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021958567747569848,
+      "loss": 4.1453,
+      "step": 22880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021955393065982022,
+      "loss": 4.0375,
+      "step": 22885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002195221798744924,
+      "loss": 4.0531,
+      "step": 22890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021949042512152707,
+      "loss": 3.9891,
+      "step": 22895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002194586664027365,
+      "loss": 3.9531,
+      "step": 22900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021942690371993308,
+      "loss": 3.9813,
+      "step": 22905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021939513707492964,
+      "loss": 4.1078,
+      "step": 22910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021936336646953894,
+      "loss": 3.9547,
+      "step": 22915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021933159190557427,
+      "loss": 4.0141,
+      "step": 22920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021929981338484895,
+      "loss": 4.0469,
+      "step": 22925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002192680309091766,
+      "loss": 4.0359,
+      "step": 22930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021923624448037105,
+      "loss": 4.0109,
+      "step": 22935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021920445410024628,
+      "loss": 4.0828,
+      "step": 22940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191726597706167,
+      "loss": 3.9703,
+      "step": 22945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021914086149329672,
+      "loss": 4.0328,
+      "step": 22950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191090592701011,
+      "loss": 4.0328,
+      "step": 22955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021907725310284484,
+      "loss": 3.925,
+      "step": 22960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000219045442993343,
+      "loss": 3.9547,
+      "step": 22965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002190136289434111,
+      "loss": 4.0016,
+      "step": 22970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021898181095486474,
+      "loss": 4.0109,
+      "step": 22975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002189499890295198,
+      "loss": 4.0266,
+      "step": 22980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021891816316919226,
+      "loss": 4.0734,
+      "step": 22985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021888633337569858,
+      "loss": 4.0156,
+      "step": 22990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021885449965085513,
+      "loss": 4.0187,
+      "step": 22995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021882266199647877,
+      "loss": 4.0375,
+      "step": 23000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021879082041438645,
+      "loss": 3.925,
+      "step": 23005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021875897490639533,
+      "loss": 4.0187,
+      "step": 23010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021872712547432296,
+      "loss": 3.9719,
+      "step": 23015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021869527211998681,
+      "loss": 4.0156,
+      "step": 23020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021866341484520492,
+      "loss": 4.0797,
+      "step": 23025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021863155365179526,
+      "loss": 4.0141,
+      "step": 23030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021859968854157628,
+      "loss": 4.0016,
+      "step": 23035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021856781951636636,
+      "loss": 3.9813,
+      "step": 23040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021853594657798446,
+      "loss": 4.0328,
+      "step": 23045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021850406972824939,
+      "loss": 4.0625,
+      "step": 23050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021847218896898046,
+      "loss": 4.0516,
+      "step": 23055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021844030430199715,
+      "loss": 3.9703,
+      "step": 23060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218408415729119,
+      "loss": 4.05,
+      "step": 23065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218376523252166,
+      "loss": 4.0953,
+      "step": 23070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021834462687295823,
+      "loss": 3.9344,
+      "step": 23075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021831272659331597,
+      "loss": 4.0828,
+      "step": 23080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021828082241505983,
+      "loss": 3.9984,
+      "step": 23085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021824891434001054,
+      "loss": 4.0578,
+      "step": 23090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021821700236998915,
+      "loss": 3.9953,
+      "step": 23095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021818508650681685,
+      "loss": 4.0203,
+      "step": 23100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021815316675231504,
+      "loss": 4.0297,
+      "step": 23105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021812124310830545,
+      "loss": 3.9047,
+      "step": 23110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021808931557660993,
+      "loss": 3.9859,
+      "step": 23115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021805738415905064,
+      "loss": 3.9891,
+      "step": 23120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021802544885744985,
+      "loss": 3.9719,
+      "step": 23125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002179935096736301,
+      "loss": 4.0453,
+      "step": 23130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021796156660941422,
+      "loss": 4.0125,
+      "step": 23135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021792961966662512,
+      "loss": 4.0141,
+      "step": 23140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021789766884708613,
+      "loss": 3.9781,
+      "step": 23145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021786571415262061,
+      "loss": 4.075,
+      "step": 23150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021783375558505222,
+      "loss": 4.0031,
+      "step": 23155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021780179314620485,
+      "loss": 4.0219,
+      "step": 23160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021776982683790257,
+      "loss": 4.0187,
+      "step": 23165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177378566619698,
+      "loss": 3.9875,
+      "step": 23170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021770588262023097,
+      "loss": 4.0563,
+      "step": 23175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021767390471451092,
+      "loss": 4.0516,
+      "step": 23180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021764192294663456,
+      "loss": 4.0266,
+      "step": 23185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002176099373184271,
+      "loss": 4.0406,
+      "step": 23190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021757794783171402,
+      "loss": 4.0484,
+      "step": 23195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021754595448832096,
+      "loss": 4.0,
+      "step": 23200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021751395729007372,
+      "loss": 4.0438,
+      "step": 23205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021748195623879843,
+      "loss": 4.0656,
+      "step": 23210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002174499513363214,
+      "loss": 3.9688,
+      "step": 23215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002174179425844691,
+      "loss": 4.0641,
+      "step": 23220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021738592998506837,
+      "loss": 4.0078,
+      "step": 23225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021735391353994607,
+      "loss": 4.1047,
+      "step": 23230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002173218932509294,
+      "loss": 3.9984,
+      "step": 23235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021728986911984582,
+      "loss": 4.0297,
+      "step": 23240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021725784114852293,
+      "loss": 4.0172,
+      "step": 23245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021722580933878851,
+      "loss": 4.075,
+      "step": 23250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021719377369247065,
+      "loss": 3.9531,
+      "step": 23255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002171617342113977,
+      "loss": 4.1063,
+      "step": 23260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021712969089739806,
+      "loss": 4.0859,
+      "step": 23265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002170976437523005,
+      "loss": 4.0859,
+      "step": 23270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002170655927779339,
+      "loss": 4.0734,
+      "step": 23275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021703353797612745,
+      "loss": 4.0328,
+      "step": 23280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021700147934871052,
+      "loss": 4.0469,
+      "step": 23285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002169694168975127,
+      "loss": 4.0109,
+      "step": 23290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021693735062436383,
+      "loss": 4.0328,
+      "step": 23295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021690528053109386,
+      "loss": 4.0391,
+      "step": 23300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021687320661953308,
+      "loss": 4.0219,
+      "step": 23305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021684112889151194,
+      "loss": 3.9547,
+      "step": 23310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021680904734886115,
+      "loss": 4.0156,
+      "step": 23315
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002167769619934115,
+      "loss": 4.0375,
+      "step": 23320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021674487282699424,
+      "loss": 4.0422,
+      "step": 23325
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021671277985144068,
+      "loss": 3.9844,
+      "step": 23330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021668068306858227,
+      "loss": 3.9875,
+      "step": 23335
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021664858248025086,
+      "loss": 4.0391,
+      "step": 23340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002166164780882784,
+      "loss": 4.0625,
+      "step": 23345
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021658436989449714,
+      "loss": 4.0766,
+      "step": 23350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021655225790073946,
+      "loss": 4.0234,
+      "step": 23355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021652014210883796,
+      "loss": 4.0906,
+      "step": 23360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021648802252062553,
+      "loss": 4.0141,
+      "step": 23365
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021645589913793525,
+      "loss": 4.0328,
+      "step": 23370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021642377196260044,
+      "loss": 4.0078,
+      "step": 23375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021639164099645445,
+      "loss": 4.0297,
+      "step": 23380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021635950624133117,
+      "loss": 3.875,
+      "step": 23385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163273676990645,
+      "loss": 4.0609,
+      "step": 23390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002162952253714885,
+      "loss": 4.0984,
+      "step": 23395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021626307926043758,
+      "loss": 4.1437,
+      "step": 23400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021623092936774637,
+      "loss": 3.9281,
+      "step": 23405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021619877569524956,
+      "loss": 3.9953,
+      "step": 23410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021616661824478228,
+      "loss": 4.0125,
+      "step": 23415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021613445701817972,
+      "loss": 4.0172,
+      "step": 23420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021610229201727727,
+      "loss": 3.9625,
+      "step": 23425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021607012324391067,
+      "loss": 4.0906,
+      "step": 23430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021603795069991574,
+      "loss": 3.9219,
+      "step": 23435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021600577438712863,
+      "loss": 3.9344,
+      "step": 23440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021597359430738556,
+      "loss": 4.0594,
+      "step": 23445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002159414104625231,
+      "loss": 3.975,
+      "step": 23450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021590922285437799,
+      "loss": 4.0156,
+      "step": 23455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021587703148478714,
+      "loss": 3.9906,
+      "step": 23460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021584483635558782,
+      "loss": 4.0203,
+      "step": 23465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021581263746861724,
+      "loss": 4.0922,
+      "step": 23470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021578043482571317,
+      "loss": 4.0094,
+      "step": 23475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021574822842871326,
+      "loss": 4.0172,
+      "step": 23480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021571601827945566,
+      "loss": 3.9734,
+      "step": 23485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156838043797785,
+      "loss": 3.9297,
+      "step": 23490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021565158673152036,
+      "loss": 4.1453,
+      "step": 23495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021561936533651978,
+      "loss": 4.0031,
+      "step": 23500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021558714019661565,
+      "loss": 4.0391,
+      "step": 23505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021555491131364715,
+      "loss": 4.0422,
+      "step": 23510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021552267868945354,
+      "loss": 4.0953,
+      "step": 23515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021549044232587432,
+      "loss": 4.1094,
+      "step": 23520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021545820222474923,
+      "loss": 4.0391,
+      "step": 23525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021542595838791823,
+      "loss": 4.0266,
+      "step": 23530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021539371081722144,
+      "loss": 3.9891,
+      "step": 23535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153614595144993,
+      "loss": 4.0687,
+      "step": 23540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153292044815923,
+      "loss": 3.9984,
+      "step": 23545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021529694572034133,
+      "loss": 4.0859,
+      "step": 23550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152646832325874,
+      "loss": 4.0094,
+      "step": 23555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021523241702017164,
+      "loss": 4.0594,
+      "step": 23560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021520014708493557,
+      "loss": 4.0797,
+      "step": 23565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021516787342872078,
+      "loss": 4.0234,
+      "step": 23570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021513559605336924,
+      "loss": 4.0469,
+      "step": 23575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021510331496072286,
+      "loss": 4.0672,
+      "step": 23580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021507103015262402,
+      "loss": 4.0172,
+      "step": 23585
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021503874163091529,
+      "loss": 3.9859,
+      "step": 23590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002150064493974392,
+      "loss": 4.0062,
+      "step": 23595
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149741534540388,
+      "loss": 4.025,
+      "step": 23600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149418538025572,
+      "loss": 4.0984,
+      "step": 23605
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021490955044483774,
+      "loss": 4.0328,
+      "step": 23610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021487724338272392,
+      "loss": 4.1,
+      "step": 23615
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021484493261805963,
+      "loss": 3.9875,
+      "step": 23620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021481261815268875,
+      "loss": 3.9641,
+      "step": 23625
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002147802999884555,
+      "loss": 4.0406,
+      "step": 23630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021474797812720425,
+      "loss": 4.0547,
+      "step": 23635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021471565257077962,
+      "loss": 3.9969,
+      "step": 23640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021468332332102652,
+      "loss": 4.0344,
+      "step": 23645
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021465099037978986,
+      "loss": 4.0187,
+      "step": 23650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021461865374891497,
+      "loss": 3.9531,
+      "step": 23655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021458631343024726,
+      "loss": 4.0062,
+      "step": 23660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021455396942563246,
+      "loss": 4.0359,
+      "step": 23665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002145216217369163,
+      "loss": 3.9875,
+      "step": 23670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021448927036594504,
+      "loss": 4.0234,
+      "step": 23675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021445691531456483,
+      "loss": 4.0703,
+      "step": 23680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021442455658462226,
+      "loss": 4.075,
+      "step": 23685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021439219417796404,
+      "loss": 4.0,
+      "step": 23690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021435982809643706,
+      "loss": 4.0703,
+      "step": 23695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002143274583418885,
+      "loss": 4.1797,
+      "step": 23700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021429508491616569,
+      "loss": 4.0328,
+      "step": 23705
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021426270782111618,
+      "loss": 3.9672,
+      "step": 23710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021423032705858772,
+      "loss": 4.0828,
+      "step": 23715
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021419794263042829,
+      "loss": 4.025,
+      "step": 23720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021416555453848606,
+      "loss": 3.9578,
+      "step": 23725
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021413316278460945,
+      "loss": 3.9859,
+      "step": 23730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141007673706471,
+      "loss": 4.0969,
+      "step": 23735
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021406836829844769,
+      "loss": 4.0609,
+      "step": 23740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021403596556986035,
+      "loss": 4.0016,
+      "step": 23745
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021400355918673426,
+      "loss": 3.9937,
+      "step": 23750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002139711491509189,
+      "loss": 4.0922,
+      "step": 23755
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021393873546426387,
+      "loss": 4.1141,
+      "step": 23760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021390631812861902,
+      "loss": 3.9516,
+      "step": 23765
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021387389714583446,
+      "loss": 3.9875,
+      "step": 23770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021384147251776037,
+      "loss": 3.9813,
+      "step": 23775
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021380904424624734,
+      "loss": 4.0297,
+      "step": 23780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021377661233314593,
+      "loss": 3.9641,
+      "step": 23785
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021374417678030717,
+      "loss": 3.9609,
+      "step": 23790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021371173758958205,
+      "loss": 4.0859,
+      "step": 23795
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002136792947628219,
+      "loss": 4.0328,
+      "step": 23800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002136468483018783,
+      "loss": 4.0172,
+      "step": 23805
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021361439820860288,
+      "loss": 4.0078,
+      "step": 23810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002135819444848476,
+      "loss": 4.0062,
+      "step": 23815
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021354948713246463,
+      "loss": 3.9953,
+      "step": 23820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002135170261533063,
+      "loss": 4.0281,
+      "step": 23825
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021348456154922517,
+      "loss": 4.0406,
+      "step": 23830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021345209332207396,
+      "loss": 4.0203,
+      "step": 23835
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021341962147370568,
+      "loss": 4.0312,
+      "step": 23840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021338714600597345,
+      "loss": 4.0062,
+      "step": 23845
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021335466692073074,
+      "loss": 4.125,
+      "step": 23850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021332218421983101,
+      "loss": 4.0719,
+      "step": 23855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021328969790512808,
+      "loss": 4.0516,
+      "step": 23860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021325720797847608,
+      "loss": 4.0125,
+      "step": 23865
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021322471444172906,
+      "loss": 4.1125,
+      "step": 23870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002131922172967415,
+      "loss": 3.9797,
+      "step": 23875
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000213159716545368,
+      "loss": 4.0594,
+      "step": 23880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021312721218946334,
+      "loss": 3.9594,
+      "step": 23885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021309470423088266,
+      "loss": 4.0359,
+      "step": 23890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021306219267148104,
+      "loss": 4.0281,
+      "step": 23895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021302967751311408,
+      "loss": 4.1016,
+      "step": 23900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021299715875763725,
+      "loss": 4.0812,
+      "step": 23905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021296463640690653,
+      "loss": 3.9844,
+      "step": 23910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021293211046277796,
+      "loss": 4.0531,
+      "step": 23915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021289958092710772,
+      "loss": 4.0547,
+      "step": 23920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021286704780175238,
+      "loss": 4.0578,
+      "step": 23925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002128345110885685,
+      "loss": 4.0984,
+      "step": 23930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021280197078941306,
+      "loss": 4.0062,
+      "step": 23935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021276942690614303,
+      "loss": 3.9844,
+      "step": 23940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002127368794406158,
+      "loss": 3.975,
+      "step": 23945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021270432839468878,
+      "loss": 4.0609,
+      "step": 23950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021267177377021963,
+      "loss": 4.0656,
+      "step": 23955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002126392155690664,
+      "loss": 4.0563,
+      "step": 23960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021260665379308703,
+      "loss": 4.1266,
+      "step": 23965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002125740884441399,
+      "loss": 3.9578,
+      "step": 23970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002125415195240835,
+      "loss": 4.0062,
+      "step": 23975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002125089470347766,
+      "loss": 4.0656,
+      "step": 23980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021247637097807797,
+      "loss": 3.9906,
+      "step": 23985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021244379135584686,
+      "loss": 4.0453,
+      "step": 23990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021241120816994252,
+      "loss": 4.0859,
+      "step": 23995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021237862142222452,
+      "loss": 4.0016,
+      "step": 24000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021234603111455261,
+      "loss": 4.1016,
+      "step": 24005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021231343724878666,
+      "loss": 4.0438,
+      "step": 24010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021228083982678683,
+      "loss": 4.0125,
+      "step": 24015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002122482388504134,
+      "loss": 4.05,
+      "step": 24020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021221563432152705,
+      "loss": 3.9109,
+      "step": 24025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021218302624198844,
+      "loss": 4.0875,
+      "step": 24030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021215041461365852,
+      "loss": 3.9766,
+      "step": 24035
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002121177994383984,
+      "loss": 4.1125,
+      "step": 24040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021208518071806948,
+      "loss": 4.0953,
+      "step": 24045
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120525584545333,
+      "loss": 4.0078,
+      "step": 24050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120199326496516,
+      "loss": 4.025,
+      "step": 24055
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021198730330528642,
+      "loss": 3.95,
+      "step": 24060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002119546704232998,
+      "loss": 3.9906,
+      "step": 24065
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002119220340055542,
+      "loss": 4.0266,
+      "step": 24070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021188939405391213,
+      "loss": 4.0687,
+      "step": 24075
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021185675057023635,
+      "loss": 3.9156,
+      "step": 24080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021182410355638986,
+      "loss": 3.9766,
+      "step": 24085
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002117914530142358,
+      "loss": 4.0359,
+      "step": 24090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021175879894563755,
+      "loss": 4.0,
+      "step": 24095
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021172614135245868,
+      "loss": 4.0109,
+      "step": 24100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000211693480236563,
+      "loss": 4.0312,
+      "step": 24105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002116608155998144,
+      "loss": 4.0938,
+      "step": 24110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021162814744407713,
+      "loss": 4.0453,
+      "step": 24115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021159547577121552,
+      "loss": 4.0391,
+      "step": 24120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002115628005830942,
+      "loss": 3.9734,
+      "step": 24125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021153012188157786,
+      "loss": 4.0078,
+      "step": 24130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021149743966853153,
+      "loss": 3.9516,
+      "step": 24135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114647539458204,
+      "loss": 3.9734,
+      "step": 24140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114320647153098,
+      "loss": 4.0516,
+      "step": 24145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021139937197886538,
+      "loss": 4.0687,
+      "step": 24150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021136667573835286,
+      "loss": 4.0969,
+      "step": 24155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021133397599563825,
+      "loss": 4.1266,
+      "step": 24160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113012727525877,
+      "loss": 3.9719,
+      "step": 24165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021126856601106758,
+      "loss": 4.0484,
+      "step": 24170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021123585577294447,
+      "loss": 4.0219,
+      "step": 24175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021120314204008516,
+      "loss": 3.9797,
+      "step": 24180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021117042481435665,
+      "loss": 4.0047,
+      "step": 24185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002111377040976261,
+      "loss": 4.0328,
+      "step": 24190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021110497989176088,
+      "loss": 4.0687,
+      "step": 24195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021107225219862852,
+      "loss": 3.9656,
+      "step": 24200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021103952102009685,
+      "loss": 4.0875,
+      "step": 24205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021100678635803377,
+      "loss": 3.9672,
+      "step": 24210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002109740482143076,
+      "loss": 4.0438,
+      "step": 24215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021094130659078656,
+      "loss": 4.15,
+      "step": 24220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021090856148933928,
+      "loss": 3.9703,
+      "step": 24225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021087581291183452,
+      "loss": 4.1125,
+      "step": 24230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021084306086014123,
+      "loss": 3.9594,
+      "step": 24235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021081030533612858,
+      "loss": 3.9766,
+      "step": 24240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021077754634166594,
+      "loss": 3.9953,
+      "step": 24245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002107447838786229,
+      "loss": 3.9438,
+      "step": 24250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021071201794886911,
+      "loss": 4.0703,
+      "step": 24255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021067924855427465,
+      "loss": 4.0594,
+      "step": 24260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002106464756967096,
+      "loss": 3.9406,
+      "step": 24265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021061369937804434,
+      "loss": 4.0125,
+      "step": 24270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105809196001494,
+      "loss": 3.9969,
+      "step": 24275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105481363648955,
+      "loss": 4.0078,
+      "step": 24280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021051534967415363,
+      "loss": 4.0312,
+      "step": 24285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021048255952979493,
+      "loss": 4.0891,
+      "step": 24290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021044976593369072,
+      "loss": 4.0391,
+      "step": 24295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002104169688877125,
+      "loss": 4.0109,
+      "step": 24300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021038416839373207,
+      "loss": 4.025,
+      "step": 24305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002103513644536213,
+      "loss": 4.0312,
+      "step": 24310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002103185570692523,
+      "loss": 4.0453,
+      "step": 24315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102857462424975,
+      "loss": 3.9875,
+      "step": 24320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021025293197522927,
+      "loss": 4.0766,
+      "step": 24325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102201142693205,
+      "loss": 4.0297,
+      "step": 24330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101872931266439,
+      "loss": 4.0312,
+      "step": 24335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101544685490727,
+      "loss": 4.0844,
+      "step": 24340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021012164053848016,
+      "loss": 3.9797,
+      "step": 24345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021008880909673981,
+      "loss": 3.9891,
+      "step": 24350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002100559742257253,
+      "loss": 3.9453,
+      "step": 24355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002100231359273106,
+      "loss": 3.9984,
+      "step": 24360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020999029420336966,
+      "loss": 4.0078,
+      "step": 24365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020995744905577684,
+      "loss": 3.9906,
+      "step": 24370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002099246004864067,
+      "loss": 4.0016,
+      "step": 24375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020989174849713377,
+      "loss": 4.1031,
+      "step": 24380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020985889308983299,
+      "loss": 4.0438,
+      "step": 24385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002098260342663794,
+      "loss": 4.0312,
+      "step": 24390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020979317202864827,
+      "loss": 3.9922,
+      "step": 24395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020976030637851498,
+      "loss": 4.0828,
+      "step": 24400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020972743731785526,
+      "loss": 4.0516,
+      "step": 24405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000209694564848545,
+      "loss": 4.0031,
+      "step": 24410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002096616889724601,
+      "loss": 4.0078,
+      "step": 24415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020962880969147685,
+      "loss": 4.0734,
+      "step": 24420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002095959270074717,
+      "loss": 4.0234,
+      "step": 24425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020956304092232124,
+      "loss": 4.0344,
+      "step": 24430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020953015143790224,
+      "loss": 4.0156,
+      "step": 24435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002094972585560918,
+      "loss": 3.9688,
+      "step": 24440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020946436227876704,
+      "loss": 3.9797,
+      "step": 24445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002094314626078054,
+      "loss": 3.9719,
+      "step": 24450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020939855954508446,
+      "loss": 4.0672,
+      "step": 24455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020936565309248196,
+      "loss": 4.0109,
+      "step": 24460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020933274325187596,
+      "loss": 4.0125,
+      "step": 24465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002092998300251445,
+      "loss": 3.9219,
+      "step": 24470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020926691341416613,
+      "loss": 4.025,
+      "step": 24475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020923399342081918,
+      "loss": 4.0234,
+      "step": 24480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020920107004698257,
+      "loss": 3.9188,
+      "step": 24485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020916814329453515,
+      "loss": 4.0656,
+      "step": 24490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020913521316535608,
+      "loss": 4.0312,
+      "step": 24495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020910227966132472,
+      "loss": 3.9547,
+      "step": 24500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002090693427843205,
+      "loss": 4.0141,
+      "step": 24505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020903640253622325,
+      "loss": 4.0453,
+      "step": 24510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020900345891891272,
+      "loss": 3.9516,
+      "step": 24515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002089705119342692,
+      "loss": 4.0734,
+      "step": 24520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002089375615841728,
+      "loss": 3.9484,
+      "step": 24525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002089046078705041,
+      "loss": 3.9656,
+      "step": 24530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002088716507951437,
+      "loss": 4.0187,
+      "step": 24535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020883869035997254,
+      "loss": 3.9594,
+      "step": 24540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020880572656687167,
+      "loss": 4.0172,
+      "step": 24545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020877275941772226,
+      "loss": 3.9797,
+      "step": 24550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020873978891440583,
+      "loss": 4.0359,
+      "step": 24555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020870681505880394,
+      "loss": 4.0203,
+      "step": 24560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020867383785279847,
+      "loss": 4.0391,
+      "step": 24565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020864085729827142,
+      "loss": 3.9859,
+      "step": 24570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020860787339710494,
+      "loss": 4.0016,
+      "step": 24575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002085748861511815,
+      "loss": 4.0281,
+      "step": 24580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020854189556238366,
+      "loss": 3.9937,
+      "step": 24585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020850890163259418,
+      "loss": 4.0406,
+      "step": 24590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020847590436369603,
+      "loss": 4.0687,
+      "step": 24595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020844290375757237,
+      "loss": 4.0203,
+      "step": 24600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020840989981610656,
+      "loss": 3.9797,
+      "step": 24605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020837689254118216,
+      "loss": 4.0547,
+      "step": 24610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020834388193468278,
+      "loss": 4.0094,
+      "step": 24615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020831086799849252,
+      "loss": 4.1281,
+      "step": 24620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020827785073449537,
+      "loss": 4.0375,
+      "step": 24625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002082448301445756,
+      "loss": 4.0438,
+      "step": 24630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002082118062306178,
+      "loss": 4.0375,
+      "step": 24635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002081787789945066,
+      "loss": 4.0625,
+      "step": 24640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020814574843812687,
+      "loss": 3.8906,
+      "step": 24645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020811271456336363,
+      "loss": 3.9969,
+      "step": 24650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002080796773721022,
+      "loss": 4.0047,
+      "step": 24655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020804663686622795,
+      "loss": 4.0406,
+      "step": 24660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020801359304762655,
+      "loss": 3.9219,
+      "step": 24665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020798054591818378,
+      "loss": 3.9797,
+      "step": 24670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020794749547978564,
+      "loss": 4.0484,
+      "step": 24675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020791444173431838,
+      "loss": 4.0531,
+      "step": 24680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020788138468366829,
+      "loss": 3.9797,
+      "step": 24685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020784832432972205,
+      "loss": 3.9969,
+      "step": 24690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002078152606743663,
+      "loss": 3.9281,
+      "step": 24695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020778219371948807,
+      "loss": 4.0687,
+      "step": 24700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020774912346697446,
+      "loss": 4.0563,
+      "step": 24705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020771604991871281,
+      "loss": 4.0266,
+      "step": 24710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002076829730765906,
+      "loss": 4.0328,
+      "step": 24715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020764989294249557,
+      "loss": 4.0016,
+      "step": 24720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002076168095183156,
+      "loss": 4.05,
+      "step": 24725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002075837228059387,
+      "loss": 4.0172,
+      "step": 24730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020755063280725324,
+      "loss": 4.0406,
+      "step": 24735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020751753952414755,
+      "loss": 4.0344,
+      "step": 24740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020748444295851033,
+      "loss": 4.0016,
+      "step": 24745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020745134311223046,
+      "loss": 4.0594,
+      "step": 24750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020741823998719685,
+      "loss": 4.0016,
+      "step": 24755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020738513358529875,
+      "loss": 4.0328,
+      "step": 24760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002073520239084255,
+      "loss": 4.0953,
+      "step": 24765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020731891095846676,
+      "loss": 4.0125,
+      "step": 24770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020728579473731223,
+      "loss": 4.0172,
+      "step": 24775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020725267524685184,
+      "loss": 4.0953,
+      "step": 24780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020721955248897576,
+      "loss": 4.05,
+      "step": 24785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020718642646557435,
+      "loss": 4.0609,
+      "step": 24790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020715329717853798,
+      "loss": 3.9281,
+      "step": 24795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020712016462975745,
+      "loss": 4.0453,
+      "step": 24800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020708702882112363,
+      "loss": 3.9859,
+      "step": 24805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070538897545275,
+      "loss": 4.0187,
+      "step": 24810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020702074743186047,
+      "loss": 4.0563,
+      "step": 24815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002069876018550138,
+      "loss": 4.0578,
+      "step": 24820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020695445302587923,
+      "loss": 4.0156,
+      "step": 24825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002069213009463485,
+      "loss": 4.0125,
+      "step": 24830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020688814561831367,
+      "loss": 3.9859,
+      "step": 24835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020685498704366687,
+      "loss": 4.0422,
+      "step": 24840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020682182522430045,
+      "loss": 4.0922,
+      "step": 24845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000206788660162107,
+      "loss": 4.0391,
+      "step": 24850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020675549185897923,
+      "loss": 3.9734,
+      "step": 24855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020672232031681003,
+      "loss": 3.9984,
+      "step": 24860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020668914553749254,
+      "loss": 3.9953,
+      "step": 24865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020665596752292008,
+      "loss": 4.0625,
+      "step": 24870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020662278627498607,
+      "loss": 4.0219,
+      "step": 24875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020658960179558417,
+      "loss": 3.9953,
+      "step": 24880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020655641408660823,
+      "loss": 4.0234,
+      "step": 24885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020652322314995228,
+      "loss": 3.9297,
+      "step": 24890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020649002898751054,
+      "loss": 4.0094,
+      "step": 24895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002064568316011774,
+      "loss": 3.9766,
+      "step": 24900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020642363099284742,
+      "loss": 3.9969,
+      "step": 24905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020639042716441534,
+      "loss": 3.9484,
+      "step": 24910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020635722011777622,
+      "loss": 4.0203,
+      "step": 24915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020632400985482502,
+      "loss": 3.9719,
+      "step": 24920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020629079637745723,
+      "loss": 3.9797,
+      "step": 24925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020625757968756818,
+      "loss": 4.0406,
+      "step": 24930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020622435978705366,
+      "loss": 3.9422,
+      "step": 24935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020619113667780947,
+      "loss": 3.9859,
+      "step": 24940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020615791036173172,
+      "loss": 4.0312,
+      "step": 24945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061246808407166,
+      "loss": 3.9953,
+      "step": 24950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002060914481166605,
+      "loss": 3.9813,
+      "step": 24955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020605821219146005,
+      "loss": 3.9,
+      "step": 24960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020602497306701202,
+      "loss": 4.0344,
+      "step": 24965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020599173074521343,
+      "loss": 3.9172,
+      "step": 24970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020595848522796128,
+      "loss": 4.0031,
+      "step": 24975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000205925236517153,
+      "loss": 3.9906,
+      "step": 24980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020589198461468607,
+      "loss": 4.0312,
+      "step": 24985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020585872952245818,
+      "loss": 4.0031,
+      "step": 24990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020582547124236723,
+      "loss": 3.9531,
+      "step": 24995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020579220977631123,
+      "loss": 4.0531,
+      "step": 25000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020575894512618844,
+      "loss": 3.9672,
+      "step": 25005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020572567729389724,
+      "loss": 4.1188,
+      "step": 25010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002056924062813363,
+      "loss": 4.0344,
+      "step": 25015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020565913209040433,
+      "loss": 3.9922,
+      "step": 25020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020562585472300035,
+      "loss": 3.9734,
+      "step": 25025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020559257418102344,
+      "loss": 4.0328,
+      "step": 25030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020555929046637295,
+      "loss": 4.0484,
+      "step": 25035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020552600358094843,
+      "loss": 4.0547,
+      "step": 25040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020549271352664948,
+      "loss": 4.0,
+      "step": 25045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020545942030537602,
+      "loss": 4.0187,
+      "step": 25050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020542612391902812,
+      "loss": 4.0406,
+      "step": 25055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020539282436950594,
+      "loss": 3.9594,
+      "step": 25060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020535952165870997,
+      "loss": 3.9516,
+      "step": 25065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002053262157885407,
+      "loss": 4.0078,
+      "step": 25070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000205292906760899,
+      "loss": 4.0359,
+      "step": 25075
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020525959457768573,
+      "loss": 3.9234,
+      "step": 25080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002052262792408021,
+      "loss": 3.9609,
+      "step": 25085
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020519296075214937,
+      "loss": 4.0594,
+      "step": 25090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020515963911362905,
+      "loss": 4.0469,
+      "step": 25095
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002051263143271428,
+      "loss": 4.0359,
+      "step": 25100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020509298639459245,
+      "loss": 4.0344,
+      "step": 25105
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020505965531788007,
+      "loss": 4.05,
+      "step": 25110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020502632109890785,
+      "loss": 3.9469,
+      "step": 25115
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020499298373957817,
+      "loss": 4.0797,
+      "step": 25120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020495964324179354,
+      "loss": 4.1469,
+      "step": 25125
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020492629960745687,
+      "loss": 4.0062,
+      "step": 25130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020489295283847091,
+      "loss": 4.0422,
+      "step": 25135
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020485960293673886,
+      "loss": 4.05,
+      "step": 25140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020482624990416395,
+      "loss": 4.0516,
+      "step": 25145
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002047928937426497,
+      "loss": 3.9734,
+      "step": 25150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020475953445409968,
+      "loss": 3.9875,
+      "step": 25155
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020472617204041776,
+      "loss": 4.0406,
+      "step": 25160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002046928065035079,
+      "loss": 4.0672,
+      "step": 25165
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020465943784527433,
+      "loss": 4.0625,
+      "step": 25170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020462606606762133,
+      "loss": 3.9984,
+      "step": 25175
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002045926911724535,
+      "loss": 4.0312,
+      "step": 25180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020455931316167548,
+      "loss": 4.0609,
+      "step": 25185
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020452593203719224,
+      "loss": 4.1203,
+      "step": 25190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020449254780090877,
+      "loss": 3.9906,
+      "step": 25195
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002044591604547303,
+      "loss": 3.9703,
+      "step": 25200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020442577000056233,
+      "loss": 3.9984,
+      "step": 25205
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002043923764403104,
+      "loss": 4.1078,
+      "step": 25210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020435897977588027,
+      "loss": 4.0719,
+      "step": 25215
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020432558000917794,
+      "loss": 4.0406,
+      "step": 25220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020429217714210954,
+      "loss": 3.9219,
+      "step": 25225
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020425877117658133,
+      "loss": 4.0781,
+      "step": 25230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020422536211449978,
+      "loss": 4.0641,
+      "step": 25235
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020419194995777162,
+      "loss": 4.0625,
+      "step": 25240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020415853470830364,
+      "loss": 4.0625,
+      "step": 25245
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020412511636800284,
+      "loss": 4.0016,
+      "step": 25250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020409169493877641,
+      "loss": 4.0062,
+      "step": 25255
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020405827042253174,
+      "loss": 4.025,
+      "step": 25260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002040248428211764,
+      "loss": 4.0453,
+      "step": 25265
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000203991412136618,
+      "loss": 3.9875,
+      "step": 25270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020395797837076456,
+      "loss": 4.0375,
+      "step": 25275
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020392454152552404,
+      "loss": 4.0609,
+      "step": 25280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020389110160280475,
+      "loss": 4.1094,
+      "step": 25285
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020385765860451508,
+      "loss": 3.9891,
+      "step": 25290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020382421253256367,
+      "loss": 4.0141,
+      "step": 25295
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002037907633888592,
+      "loss": 4.0703,
+      "step": 25300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002037573111753107,
+      "loss": 4.0141,
+      "step": 25305
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020372385589382725,
+      "loss": 4.0078,
+      "step": 25310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020369039754631815,
+      "loss": 4.0484,
+      "step": 25315
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020365693613469293,
+      "loss": 3.9484,
+      "step": 25320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020362347166086114,
+      "loss": 3.8984,
+      "step": 25325
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020359000412673264,
+      "loss": 4.0516,
+      "step": 25330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020355653353421746,
+      "loss": 3.9391,
+      "step": 25335
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002035230598852257,
+      "loss": 4.1016,
+      "step": 25340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020348958318166778,
+      "loss": 4.05,
+      "step": 25345
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002034561034254542,
+      "loss": 4.0469,
+      "step": 25350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002034226206184956,
+      "loss": 3.9953,
+      "step": 25355
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002033891347627029,
+      "loss": 3.9438,
+      "step": 25360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020335564585998714,
+      "loss": 4.0281,
+      "step": 25365
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020332215391225949,
+      "loss": 4.0391,
+      "step": 25370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020328865892143137,
+      "loss": 3.9328,
+      "step": 25375
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020325516088941437,
+      "loss": 3.9844,
+      "step": 25380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002032216598181202,
+      "loss": 3.9156,
+      "step": 25385
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020318815570946074,
+      "loss": 3.9547,
+      "step": 25390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020315464856534811,
+      "loss": 3.9656,
+      "step": 25395
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020312113838769457,
+      "loss": 4.0297,
+      "step": 25400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020308762517841247,
+      "loss": 4.0406,
+      "step": 25405
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020305410893941455,
+      "loss": 3.9453,
+      "step": 25410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020302058967261346,
+      "loss": 4.0719,
+      "step": 25415
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020298706737992227,
+      "loss": 3.9859,
+      "step": 25420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020295354206325398,
+      "loss": 3.9094,
+      "step": 25425
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020292001372452194,
+      "loss": 4.0609,
+      "step": 25430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002028864823656396,
+      "loss": 3.9047,
+      "step": 25435
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020285294798852058,
+      "loss": 4.0563,
+      "step": 25440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002028194105950788,
+      "loss": 4.0812,
+      "step": 25445
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020278587018722808,
+      "loss": 4.0703,
+      "step": 25450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027523267668827,
+      "loss": 4.0687,
+      "step": 25455
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027187803359569,
+      "loss": 4.0641,
+      "step": 25460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020268523089636526,
+      "loss": 4.0516,
+      "step": 25465
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020265167845002238,
+      "loss": 4.0656,
+      "step": 25470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020261812299884316,
+      "loss": 3.9625,
+      "step": 25475
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020258456454474253,
+      "loss": 3.9797,
+      "step": 25480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020255100308963573,
+      "loss": 4.0563,
+      "step": 25485
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020251743863543815,
+      "loss": 4.0031,
+      "step": 25490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020248387118406523,
+      "loss": 3.9797,
+      "step": 25495
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020245030073743275,
+      "loss": 4.0359,
+      "step": 25500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020241672729745655,
+      "loss": 4.15,
+      "step": 25505
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020238315086605269,
+      "loss": 4.0766,
+      "step": 25510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020234957144513727,
+      "loss": 3.9922,
+      "step": 25515
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020231598903662682,
+      "loss": 4.0563,
+      "step": 25520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020228240364243778,
+      "loss": 4.0094,
+      "step": 25525
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020224881526448692,
+      "loss": 3.9484,
+      "step": 25530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020221522390469118,
+      "loss": 4.0078,
+      "step": 25535
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020218162956496752,
+      "loss": 4.0906,
+      "step": 25540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020214803224723326,
+      "loss": 4.0266,
+      "step": 25545
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020211443195340575,
+      "loss": 3.9844,
+      "step": 25550
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002020808286854026,
+      "loss": 3.9297,
+      "step": 25555
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020204722244514147,
+      "loss": 4.0531,
+      "step": 25560
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020201361323454034,
+      "loss": 3.9875,
+      "step": 25565
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020198000105551737,
+      "loss": 4.0422,
+      "step": 25570
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020194638590999062,
+      "loss": 4.0219,
+      "step": 25575
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002019127677998787,
+      "loss": 3.9937,
+      "step": 25580
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020187914672710003,
+      "loss": 4.0734,
+      "step": 25585
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020184552269357346,
+      "loss": 4.1047,
+      "step": 25590
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020181189570121793,
+      "loss": 4.0516,
+      "step": 25595
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002017782657519525,
+      "loss": 3.9937,
+      "step": 25600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020174463284769646,
+      "loss": 4.0156,
+      "step": 25605
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020171099699036924,
+      "loss": 4.025,
+      "step": 25610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002016773581818904,
+      "loss": 3.9969,
+      "step": 25615
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002016437164241797,
+      "loss": 4.0203,
+      "step": 25620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020161007171915724,
+      "loss": 4.1234,
+      "step": 25625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015764240687429,
+      "loss": 3.9906,
+      "step": 25630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015427734748571,
+      "loss": 3.9391,
+      "step": 25635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020150911993942024,
+      "loss": 3.9844,
+      "step": 25640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020147546346435295,
+      "loss": 3.9156,
+      "step": 25645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020144180405157598,
+      "loss": 4.0,
+      "step": 25650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020140814170301025,
+      "loss": 4.0312,
+      "step": 25655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020137447642057694,
+      "loss": 4.0219,
+      "step": 25660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002013408082061973,
+      "loss": 4.0328,
+      "step": 25665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020130713706179279,
+      "loss": 3.9047,
+      "step": 25670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000201273462989285,
+      "loss": 4.075,
+      "step": 25675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020123978599059574,
+      "loss": 4.0141,
+      "step": 25680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002012061060676469,
+      "loss": 4.0109,
+      "step": 25685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020117242322236073,
+      "loss": 4.0656,
+      "step": 25690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020113873745665936,
+      "loss": 3.9281,
+      "step": 25695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020110504877246535,
+      "loss": 4.0938,
+      "step": 25700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020107135717170125,
+      "loss": 3.9266,
+      "step": 25705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020103766265628983,
+      "loss": 3.9781,
+      "step": 25710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020100396522815416,
+      "loss": 4.0047,
+      "step": 25715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002009702648892172,
+      "loss": 3.9578,
+      "step": 25720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020093656164140233,
+      "loss": 4.0219,
+      "step": 25725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020090285548663296,
+      "loss": 3.9891,
+      "step": 25730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020086914642683274,
+      "loss": 3.9672,
+      "step": 25735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020083543446392538,
+      "loss": 4.0922,
+      "step": 25740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002008017195998349,
+      "loss": 4.0109,
+      "step": 25745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020076800183648538,
+      "loss": 4.0,
+      "step": 25750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020073428117580106,
+      "loss": 3.9906,
+      "step": 25755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020070055761970646,
+      "loss": 3.8766,
+      "step": 25760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020066683117012609,
+      "loss": 3.9781,
+      "step": 25765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020063310182898483,
+      "loss": 4.0781,
+      "step": 25770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020059936959820754,
+      "loss": 4.0219,
+      "step": 25775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020056563447971935,
+      "loss": 3.9969,
+      "step": 25780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002005318964754455,
+      "loss": 3.9781,
+      "step": 25785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020049815558731144,
+      "loss": 4.0172,
+      "step": 25790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020046441181724277,
+      "loss": 4.0656,
+      "step": 25795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020043066516716525,
+      "loss": 4.0078,
+      "step": 25800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002003969156390048,
+      "loss": 4.0594,
+      "step": 25805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020036316323468752,
+      "loss": 3.9781,
+      "step": 25810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020032940795613964,
+      "loss": 3.9813,
+      "step": 25815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002002956498052876,
+      "loss": 4.0031,
+      "step": 25820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020026188878405793,
+      "loss": 3.9641,
+      "step": 25825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020022812489437747,
+      "loss": 4.0312,
+      "step": 25830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020019435813817306,
+      "loss": 4.0328,
+      "step": 25835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020016058851737175,
+      "loss": 4.0469,
+      "step": 25840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020012681603390084,
+      "loss": 4.0547,
+      "step": 25845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002000930406896877,
+      "loss": 3.9813,
+      "step": 25850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020005926248665987,
+      "loss": 3.9266,
+      "step": 25855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020002548142674515,
+      "loss": 3.9688,
+      "step": 25860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019999169751187134,
+      "loss": 3.9688,
+      "step": 25865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019995791074396653,
+      "loss": 3.9875,
+      "step": 25870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019992412112495895,
+      "loss": 3.9406,
+      "step": 25875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019989032865677694,
+      "loss": 3.9922,
+      "step": 25880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019985653334134905,
+      "loss": 3.9641,
+      "step": 25885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019982273518060397,
+      "loss": 4.0312,
+      "step": 25890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019978893417647067,
+      "loss": 3.8922,
+      "step": 25895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019975513033087798,
+      "loss": 3.9734,
+      "step": 25900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019972132364575528,
+      "loss": 3.9469,
+      "step": 25905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001996875141230318,
+      "loss": 3.9766,
+      "step": 25910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019965370176463712,
+      "loss": 4.0641,
+      "step": 25915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019961988657250085,
+      "loss": 3.9953,
+      "step": 25920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019958606854855293,
+      "loss": 4.0312,
+      "step": 25925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001995522476947232,
+      "loss": 4.0531,
+      "step": 25930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019951842401294198,
+      "loss": 3.95,
+      "step": 25935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019948459750513948,
+      "loss": 3.9969,
+      "step": 25940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019945076817324625,
+      "loss": 3.9688,
+      "step": 25945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019941693601919293,
+      "loss": 4.0359,
+      "step": 25950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019938310104491023,
+      "loss": 4.0203,
+      "step": 25955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019934926325232928,
+      "loss": 3.9688,
+      "step": 25960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000199315422643381,
+      "loss": 4.0656,
+      "step": 25965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019928157921999688,
+      "loss": 4.0016,
+      "step": 25970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019924773298410825,
+      "loss": 4.1188,
+      "step": 25975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019921388393764672,
+      "loss": 4.0391,
+      "step": 25980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019918003208254411,
+      "loss": 4.0,
+      "step": 25985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001991461774207323,
+      "loss": 4.0047,
+      "step": 25990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019911231995414343,
+      "loss": 3.9422,
+      "step": 25995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001990784596847097,
+      "loss": 3.9797,
+      "step": 26000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019904459661436354,
+      "loss": 4.1,
+      "step": 26005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019901073074503754,
+      "loss": 4.025,
+      "step": 26010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019897686207866438,
+      "loss": 4.0453,
+      "step": 26015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019894299061717696,
+      "loss": 3.9531,
+      "step": 26020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019890911636250835,
+      "loss": 4.0031,
+      "step": 26025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019887523931659177,
+      "loss": 4.0094,
+      "step": 26030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019884135948136056,
+      "loss": 3.9703,
+      "step": 26035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019880747685874824,
+      "loss": 4.0312,
+      "step": 26040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001987735914506885,
+      "loss": 4.0359,
+      "step": 26045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019873970325911522,
+      "loss": 4.0187,
+      "step": 26050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019870581228596232,
+      "loss": 4.0187,
+      "step": 26055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019867191853316406,
+      "loss": 4.0875,
+      "step": 26060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019863802200265464,
+      "loss": 4.0516,
+      "step": 26065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019860412269636865,
+      "loss": 4.0047,
+      "step": 26070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001985702206162407,
+      "loss": 4.0031,
+      "step": 26075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019853631576420555,
+      "loss": 4.0062,
+      "step": 26080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019850240814219818,
+      "loss": 4.0563,
+      "step": 26085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019846849775215371,
+      "loss": 4.0438,
+      "step": 26090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019843458459600736,
+      "loss": 4.0281,
+      "step": 26095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019840066867569462,
+      "loss": 3.9875,
+      "step": 26100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019836674999315108,
+      "loss": 3.95,
+      "step": 26105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019833282855031236,
+      "loss": 4.0031,
+      "step": 26110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019829890434911452,
+      "loss": 4.0422,
+      "step": 26115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019826497739149353,
+      "loss": 3.9844,
+      "step": 26120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001982310476793856,
+      "loss": 4.1031,
+      "step": 26125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019819711521472714,
+      "loss": 4.0641,
+      "step": 26130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019816317999945467,
+      "loss": 4.0828,
+      "step": 26135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019812924203550488,
+      "loss": 4.0563,
+      "step": 26140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001980953013248146,
+      "loss": 3.8781,
+      "step": 26145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001980613578693208,
+      "loss": 3.975,
+      "step": 26150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019802741167096065,
+      "loss": 4.0438,
+      "step": 26155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019799346273167152,
+      "loss": 4.0516,
+      "step": 26160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019795951105339084,
+      "loss": 3.9531,
+      "step": 26165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001979255566380562,
+      "loss": 4.0016,
+      "step": 26170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019789159948760548,
+      "loss": 4.0391,
+      "step": 26175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019785763960397646,
+      "loss": 3.9875,
+      "step": 26180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019782367698910743,
+      "loss": 4.0531,
+      "step": 26185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001977897116449365,
+      "loss": 4.0031,
+      "step": 26190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001977557435734021,
+      "loss": 4.0438,
+      "step": 26195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001977217727764428,
+      "loss": 4.0047,
+      "step": 26200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001976877992559973,
+      "loss": 4.0203,
+      "step": 26205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019765382301400458,
+      "loss": 4.0453,
+      "step": 26210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019761984405240347,
+      "loss": 4.0078,
+      "step": 26215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019758586237313334,
+      "loss": 4.0234,
+      "step": 26220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001975518779781334,
+      "loss": 3.9344,
+      "step": 26225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019751789086934325,
+      "loss": 4.0406,
+      "step": 26230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019748390104870243,
+      "loss": 3.9547,
+      "step": 26235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019744990851815083,
+      "loss": 3.9984,
+      "step": 26240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019741591327962835,
+      "loss": 4.1172,
+      "step": 26245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001973819153350751,
+      "loss": 4.0125,
+      "step": 26250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019734791468643142,
+      "loss": 3.9594,
+      "step": 26255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019731391133563764,
+      "loss": 4.0,
+      "step": 26260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001972799052846344,
+      "loss": 4.0094,
+      "step": 26265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019724589653536236,
+      "loss": 4.0344,
+      "step": 26270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001972118850897625,
+      "loss": 3.9984,
+      "step": 26275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019717787094977577,
+      "loss": 4.0422,
+      "step": 26280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019714385411734339,
+      "loss": 4.0062,
+      "step": 26285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001971098345944067,
+      "loss": 4.0125,
+      "step": 26290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001970758123829072,
+      "loss": 4.0203,
+      "step": 26295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001970417874847866,
+      "loss": 3.9531,
+      "step": 26300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019700775990198654,
+      "loss": 3.9406,
+      "step": 26305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019697372963644919,
+      "loss": 3.9891,
+      "step": 26310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019693969669011647,
+      "loss": 4.075,
+      "step": 26315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019690566106493078,
+      "loss": 4.0641,
+      "step": 26320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019687162276283446,
+      "loss": 4.0703,
+      "step": 26325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019683758178577014,
+      "loss": 3.9547,
+      "step": 26330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019680353813568047,
+      "loss": 4.0828,
+      "step": 26335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019676949181450838,
+      "loss": 3.9109,
+      "step": 26340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019673544282419685,
+      "loss": 4.0719,
+      "step": 26345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019670139116668913,
+      "loss": 4.0094,
+      "step": 26350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019666733684392846,
+      "loss": 3.9844,
+      "step": 26355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001966332798578584,
+      "loss": 4.0797,
+      "step": 26360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019659922021042256,
+      "loss": 4.0031,
+      "step": 26365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001965651579035647,
+      "loss": 4.0078,
+      "step": 26370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019653109293922883,
+      "loss": 3.9781,
+      "step": 26375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019649702531935895,
+      "loss": 4.1188,
+      "step": 26380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019646295504589933,
+      "loss": 4.1141,
+      "step": 26385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019642888212079447,
+      "loss": 3.8672,
+      "step": 26390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019639480654598873,
+      "loss": 4.0187,
+      "step": 26395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019636072832342698,
+      "loss": 4.0328,
+      "step": 26400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019632664745505393,
+      "loss": 4.0672,
+      "step": 26405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019629256394281468,
+      "loss": 4.0,
+      "step": 26410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019625847778865428,
+      "loss": 3.9703,
+      "step": 26415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019622438899451815,
+      "loss": 4.0344,
+      "step": 26420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019619029756235166,
+      "loss": 4.0078,
+      "step": 26425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019615620349410038,
+      "loss": 3.9609,
+      "step": 26430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019612210679171015,
+      "loss": 4.0469,
+      "step": 26435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019608800745712685,
+      "loss": 3.9875,
+      "step": 26440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960539054922965,
+      "loss": 4.0109,
+      "step": 26445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960198008991653,
+      "loss": 3.9156,
+      "step": 26450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019598569367967964,
+      "loss": 4.0625,
+      "step": 26455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195951583835786,
+      "loss": 4.0641,
+      "step": 26460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019591747136943105,
+      "loss": 4.0766,
+      "step": 26465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019588335628256155,
+      "loss": 4.0641,
+      "step": 26470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019584923857712447,
+      "loss": 4.0922,
+      "step": 26475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019581511825506696,
+      "loss": 3.9734,
+      "step": 26480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019578099531833619,
+      "loss": 3.9875,
+      "step": 26485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019574686976887965,
+      "loss": 4.0938,
+      "step": 26490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001957127416086448,
+      "loss": 4.0703,
+      "step": 26495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019567861083957938,
+      "loss": 3.9937,
+      "step": 26500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001956444774636312,
+      "loss": 4.0328,
+      "step": 26505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019561034148274833,
+      "loss": 3.9734,
+      "step": 26510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019557620289887884,
+      "loss": 4.0422,
+      "step": 26515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019554206171397103,
+      "loss": 3.9828,
+      "step": 26520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019550791792997342,
+      "loss": 4.0062,
+      "step": 26525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019547377154883447,
+      "loss": 3.9984,
+      "step": 26530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019543962257250304,
+      "loss": 4.0328,
+      "step": 26535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019540547100292796,
+      "loss": 4.0125,
+      "step": 26540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019537131684205827,
+      "loss": 3.9797,
+      "step": 26545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019533716009184315,
+      "loss": 4.1281,
+      "step": 26550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019530300075423193,
+      "loss": 4.0953,
+      "step": 26555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019526883883117402,
+      "loss": 4.0938,
+      "step": 26560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019523467432461917,
+      "loss": 4.0953,
+      "step": 26565
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019520050723651709,
+      "loss": 4.0047,
+      "step": 26570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001951663375688177,
+      "loss": 4.0875,
+      "step": 26575
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195132165323471,
+      "loss": 4.0422,
+      "step": 26580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019509799050242734,
+      "loss": 3.9984,
+      "step": 26585
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019506381310763695,
+      "loss": 3.9547,
+      "step": 26590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019502963314105043,
+      "loss": 3.9875,
+      "step": 26595
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001949954506046184,
+      "loss": 4.0,
+      "step": 26600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019496126550029168,
+      "loss": 3.9984,
+      "step": 26605
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019492707783002117,
+      "loss": 4.0328,
+      "step": 26610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019489288759575802,
+      "loss": 3.9688,
+      "step": 26615
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001948586947994534,
+      "loss": 4.0312,
+      "step": 26620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019482449944305879,
+      "loss": 3.9375,
+      "step": 26625
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019479030152852561,
+      "loss": 4.0563,
+      "step": 26630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019475610105780565,
+      "loss": 3.9859,
+      "step": 26635
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019472189803285063,
+      "loss": 3.9734,
+      "step": 26640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019468769245561263,
+      "loss": 3.9844,
+      "step": 26645
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001946534843280437,
+      "loss": 4.0078,
+      "step": 26650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019461927365209603,
+      "loss": 3.9969,
+      "step": 26655
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001945850604297222,
+      "loss": 3.9641,
+      "step": 26660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019455084466287462,
+      "loss": 3.9969,
+      "step": 26665
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019451662635350608,
+      "loss": 4.0031,
+      "step": 26670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001944824055035693,
+      "loss": 4.0172,
+      "step": 26675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019444818211501744,
+      "loss": 4.0609,
+      "step": 26680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019441395618980344,
+      "loss": 3.9844,
+      "step": 26685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019437972772988075,
+      "loss": 4.0031,
+      "step": 26690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019434549673720268,
+      "loss": 3.8625,
+      "step": 26695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019431126321372278,
+      "loss": 4.0016,
+      "step": 26700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019427702716139493,
+      "loss": 4.0922,
+      "step": 26705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019424278858217275,
+      "loss": 4.0,
+      "step": 26710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019420854747801043,
+      "loss": 3.9813,
+      "step": 26715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194174303850862,
+      "loss": 4.0141,
+      "step": 26720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001941400577026818,
+      "loss": 4.0391,
+      "step": 26725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019410580903542424,
+      "loss": 4.0625,
+      "step": 26730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019407155785104393,
+      "loss": 3.9984,
+      "step": 26735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001940373041514955,
+      "loss": 4.0047,
+      "step": 26740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001940030479387339,
+      "loss": 3.9719,
+      "step": 26745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019396878921471413,
+      "loss": 3.9453,
+      "step": 26750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019393452798139126,
+      "loss": 3.9562,
+      "step": 26755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019390026424072072,
+      "loss": 3.9547,
+      "step": 26760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938659979946578,
+      "loss": 3.9937,
+      "step": 26765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019383172924515815,
+      "loss": 4.0563,
+      "step": 26770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001937974579941775,
+      "loss": 4.0547,
+      "step": 26775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019376318424367168,
+      "loss": 4.0563,
+      "step": 26780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001937289079955967,
+      "loss": 4.0641,
+      "step": 26785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001936946292519087,
+      "loss": 4.0266,
+      "step": 26790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019366034801456404,
+      "loss": 4.1047,
+      "step": 26795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019362606428551907,
+      "loss": 4.0516,
+      "step": 26800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019359177806673044,
+      "loss": 4.0578,
+      "step": 26805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001935574893601548,
+      "loss": 4.0578,
+      "step": 26810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019352319816774903,
+      "loss": 4.0062,
+      "step": 26815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019348890449147015,
+      "loss": 4.0469,
+      "step": 26820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019345460833327527,
+      "loss": 4.0094,
+      "step": 26825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019342030969512174,
+      "loss": 3.9312,
+      "step": 26830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019338600857896692,
+      "loss": 4.0109,
+      "step": 26835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019335170498676837,
+      "loss": 4.0812,
+      "step": 26840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933173989204839,
+      "loss": 4.0297,
+      "step": 26845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019328309038207125,
+      "loss": 4.0406,
+      "step": 26850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019324877937348844,
+      "loss": 3.8625,
+      "step": 26855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019321446589669366,
+      "loss": 4.0219,
+      "step": 26860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019318014995364509,
+      "loss": 4.0938,
+      "step": 26865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019314583154630125,
+      "loss": 3.9844,
+      "step": 26870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001931115106766206,
+      "loss": 4.0328,
+      "step": 26875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019307718734656187,
+      "loss": 4.0297,
+      "step": 26880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019304286155808392,
+      "loss": 4.05,
+      "step": 26885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001930085333131457,
+      "loss": 4.0281,
+      "step": 26890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019297420261370636,
+      "loss": 3.9484,
+      "step": 26895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019293986946172507,
+      "loss": 4.0516,
+      "step": 26900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019290553385916133,
+      "loss": 3.8578,
+      "step": 26905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019287119580797463,
+      "loss": 4.0141,
+      "step": 26910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019283685531012461,
+      "loss": 4.0422,
+      "step": 26915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019280251236757117,
+      "loss": 4.0172,
+      "step": 26920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019276816698227415,
+      "loss": 4.0391,
+      "step": 26925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001927338191561938,
+      "loss": 4.0219,
+      "step": 26930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001926994688912902,
+      "loss": 3.9891,
+      "step": 26935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019266511618952386,
+      "loss": 4.0094,
+      "step": 26940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019263076105285513,
+      "loss": 3.9469,
+      "step": 26945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019259640348324482,
+      "loss": 4.0062,
+      "step": 26950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019256204348265358,
+      "loss": 4.0031,
+      "step": 26955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001925276810530425,
+      "loss": 3.9922,
+      "step": 26960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019249331619637246,
+      "loss": 3.9547,
+      "step": 26965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001924589489146048,
+      "loss": 4.0156,
+      "step": 26970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019242457920970086,
+      "loss": 4.0203,
+      "step": 26975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923902070836221,
+      "loss": 3.9406,
+      "step": 26980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923558325383301,
+      "loss": 3.9562,
+      "step": 26985
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923214555757866,
+      "loss": 4.0703,
+      "step": 26990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019228707619795365,
+      "loss": 4.0344,
+      "step": 26995
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019225269440679307,
+      "loss": 4.0094,
+      "step": 27000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019221831020426722,
+      "loss": 4.0266,
+      "step": 27005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019218392359233828,
+      "loss": 4.0484,
+      "step": 27010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019214953457296878,
+      "loss": 4.0406,
+      "step": 27015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019211514314812129,
+      "loss": 4.0906,
+      "step": 27020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019208074931975844,
+      "loss": 4.0516,
+      "step": 27025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019204635308984328,
+      "loss": 3.95,
+      "step": 27030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019201195446033858,
+      "loss": 4.0172,
+      "step": 27035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019197755343320768,
+      "loss": 4.0109,
+      "step": 27040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019194315001041368,
+      "loss": 3.9641,
+      "step": 27045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001919087441939201,
+      "loss": 4.0797,
+      "step": 27050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001918743359856904,
+      "loss": 4.0547,
+      "step": 27055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019183992538768832,
+      "loss": 4.0875,
+      "step": 27060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019180551240187768,
+      "loss": 3.8719,
+      "step": 27065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019177109703022238,
+      "loss": 4.0594,
+      "step": 27070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019173667927468654,
+      "loss": 3.9594,
+      "step": 27075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019170225913723443,
+      "loss": 3.9078,
+      "step": 27080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019166783661983032,
+      "loss": 4.0172,
+      "step": 27085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019163341172443875,
+      "loss": 3.9984,
+      "step": 27090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001915989844530244,
+      "loss": 4.0375,
+      "step": 27095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019156455480755198,
+      "loss": 3.9453,
+      "step": 27100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019153012278998638,
+      "loss": 3.9344,
+      "step": 27105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019149568840229268,
+      "loss": 4.1188,
+      "step": 27110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019146125164643603,
+      "loss": 4.0594,
+      "step": 27115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019142681252438175,
+      "loss": 3.9813,
+      "step": 27120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913923710380953,
+      "loss": 3.9828,
+      "step": 27125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913579271895422,
+      "loss": 3.9734,
+      "step": 27130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019132348098068822,
+      "loss": 3.9625,
+      "step": 27135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019128903241349922,
+      "loss": 3.9219,
+      "step": 27140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001912545814899411,
+      "loss": 4.1203,
+      "step": 27145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019122012821198004,
+      "loss": 4.0078,
+      "step": 27150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019118567258158233,
+      "loss": 3.9391,
+      "step": 27155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019115121460071426,
+      "loss": 4.0281,
+      "step": 27160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019111675427134245,
+      "loss": 4.0734,
+      "step": 27165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019108229159543348,
+      "loss": 4.0563,
+      "step": 27170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019104782657495414,
+      "loss": 4.0047,
+      "step": 27175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001910133592118714,
+      "loss": 3.9078,
+      "step": 27180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019097888950815232,
+      "loss": 4.0859,
+      "step": 27185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000190944417465764,
+      "loss": 4.0156,
+      "step": 27190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019090994308667384,
+      "loss": 4.0953,
+      "step": 27195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019087546637284932,
+      "loss": 3.9516,
+      "step": 27200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019084098732625792,
+      "loss": 4.0359,
+      "step": 27205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001908065059488675,
+      "loss": 3.9875,
+      "step": 27210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001907720222426458,
+      "loss": 4.0375,
+      "step": 27215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001907375362095609,
+      "loss": 4.0297,
+      "step": 27220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019070304785158082,
+      "loss": 4.0078,
+      "step": 27225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019066855717067392,
+      "loss": 4.0297,
+      "step": 27230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001906340641688085,
+      "loss": 4.0141,
+      "step": 27235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001905995688479531,
+      "loss": 4.0578,
+      "step": 27240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019056507121007645,
+      "loss": 3.9484,
+      "step": 27245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001905305712571472,
+      "loss": 4.0234,
+      "step": 27250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001904960689911344,
+      "loss": 4.0484,
+      "step": 27255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000190461564414007,
+      "loss": 4.0344,
+      "step": 27260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019042705752773422,
+      "loss": 4.0172,
+      "step": 27265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019039254833428534,
+      "loss": 3.9391,
+      "step": 27270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019035803683562988,
+      "loss": 3.9625,
+      "step": 27275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001903235230337373,
+      "loss": 3.9828,
+      "step": 27280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019028900693057738,
+      "loss": 4.0344,
+      "step": 27285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019025448852812,
+      "loss": 3.9828,
+      "step": 27290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019021996782833501,
+      "loss": 4.0625,
+      "step": 27295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019018544483319263,
+      "loss": 4.0625,
+      "step": 27300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019015091954466298,
+      "loss": 4.0031,
+      "step": 27305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001901163919647165,
+      "loss": 4.0047,
+      "step": 27310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019008186209532366,
+      "loss": 4.0563,
+      "step": 27315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019004732993845508,
+      "loss": 3.8953,
+      "step": 27320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019001279549608153,
+      "loss": 3.9984,
+      "step": 27325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018997825877017384,
+      "loss": 4.1297,
+      "step": 27330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018994371976270306,
+      "loss": 3.9891,
+      "step": 27335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018990917847564042,
+      "loss": 3.9312,
+      "step": 27340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000189874634910957,
+      "loss": 4.0266,
+      "step": 27345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001898400890706244,
+      "loss": 4.1219,
+      "step": 27350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018980554095661403,
+      "loss": 4.0,
+      "step": 27355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001897709905708976,
+      "loss": 3.9922,
+      "step": 27360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018973643791544696,
+      "loss": 4.0656,
+      "step": 27365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018970188299223387,
+      "loss": 4.0703,
+      "step": 27370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018966732580323053,
+      "loss": 4.0453,
+      "step": 27375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001896327663504091,
+      "loss": 4.0328,
+      "step": 27380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018959820463574183,
+      "loss": 3.9141,
+      "step": 27385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018956364066120124,
+      "loss": 3.95,
+      "step": 27390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018952907442875983,
+      "loss": 4.0797,
+      "step": 27395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018949450594039034,
+      "loss": 3.9156,
+      "step": 27400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018945993519806556,
+      "loss": 4.0016,
+      "step": 27405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018942536220375853,
+      "loss": 4.0781,
+      "step": 27410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001893907869594422,
+      "loss": 4.0547,
+      "step": 27415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018935620946708987,
+      "loss": 3.9688,
+      "step": 27420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018932162972867488,
+      "loss": 3.9781,
+      "step": 27425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018928704774617068,
+      "loss": 4.1,
+      "step": 27430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018925246352155092,
+      "loss": 4.0484,
+      "step": 27435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018921787705678924,
+      "loss": 4.025,
+      "step": 27440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018918328835385953,
+      "loss": 4.0031,
+      "step": 27445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018914869741473576,
+      "loss": 3.9969,
+      "step": 27450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001891141042413921,
+      "loss": 3.9781,
+      "step": 27455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018907950883580263,
+      "loss": 4.075,
+      "step": 27460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018904491119994186,
+      "loss": 4.0094,
+      "step": 27465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001890103113357843,
+      "loss": 3.9797,
+      "step": 27470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018897570924530447,
+      "loss": 4.0984,
+      "step": 27475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018894110493047713,
+      "loss": 4.0516,
+      "step": 27480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001889064983932772,
+      "loss": 4.0328,
+      "step": 27485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018887188963567967,
+      "loss": 3.9312,
+      "step": 27490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001888372786596596,
+      "loss": 3.9844,
+      "step": 27495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018880266546719235,
+      "loss": 4.0469,
+      "step": 27500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018876805006025316,
+      "loss": 4.0016,
+      "step": 27505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018873343244081762,
+      "loss": 3.9438,
+      "step": 27510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018869881261086141,
+      "loss": 4.0109,
+      "step": 27515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018866419057236017,
+      "loss": 4.0203,
+      "step": 27520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018862956632728989,
+      "loss": 4.0016,
+      "step": 27525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018859493987762648,
+      "loss": 4.0156,
+      "step": 27530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018856031122534616,
+      "loss": 3.975,
+      "step": 27535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018852568037242514,
+      "loss": 3.9813,
+      "step": 27540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018849104732083987,
+      "loss": 3.9891,
+      "step": 27545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018845641207256676,
+      "loss": 4.0375,
+      "step": 27550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001884217746295825,
+      "loss": 3.9703,
+      "step": 27555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883871349938639,
+      "loss": 3.9906,
+      "step": 27560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018835249316738773,
+      "loss": 3.9766,
+      "step": 27565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883178491521311,
+      "loss": 4.0359,
+      "step": 27570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018828320295007114,
+      "loss": 3.9719,
+      "step": 27575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018824855456318505,
+      "loss": 3.9,
+      "step": 27580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018821390399345028,
+      "loss": 3.8891,
+      "step": 27585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018817925124284436,
+      "loss": 3.9391,
+      "step": 27590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018814459631334485,
+      "loss": 4.0172,
+      "step": 27595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018810993920692957,
+      "loss": 4.0016,
+      "step": 27600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018807527992557635,
+      "loss": 4.0422,
+      "step": 27605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018804061847126324,
+      "loss": 4.0312,
+      "step": 27610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880059548459684,
+      "loss": 4.0969,
+      "step": 27615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018797128905166996,
+      "loss": 3.9969,
+      "step": 27620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018793662109034646,
+      "loss": 3.9594,
+      "step": 27625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018790195096397632,
+      "loss": 4.0484,
+      "step": 27630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018786727867453818,
+      "loss": 3.9688,
+      "step": 27635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018783260422401078,
+      "loss": 3.9578,
+      "step": 27640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018779792761437297,
+      "loss": 4.0219,
+      "step": 27645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018776324884760387,
+      "loss": 3.9969,
+      "step": 27650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018772856792568244,
+      "loss": 3.9797,
+      "step": 27655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018769388485058804,
+      "loss": 3.9875,
+      "step": 27660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876591996243,
+      "loss": 3.9312,
+      "step": 27665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018762451224879777,
+      "loss": 4.0,
+      "step": 27670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018758982272606097,
+      "loss": 3.9562,
+      "step": 27675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875551310580694,
+      "loss": 3.9625,
+      "step": 27680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018752043724680289,
+      "loss": 3.9844,
+      "step": 27685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018748574129424133,
+      "loss": 4.0547,
+      "step": 27690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187451043202365,
+      "loss": 4.0,
+      "step": 27695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001874163429731539,
+      "loss": 3.9875,
+      "step": 27700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018738164060858858,
+      "loss": 4.0281,
+      "step": 27705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001873469361106494,
+      "loss": 4.0391,
+      "step": 27710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187312229481317,
+      "loss": 4.0203,
+      "step": 27715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187277520722572,
+      "loss": 3.9422,
+      "step": 27720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018724280983639538,
+      "loss": 3.9953,
+      "step": 27725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018720809682476792,
+      "loss": 3.9844,
+      "step": 27730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018717338168967081,
+      "loss": 4.0281,
+      "step": 27735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018713866443308523,
+      "loss": 3.9828,
+      "step": 27740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018710394505699248,
+      "loss": 3.9672,
+      "step": 27745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187069223563374,
+      "loss": 3.9937,
+      "step": 27750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018703449995421137,
+      "loss": 3.9891,
+      "step": 27755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018699977423148624,
+      "loss": 4.0344,
+      "step": 27760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001869650463971804,
+      "loss": 4.0234,
+      "step": 27765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018693031645327586,
+      "loss": 3.9469,
+      "step": 27770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018689558440175455,
+      "loss": 4.0453,
+      "step": 27775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018686085024459864,
+      "loss": 3.9766,
+      "step": 27780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001868261139837905,
+      "loss": 3.9734,
+      "step": 27785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018679137562131246,
+      "loss": 4.0609,
+      "step": 27790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001867566351591471,
+      "loss": 4.0125,
+      "step": 27795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018672189259927698,
+      "loss": 4.0656,
+      "step": 27800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018668714794368491,
+      "loss": 3.9906,
+      "step": 27805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018665240119435377,
+      "loss": 3.9422,
+      "step": 27810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018661765235326655,
+      "loss": 4.0125,
+      "step": 27815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018658290142240641,
+      "loss": 3.8844,
+      "step": 27820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001865481484037565,
+      "loss": 4.0047,
+      "step": 27825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018651339329930024,
+      "loss": 3.9797,
+      "step": 27830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018647863611102113,
+      "loss": 4.0859,
+      "step": 27835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018644387684090273,
+      "loss": 4.0328,
+      "step": 27840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018640911549092876,
+      "loss": 3.9937,
+      "step": 27845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018637435206308306,
+      "loss": 3.9891,
+      "step": 27850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018633958655934953,
+      "loss": 3.9688,
+      "step": 27855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018630481898171235,
+      "loss": 3.9875,
+      "step": 27860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018627004933215562,
+      "loss": 4.0375,
+      "step": 27865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018623527761266366,
+      "loss": 3.9672,
+      "step": 27870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018620050382522094,
+      "loss": 4.0172,
+      "step": 27875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018616572797181192,
+      "loss": 4.0594,
+      "step": 27880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001861309500544214,
+      "loss": 3.9406,
+      "step": 27885
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018609617007503401,
+      "loss": 3.9234,
+      "step": 27890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018606138803563472,
+      "loss": 3.9937,
+      "step": 27895
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018602660393820852,
+      "loss": 4.0516,
+      "step": 27900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018599181778474058,
+      "loss": 4.1016,
+      "step": 27905
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001859570295772161,
+      "loss": 4.1078,
+      "step": 27910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001859222393176205,
+      "loss": 3.9688,
+      "step": 27915
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001858874470079392,
+      "loss": 3.9359,
+      "step": 27920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018585265265015787,
+      "loss": 3.9156,
+      "step": 27925
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018581785624626217,
+      "loss": 4.0172,
+      "step": 27930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018578305779823795,
+      "loss": 4.0125,
+      "step": 27935
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001857482573080712,
+      "loss": 3.9844,
+      "step": 27940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018571345477774793,
+      "loss": 4.0234,
+      "step": 27945
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018567865020925438,
+      "loss": 3.9438,
+      "step": 27950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018564384360457678,
+      "loss": 3.9578,
+      "step": 27955
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018560903496570158,
+      "loss": 3.9438,
+      "step": 27960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018557422429461536,
+      "loss": 4.0375,
+      "step": 27965
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018553941159330469,
+      "loss": 3.9219,
+      "step": 27970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018550459686375638,
+      "loss": 3.9984,
+      "step": 27975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001854697801079573,
+      "loss": 3.9125,
+      "step": 27980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018543496132789446,
+      "loss": 4.0125,
+      "step": 27985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018540014052555493,
+      "loss": 3.9562,
+      "step": 27990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018536531770292598,
+      "loss": 4.0906,
+      "step": 27995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001853304928619949,
+      "loss": 4.0406,
+      "step": 28000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018529566600474921,
+      "loss": 4.0328,
+      "step": 28005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018526083713317647,
+      "loss": 3.9797,
+      "step": 28010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018522600624926435,
+      "loss": 3.9516,
+      "step": 28015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018519117335500062,
+      "loss": 4.1266,
+      "step": 28020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018515633845237324,
+      "loss": 4.0328,
+      "step": 28025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001851215015433703,
+      "loss": 3.9375,
+      "step": 28030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018508666262997977,
+      "loss": 3.9766,
+      "step": 28035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001850518217141901,
+      "loss": 3.9672,
+      "step": 28040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018501697879798952,
+      "loss": 4.0922,
+      "step": 28045
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018498213388336656,
+      "loss": 3.9703,
+      "step": 28050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018494728697230993,
+      "loss": 4.025,
+      "step": 28055
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018491243806680822,
+      "loss": 3.9781,
+      "step": 28060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018487758716885026,
+      "loss": 3.8813,
+      "step": 28065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001848427342804251,
+      "loss": 4.0297,
+      "step": 28070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018480787940352168,
+      "loss": 4.0484,
+      "step": 28075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018477302254012924,
+      "loss": 4.0359,
+      "step": 28080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018473816369223703,
+      "loss": 4.0531,
+      "step": 28085
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001847033028618345,
+      "loss": 3.9813,
+      "step": 28090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846684400509111,
+      "loss": 4.025,
+      "step": 28095
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846335752614565,
+      "loss": 4.0484,
+      "step": 28100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018459870849546036,
+      "loss": 4.0359,
+      "step": 28105
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018456383975491266,
+      "loss": 4.0359,
+      "step": 28110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018452896904180323,
+      "loss": 4.0203,
+      "step": 28115
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018449409635812225,
+      "loss": 3.9922,
+      "step": 28120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018445922170585985,
+      "loss": 3.9125,
+      "step": 28125
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018442434508700634,
+      "loss": 3.9516,
+      "step": 28130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001843894665035521,
+      "loss": 4.0406,
+      "step": 28135
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018435458595748774,
+      "loss": 3.9656,
+      "step": 28140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018431970345080383,
+      "loss": 3.9281,
+      "step": 28145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018428481898549108,
+      "loss": 4.0125,
+      "step": 28150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018424993256354046,
+      "loss": 3.9969,
+      "step": 28155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842150441869429,
+      "loss": 4.0484,
+      "step": 28160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001841801538576894,
+      "loss": 4.0187,
+      "step": 28165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018414526157777122,
+      "loss": 3.9734,
+      "step": 28170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018411036734917975,
+      "loss": 3.9984,
+      "step": 28175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018407547117390625,
+      "loss": 4.0359,
+      "step": 28180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018404057305394233,
+      "loss": 3.9781,
+      "step": 28185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018400567299127965,
+      "loss": 3.9922,
+      "step": 28190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001839707709879099,
+      "loss": 4.0187,
+      "step": 28195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018393586704582498,
+      "loss": 4.0203,
+      "step": 28200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018390096116701682,
+      "loss": 4.0375,
+      "step": 28205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001838660533534776,
+      "loss": 4.0328,
+      "step": 28210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001838311436071994,
+      "loss": 4.1125,
+      "step": 28215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018379623193017457,
+      "loss": 4.0062,
+      "step": 28220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018376131832439552,
+      "loss": 4.025,
+      "step": 28225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018372640279185477,
+      "loss": 3.9797,
+      "step": 28230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000183691485334545,
+      "loss": 4.0172,
+      "step": 28235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018365656595445884,
+      "loss": 4.0594,
+      "step": 28240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018362164465358927,
+      "loss": 3.8906,
+      "step": 28245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018358672143392918,
+      "loss": 4.0375,
+      "step": 28250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018355179629747165,
+      "loss": 3.9828,
+      "step": 28255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018351686924620987,
+      "loss": 4.0719,
+      "step": 28260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834819402821371,
+      "loss": 4.0094,
+      "step": 28265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834470094072468,
+      "loss": 3.9062,
+      "step": 28270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018341207662353243,
+      "loss": 4.0125,
+      "step": 28275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018337714193298764,
+      "loss": 4.1453,
+      "step": 28280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001833422053376061,
+      "loss": 3.9797,
+      "step": 28285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018330726683938174,
+      "loss": 3.9891,
+      "step": 28290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832723264403084,
+      "loss": 4.0484,
+      "step": 28295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832373841423802,
+      "loss": 4.0609,
+      "step": 28300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018320243994759125,
+      "loss": 4.0469,
+      "step": 28305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018316749385793586,
+      "loss": 3.9875,
+      "step": 28310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018313254587540843,
+      "loss": 4.0016,
+      "step": 28315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018309759600200338,
+      "loss": 4.0281,
+      "step": 28320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001830626442397153,
+      "loss": 3.9469,
+      "step": 28325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018302769059053899,
+      "loss": 4.0234,
+      "step": 28330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018299273505646913,
+      "loss": 4.0094,
+      "step": 28335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001829577776395007,
+      "loss": 3.9484,
+      "step": 28340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018292281834162878,
+      "loss": 3.9609,
+      "step": 28345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018288785716484834,
+      "loss": 3.9562,
+      "step": 28350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828528941111548,
+      "loss": 4.0344,
+      "step": 28355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018281792918254338,
+      "loss": 4.0141,
+      "step": 28360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018278296238100956,
+      "loss": 3.9813,
+      "step": 28365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018274799370854896,
+      "loss": 4.0359,
+      "step": 28370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827130231671572,
+      "loss": 4.0312,
+      "step": 28375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018267805075883003,
+      "loss": 3.9359,
+      "step": 28380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018264307648556332,
+      "loss": 3.9078,
+      "step": 28385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018260810034935313,
+      "loss": 3.9656,
+      "step": 28390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001825731223521955,
+      "loss": 4.0281,
+      "step": 28395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018253814249608662,
+      "loss": 3.9375,
+      "step": 28400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018250316078302278,
+      "loss": 4.0156,
+      "step": 28405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018246817721500043,
+      "loss": 4.0531,
+      "step": 28410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001824331917940161,
+      "loss": 4.0641,
+      "step": 28415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018239820452206634,
+      "loss": 4.0109,
+      "step": 28420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018236321540114798,
+      "loss": 3.8484,
+      "step": 28425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018232822443325775,
+      "loss": 4.1,
+      "step": 28430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822932316203927,
+      "loss": 4.0578,
+      "step": 28435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001822582369645497,
+      "loss": 3.9906,
+      "step": 28440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018222324046772605,
+      "loss": 3.9062,
+      "step": 28445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018218824213191896,
+      "loss": 3.9688,
+      "step": 28450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018215324195912575,
+      "loss": 4.0219,
+      "step": 28455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182118239951344,
+      "loss": 4.0281,
+      "step": 28460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001820832361105711,
+      "loss": 4.0297,
+      "step": 28465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018204823043880487,
+      "loss": 4.0531,
+      "step": 28470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018201322293804305,
+      "loss": 3.975,
+      "step": 28475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018197821361028347,
+      "loss": 4.0453,
+      "step": 28480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018194320245752416,
+      "loss": 3.9688,
+      "step": 28485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018190818948176323,
+      "loss": 4.0656,
+      "step": 28490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818731746849988,
+      "loss": 4.0453,
+      "step": 28495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018183815806922923,
+      "loss": 4.0219,
+      "step": 28500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018180313963645292,
+      "loss": 4.0297,
+      "step": 28505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018176811938866834,
+      "loss": 3.9312,
+      "step": 28510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018173309732787417,
+      "loss": 3.9906,
+      "step": 28515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000181698073456069,
+      "loss": 4.0062,
+      "step": 28520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018166304777525176,
+      "loss": 4.0328,
+      "step": 28525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018162802028742132,
+      "loss": 3.95,
+      "step": 28530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018159299099457674,
+      "loss": 3.8984,
+      "step": 28535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018155795989871707,
+      "loss": 4.0781,
+      "step": 28540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018152292700184157,
+      "loss": 3.9594,
+      "step": 28545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018148789230594962,
+      "loss": 3.8641,
+      "step": 28550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001814528558130406,
+      "loss": 4.0141,
+      "step": 28555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001814178175251141,
+      "loss": 3.9484,
+      "step": 28560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018138277744416968,
+      "loss": 3.9703,
+      "step": 28565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018134773557220712,
+      "loss": 3.9234,
+      "step": 28570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001813126919112263,
+      "loss": 3.9562,
+      "step": 28575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018127764646322706,
+      "loss": 4.0328,
+      "step": 28580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018124259923020958,
+      "loss": 4.0281,
+      "step": 28585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018120755021417393,
+      "loss": 3.9297,
+      "step": 28590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018117249941712035,
+      "loss": 4.0234,
+      "step": 28595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018113744684104924,
+      "loss": 3.9594,
+      "step": 28600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018110239248796107,
+      "loss": 3.9906,
+      "step": 28605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018106733635985632,
+      "loss": 3.9766,
+      "step": 28610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018103227845873567,
+      "loss": 4.0219,
+      "step": 28615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001809972187865999,
+      "loss": 4.1375,
+      "step": 28620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001809621573454499,
+      "loss": 4.0156,
+      "step": 28625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018092709413728656,
+      "loss": 3.9609,
+      "step": 28630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018089202916411097,
+      "loss": 4.1312,
+      "step": 28635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018085696242792432,
+      "loss": 4.0156,
+      "step": 28640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018082189393072784,
+      "loss": 4.0531,
+      "step": 28645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018078682367452287,
+      "loss": 4.0016,
+      "step": 28650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018075175166131093,
+      "loss": 4.0281,
+      "step": 28655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018071667789309356,
+      "loss": 4.0047,
+      "step": 28660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018068160237187235,
+      "loss": 3.9813,
+      "step": 28665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018064652509964922,
+      "loss": 4.0672,
+      "step": 28670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001806114460784259,
+      "loss": 4.075,
+      "step": 28675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018057636531020437,
+      "loss": 3.9188,
+      "step": 28680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018054128279698677,
+      "loss": 4.0359,
+      "step": 28685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001805061985407752,
+      "loss": 4.0016,
+      "step": 28690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018047111254357193,
+      "loss": 4.0078,
+      "step": 28695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018043602480737931,
+      "loss": 3.9672,
+      "step": 28700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018040093533419987,
+      "loss": 4.0266,
+      "step": 28705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018036584412603604,
+      "loss": 3.9922,
+      "step": 28710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018033075118489063,
+      "loss": 4.0172,
+      "step": 28715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018029565651276626,
+      "loss": 4.0312,
+      "step": 28720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018026056011166585,
+      "loss": 4.0359,
+      "step": 28725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001802254619835924,
+      "loss": 3.9781,
+      "step": 28730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018019036213054888,
+      "loss": 4.0438,
+      "step": 28735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018015526055453851,
+      "loss": 4.0203,
+      "step": 28740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018012015725756448,
+      "loss": 3.9688,
+      "step": 28745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018008505224163018,
+      "loss": 4.0094,
+      "step": 28750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018004994550873904,
+      "loss": 3.9484,
+      "step": 28755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018001483706089463,
+      "loss": 3.9391,
+      "step": 28760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017997972690010053,
+      "loss": 3.9547,
+      "step": 28765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001799446150283605,
+      "loss": 4.0375,
+      "step": 28770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001799095014476785,
+      "loss": 3.9359,
+      "step": 28775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017987438616005828,
+      "loss": 3.9609,
+      "step": 28780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000179839269167504,
+      "loss": 3.9813,
+      "step": 28785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017980415047201973,
+      "loss": 4.0016,
+      "step": 28790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017976903007560973,
+      "loss": 3.9922,
+      "step": 28795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001797339079802783,
+      "loss": 4.0578,
+      "step": 28800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001796987841880299,
+      "loss": 3.9781,
+      "step": 28805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000179663658700869,
+      "loss": 3.9047,
+      "step": 28810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017962853152080017,
+      "loss": 3.9875,
+      "step": 28815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017959340264982822,
+      "loss": 3.9438,
+      "step": 28820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017955827208995798,
+      "loss": 4.0141,
+      "step": 28825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017952313984319424,
+      "loss": 4.05,
+      "step": 28830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001794880059115421,
+      "loss": 3.9625,
+      "step": 28835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001794528702970066,
+      "loss": 4.0578,
+      "step": 28840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017941773300159296,
+      "loss": 4.0125,
+      "step": 28845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017938259402730644,
+      "loss": 4.0578,
+      "step": 28850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017934745337615246,
+      "loss": 4.0516,
+      "step": 28855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017931231105013646,
+      "loss": 3.9688,
+      "step": 28860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017927716705126406,
+      "loss": 4.0766,
+      "step": 28865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017924202138154088,
+      "loss": 4.1031,
+      "step": 28870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017920687404297274,
+      "loss": 4.0469,
+      "step": 28875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017917172503756546,
+      "loss": 3.9422,
+      "step": 28880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017913657436732504,
+      "loss": 4.0094,
+      "step": 28885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017910142203425754,
+      "loss": 3.8969,
+      "step": 28890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017906626804036903,
+      "loss": 3.9312,
+      "step": 28895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001790311123876658,
+      "loss": 4.0141,
+      "step": 28900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017899595507815414,
+      "loss": 3.9344,
+      "step": 28905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001789607961138406,
+      "loss": 4.0031,
+      "step": 28910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001789256354967316,
+      "loss": 3.9719,
+      "step": 28915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017889047322883382,
+      "loss": 3.9438,
+      "step": 28920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788553093121539,
+      "loss": 3.9688,
+      "step": 28925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788201437486987,
+      "loss": 4.0984,
+      "step": 28930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787849765404751,
+      "loss": 3.9594,
+      "step": 28935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017874980768949015,
+      "loss": 3.9719,
+      "step": 28940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787146371977509,
+      "loss": 3.9969,
+      "step": 28945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001786794650672645,
+      "loss": 4.1016,
+      "step": 28950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017864429130003828,
+      "loss": 4.0266,
+      "step": 28955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017860911589807958,
+      "loss": 3.9734,
+      "step": 28960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017857393886339592,
+      "loss": 4.0344,
+      "step": 28965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017853876019799475,
+      "loss": 3.9391,
+      "step": 28970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001785035799038838,
+      "loss": 4.0531,
+      "step": 28975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001784683979830708,
+      "loss": 3.9969,
+      "step": 28980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017843321443756358,
+      "loss": 3.9719,
+      "step": 28985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017839802926937007,
+      "loss": 4.0516,
+      "step": 28990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017836284248049823,
+      "loss": 3.9172,
+      "step": 28995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001783276540729563,
+      "loss": 4.0297,
+      "step": 29000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001782924640487524,
+      "loss": 4.0172,
+      "step": 29005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017825727240989488,
+      "loss": 4.0312,
+      "step": 29010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017822207915839202,
+      "loss": 4.0312,
+      "step": 29015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017818688429625248,
+      "loss": 4.0031,
+      "step": 29020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017815168782548462,
+      "loss": 4.1031,
+      "step": 29025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001781164897480973,
+      "loss": 4.0438,
+      "step": 29030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017808129006609917,
+      "loss": 4.0297,
+      "step": 29035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001780460887814991,
+      "loss": 4.0281,
+      "step": 29040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000178010885896306,
+      "loss": 4.0266,
+      "step": 29045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000177975681412529,
+      "loss": 3.9953,
+      "step": 29050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017794047533217712,
+      "loss": 4.0594,
+      "step": 29055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017790526765725962,
+      "loss": 3.9859,
+      "step": 29060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017787005838978582,
+      "loss": 4.0563,
+      "step": 29065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017783484753176512,
+      "loss": 4.1047,
+      "step": 29070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017779963508520694,
+      "loss": 4.0328,
+      "step": 29075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001777644210521209,
+      "loss": 3.9937,
+      "step": 29080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017772920543451668,
+      "loss": 3.9625,
+      "step": 29085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017769398823440404,
+      "loss": 3.9766,
+      "step": 29090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001776587694537928,
+      "loss": 3.9438,
+      "step": 29095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017762354909469294,
+      "loss": 3.9531,
+      "step": 29100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017758832715911445,
+      "loss": 3.9391,
+      "step": 29105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001775531036490675,
+      "loss": 4.0094,
+      "step": 29110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001775178785665622,
+      "loss": 4.0516,
+      "step": 29115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000177482651913609,
+      "loss": 3.95,
+      "step": 29120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017744742369221812,
+      "loss": 4.0203,
+      "step": 29125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017741219390440013,
+      "loss": 4.0109,
+      "step": 29130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017737696255216564,
+      "loss": 3.9484,
+      "step": 29135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001773417296375252,
+      "loss": 4.0469,
+      "step": 29140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001773064951624897,
+      "loss": 3.9906,
+      "step": 29145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001772712591290698,
+      "loss": 4.1359,
+      "step": 29150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017723602153927654,
+      "loss": 4.1031,
+      "step": 29155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017720078239512088,
+      "loss": 3.8875,
+      "step": 29160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017716554169861399,
+      "loss": 3.9688,
+      "step": 29165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017713029945176694,
+      "loss": 3.9891,
+      "step": 29170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770950556565911,
+      "loss": 4.0359,
+      "step": 29175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017705981031509787,
+      "loss": 4.0281,
+      "step": 29180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017702456342929862,
+      "loss": 3.9953,
+      "step": 29185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017698931500120493,
+      "loss": 3.9125,
+      "step": 29190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017695406503282843,
+      "loss": 3.9891,
+      "step": 29195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017691881352618083,
+      "loss": 3.9813,
+      "step": 29200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017688356048327394,
+      "loss": 3.9734,
+      "step": 29205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017684830590611967,
+      "loss": 4.0422,
+      "step": 29210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017681304979673,
+      "loss": 3.9984,
+      "step": 29215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017677779215711695,
+      "loss": 3.9969,
+      "step": 29220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017674253298929275,
+      "loss": 4.0469,
+      "step": 29225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001767072722952696,
+      "loss": 4.0172,
+      "step": 29230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017667201007705985,
+      "loss": 4.0078,
+      "step": 29235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017663674633667587,
+      "loss": 4.0031,
+      "step": 29240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001766014810761303,
+      "loss": 4.0547,
+      "step": 29245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017656621429743555,
+      "loss": 3.9547,
+      "step": 29250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017653094600260442,
+      "loss": 4.0453,
+      "step": 29255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017649567619364964,
+      "loss": 3.9578,
+      "step": 29260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017646040487258403,
+      "loss": 3.9672,
+      "step": 29265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001764251320414206,
+      "loss": 4.0469,
+      "step": 29270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001763898577021723,
+      "loss": 4.0219,
+      "step": 29275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017635458185685232,
+      "loss": 3.9906,
+      "step": 29280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017631930450747374,
+      "loss": 3.9062,
+      "step": 29285
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017628402565605,
+      "loss": 4.0719,
+      "step": 29290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017624874530459432,
+      "loss": 3.9766,
+      "step": 29295
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001762134634551202,
+      "loss": 3.9797,
+      "step": 29300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001761781801096412,
+      "loss": 3.9797,
+      "step": 29305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017614289527017094,
+      "loss": 4.0328,
+      "step": 29310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017610760893872312,
+      "loss": 4.0234,
+      "step": 29315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017607232111731158,
+      "loss": 3.9609,
+      "step": 29320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001760370318079501,
+      "loss": 3.9406,
+      "step": 29325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017600174101265275,
+      "loss": 4.0875,
+      "step": 29330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001759664487334335,
+      "loss": 3.9766,
+      "step": 29335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001759311549723065,
+      "loss": 3.9906,
+      "step": 29340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017589585973128602,
+      "loss": 4.0016,
+      "step": 29345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001758605630123863,
+      "loss": 4.0016,
+      "step": 29350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017582526481762175,
+      "loss": 3.925,
+      "step": 29355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017578996514900684,
+      "loss": 4.0734,
+      "step": 29360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017575466400855615,
+      "loss": 3.9422,
+      "step": 29365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001757193613982843,
+      "loss": 3.9609,
+      "step": 29370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000175684057320206,
+      "loss": 3.9109,
+      "step": 29375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001756487517763361,
+      "loss": 4.0094,
+      "step": 29380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017561344476868944,
+      "loss": 4.0094,
+      "step": 29385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017557813629928102,
+      "loss": 3.9547,
+      "step": 29390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017554282637012583,
+      "loss": 4.0078,
+      "step": 29395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017550751498323914,
+      "loss": 3.9547,
+      "step": 29400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017547220214063613,
+      "loss": 3.9578,
+      "step": 29405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000175436887844332,
+      "loss": 4.0344,
+      "step": 29410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001754015720963423,
+      "loss": 4.0297,
+      "step": 29415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753662548986824,
+      "loss": 4.0531,
+      "step": 29420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753309362533679,
+      "loss": 4.0531,
+      "step": 29425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001752956161624144,
+      "loss": 4.0047,
+      "step": 29430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017526029462783767,
+      "loss": 3.9969,
+      "step": 29435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017522497165165342,
+      "loss": 3.9453,
+      "step": 29440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017518964723587765,
+      "loss": 3.9406,
+      "step": 29445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001751543213825263,
+      "loss": 3.9734,
+      "step": 29450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017511899409361535,
+      "loss": 3.9937,
+      "step": 29455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017508366537116109,
+      "loss": 4.0062,
+      "step": 29460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017504833521717953,
+      "loss": 4.0328,
+      "step": 29465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017501300363368706,
+      "loss": 3.9891,
+      "step": 29470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017497767062270007,
+      "loss": 4.1047,
+      "step": 29475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017494233618623505,
+      "loss": 4.025,
+      "step": 29480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017490700032630848,
+      "loss": 4.0062,
+      "step": 29485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017487166304493693,
+      "loss": 3.9984,
+      "step": 29490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017483632434413723,
+      "loss": 3.9719,
+      "step": 29495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017480098422592607,
+      "loss": 3.9797,
+      "step": 29500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001747656426923204,
+      "loss": 3.9781,
+      "step": 29505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001747302997453371,
+      "loss": 4.0031,
+      "step": 29510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017469495538699318,
+      "loss": 3.9953,
+      "step": 29515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017465960961930577,
+      "loss": 4.0,
+      "step": 29520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017462426244429206,
+      "loss": 3.9453,
+      "step": 29525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017458891386396932,
+      "loss": 4.0609,
+      "step": 29530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017455356388035488,
+      "loss": 4.0344,
+      "step": 29535
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017451821249546624,
+      "loss": 3.9828,
+      "step": 29540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017448285971132077,
+      "loss": 4.0312,
+      "step": 29545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017444750552993618,
+      "loss": 4.0906,
+      "step": 29550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017441214995333006,
+      "loss": 4.0687,
+      "step": 29555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001743767929835202,
+      "loss": 4.0187,
+      "step": 29560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017434143462252442,
+      "loss": 4.0328,
+      "step": 29565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017430607487236056,
+      "loss": 3.9422,
+      "step": 29570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001742707137350467,
+      "loss": 4.0516,
+      "step": 29575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017423535121260087,
+      "loss": 4.0328,
+      "step": 29580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741999873070412,
+      "loss": 4.1312,
+      "step": 29585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017416462202038592,
+      "loss": 3.9828,
+      "step": 29590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017412925535465333,
+      "loss": 4.0391,
+      "step": 29595
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001740938873118618,
+      "loss": 3.9734,
+      "step": 29600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017405851789402985,
+      "loss": 4.0578,
+      "step": 29605
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001740231471031759,
+      "loss": 3.9734,
+      "step": 29610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017398777494131865,
+      "loss": 3.9312,
+      "step": 29615
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017395240141047675,
+      "loss": 4.0547,
+      "step": 29620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017391702651266898,
+      "loss": 4.0391,
+      "step": 29625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017388165024991424,
+      "loss": 3.9922,
+      "step": 29630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738462726242314,
+      "loss": 3.9234,
+      "step": 29635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738108936376395,
+      "loss": 4.0297,
+      "step": 29640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017377551329215754,
+      "loss": 4.0766,
+      "step": 29645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017374013158980478,
+      "loss": 3.9672,
+      "step": 29650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001737047485326004,
+      "loss": 3.9625,
+      "step": 29655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017366936412256377,
+      "loss": 4.0469,
+      "step": 29660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001736339783617142,
+      "loss": 3.9937,
+      "step": 29665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017359859125207124,
+      "loss": 3.9297,
+      "step": 29670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735632027956544,
+      "loss": 4.05,
+      "step": 29675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017352781299448327,
+      "loss": 4.0172,
+      "step": 29680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734924218505776,
+      "loss": 3.9516,
+      "step": 29685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017345702936595714,
+      "loss": 4.025,
+      "step": 29690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017342163554264178,
+      "loss": 4.0375,
+      "step": 29695
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017338624038265138,
+      "loss": 3.9984,
+      "step": 29700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000173350843888006,
+      "loss": 4.0375,
+      "step": 29705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001733154460607257,
+      "loss": 3.9562,
+      "step": 29710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017328004690283063,
+      "loss": 4.0578,
+      "step": 29715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017324464641634108,
+      "loss": 4.0344,
+      "step": 29720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017320924460327726,
+      "loss": 4.0422,
+      "step": 29725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017317384146565967,
+      "loss": 3.9219,
+      "step": 29730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017313843700550867,
+      "loss": 3.975,
+      "step": 29735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017310303122484484,
+      "loss": 3.9844,
+      "step": 29740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001730676241256888,
+      "loss": 3.9672,
+      "step": 29745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017303221571006122,
+      "loss": 3.9672,
+      "step": 29750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017299680597998284,
+      "loss": 4.0234,
+      "step": 29755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017296139493747456,
+      "loss": 4.0141,
+      "step": 29760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017292598258455725,
+      "loss": 3.9328,
+      "step": 29765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017289056892325186,
+      "loss": 3.9516,
+      "step": 29770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017285515395557954,
+      "loss": 4.0219,
+      "step": 29775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017281973768356137,
+      "loss": 4.0484,
+      "step": 29780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017278432010921856,
+      "loss": 4.0062,
+      "step": 29785
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017274890123457237,
+      "loss": 3.9828,
+      "step": 29790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017271348106164423,
+      "loss": 4.05,
+      "step": 29795
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017267805959245554,
+      "loss": 3.9922,
+      "step": 29800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017264263682902775,
+      "loss": 4.0469,
+      "step": 29805
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001726072127733825,
+      "loss": 3.9781,
+      "step": 29810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017257178742754146,
+      "loss": 3.9781,
+      "step": 29815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017253636079352636,
+      "loss": 3.9219,
+      "step": 29820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017250093287335896,
+      "loss": 3.9781,
+      "step": 29825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724655036690611,
+      "loss": 4.0422,
+      "step": 29830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017243007318265482,
+      "loss": 3.9547,
+      "step": 29835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001723946414161621,
+      "loss": 4.0047,
+      "step": 29840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000172359208371605,
+      "loss": 3.975,
+      "step": 29845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017232377405100574,
+      "loss": 4.0438,
+      "step": 29850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017228833845638658,
+      "loss": 3.9328,
+      "step": 29855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001722529015897697,
+      "loss": 4.0047,
+      "step": 29860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017221746345317764,
+      "loss": 3.9141,
+      "step": 29865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001721820240486328,
+      "loss": 3.9531,
+      "step": 29870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017214658337815769,
+      "loss": 3.9688,
+      "step": 29875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017211114144377495,
+      "loss": 4.0203,
+      "step": 29880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017207569824750723,
+      "loss": 3.9578,
+      "step": 29885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017204025379137722,
+      "loss": 4.0344,
+      "step": 29890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017200480807740785,
+      "loss": 3.9969,
+      "step": 29895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017196936110762197,
+      "loss": 3.9391,
+      "step": 29900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001719339128840425,
+      "loss": 4.0062,
+      "step": 29905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017189846340869255,
+      "loss": 4.0375,
+      "step": 29910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017186301268359513,
+      "loss": 3.9016,
+      "step": 29915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001718275607107735,
+      "loss": 4.0219,
+      "step": 29920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017179210749225084,
+      "loss": 4.025,
+      "step": 29925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017175665303005054,
+      "loss": 3.9875,
+      "step": 29930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017172119732619592,
+      "loss": 4.0672,
+      "step": 29935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017168574038271047,
+      "loss": 3.8828,
+      "step": 29940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017165028220161773,
+      "loss": 4.0141,
+      "step": 29945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001716148227849413,
+      "loss": 3.9422,
+      "step": 29950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001715793621347049,
+      "loss": 3.8312,
+      "step": 29955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017154390025293215,
+      "loss": 3.9547,
+      "step": 29960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017150843714164695,
+      "loss": 3.9469,
+      "step": 29965
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017147297280287316,
+      "loss": 3.9969,
+      "step": 29970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017143750723863475,
+      "loss": 3.9859,
+      "step": 29975
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017140204045095573,
+      "loss": 3.9125,
+      "step": 29980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001713665724418602,
+      "loss": 3.9891,
+      "step": 29985
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001713311032133723,
+      "loss": 3.9906,
+      "step": 29990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712956327675163,
+      "loss": 3.95,
+      "step": 29995
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017126016110631648,
+      "loss": 4.0094,
+      "step": 30000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712246882317972,
+      "loss": 3.9703,
+      "step": 30005
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017118921414598293,
+      "loss": 3.9297,
+      "step": 30010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017115373885089814,
+      "loss": 3.9766,
+      "step": 30015
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017111826234856745,
+      "loss": 4.0438,
+      "step": 30020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001710827846410155,
+      "loss": 3.9531,
+      "step": 30025
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017104730573026694,
+      "loss": 4.0031,
+      "step": 30030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017101182561834669,
+      "loss": 3.9328,
+      "step": 30035
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017097634430727948,
+      "loss": 3.8906,
+      "step": 30040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001709408617990903,
+      "loss": 3.9781,
+      "step": 30045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017090537809580405,
+      "loss": 4.0375,
+      "step": 30050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001708698931994459,
+      "loss": 4.0703,
+      "step": 30055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017083440711204096,
+      "loss": 3.9094,
+      "step": 30060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017079891983561436,
+      "loss": 3.95,
+      "step": 30065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017076343137219138,
+      "loss": 4.0156,
+      "step": 30070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017072794172379738,
+      "loss": 3.9719,
+      "step": 30075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017069245089245777,
+      "loss": 4.0141,
+      "step": 30080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017065695888019797,
+      "loss": 4.0391,
+      "step": 30085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017062146568904354,
+      "loss": 3.9906,
+      "step": 30090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017058597132102007,
+      "loss": 3.9391,
+      "step": 30095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017055047577815323,
+      "loss": 3.9125,
+      "step": 30100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017051497906246874,
+      "loss": 4.0,
+      "step": 30105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017047948117599243,
+      "loss": 3.9734,
+      "step": 30110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017044398212075014,
+      "loss": 4.0078,
+      "step": 30115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001704084818987678,
+      "loss": 4.0109,
+      "step": 30120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017037298051207148,
+      "loss": 3.9312,
+      "step": 30125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017033747796268718,
+      "loss": 3.9594,
+      "step": 30130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017030197425264105,
+      "loss": 3.9188,
+      "step": 30135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001702664693839593,
+      "loss": 4.0844,
+      "step": 30140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017023096335866815,
+      "loss": 4.0375,
+      "step": 30145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000170195456178794,
+      "loss": 4.0953,
+      "step": 30150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701599478463632,
+      "loss": 3.9484,
+      "step": 30155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701244383634022,
+      "loss": 3.8922,
+      "step": 30160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001700889277319376,
+      "loss": 3.9953,
+      "step": 30165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017005341595399597,
+      "loss": 3.9422,
+      "step": 30170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017001790303160393,
+      "loss": 4.0312,
+      "step": 30175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016998238896678825,
+      "loss": 3.9609,
+      "step": 30180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016994687376157565,
+      "loss": 4.0641,
+      "step": 30185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001699113574179931,
+      "loss": 4.0484,
+      "step": 30190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016987583993806742,
+      "loss": 3.9875,
+      "step": 30195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001698403213238257,
+      "loss": 3.9781,
+      "step": 30200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016980480157729484,
+      "loss": 4.0141,
+      "step": 30205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016976928070050205,
+      "loss": 4.0047,
+      "step": 30210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016973375869547454,
+      "loss": 3.9859,
+      "step": 30215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001696982355642395,
+      "loss": 4.0156,
+      "step": 30220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016966271130882426,
+      "loss": 3.9172,
+      "step": 30225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001696271859312562,
+      "loss": 3.9844,
+      "step": 30230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016959165943356276,
+      "loss": 4.0156,
+      "step": 30235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016955613181777137,
+      "loss": 4.0516,
+      "step": 30240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001695206030859097,
+      "loss": 4.0375,
+      "step": 30245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694850732400053,
+      "loss": 3.9469,
+      "step": 30250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694495422820859,
+      "loss": 4.0234,
+      "step": 30255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016941401021417925,
+      "loss": 3.975,
+      "step": 30260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016937847703831315,
+      "loss": 4.025,
+      "step": 30265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016934294275651552,
+      "loss": 3.9781,
+      "step": 30270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016930740737081422,
+      "loss": 3.9922,
+      "step": 30275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001692718708832374,
+      "loss": 3.9813,
+      "step": 30280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016923633329581298,
+      "loss": 3.9203,
+      "step": 30285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016920079461056918,
+      "loss": 3.9312,
+      "step": 30290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016916525482953417,
+      "loss": 3.975,
+      "step": 30295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016912971395473625,
+      "loss": 3.9547,
+      "step": 30300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016909417198820369,
+      "loss": 3.9141,
+      "step": 30305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001690586289319649,
+      "loss": 3.9516,
+      "step": 30310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001690230847880483,
+      "loss": 3.9391,
+      "step": 30315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689875395584824,
+      "loss": 3.9906,
+      "step": 30320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689519932452958,
+      "loss": 4.075,
+      "step": 30325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016891644585051712,
+      "loss": 4.0266,
+      "step": 30330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016888089737617507,
+      "loss": 3.9141,
+      "step": 30335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016884534782429835,
+      "loss": 3.9375,
+      "step": 30340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001688097971969158,
+      "loss": 4.0047,
+      "step": 30345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016877424549605635,
+      "loss": 3.9625,
+      "step": 30350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016873869272374887,
+      "loss": 4.1203,
+      "step": 30355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016870313888202242,
+      "loss": 3.9937,
+      "step": 30360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000168667583972906,
+      "loss": 4.0234,
+      "step": 30365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001686320279984288,
+      "loss": 3.975,
+      "step": 30370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016859647096061992,
+      "loss": 3.9641,
+      "step": 30375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016856091286150866,
+      "loss": 4.0719,
+      "step": 30380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001685253537031243,
+      "loss": 4.0391,
+      "step": 30385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016848979348749621,
+      "loss": 3.9781,
+      "step": 30390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016845423221665385,
+      "loss": 3.9141,
+      "step": 30395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016841866989262665,
+      "loss": 4.0141,
+      "step": 30400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001683831065174442,
+      "loss": 3.9188,
+      "step": 30405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001683475420931361,
+      "loss": 3.9859,
+      "step": 30410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016831197662173198,
+      "loss": 3.9891,
+      "step": 30415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001682764101052616,
+      "loss": 3.9859,
+      "step": 30420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016824084254575474,
+      "loss": 4.0234,
+      "step": 30425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001682052739452412,
+      "loss": 3.9922,
+      "step": 30430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001681697043057509,
+      "loss": 3.9656,
+      "step": 30435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001681341336293139,
+      "loss": 4.1031,
+      "step": 30440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016809856191796007,
+      "loss": 3.9859,
+      "step": 30445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001680629891737196,
+      "loss": 4.0453,
+      "step": 30450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016802741539862257,
+      "loss": 3.9766,
+      "step": 30455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016799184059469925,
+      "loss": 3.9625,
+      "step": 30460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001679562647639798,
+      "loss": 3.9922,
+      "step": 30465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016792068790849463,
+      "loss": 3.9859,
+      "step": 30470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016788511003027403,
+      "loss": 3.9906,
+      "step": 30475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016784953113134848,
+      "loss": 3.9781,
+      "step": 30480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001678139512137485,
+      "loss": 4.0406,
+      "step": 30485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016777837027950456,
+      "loss": 3.9281,
+      "step": 30490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016774278833064736,
+      "loss": 3.9719,
+      "step": 30495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677072053692075,
+      "loss": 3.9219,
+      "step": 30500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001676716213972157,
+      "loss": 3.9875,
+      "step": 30505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016763603641670278,
+      "loss": 4.0156,
+      "step": 30510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016760045042969954,
+      "loss": 4.0891,
+      "step": 30515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001675648634382369,
+      "loss": 3.9234,
+      "step": 30520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001675292754443458,
+      "loss": 3.9797,
+      "step": 30525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001674936864500573,
+      "loss": 3.9188,
+      "step": 30530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001674580964574024,
+      "loss": 3.9469,
+      "step": 30535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016742250546841227,
+      "loss": 4.0062,
+      "step": 30540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016738691348511803,
+      "loss": 4.0359,
+      "step": 30545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016735132050955094,
+      "loss": 3.9813,
+      "step": 30550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016731572654374237,
+      "loss": 3.9062,
+      "step": 30555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016728013158972359,
+      "loss": 4.0109,
+      "step": 30560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016724453564952605,
+      "loss": 3.9984,
+      "step": 30565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001672089387251812,
+      "loss": 3.9625,
+      "step": 30570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016717334081872053,
+      "loss": 3.9578,
+      "step": 30575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016713774193217564,
+      "loss": 3.9625,
+      "step": 30580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001671021420675782,
+      "loss": 3.9328,
+      "step": 30585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001670665412269598,
+      "loss": 4.0094,
+      "step": 30590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016703093941235233,
+      "loss": 3.9984,
+      "step": 30595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016699533662578744,
+      "loss": 4.0062,
+      "step": 30600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016695973286929707,
+      "loss": 4.0187,
+      "step": 30605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016692412814491307,
+      "loss": 3.9359,
+      "step": 30610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016688852245466747,
+      "loss": 4.0141,
+      "step": 30615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016685291580059226,
+      "loss": 3.9719,
+      "step": 30620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001668173081847195,
+      "loss": 3.9781,
+      "step": 30625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001667816996090814,
+      "loss": 3.9594,
+      "step": 30630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016674609007571003,
+      "loss": 3.9469,
+      "step": 30635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016671047958663774,
+      "loss": 4.0563,
+      "step": 30640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001666748681438967,
+      "loss": 4.0672,
+      "step": 30645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016663925574951938,
+      "loss": 3.9906,
+      "step": 30650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001666036424055381,
+      "loss": 4.0297,
+      "step": 30655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001665680281139853,
+      "loss": 4.0109,
+      "step": 30660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016653241287689361,
+      "loss": 4.05,
+      "step": 30665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664967966962955,
+      "loss": 3.9922,
+      "step": 30670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016646117957422365,
+      "loss": 3.975,
+      "step": 30675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016642556151271062,
+      "loss": 3.9562,
+      "step": 30680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016638994251378926,
+      "loss": 4.1063,
+      "step": 30685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016635432257949224,
+      "loss": 4.0031,
+      "step": 30690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016631870171185249,
+      "loss": 4.0672,
+      "step": 30695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016628307991290288,
+      "loss": 4.1031,
+      "step": 30700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016624745718467625,
+      "loss": 4.0047,
+      "step": 30705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016621183352920577,
+      "loss": 4.0031,
+      "step": 30710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016617620894852434,
+      "loss": 3.9797,
+      "step": 30715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001661405834446651,
+      "loss": 3.9891,
+      "step": 30720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001661049570196612,
+      "loss": 3.9875,
+      "step": 30725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016606932967554587,
+      "loss": 3.9578,
+      "step": 30730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001660337014143523,
+      "loss": 4.0312,
+      "step": 30735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659980722381139,
+      "loss": 3.9641,
+      "step": 30740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016596244214886393,
+      "loss": 4.0234,
+      "step": 30745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016592681114863583,
+      "loss": 4.0469,
+      "step": 30750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016589117923946316,
+      "loss": 3.9953,
+      "step": 30755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016585554642337924,
+      "loss": 3.8922,
+      "step": 30760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016581991270241787,
+      "loss": 3.9266,
+      "step": 30765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016578427807861247,
+      "loss": 3.9359,
+      "step": 30770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016574864255399687,
+      "loss": 4.0406,
+      "step": 30775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016571300613060465,
+      "loss": 3.9578,
+      "step": 30780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016567736881046975,
+      "loss": 3.9641,
+      "step": 30785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656417305956258,
+      "loss": 3.9188,
+      "step": 30790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656060914881068,
+      "loss": 3.9188,
+      "step": 30795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016557045148994667,
+      "loss": 3.9344,
+      "step": 30800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016553481060317934,
+      "loss": 3.9047,
+      "step": 30805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016549916882983888,
+      "loss": 3.9078,
+      "step": 30810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001654635261719594,
+      "loss": 4.0391,
+      "step": 30815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016542788263157492,
+      "loss": 4.0172,
+      "step": 30820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016539223821071972,
+      "loss": 4.1109,
+      "step": 30825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000165356592911428,
+      "loss": 3.975,
+      "step": 30830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016532094673573408,
+      "loss": 3.9453,
+      "step": 30835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016528529968567218,
+      "loss": 3.9562,
+      "step": 30840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001652496517632768,
+      "loss": 3.9641,
+      "step": 30845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001652140029705823,
+      "loss": 3.9594,
+      "step": 30850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001651783533096232,
+      "loss": 4.0344,
+      "step": 30855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000165142702782434,
+      "loss": 4.0812,
+      "step": 30860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016510705139104932,
+      "loss": 3.9625,
+      "step": 30865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016507139913750373,
+      "loss": 4.0078,
+      "step": 30870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016503574602383197,
+      "loss": 4.0109,
+      "step": 30875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016500009205206866,
+      "loss": 3.9641,
+      "step": 30880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001649644372242487,
+      "loss": 3.9766,
+      "step": 30885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016492878154240689,
+      "loss": 3.9344,
+      "step": 30890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016489312500857803,
+      "loss": 4.0203,
+      "step": 30895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001648574676247971,
+      "loss": 3.8969,
+      "step": 30900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016482180939309904,
+      "loss": 4.0078,
+      "step": 30905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016478615031551892,
+      "loss": 3.9625,
+      "step": 30910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016475049039409176,
+      "loss": 4.0125,
+      "step": 30915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001647148296308527,
+      "loss": 4.0828,
+      "step": 30920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016467916802783688,
+      "loss": 4.0062,
+      "step": 30925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001646435055870795,
+      "loss": 3.9438,
+      "step": 30930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001646078423106159,
+      "loss": 4.025,
+      "step": 30935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016457217820048126,
+      "loss": 4.0781,
+      "step": 30940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016453651325871104,
+      "loss": 3.975,
+      "step": 30945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016450084748734063,
+      "loss": 3.9672,
+      "step": 30950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016446518088840542,
+      "loss": 3.9047,
+      "step": 30955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016442951346394094,
+      "loss": 3.9984,
+      "step": 30960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016439384521598274,
+      "loss": 3.9719,
+      "step": 30965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001643581761465664,
+      "loss": 3.9359,
+      "step": 30970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016432250625772756,
+      "loss": 4.0094,
+      "step": 30975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016428683555150194,
+      "loss": 3.8453,
+      "step": 30980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001642511640299252,
+      "loss": 4.0563,
+      "step": 30985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016421549169503319,
+      "loss": 3.9516,
+      "step": 30990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016417981854886164,
+      "loss": 3.9469,
+      "step": 30995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016414414459344653,
+      "loss": 4.0266,
+      "step": 31000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001641084698308237,
+      "loss": 4.0219,
+      "step": 31005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016407279426302912,
+      "loss": 3.9922,
+      "step": 31010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016403711789209882,
+      "loss": 3.9281,
+      "step": 31015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016400144072006886,
+      "loss": 3.9703,
+      "step": 31020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016396576274897533,
+      "loss": 4.0547,
+      "step": 31025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016393008398085435,
+      "loss": 3.9844,
+      "step": 31030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016389440441774217,
+      "loss": 3.9813,
+      "step": 31035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016385872406167492,
+      "loss": 4.0234,
+      "step": 31040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016382304291468898,
+      "loss": 4.0203,
+      "step": 31045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016378736097882068,
+      "loss": 4.0672,
+      "step": 31050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016375167825610632,
+      "loss": 4.0297,
+      "step": 31055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016371599474858238,
+      "loss": 3.9984,
+      "step": 31060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016368031045828528,
+      "loss": 3.8953,
+      "step": 31065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016364462538725157,
+      "loss": 3.8984,
+      "step": 31070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016360893953751776,
+      "loss": 3.9172,
+      "step": 31075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016357325291112043,
+      "loss": 3.8797,
+      "step": 31080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016353756551009625,
+      "loss": 4.0266,
+      "step": 31085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016350187733648193,
+      "loss": 3.9922,
+      "step": 31090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016346618839231412,
+      "loss": 3.9953,
+      "step": 31095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001634304986796297,
+      "loss": 4.0281,
+      "step": 31100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016339480820046537,
+      "loss": 4.0438,
+      "step": 31105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016335911695685806,
+      "loss": 4.0141,
+      "step": 31110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001633234249508447,
+      "loss": 3.9172,
+      "step": 31115
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016328773218446212,
+      "loss": 3.9344,
+      "step": 31120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016325203865974743,
+      "loss": 4.0047,
+      "step": 31125
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001632163443787376,
+      "loss": 4.0172,
+      "step": 31130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016318064934346972,
+      "loss": 4.0438,
+      "step": 31135
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016314495355598092,
+      "loss": 3.9703,
+      "step": 31140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016310925701830834,
+      "loss": 3.9844,
+      "step": 31145
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630735597324892,
+      "loss": 4.0187,
+      "step": 31150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016303786170056072,
+      "loss": 3.9656,
+      "step": 31155
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016300216292456025,
+      "loss": 3.8906,
+      "step": 31160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016296646340652508,
+      "loss": 3.9828,
+      "step": 31165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001629307631484926,
+      "loss": 4.0469,
+      "step": 31170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016289506215250017,
+      "loss": 3.9719,
+      "step": 31175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016285936042058537,
+      "loss": 3.975,
+      "step": 31180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016282365795478558,
+      "loss": 3.9937,
+      "step": 31185
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001627879547571384,
+      "loss": 4.0547,
+      "step": 31190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001627522508296814,
+      "loss": 3.9609,
+      "step": 31195
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016271654617445221,
+      "loss": 4.0109,
+      "step": 31200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016268084079348855,
+      "loss": 3.9906,
+      "step": 31205
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016264513468882801,
+      "loss": 3.9391,
+      "step": 31210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016260942786250847,
+      "loss": 4.0141,
+      "step": 31215
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016257372031656765,
+      "loss": 3.8547,
+      "step": 31220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016253801205304343,
+      "loss": 3.9937,
+      "step": 31225
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001625023030739736,
+      "loss": 4.0156,
+      "step": 31230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016246659338139613,
+      "loss": 3.9937,
+      "step": 31235
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016243088297734898,
+      "loss": 4.0672,
+      "step": 31240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016239517186387016,
+      "loss": 3.9953,
+      "step": 31245
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016235946004299767,
+      "loss": 4.0109,
+      "step": 31250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001623237475167696,
+      "loss": 3.9734,
+      "step": 31255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001622880342872241,
+      "loss": 3.9875,
+      "step": 31260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016225232035639922,
+      "loss": 4.0469,
+      "step": 31265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016221660572633332,
+      "loss": 3.8641,
+      "step": 31270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016218089039906448,
+      "loss": 3.9875,
+      "step": 31275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016214517437663112,
+      "loss": 4.0016,
+      "step": 31280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001621094576610714,
+      "loss": 3.8859,
+      "step": 31285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016207374025442383,
+      "loss": 4.0187,
+      "step": 31290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016203802215872674,
+      "loss": 3.9984,
+      "step": 31295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016200230337601852,
+      "loss": 4.0109,
+      "step": 31300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001619665839083377,
+      "loss": 3.9469,
+      "step": 31305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001619308637577228,
+      "loss": 3.9344,
+      "step": 31310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016189514292621235,
+      "loss": 3.9734,
+      "step": 31315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016185942141584494,
+      "loss": 4.0203,
+      "step": 31320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001618236992286592,
+      "loss": 3.8984,
+      "step": 31325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016178797636669378,
+      "loss": 4.0469,
+      "step": 31330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016175225283198745,
+      "loss": 4.0406,
+      "step": 31335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001617165286265789,
+      "loss": 4.0516,
+      "step": 31340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016168080375250694,
+      "loss": 4.025,
+      "step": 31345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001616450782118104,
+      "loss": 3.9328,
+      "step": 31350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001616093520065281,
+      "loss": 3.9984,
+      "step": 31355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016157362513869898,
+      "loss": 3.9172,
+      "step": 31360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016153789761036194,
+      "loss": 4.0,
+      "step": 31365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000161502169423556,
+      "loss": 3.9688,
+      "step": 31370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016146644058032006,
+      "loss": 4.0453,
+      "step": 31375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016143071108269327,
+      "loss": 3.9656,
+      "step": 31380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016139498093271476,
+      "loss": 4.0172,
+      "step": 31385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016135925013242353,
+      "loss": 3.9891,
+      "step": 31390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016132351868385886,
+      "loss": 3.9922,
+      "step": 31395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001612877865890598,
+      "loss": 3.9391,
+      "step": 31400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016125205385006576,
+      "loss": 3.9891,
+      "step": 31405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016121632046891586,
+      "loss": 3.9484,
+      "step": 31410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001611805864476495,
+      "loss": 3.95,
+      "step": 31415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016114485178830593,
+      "loss": 3.9891,
+      "step": 31420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016110911649292467,
+      "loss": 3.9984,
+      "step": 31425
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016107338056354503,
+      "loss": 3.9031,
+      "step": 31430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016103764400220649,
+      "loss": 3.9266,
+      "step": 31435
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016100190681094854,
+      "loss": 4.0094,
+      "step": 31440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001609661689918107,
+      "loss": 3.9516,
+      "step": 31445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016093043054683258,
+      "loss": 4.0031,
+      "step": 31450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016089469147805366,
+      "loss": 3.9344,
+      "step": 31455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001608589517875137,
+      "loss": 3.9875,
+      "step": 31460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016082321147725228,
+      "loss": 3.9672,
+      "step": 31465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016078747054930917,
+      "loss": 3.9906,
+      "step": 31470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016075172900572405,
+      "loss": 4.0406,
+      "step": 31475
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016071598684853672,
+      "loss": 3.9641,
+      "step": 31480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160680244079787,
+      "loss": 3.975,
+      "step": 31485
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001606445007015147,
+      "loss": 3.9344,
+      "step": 31490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016060875671575976,
+      "loss": 3.8859,
+      "step": 31495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160573012124562,
+      "loss": 3.9547,
+      "step": 31500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016053726692996144,
+      "loss": 3.9578,
+      "step": 31505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016050152113399803,
+      "loss": 4.0281,
+      "step": 31510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016046577473871179,
+      "loss": 4.0453,
+      "step": 31515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001604300277461428,
+      "loss": 4.1437,
+      "step": 31520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016039428015833111,
+      "loss": 3.9422,
+      "step": 31525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016035853197731684,
+      "loss": 3.9016,
+      "step": 31530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016032278320514013,
+      "loss": 4.025,
+      "step": 31535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001602870338438412,
+      "loss": 3.9688,
+      "step": 31540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016025128389546032,
+      "loss": 4.0187,
+      "step": 31545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016021553336203757,
+      "loss": 4.0,
+      "step": 31550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001601797822456134,
+      "loss": 4.1531,
+      "step": 31555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016014403054822807,
+      "loss": 3.9859,
+      "step": 31560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016010827827192192,
+      "loss": 3.9641,
+      "step": 31565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001600725254187354,
+      "loss": 3.975,
+      "step": 31570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016003677199070884,
+      "loss": 4.0156,
+      "step": 31575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016000101798988275,
+      "loss": 4.0438,
+      "step": 31580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015996526341829762,
+      "loss": 4.0656,
+      "step": 31585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015992950827799392,
+      "loss": 4.0047,
+      "step": 31590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015989375257101222,
+      "loss": 4.0625,
+      "step": 31595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001598579962993931,
+      "loss": 3.8609,
+      "step": 31600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015982223946517718,
+      "loss": 4.0438,
+      "step": 31605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015978648207040514,
+      "loss": 3.9781,
+      "step": 31610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015975072411711757,
+      "loss": 3.9094,
+      "step": 31615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001597149656073553,
+      "loss": 3.9922,
+      "step": 31620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015967920654315896,
+      "loss": 3.9859,
+      "step": 31625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001596434469265694,
+      "loss": 4.0062,
+      "step": 31630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015960768675962733,
+      "loss": 3.9469,
+      "step": 31635
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015957192604437373,
+      "loss": 4.0594,
+      "step": 31640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001595361647828493,
+      "loss": 3.9937,
+      "step": 31645
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015950040297709505,
+      "loss": 4.0031,
+      "step": 31650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015946464062915197,
+      "loss": 3.9703,
+      "step": 31655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015942887774106083,
+      "loss": 4.0797,
+      "step": 31660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001593931143148628,
+      "loss": 3.9516,
+      "step": 31665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001593573503525988,
+      "loss": 3.9453,
+      "step": 31670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001593215858563099,
+      "loss": 3.9781,
+      "step": 31675
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001592858208280372,
+      "loss": 4.0406,
+      "step": 31680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015925005526982186,
+      "loss": 4.0047,
+      "step": 31685
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001592142891837049,
+      "loss": 3.9422,
+      "step": 31690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591785225717276,
+      "loss": 3.9875,
+      "step": 31695
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591427554359311,
+      "loss": 3.9844,
+      "step": 31700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591069877783567,
+      "loss": 3.9234,
+      "step": 31705
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015907121960104565,
+      "loss": 3.9438,
+      "step": 31710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015903545090603918,
+      "loss": 3.9937,
+      "step": 31715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015899968169537867,
+      "loss": 4.0281,
+      "step": 31720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589639119711054,
+      "loss": 3.9406,
+      "step": 31725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015892814173526092,
+      "loss": 3.9797,
+      "step": 31730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015889237098988643,
+      "loss": 3.9344,
+      "step": 31735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001588565997370235,
+      "loss": 4.0187,
+      "step": 31740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015882082797871361,
+      "loss": 3.9953,
+      "step": 31745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015878505571699817,
+      "loss": 3.9562,
+      "step": 31750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015874928295391878,
+      "loss": 3.9766,
+      "step": 31755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015871350969151693,
+      "loss": 3.9766,
+      "step": 31760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015867773593183427,
+      "loss": 4.0531,
+      "step": 31765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015864196167691237,
+      "loss": 3.95,
+      "step": 31770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001586061869287929,
+      "loss": 4.0484,
+      "step": 31775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015857041168951749,
+      "loss": 3.8891,
+      "step": 31780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015853463596112786,
+      "loss": 4.0547,
+      "step": 31785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015849885974566577,
+      "loss": 3.9172,
+      "step": 31790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001584630830451729,
+      "loss": 3.9734,
+      "step": 31795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000158427305861691,
+      "loss": 4.0078,
+      "step": 31800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015839152819726207,
+      "loss": 4.0531,
+      "step": 31805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015835575005392774,
+      "loss": 4.0266,
+      "step": 31810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015831997143372993,
+      "loss": 4.0094,
+      "step": 31815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001582841923387106,
+      "loss": 4.0391,
+      "step": 31820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015824841277091157,
+      "loss": 3.9813,
+      "step": 31825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015821263273237482,
+      "loss": 3.9391,
+      "step": 31830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015817685222514237,
+      "loss": 3.9266,
+      "step": 31835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015814107125125616,
+      "loss": 3.9609,
+      "step": 31840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015810528981275826,
+      "loss": 3.975,
+      "step": 31845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015806950791169062,
+      "loss": 4.0109,
+      "step": 31850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015803372555009545,
+      "loss": 4.0641,
+      "step": 31855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015799794273001474,
+      "loss": 3.9453,
+      "step": 31860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001579621594534907,
+      "loss": 3.9297,
+      "step": 31865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001579263757225654,
+      "loss": 3.9406,
+      "step": 31870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015789059153928112,
+      "loss": 3.9719,
+      "step": 31875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015785480690568002,
+      "loss": 3.9516,
+      "step": 31880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001578190218238043,
+      "loss": 3.9422,
+      "step": 31885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001577832362956963,
+      "loss": 3.9438,
+      "step": 31890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015774745032339822,
+      "loss": 3.8766,
+      "step": 31895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015771166390895239,
+      "loss": 3.9266,
+      "step": 31900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001576758770544012,
+      "loss": 4.0359,
+      "step": 31905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001576400897617869,
+      "loss": 3.9422,
+      "step": 31910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000157604302033152,
+      "loss": 3.9578,
+      "step": 31915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001575685138705388,
+      "loss": 3.9125,
+      "step": 31920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015753272527598984,
+      "loss": 4.0125,
+      "step": 31925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015749693625154745,
+      "loss": 4.0062,
+      "step": 31930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015746114679925425,
+      "loss": 3.9656,
+      "step": 31935
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015742535692115265,
+      "loss": 4.0062,
+      "step": 31940
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015738956661928523,
+      "loss": 4.0969,
+      "step": 31945
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015735377589569448,
+      "loss": 3.9859,
+      "step": 31950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001573179847524231,
+      "loss": 3.9859,
+      "step": 31955
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015728219319151358,
+      "loss": 3.9859,
+      "step": 31960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015724640121500858,
+      "loss": 4.0578,
+      "step": 31965
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015721060882495082,
+      "loss": 3.9203,
+      "step": 31970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015717481602338284,
+      "loss": 3.8953,
+      "step": 31975
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001571390228123475,
+      "loss": 3.975,
+      "step": 31980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015710322919388738,
+      "loss": 3.9547,
+      "step": 31985
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001570674351700453,
+      "loss": 3.9594,
+      "step": 31990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015703164074286406,
+      "loss": 3.975,
+      "step": 31995
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015699584591438636,
+      "loss": 4.0406,
+      "step": 32000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015696005068665506,
+      "loss": 4.0219,
+      "step": 32005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015692425506171298,
+      "loss": 3.9562,
+      "step": 32010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000156888459041603,
+      "loss": 3.9719,
+      "step": 32015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015685266262836802,
+      "loss": 3.975,
+      "step": 32020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015681686582405093,
+      "loss": 4.0344,
+      "step": 32025
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001567810686306946,
+      "loss": 3.9188,
+      "step": 32030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015674527105034207,
+      "loss": 4.0438,
+      "step": 32035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015670947308503625,
+      "loss": 3.9578,
+      "step": 32040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015667367473682015,
+      "loss": 4.0297,
+      "step": 32045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566378760077368,
+      "loss": 4.0219,
+      "step": 32050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015660207689982924,
+      "loss": 4.0406,
+      "step": 32055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015656627741514047,
+      "loss": 3.9078,
+      "step": 32060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015653047755571366,
+      "loss": 4.1141,
+      "step": 32065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015649467732359185,
+      "loss": 3.9578,
+      "step": 32070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015645887672081818,
+      "loss": 4.0156,
+      "step": 32075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001564230757494358,
+      "loss": 3.9875,
+      "step": 32080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015638727441148783,
+      "loss": 4.0031,
+      "step": 32085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001563514727090175,
+      "loss": 4.0484,
+      "step": 32090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015631567064406803,
+      "loss": 3.9391,
+      "step": 32095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015627986821868263,
+      "loss": 3.9625,
+      "step": 32100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015624406543490456,
+      "loss": 3.9531,
+      "step": 32105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015620826229477706,
+      "loss": 3.8891,
+      "step": 32110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015617245880034344,
+      "loss": 4.1219,
+      "step": 32115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015613665495364698,
+      "loss": 4.0031,
+      "step": 32120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000156100850756731,
+      "loss": 4.0687,
+      "step": 32125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015606504621163894,
+      "loss": 3.975,
+      "step": 32130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001560292413204141,
+      "loss": 3.9875,
+      "step": 32135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559934360850998,
+      "loss": 4.0672,
+      "step": 32140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559576305077396,
+      "loss": 4.0875,
+      "step": 32145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015592182459037686,
+      "loss": 3.9312,
+      "step": 32150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015588601833505496,
+      "loss": 3.9937,
+      "step": 32155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001558502117438175,
+      "loss": 3.9391,
+      "step": 32160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001558144048187078,
+      "loss": 4.0594,
+      "step": 32165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015577859756176954,
+      "loss": 3.9453,
+      "step": 32170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015574278997504607,
+      "loss": 4.0453,
+      "step": 32175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001557069820605811,
+      "loss": 4.0734,
+      "step": 32180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001556711738204181,
+      "loss": 4.0594,
+      "step": 32185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015563536525660062,
+      "loss": 4.0234,
+      "step": 32190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015559955637117236,
+      "loss": 4.0141,
+      "step": 32195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015556374716617683,
+      "loss": 4.0125,
+      "step": 32200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015552793764365773,
+      "loss": 4.0,
+      "step": 32205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015549212780565868,
+      "loss": 3.9703,
+      "step": 32210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015545631765422342,
+      "loss": 3.85,
+      "step": 32215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015542050719139554,
+      "loss": 3.9969,
+      "step": 32220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015538469641921882,
+      "loss": 3.9875,
+      "step": 32225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001553488853397369,
+      "loss": 3.9937,
+      "step": 32230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015531307395499357,
+      "loss": 4.0422,
+      "step": 32235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001552772622670327,
+      "loss": 3.9625,
+      "step": 32240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015524145027789784,
+      "loss": 4.0531,
+      "step": 32245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000155205637989633,
+      "loss": 3.9312,
+      "step": 32250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015516982540428187,
+      "loss": 3.9406,
+      "step": 32255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551340125238883,
+      "loss": 3.8719,
+      "step": 32260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015509819935049612,
+      "loss": 3.975,
+      "step": 32265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015506238588614923,
+      "loss": 4.0031,
+      "step": 32270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001550265721328915,
+      "loss": 3.9969,
+      "step": 32275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001549907580927668,
+      "loss": 4.0,
+      "step": 32280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015495494376781906,
+      "loss": 4.0406,
+      "step": 32285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015491912916009218,
+      "loss": 4.0641,
+      "step": 32290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015488331427163012,
+      "loss": 3.9516,
+      "step": 32295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015484749910447688,
+      "loss": 3.9359,
+      "step": 32300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015481168366067636,
+      "loss": 3.9781,
+      "step": 32305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001547758679422726,
+      "loss": 3.9797,
+      "step": 32310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015474005195130963,
+      "loss": 3.9625,
+      "step": 32315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015470423568983142,
+      "loss": 3.9891,
+      "step": 32320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015466841915988202,
+      "loss": 4.0078,
+      "step": 32325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015463260236350546,
+      "loss": 4.0031,
+      "step": 32330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015459678530274585,
+      "loss": 3.9219,
+      "step": 32335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001545609679796473,
+      "loss": 3.9141,
+      "step": 32340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015452515039625383,
+      "loss": 3.925,
+      "step": 32345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015448933255460961,
+      "loss": 3.9781,
+      "step": 32350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015445351445675872,
+      "loss": 3.9859,
+      "step": 32355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015441769610474536,
+      "loss": 4.0531,
+      "step": 32360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015438187750061366,
+      "loss": 3.9266,
+      "step": 32365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015434605864640775,
+      "loss": 3.9156,
+      "step": 32370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015431023954417188,
+      "loss": 3.9531,
+      "step": 32375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001542744201959502,
+      "loss": 3.9203,
+      "step": 32380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015423860060378698,
+      "loss": 4.0391,
+      "step": 32385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015420278076972636,
+      "loss": 3.8969,
+      "step": 32390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001541669606958127,
+      "loss": 4.0172,
+      "step": 32395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015413114038409015,
+      "loss": 4.0359,
+      "step": 32400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015409531983660306,
+      "loss": 3.9469,
+      "step": 32405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015405949905539565,
+      "loss": 4.0359,
+      "step": 32410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001540236780425122,
+      "loss": 4.025,
+      "step": 32415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015398785679999712,
+      "loss": 3.8859,
+      "step": 32420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015395203532989464,
+      "loss": 3.9562,
+      "step": 32425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001539162136342491,
+      "loss": 3.9719,
+      "step": 32430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015388039171510492,
+      "loss": 4.0922,
+      "step": 32435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001538445695745064,
+      "loss": 3.9672,
+      "step": 32440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015380874721449787,
+      "loss": 3.9781,
+      "step": 32445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015377292463712383,
+      "loss": 3.9344,
+      "step": 32450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015373710184442854,
+      "loss": 3.9688,
+      "step": 32455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015370127883845654,
+      "loss": 4.0,
+      "step": 32460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015366545562125222,
+      "loss": 3.8984,
+      "step": 32465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015362963219485995,
+      "loss": 3.9531,
+      "step": 32470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015359380856132425,
+      "loss": 4.0062,
+      "step": 32475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015355798472268953,
+      "loss": 3.9547,
+      "step": 32480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015352216068100028,
+      "loss": 3.9859,
+      "step": 32485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000153486336438301,
+      "loss": 3.9391,
+      "step": 32490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015345051199663616,
+      "loss": 3.9859,
+      "step": 32495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015341468735805024,
+      "loss": 3.9656,
+      "step": 32500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001533788625245878,
+      "loss": 4.0203,
+      "step": 32505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015334303749829336,
+      "loss": 4.0281,
+      "step": 32510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015330721228121142,
+      "loss": 3.9484,
+      "step": 32515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015327138687538656,
+      "loss": 3.9531,
+      "step": 32520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015323556128286336,
+      "loss": 4.0703,
+      "step": 32525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015319973550568635,
+      "loss": 3.9531,
+      "step": 32530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015316390954590013,
+      "loss": 3.9484,
+      "step": 32535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015312808340554927,
+      "loss": 4.0078,
+      "step": 32540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001530922570866784,
+      "loss": 3.9719,
+      "step": 32545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015305643059133211,
+      "loss": 3.8375,
+      "step": 32550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015302060392155504,
+      "loss": 3.9953,
+      "step": 32555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015298477707939183,
+      "loss": 3.9438,
+      "step": 32560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001529489500668871,
+      "loss": 4.0125,
+      "step": 32565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015291312288608547,
+      "loss": 3.9766,
+      "step": 32570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015287729553903166,
+      "loss": 3.9547,
+      "step": 32575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015284146802777032,
+      "loss": 4.0297,
+      "step": 32580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015280564035434613,
+      "loss": 3.9766,
+      "step": 32585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015276981252080376,
+      "loss": 3.9391,
+      "step": 32590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015273398452918791,
+      "loss": 3.8875,
+      "step": 32595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015269815638154332,
+      "loss": 4.0156,
+      "step": 32600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015266232807991469,
+      "loss": 3.9328,
+      "step": 32605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015262649962634672,
+      "loss": 3.9969,
+      "step": 32610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015259067102288415,
+      "loss": 4.0531,
+      "step": 32615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015255484227157177,
+      "loss": 3.975,
+      "step": 32620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015251901337445432,
+      "loss": 4.0344,
+      "step": 32625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001524831843335765,
+      "loss": 3.925,
+      "step": 32630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015244735515098308,
+      "loss": 4.0156,
+      "step": 32635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001524115258287189,
+      "loss": 3.9734,
+      "step": 32640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015237569636882868,
+      "loss": 3.9625,
+      "step": 32645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015233986677335728,
+      "loss": 3.9531,
+      "step": 32650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015230403704434944,
+      "loss": 3.9813,
+      "step": 32655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015226820718384996,
+      "loss": 3.9531,
+      "step": 32660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015223237719390374,
+      "loss": 4.0187,
+      "step": 32665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015219654707655548,
+      "loss": 3.9203,
+      "step": 32670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001521607168338501,
+      "loss": 4.0359,
+      "step": 32675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015212488646783236,
+      "loss": 4.0359,
+      "step": 32680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015208905598054716,
+      "loss": 3.9578,
+      "step": 32685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015205322537403934,
+      "loss": 3.8797,
+      "step": 32690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015201739465035374,
+      "loss": 3.9547,
+      "step": 32695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015198156381153525,
+      "loss": 3.9469,
+      "step": 32700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015194573285962873,
+      "loss": 3.9016,
+      "step": 32705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015190990179667904,
+      "loss": 3.9609,
+      "step": 32710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015187407062473103,
+      "loss": 4.0219,
+      "step": 32715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001518382393458297,
+      "loss": 4.0516,
+      "step": 32720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015180240796201983,
+      "loss": 3.9828,
+      "step": 32725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015176657647534636,
+      "loss": 3.9625,
+      "step": 32730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015173074488785418,
+      "loss": 3.9797,
+      "step": 32735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015169491320158828,
+      "loss": 3.9453,
+      "step": 32740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001516590814185935,
+      "loss": 3.9422,
+      "step": 32745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015162324954091477,
+      "loss": 3.9406,
+      "step": 32750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015158741757059703,
+      "loss": 4.0547,
+      "step": 32755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015155158550968524,
+      "loss": 3.9797,
+      "step": 32760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001515157533602243,
+      "loss": 4.0141,
+      "step": 32765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015147992112425916,
+      "loss": 4.0141,
+      "step": 32770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015144408880383476,
+      "loss": 4.0422,
+      "step": 32775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015140825640099612,
+      "loss": 3.8922,
+      "step": 32780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015137242391778812,
+      "loss": 4.0094,
+      "step": 32785
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015133659135625575,
+      "loss": 3.9672,
+      "step": 32790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015130075871844402,
+      "loss": 3.95,
+      "step": 32795
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001512649260063978,
+      "loss": 3.9297,
+      "step": 32800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001512290932221622,
+      "loss": 4.0016,
+      "step": 32805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015119326036778206,
+      "loss": 3.9391,
+      "step": 32810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001511574274453025,
+      "loss": 3.9891,
+      "step": 32815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015112159445676842,
+      "loss": 4.0187,
+      "step": 32820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015108576140422485,
+      "loss": 3.9953,
+      "step": 32825
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015104992828971674,
+      "loss": 4.0203,
+      "step": 32830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015101409511528914,
+      "loss": 4.0266,
+      "step": 32835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015097826188298702,
+      "loss": 3.9891,
+      "step": 32840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015094242859485543,
+      "loss": 3.9438,
+      "step": 32845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015090659525293933,
+      "loss": 3.9797,
+      "step": 32850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015087076185928377,
+      "loss": 3.9344,
+      "step": 32855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015083492841593373,
+      "loss": 3.9734,
+      "step": 32860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015079909492493425,
+      "loss": 3.8937,
+      "step": 32865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015076326138833033,
+      "loss": 3.9562,
+      "step": 32870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000150727427808167,
+      "loss": 3.9609,
+      "step": 32875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015069159418648937,
+      "loss": 3.9766,
+      "step": 32880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015065576052534227,
+      "loss": 3.9734,
+      "step": 32885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015061992682677099,
+      "loss": 3.9266,
+      "step": 32890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015058409309282034,
+      "loss": 3.9562,
+      "step": 32895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015054825932553546,
+      "loss": 3.9719,
+      "step": 32900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015051242552696135,
+      "loss": 3.9266,
+      "step": 32905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015047659169914305,
+      "loss": 4.0797,
+      "step": 32910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001504407578441257,
+      "loss": 4.0734,
+      "step": 32915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015040492396395418,
+      "loss": 3.9922,
+      "step": 32920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015036909006067365,
+      "loss": 3.9781,
+      "step": 32925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015033325613632908,
+      "loss": 3.9438,
+      "step": 32930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001502974221929656,
+      "loss": 3.9594,
+      "step": 32935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015026158823262816,
+      "loss": 3.9344,
+      "step": 32940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001502257542573619,
+      "loss": 3.9688,
+      "step": 32945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015018992026921175,
+      "loss": 4.05,
+      "step": 32950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001501540862702229,
+      "loss": 4.0219,
+      "step": 32955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001501182522624403,
+      "loss": 3.9406,
+      "step": 32960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015008241824790902,
+      "loss": 3.9234,
+      "step": 32965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015004658422867415,
+      "loss": 3.9453,
+      "step": 32970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001500107502067807,
+      "loss": 4.0469,
+      "step": 32975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014997491618427375,
+      "loss": 3.95,
+      "step": 32980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014993908216319834,
+      "loss": 3.9813,
+      "step": 32985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014990324814559953,
+      "loss": 4.0328,
+      "step": 32990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014986741413352227,
+      "loss": 3.9578,
+      "step": 32995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001498315801290118,
+      "loss": 3.9844,
+      "step": 33000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014979574613411303,
+      "loss": 4.0,
+      "step": 33005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014975991215087107,
+      "loss": 3.9734,
+      "step": 33010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014972407818133088,
+      "loss": 3.9766,
+      "step": 33015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014968824422753762,
+      "loss": 3.9531,
+      "step": 33020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001496524102915363,
+      "loss": 3.925,
+      "step": 33025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014961657637537192,
+      "loss": 3.9266,
+      "step": 33030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001495807424810896,
+      "loss": 3.9391,
+      "step": 33035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014954490861073436,
+      "loss": 3.9547,
+      "step": 33040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014950907476635119,
+      "loss": 4.0078,
+      "step": 33045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014947324094998515,
+      "loss": 4.0359,
+      "step": 33050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014943740716368136,
+      "loss": 3.9937,
+      "step": 33055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014940157340948476,
+      "loss": 3.9531,
+      "step": 33060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014936573968944047,
+      "loss": 4.0281,
+      "step": 33065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014932990600559343,
+      "loss": 4.0016,
+      "step": 33070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001492940723599887,
+      "loss": 3.9359,
+      "step": 33075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001492582387546714,
+      "loss": 3.9094,
+      "step": 33080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014922240519168645,
+      "loss": 3.9516,
+      "step": 33085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014918657167307887,
+      "loss": 3.9781,
+      "step": 33090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001491507382008938,
+      "loss": 3.9875,
+      "step": 33095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014911490477717616,
+      "loss": 4.0891,
+      "step": 33100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001490790714039709,
+      "loss": 3.9734,
+      "step": 33105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014904323808332323,
+      "loss": 4.0672,
+      "step": 33110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014900740481727805,
+      "loss": 3.95,
+      "step": 33115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014897157160788038,
+      "loss": 4.0,
+      "step": 33120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014893573845717513,
+      "loss": 4.0187,
+      "step": 33125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014889990536720745,
+      "loss": 3.8766,
+      "step": 33130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014886407234002228,
+      "loss": 3.9562,
+      "step": 33135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014882823937766455,
+      "loss": 3.9547,
+      "step": 33140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001487924064821794,
+      "loss": 4.0641,
+      "step": 33145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001487565736556117,
+      "loss": 4.0406,
+      "step": 33150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014872074090000647,
+      "loss": 3.9734,
+      "step": 33155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014868490821740862,
+      "loss": 3.9,
+      "step": 33160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014864907560986325,
+      "loss": 4.0172,
+      "step": 33165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014861324307941526,
+      "loss": 3.9875,
+      "step": 33170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014857741062810956,
+      "loss": 3.9859,
+      "step": 33175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014854157825799124,
+      "loss": 4.0219,
+      "step": 33180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001485057459711052,
+      "loss": 3.8328,
+      "step": 33185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014846991376949635,
+      "loss": 3.9547,
+      "step": 33190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014843408165520963,
+      "loss": 3.9813,
+      "step": 33195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014839824963029007,
+      "loss": 4.0031,
+      "step": 33200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014836241769678257,
+      "loss": 3.9859,
+      "step": 33205
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014832658585673203,
+      "loss": 3.9875,
+      "step": 33210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014829075411218335,
+      "loss": 4.0047,
+      "step": 33215
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001482549224651816,
+      "loss": 3.9859,
+      "step": 33220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014821909091777152,
+      "loss": 4.0297,
+      "step": 33225
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014818325947199806,
+      "loss": 4.0203,
+      "step": 33230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014814742812990623,
+      "loss": 3.9781,
+      "step": 33235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014811159689354084,
+      "loss": 3.9172,
+      "step": 33240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014807576576494682,
+      "loss": 4.0469,
+      "step": 33245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014803993474616896,
+      "loss": 4.0094,
+      "step": 33250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014800410383925228,
+      "loss": 3.9594,
+      "step": 33255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014796827304624157,
+      "loss": 3.9969,
+      "step": 33260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014793244236918166,
+      "loss": 4.0,
+      "step": 33265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014789661181011752,
+      "loss": 4.0578,
+      "step": 33270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014786078137109398,
+      "loss": 4.0031,
+      "step": 33275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014782495105415577,
+      "loss": 4.0156,
+      "step": 33280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014778912086134784,
+      "loss": 3.9625,
+      "step": 33285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014775329079471503,
+      "loss": 4.0453,
+      "step": 33290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001477174608563021,
+      "loss": 4.0,
+      "step": 33295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014768163104815393,
+      "loss": 3.9406,
+      "step": 33300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014764580137231528,
+      "loss": 3.9438,
+      "step": 33305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014760997183083094,
+      "loss": 3.9219,
+      "step": 33310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001475741424257458,
+      "loss": 4.0078,
+      "step": 33315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014753831315910453,
+      "loss": 3.9844,
+      "step": 33320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014750248403295197,
+      "loss": 3.9891,
+      "step": 33325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014746665504933296,
+      "loss": 3.9406,
+      "step": 33330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014743082621029215,
+      "loss": 3.9438,
+      "step": 33335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014739499751787426,
+      "loss": 3.9625,
+      "step": 33340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014735916897412423,
+      "loss": 4.0328,
+      "step": 33345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014732334058108665,
+      "loss": 3.9797,
+      "step": 33350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001472875123408062,
+      "loss": 4.0187,
+      "step": 33355
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014725168425532776,
+      "loss": 3.9688,
+      "step": 33360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014721585632669596,
+      "loss": 4.0297,
+      "step": 33365
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014718002855695553,
+      "loss": 3.9172,
+      "step": 33370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014714420094815106,
+      "loss": 4.0125,
+      "step": 33375
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001471083735023274,
+      "loss": 4.0047,
+      "step": 33380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014707254622152913,
+      "loss": 3.9125,
+      "step": 33385
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014703671910780091,
+      "loss": 4.0266,
+      "step": 33390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014700089216318737,
+      "loss": 3.9766,
+      "step": 33395
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014696506538973327,
+      "loss": 3.9312,
+      "step": 33400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014692923878948318,
+      "loss": 3.9875,
+      "step": 33405
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001468934123644817,
+      "loss": 3.9453,
+      "step": 33410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001468575861167735,
+      "loss": 3.9766,
+      "step": 33415
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014682176004840317,
+      "loss": 4.0375,
+      "step": 33420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014678593416141527,
+      "loss": 3.9328,
+      "step": 33425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001467501084578544,
+      "loss": 3.9859,
+      "step": 33430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001467142829397652,
+      "loss": 4.0016,
+      "step": 33435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001466784576091922,
+      "loss": 3.9125,
+      "step": 33440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014664263246817988,
+      "loss": 4.0609,
+      "step": 33445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014660680751877288,
+      "loss": 4.0281,
+      "step": 33450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014657098276301574,
+      "loss": 3.9656,
+      "step": 33455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014653515820295293,
+      "loss": 4.0,
+      "step": 33460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014649933384062894,
+      "loss": 3.975,
+      "step": 33465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001464635096780883,
+      "loss": 3.9359,
+      "step": 33470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014642768571737557,
+      "loss": 3.9562,
+      "step": 33475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014639186196053515,
+      "loss": 3.95,
+      "step": 33480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014635603840961145,
+      "loss": 3.9672,
+      "step": 33485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014632021506664903,
+      "loss": 3.9781,
+      "step": 33490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001462843919336923,
+      "loss": 3.9859,
+      "step": 33495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014624856901278562,
+      "loss": 4.0375,
+      "step": 33500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014621274630597354,
+      "loss": 3.9813,
+      "step": 33505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014617692381530039,
+      "loss": 3.95,
+      "step": 33510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014614110154281053,
+      "loss": 3.9375,
+      "step": 33515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014610527949054838,
+      "loss": 4.0031,
+      "step": 33520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014606945766055835,
+      "loss": 4.0031,
+      "step": 33525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001460336360548847,
+      "loss": 3.9984,
+      "step": 33530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014599781467557185,
+      "loss": 4.0203,
+      "step": 33535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001459619935246641,
+      "loss": 3.9688,
+      "step": 33540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014592617260420577,
+      "loss": 3.9406,
+      "step": 33545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014589035191624118,
+      "loss": 3.9953,
+      "step": 33550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014585453146281457,
+      "loss": 3.9312,
+      "step": 33555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014581871124597025,
+      "loss": 3.9828,
+      "step": 33560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001457828912677525,
+      "loss": 3.8578,
+      "step": 33565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014574707153020559,
+      "loss": 3.9813,
+      "step": 33570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014571125203537367,
+      "loss": 4.0078,
+      "step": 33575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014567543278530098,
+      "loss": 3.9359,
+      "step": 33580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014563961378203182,
+      "loss": 4.0,
+      "step": 33585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001456037950276103,
+      "loss": 3.9516,
+      "step": 33590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014556797652408058,
+      "loss": 3.9766,
+      "step": 33595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014553215827348696,
+      "loss": 3.9188,
+      "step": 33600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014549634027787343,
+      "loss": 4.0062,
+      "step": 33605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014546052253928417,
+      "loss": 3.9703,
+      "step": 33610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001454247050597634,
+      "loss": 3.9312,
+      "step": 33615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014538888784135513,
+      "loss": 4.0125,
+      "step": 33620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001453530708861034,
+      "loss": 3.9719,
+      "step": 33625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014531725419605248,
+      "loss": 3.9422,
+      "step": 33630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001452814377732463,
+      "loss": 3.9625,
+      "step": 33635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014524562161972887,
+      "loss": 3.9469,
+      "step": 33640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014520980573754427,
+      "loss": 3.9344,
+      "step": 33645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014517399012873656,
+      "loss": 4.0016,
+      "step": 33650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001451381747953497,
+      "loss": 3.95,
+      "step": 33655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014510235973942768,
+      "loss": 3.9719,
+      "step": 33660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014506654496301437,
+      "loss": 4.0125,
+      "step": 33665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001450307304681539,
+      "loss": 4.0062,
+      "step": 33670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014499491625689012,
+      "loss": 4.0312,
+      "step": 33675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001449591023312669,
+      "loss": 3.9531,
+      "step": 33680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014492328869332823,
+      "loss": 3.8094,
+      "step": 33685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000144887475345118,
+      "loss": 3.9797,
+      "step": 33690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014485166228868002,
+      "loss": 3.8531,
+      "step": 33695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014481584952605813,
+      "loss": 4.0312,
+      "step": 33700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014478003705929626,
+      "loss": 3.8828,
+      "step": 33705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014474422489043818,
+      "loss": 3.9641,
+      "step": 33710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001447084130215276,
+      "loss": 3.9047,
+      "step": 33715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014467260145460852,
+      "loss": 3.9156,
+      "step": 33720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014463679019172454,
+      "loss": 4.0094,
+      "step": 33725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001446009792349195,
+      "loss": 3.9438,
+      "step": 33730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000144565168586237,
+      "loss": 4.0172,
+      "step": 33735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014452935824772092,
+      "loss": 4.0203,
+      "step": 33740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014449354822141492,
+      "loss": 4.0156,
+      "step": 33745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001444577385093626,
+      "loss": 3.9969,
+      "step": 33750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014442192911360768,
+      "loss": 3.9953,
+      "step": 33755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014438612003619383,
+      "loss": 3.9828,
+      "step": 33760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014435031127916465,
+      "loss": 3.9891,
+      "step": 33765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014431450284456372,
+      "loss": 3.9266,
+      "step": 33770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001442786947344347,
+      "loss": 3.8906,
+      "step": 33775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014424288695082113,
+      "loss": 3.9922,
+      "step": 33780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001442070794957665,
+      "loss": 3.9875,
+      "step": 33785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014417127237131443,
+      "loss": 3.9953,
+      "step": 33790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014413546557950845,
+      "loss": 4.0078,
+      "step": 33795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014409965912239196,
+      "loss": 3.9531,
+      "step": 33800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001440638530020085,
+      "loss": 4.0078,
+      "step": 33805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014402804722040156,
+      "loss": 4.0047,
+      "step": 33810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014399224177961452,
+      "loss": 3.9594,
+      "step": 33815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014395643668169082,
+      "loss": 3.9828,
+      "step": 33820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014392063192867388,
+      "loss": 3.9734,
+      "step": 33825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014388482752260702,
+      "loss": 3.9031,
+      "step": 33830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014384902346553367,
+      "loss": 4.0234,
+      "step": 33835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014381321975949718,
+      "loss": 4.0078,
+      "step": 33840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014377741640654075,
+      "loss": 3.9594,
+      "step": 33845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014374161340870783,
+      "loss": 4.0281,
+      "step": 33850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014370581076804165,
+      "loss": 4.0109,
+      "step": 33855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014367000848658536,
+      "loss": 3.9422,
+      "step": 33860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001436342065663824,
+      "loss": 3.8813,
+      "step": 33865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014359840500947587,
+      "loss": 3.9688,
+      "step": 33870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000143562603817909,
+      "loss": 3.9031,
+      "step": 33875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014352680299372487,
+      "loss": 3.9672,
+      "step": 33880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014349100253896677,
+      "loss": 3.9344,
+      "step": 33885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001434552024556778,
+      "loss": 3.975,
+      "step": 33890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000143419402745901,
+      "loss": 4.0906,
+      "step": 33895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014338360341167956,
+      "loss": 4.0797,
+      "step": 33900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014334780445505655,
+      "loss": 3.9141,
+      "step": 33905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014331200587807495,
+      "loss": 4.0016,
+      "step": 33910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014327620768277776,
+      "loss": 3.9953,
+      "step": 33915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014324040987120812,
+      "loss": 3.9578,
+      "step": 33920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014320461244540894,
+      "loss": 3.9016,
+      "step": 33925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014316881540742316,
+      "loss": 4.0078,
+      "step": 33930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014313301875929372,
+      "loss": 3.9703,
+      "step": 33935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001430972225030636,
+      "loss": 3.8953,
+      "step": 33940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014306142664077568,
+      "loss": 3.9281,
+      "step": 33945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014302563117447273,
+      "loss": 3.9953,
+      "step": 33950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014298983610619775,
+      "loss": 3.8984,
+      "step": 33955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014295404143799349,
+      "loss": 3.9625,
+      "step": 33960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014291824717190278,
+      "loss": 4.0,
+      "step": 33965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001428824533099683,
+      "loss": 3.9859,
+      "step": 33970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014284665985423302,
+      "loss": 3.9672,
+      "step": 33975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014281086680673947,
+      "loss": 3.9484,
+      "step": 33980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014277507416953042,
+      "loss": 4.0531,
+      "step": 33985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001427392819446487,
+      "loss": 3.95,
+      "step": 33990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001427034901341368,
+      "loss": 3.9688,
+      "step": 33995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014266769874003742,
+      "loss": 4.0031,
+      "step": 34000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014263190776439314,
+      "loss": 4.0172,
+      "step": 34005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014259611720924664,
+      "loss": 3.9813,
+      "step": 34010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014256032707664047,
+      "loss": 3.8797,
+      "step": 34015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001425245373686171,
+      "loss": 4.0031,
+      "step": 34020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014248874808721912,
+      "loss": 3.9641,
+      "step": 34025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000142452959234489,
+      "loss": 4.0031,
+      "step": 34030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014241717081246925,
+      "loss": 4.0141,
+      "step": 34035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014238138282320225,
+      "loss": 3.9797,
+      "step": 34040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014234559526873051,
+      "loss": 3.9328,
+      "step": 34045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014230980815109639,
+      "loss": 3.9234,
+      "step": 34050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014227402147234228,
+      "loss": 4.0312,
+      "step": 34055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014223823523451046,
+      "loss": 4.0563,
+      "step": 34060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001422024494396433,
+      "loss": 3.9828,
+      "step": 34065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014216666408978313,
+      "loss": 3.8984,
+      "step": 34070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001421308791869722,
+      "loss": 4.0,
+      "step": 34075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014209509473325272,
+      "loss": 3.9516,
+      "step": 34080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014205931073066702,
+      "loss": 4.0563,
+      "step": 34085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001420235271812572,
+      "loss": 4.025,
+      "step": 34090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014198774408706537,
+      "loss": 4.0,
+      "step": 34095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014195196145013383,
+      "loss": 4.0328,
+      "step": 34100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014191617927250465,
+      "loss": 4.0078,
+      "step": 34105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014188039755621987,
+      "loss": 3.9688,
+      "step": 34110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014184461630332155,
+      "loss": 3.9609,
+      "step": 34115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001418088355158518,
+      "loss": 3.9734,
+      "step": 34120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014177305519585262,
+      "loss": 3.9328,
+      "step": 34125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001417372753453659,
+      "loss": 3.9547,
+      "step": 34130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014170149596643374,
+      "loss": 3.9641,
+      "step": 34135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000141665717061098,
+      "loss": 4.0844,
+      "step": 34140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001416299386314006,
+      "loss": 4.0156,
+      "step": 34145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014159416067938336,
+      "loss": 3.975,
+      "step": 34150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014155838320708823,
+      "loss": 3.9969,
+      "step": 34155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014152260621655698,
+      "loss": 3.9719,
+      "step": 34160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014148682970983135,
+      "loss": 4.0078,
+      "step": 34165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014145105368895322,
+      "loss": 4.0563,
+      "step": 34170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001414152781559643,
+      "loss": 4.0359,
+      "step": 34175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001413795031129063,
+      "loss": 3.9422,
+      "step": 34180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001413437285618208,
+      "loss": 3.9094,
+      "step": 34185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014130795450474962,
+      "loss": 4.0109,
+      "step": 34190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014127218094373432,
+      "loss": 4.0125,
+      "step": 34195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001412364078808165,
+      "loss": 4.1031,
+      "step": 34200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014120063531803764,
+      "loss": 4.0094,
+      "step": 34205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014116486325743946,
+      "loss": 3.9094,
+      "step": 34210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001411290917010634,
+      "loss": 3.9453,
+      "step": 34215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014109332065095082,
+      "loss": 3.9844,
+      "step": 34220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001410575501091434,
+      "loss": 3.9891,
+      "step": 34225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014102178007768243,
+      "loss": 4.0328,
+      "step": 34230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014098601055860933,
+      "loss": 3.8844,
+      "step": 34235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014095024155396545,
+      "loss": 4.0422,
+      "step": 34240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001409144730657922,
+      "loss": 4.0094,
+      "step": 34245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014087870509613086,
+      "loss": 3.9109,
+      "step": 34250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014084293764702262,
+      "loss": 4.0172,
+      "step": 34255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014080717072050886,
+      "loss": 4.0656,
+      "step": 34260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001407714043186308,
+      "loss": 4.0031,
+      "step": 34265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014073563844342954,
+      "loss": 4.0203,
+      "step": 34270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014069987309694627,
+      "loss": 3.9328,
+      "step": 34275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014066410828122218,
+      "loss": 3.9953,
+      "step": 34280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001406283439982983,
+      "loss": 3.9375,
+      "step": 34285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014059258025021578,
+      "loss": 3.9828,
+      "step": 34290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014055681703901556,
+      "loss": 3.9359,
+      "step": 34295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001405210543667387,
+      "loss": 3.9594,
+      "step": 34300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014048529223542624,
+      "loss": 3.9969,
+      "step": 34305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014044953064711903,
+      "loss": 3.8906,
+      "step": 34310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014041376960385803,
+      "loss": 4.0172,
+      "step": 34315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014037800910768415,
+      "loss": 4.0109,
+      "step": 34320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014034224916063823,
+      "loss": 3.925,
+      "step": 34325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014030648976476098,
+      "loss": 3.8922,
+      "step": 34330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001402707309220934,
+      "loss": 4.0016,
+      "step": 34335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014023497263467613,
+      "loss": 4.0734,
+      "step": 34340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014019921490454986,
+      "loss": 4.0062,
+      "step": 34345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001401634577337554,
+      "loss": 3.9125,
+      "step": 34350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014012770112433335,
+      "loss": 4.0656,
+      "step": 34355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014009194507832434,
+      "loss": 4.0844,
+      "step": 34360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014005618959776894,
+      "loss": 4.0,
+      "step": 34365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014002043468470783,
+      "loss": 3.9719,
+      "step": 34370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013998468034118146,
+      "loss": 3.9516,
+      "step": 34375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013994892656923031,
+      "loss": 3.9453,
+      "step": 34380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001399131733708949,
+      "loss": 3.9719,
+      "step": 34385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013987742074821568,
+      "loss": 3.9625,
+      "step": 34390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013984166870323304,
+      "loss": 3.8906,
+      "step": 34395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013980591723798729,
+      "loss": 4.0141,
+      "step": 34400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001397701663545189,
+      "loss": 3.9375,
+      "step": 34405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013973441605486808,
+      "loss": 3.8969,
+      "step": 34410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013969866634107515,
+      "loss": 3.9453,
+      "step": 34415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013966291721518025,
+      "loss": 3.925,
+      "step": 34420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013962716867922374,
+      "loss": 4.0094,
+      "step": 34425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001395914207352457,
+      "loss": 4.0391,
+      "step": 34430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001395556733852862,
+      "loss": 4.0047,
+      "step": 34435
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013951992663138553,
+      "loss": 3.95,
+      "step": 34440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013948418047558366,
+      "loss": 3.9734,
+      "step": 34445
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001394484349199206,
+      "loss": 3.9688,
+      "step": 34450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001394126899664363,
+      "loss": 4.0,
+      "step": 34455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001393769456171709,
+      "loss": 3.9625,
+      "step": 34460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013934120187416424,
+      "loss": 4.0172,
+      "step": 34465
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013930545873945622,
+      "loss": 3.9672,
+      "step": 34470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001392697162150866,
+      "loss": 3.9344,
+      "step": 34475
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013923397430309542,
+      "loss": 3.9016,
+      "step": 34480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013919823300552237,
+      "loss": 3.9797,
+      "step": 34485
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001391624923244071,
+      "loss": 3.9578,
+      "step": 34490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013912675226178953,
+      "loss": 3.9891,
+      "step": 34495
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013909101281970926,
+      "loss": 4.025,
+      "step": 34500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001390552740002059,
+      "loss": 4.0297,
+      "step": 34505
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001390195358053191,
+      "loss": 3.9734,
+      "step": 34510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001389837982370885,
+      "loss": 4.0406,
+      "step": 34515
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013894806129755355,
+      "loss": 3.9859,
+      "step": 34520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001389123249887538,
+      "loss": 3.9641,
+      "step": 34525
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013887658931272881,
+      "loss": 4.0563,
+      "step": 34530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013884085427151788,
+      "loss": 4.0438,
+      "step": 34535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001388051198671605,
+      "loss": 3.9625,
+      "step": 34540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000138769386101696,
+      "loss": 3.9906,
+      "step": 34545
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013873365297716373,
+      "loss": 3.8859,
+      "step": 34550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013869792049560296,
+      "loss": 4.0109,
+      "step": 34555
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013866218865905296,
+      "loss": 4.0312,
+      "step": 34560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013862645746955295,
+      "loss": 4.0359,
+      "step": 34565
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013859072692914208,
+      "loss": 3.9594,
+      "step": 34570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013855499703985957,
+      "loss": 3.9484,
+      "step": 34575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013851926780374446,
+      "loss": 3.8984,
+      "step": 34580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013848353922283585,
+      "loss": 3.8922,
+      "step": 34585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001384478112991728,
+      "loss": 4.0469,
+      "step": 34590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013841208403479426,
+      "loss": 3.9266,
+      "step": 34595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013837635743173913,
+      "loss": 4.0312,
+      "step": 34600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001383406314920465,
+      "loss": 3.9859,
+      "step": 34605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013830490621775512,
+      "loss": 3.9672,
+      "step": 34610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013826918161090383,
+      "loss": 4.0344,
+      "step": 34615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013823345767353154,
+      "loss": 4.0203,
+      "step": 34620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013819773440767693,
+      "loss": 3.9891,
+      "step": 34625
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013816201181537878,
+      "loss": 3.8531,
+      "step": 34630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001381262898986757,
+      "loss": 3.9297,
+      "step": 34635
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013809056865960647,
+      "loss": 3.9359,
+      "step": 34640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001380548481002096,
+      "loss": 3.9875,
+      "step": 34645
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013801912822252374,
+      "loss": 3.9109,
+      "step": 34650
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001379834090285873,
+      "loss": 3.9484,
+      "step": 34655
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013794769052043895,
+      "loss": 3.9406,
+      "step": 34660
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013791197270011704,
+      "loss": 3.9719,
+      "step": 34665
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013787625556966,
+      "loss": 4.0109,
+      "step": 34670
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013784053913110624,
+      "loss": 4.0016,
+      "step": 34675
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001378048233864941,
+      "loss": 3.9609,
+      "step": 34680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001377691083378619,
+      "loss": 4.0078,
+      "step": 34685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013773339398724774,
+      "loss": 3.9234,
+      "step": 34690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013769768033669008,
+      "loss": 3.9656,
+      "step": 34695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013766196738822697,
+      "loss": 3.9906,
+      "step": 34700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001376262551438965,
+      "loss": 3.9875,
+      "step": 34705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013759054360573694,
+      "loss": 3.9656,
+      "step": 34710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013755483277578623,
+      "loss": 3.9391,
+      "step": 34715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001375191226560824,
+      "loss": 4.0047,
+      "step": 34720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013748341324866344,
+      "loss": 3.9547,
+      "step": 34725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001374477045555673,
+      "loss": 3.9922,
+      "step": 34730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001374119965788319,
+      "loss": 3.9781,
+      "step": 34735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013737628932049506,
+      "loss": 3.9969,
+      "step": 34740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013734058278259454,
+      "loss": 3.975,
+      "step": 34745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013730487696716828,
+      "loss": 3.9813,
+      "step": 34750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013726917187625387,
+      "loss": 3.9844,
+      "step": 34755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000137233467511889,
+      "loss": 3.9172,
+      "step": 34760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013719776387611143,
+      "loss": 3.9609,
+      "step": 34765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013716206097095875,
+      "loss": 3.9703,
+      "step": 34770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013712635879846842,
+      "loss": 3.9047,
+      "step": 34775
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013709065736067805,
+      "loss": 3.9453,
+      "step": 34780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013705495665962514,
+      "loss": 3.9734,
+      "step": 34785
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013701925669734708,
+      "loss": 3.9141,
+      "step": 34790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013698355747588127,
+      "loss": 3.9969,
+      "step": 34795
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013694785899726514,
+      "loss": 3.9734,
+      "step": 34800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001369121612635359,
+      "loss": 3.9328,
+      "step": 34805
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013687646427673096,
+      "loss": 3.9656,
+      "step": 34810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013684076803888742,
+      "loss": 3.9688,
+      "step": 34815
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001368050725520425,
+      "loss": 3.8906,
+      "step": 34820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001367693778182334,
+      "loss": 4.0109,
+      "step": 34825
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001367336838394972,
+      "loss": 3.8734,
+      "step": 34830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013669799061787088,
+      "loss": 3.9312,
+      "step": 34835
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013666229815539156,
+      "loss": 4.0469,
+      "step": 34840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013662660645409619,
+      "loss": 3.9188,
+      "step": 34845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001365909155160216,
+      "loss": 3.9859,
+      "step": 34850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013655522534320485,
+      "loss": 4.0109,
+      "step": 34855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013651953593768267,
+      "loss": 3.9328,
+      "step": 34860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001364838473014919,
+      "loss": 4.0344,
+      "step": 34865
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013644815943666914,
+      "loss": 3.9719,
+      "step": 34870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013641247234525135,
+      "loss": 4.0,
+      "step": 34875
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013637678602927507,
+      "loss": 4.0203,
+      "step": 34880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013634110049077683,
+      "loss": 3.9984,
+      "step": 34885
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001363054157317934,
+      "loss": 3.9203,
+      "step": 34890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001362697317543612,
+      "loss": 3.9813,
+      "step": 34895
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013623404856051674,
+      "loss": 3.9703,
+      "step": 34900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001361983661522964,
+      "loss": 4.0141,
+      "step": 34905
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001361626845317367,
+      "loss": 4.0094,
+      "step": 34910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001361270037008739,
+      "loss": 3.9344,
+      "step": 34915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013609132366174437,
+      "loss": 3.8828,
+      "step": 34920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001360556444163843,
+      "loss": 3.9656,
+      "step": 34925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013601996596682993,
+      "loss": 3.8594,
+      "step": 34930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001359842883151175,
+      "loss": 4.0031,
+      "step": 34935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013594861146328305,
+      "loss": 3.8672,
+      "step": 34940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013591293541336273,
+      "loss": 4.0438,
+      "step": 34945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013587726016739258,
+      "loss": 3.9906,
+      "step": 34950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013584158572740854,
+      "loss": 3.9281,
+      "step": 34955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001358059120954465,
+      "loss": 3.9844,
+      "step": 34960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013577023927354248,
+      "loss": 4.0484,
+      "step": 34965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013573456726373228,
+      "loss": 3.9484,
+      "step": 34970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013569889606805164,
+      "loss": 3.9516,
+      "step": 34975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013566322568853647,
+      "loss": 3.9281,
+      "step": 34980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013562755612722235,
+      "loss": 3.9547,
+      "step": 34985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013559188738614504,
+      "loss": 4.05,
+      "step": 34990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013555621946734,
+      "loss": 3.8953,
+      "step": 34995
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000135520552372843,
+      "loss": 4.0516,
+      "step": 35000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013548488610468947,
+      "loss": 4.0172,
+      "step": 35005
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013544922066491486,
+      "loss": 3.8797,
+      "step": 35010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013541355605555463,
+      "loss": 4.0344,
+      "step": 35015
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013537789227864417,
+      "loss": 3.9234,
+      "step": 35020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013534222933621882,
+      "loss": 3.9813,
+      "step": 35025
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013530656723031382,
+      "loss": 3.9219,
+      "step": 35030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001352709059629645,
+      "loss": 3.8953,
+      "step": 35035
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013523524553620597,
+      "loss": 3.9438,
+      "step": 35040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013519958595207342,
+      "loss": 3.9453,
+      "step": 35045
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001351639272126019,
+      "loss": 4.0062,
+      "step": 35050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013512826931982652,
+      "loss": 3.9625,
+      "step": 35055
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013509261227578223,
+      "loss": 3.9672,
+      "step": 35060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013505695608250398,
+      "loss": 3.8594,
+      "step": 35065
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013502130074202667,
+      "loss": 3.9688,
+      "step": 35070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013498564625638522,
+      "loss": 4.0078,
+      "step": 35075
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013494999262761442,
+      "loss": 3.9656,
+      "step": 35080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013491433985774888,
+      "loss": 3.9078,
+      "step": 35085
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001348786879488235,
+      "loss": 3.9375,
+      "step": 35090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001348430369028729,
+      "loss": 4.0578,
+      "step": 35095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013480738672193158,
+      "loss": 3.9609,
+      "step": 35100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013477173740803413,
+      "loss": 3.9562,
+      "step": 35105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013473608896321518,
+      "loss": 3.9813,
+      "step": 35110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001347004413895091,
+      "loss": 3.9312,
+      "step": 35115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013466479468895022,
+      "loss": 4.0125,
+      "step": 35120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013462914886357307,
+      "loss": 4.0219,
+      "step": 35125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013459350391541188,
+      "loss": 3.9625,
+      "step": 35130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001345578598465009,
+      "loss": 3.9281,
+      "step": 35135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013452221665887428,
+      "loss": 3.9234,
+      "step": 35140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001344865743545663,
+      "loss": 4.0266,
+      "step": 35145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013445093293561108,
+      "loss": 4.0312,
+      "step": 35150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013441529240404247,
+      "loss": 3.9703,
+      "step": 35155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013437965276189476,
+      "loss": 3.9094,
+      "step": 35160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013434401401120173,
+      "loss": 3.9578,
+      "step": 35165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013430837615399738,
+      "loss": 4.0312,
+      "step": 35170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001342727391923154,
+      "loss": 3.9188,
+      "step": 35175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013423710312818982,
+      "loss": 3.9891,
+      "step": 35180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013420146796365426,
+      "loss": 3.9344,
+      "step": 35185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013416583370074245,
+      "loss": 4.0734,
+      "step": 35190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000134130200341488,
+      "loss": 4.0141,
+      "step": 35195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013409456788792459,
+      "loss": 3.9188,
+      "step": 35200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001340589363420857,
+      "loss": 4.05,
+      "step": 35205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013402330570600482,
+      "loss": 3.95,
+      "step": 35210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001339876759817155,
+      "loss": 3.9719,
+      "step": 35215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013395204717125104,
+      "loss": 3.9734,
+      "step": 35220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001339164192766448,
+      "loss": 3.9609,
+      "step": 35225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013388079229993002,
+      "loss": 3.95,
+      "step": 35230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013384516624314,
+      "loss": 3.9797,
+      "step": 35235
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013380954110830793,
+      "loss": 3.9359,
+      "step": 35240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013377391689746684,
+      "loss": 3.8891,
+      "step": 35245
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013373829361264995,
+      "loss": 3.9703,
+      "step": 35250
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013370267125589023,
+      "loss": 4.0438,
+      "step": 35255
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013366704982922058,
+      "loss": 3.9484,
+      "step": 35260
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013363142933467398,
+      "loss": 3.9172,
+      "step": 35265
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001335958097742833,
+      "loss": 3.9391,
+      "step": 35270
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013356019115008132,
+      "loss": 3.9406,
+      "step": 35275
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013352457346410088,
+      "loss": 3.9672,
+      "step": 35280
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013348895671837455,
+      "loss": 4.0547,
+      "step": 35285
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013345334091493507,
+      "loss": 3.9734,
+      "step": 35290
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013341772605581504,
+      "loss": 3.9906,
+      "step": 35295
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013338211214304697,
+      "loss": 3.9422,
+      "step": 35300
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013334649917866336,
+      "loss": 3.9531,
+      "step": 35305
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001333108871646967,
+      "loss": 4.025,
+      "step": 35310
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013327527610317927,
+      "loss": 3.9719,
+      "step": 35315
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013323966599614346,
+      "loss": 4.0406,
+      "step": 35320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013320405684562152,
+      "loss": 3.8594,
+      "step": 35325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001331684486536457,
+      "loss": 4.0016,
+      "step": 35330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013313284142224812,
+      "loss": 3.95,
+      "step": 35335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001330972351534609,
+      "loss": 3.9781,
+      "step": 35340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013306162984931614,
+      "loss": 4.0656,
+      "step": 35345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001330260255118458,
+      "loss": 4.0391,
+      "step": 35350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013299042214308173,
+      "loss": 4.0281,
+      "step": 35355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000132954819745056,
+      "loss": 4.0094,
+      "step": 35360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013291921831980034,
+      "loss": 3.9578,
+      "step": 35365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013288361786934657,
+      "loss": 4.0016,
+      "step": 35370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001328480183957263,
+      "loss": 3.9594,
+      "step": 35375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013281241990097134,
+      "loss": 3.9391,
+      "step": 35380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013277682238711322,
+      "loss": 3.9438,
+      "step": 35385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013274122585618344,
+      "loss": 4.0203,
+      "step": 35390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013270563031021366,
+      "loss": 3.9766,
+      "step": 35395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001326700357512352,
+      "loss": 3.975,
+      "step": 35400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001326344421812795,
+      "loss": 3.9906,
+      "step": 35405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013259884960237778,
+      "loss": 4.0297,
+      "step": 35410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013256325801656147,
+      "loss": 3.9453,
+      "step": 35415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001325276674258617,
+      "loss": 3.9781,
+      "step": 35420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013249207783230956,
+      "loss": 3.9719,
+      "step": 35425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001324564892379363,
+      "loss": 3.9984,
+      "step": 35430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001324209016447729,
+      "loss": 4.0047,
+      "step": 35435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013238531505485035,
+      "loss": 3.9359,
+      "step": 35440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013234972947019947,
+      "loss": 4.0187,
+      "step": 35445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013231414489285132,
+      "loss": 3.9688,
+      "step": 35450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013227856132483665,
+      "loss": 3.9906,
+      "step": 35455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013224297876818615,
+      "loss": 3.9453,
+      "step": 35460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013220739722493054,
+      "loss": 3.9375,
+      "step": 35465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013217181669710054,
+      "loss": 3.9609,
+      "step": 35470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013213623718672668,
+      "loss": 3.9688,
+      "step": 35475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013210065869583942,
+      "loss": 3.8094,
+      "step": 35480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013206508122646935,
+      "loss": 4.0297,
+      "step": 35485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013202950478064684,
+      "loss": 4.0406,
+      "step": 35490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001319939293604022,
+      "loss": 3.9656,
+      "step": 35495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013195835496776574,
+      "loss": 3.9516,
+      "step": 35500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013192278160476773,
+      "loss": 3.9672,
+      "step": 35505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013188720927343833,
+      "loss": 4.0203,
+      "step": 35510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013185163797580757,
+      "loss": 4.0359,
+      "step": 35515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013181606771390568,
+      "loss": 3.9953,
+      "step": 35520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013178049848976253,
+      "loss": 3.9547,
+      "step": 35525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013174493030540806,
+      "loss": 3.8781,
+      "step": 35530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013170936316287218,
+      "loss": 4.0328,
+      "step": 35535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013167379706418474,
+      "loss": 3.9609,
+      "step": 35540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013163823201137547,
+      "loss": 3.9156,
+      "step": 35545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013160266800647406,
+      "loss": 4.0547,
+      "step": 35550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013156710505151016,
+      "loss": 3.9656,
+      "step": 35555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013153154314851333,
+      "loss": 4.0578,
+      "step": 35560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001314959822995132,
+      "loss": 4.0438,
+      "step": 35565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013146042250653902,
+      "loss": 4.0062,
+      "step": 35570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013142486377162039,
+      "loss": 3.9188,
+      "step": 35575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001313893060967866,
+      "loss": 3.9359,
+      "step": 35580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013135374948406692,
+      "loss": 3.9516,
+      "step": 35585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013131819393549045,
+      "loss": 3.9813,
+      "step": 35590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013128263945308655,
+      "loss": 4.0328,
+      "step": 35595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013124708603888421,
+      "loss": 3.8672,
+      "step": 35600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001312115336949124,
+      "loss": 3.9609,
+      "step": 35605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013117598242320028,
+      "loss": 4.0781,
+      "step": 35610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013114043222577665,
+      "loss": 3.9312,
+      "step": 35615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013110488310467036,
+      "loss": 3.9047,
+      "step": 35620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013106933506191014,
+      "loss": 3.9719,
+      "step": 35625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013103378809952487,
+      "loss": 3.9625,
+      "step": 35630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013099824221954315,
+      "loss": 3.7984,
+      "step": 35635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001309626974239936,
+      "loss": 3.9906,
+      "step": 35640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013092715371490463,
+      "loss": 3.9922,
+      "step": 35645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013089161109430496,
+      "loss": 3.9547,
+      "step": 35650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013085606956422287,
+      "loss": 4.0187,
+      "step": 35655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013082052912668666,
+      "loss": 3.9828,
+      "step": 35660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001307849897837248,
+      "loss": 4.0328,
+      "step": 35665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013074945153736538,
+      "loss": 3.9625,
+      "step": 35670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013071391438963665,
+      "loss": 4.0094,
+      "step": 35675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001306783783425666,
+      "loss": 3.9891,
+      "step": 35680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013064284339818345,
+      "loss": 3.9609,
+      "step": 35685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013060730955851508,
+      "loss": 3.9375,
+      "step": 35690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013057177682558938,
+      "loss": 4.0484,
+      "step": 35695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001305362452014343,
+      "loss": 3.9641,
+      "step": 35700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001305007146880776,
+      "loss": 3.9328,
+      "step": 35705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000130465185287547,
+      "loss": 4.0062,
+      "step": 35710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013042965700187006,
+      "loss": 3.9828,
+      "step": 35715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013039412983307456,
+      "loss": 3.9547,
+      "step": 35720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013035860378318799,
+      "loss": 3.9578,
+      "step": 35725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001303230788542378,
+      "loss": 3.9672,
+      "step": 35730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001302875550482513,
+      "loss": 3.9516,
+      "step": 35735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000130252032367256,
+      "loss": 4.0094,
+      "step": 35740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013021651081327916,
+      "loss": 3.975,
+      "step": 35745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013018099038834786,
+      "loss": 3.9641,
+      "step": 35750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013014547109448945,
+      "loss": 3.9125,
+      "step": 35755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001301099529337309,
+      "loss": 3.9844,
+      "step": 35760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300744359080992,
+      "loss": 4.0094,
+      "step": 35765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300389200196214,
+      "loss": 4.0156,
+      "step": 35770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300034052703244,
+      "loss": 3.8891,
+      "step": 35775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012996789166223492,
+      "loss": 3.9875,
+      "step": 35780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001299323791973798,
+      "loss": 3.9969,
+      "step": 35785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001298968678777858,
+      "loss": 3.8984,
+      "step": 35790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012986135770547946,
+      "loss": 3.925,
+      "step": 35795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001298258486824874,
+      "loss": 4.1094,
+      "step": 35800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012979034081083606,
+      "loss": 4.0047,
+      "step": 35805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012975483409255194,
+      "loss": 4.075,
+      "step": 35810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012971932852966138,
+      "loss": 4.0187,
+      "step": 35815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012968382412419075,
+      "loss": 3.9141,
+      "step": 35820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012964832087816616,
+      "loss": 3.8922,
+      "step": 35825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012961281879361392,
+      "loss": 3.9219,
+      "step": 35830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012957731787256008,
+      "loss": 3.9047,
+      "step": 35835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001295418181170306,
+      "loss": 3.9641,
+      "step": 35840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012950631952905164,
+      "loss": 3.8922,
+      "step": 35845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012947082211064897,
+      "loss": 4.0234,
+      "step": 35850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001294353258638485,
+      "loss": 3.9656,
+      "step": 35855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012939983079067585,
+      "loss": 3.9469,
+      "step": 35860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012936433689315696,
+      "loss": 4.0172,
+      "step": 35865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012932884417331733,
+      "loss": 3.9625,
+      "step": 35870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012929335263318247,
+      "loss": 4.0125,
+      "step": 35875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012925786227477808,
+      "loss": 3.9141,
+      "step": 35880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012922237310012948,
+      "loss": 3.9656,
+      "step": 35885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012918688511126205,
+      "loss": 3.9328,
+      "step": 35890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012915139831020102,
+      "loss": 3.9844,
+      "step": 35895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012911591269897178,
+      "loss": 3.9297,
+      "step": 35900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012908042827959942,
+      "loss": 4.0719,
+      "step": 35905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012904494505410903,
+      "loss": 3.9578,
+      "step": 35910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001290094630245256,
+      "loss": 3.9109,
+      "step": 35915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012897398219287418,
+      "loss": 4.0375,
+      "step": 35920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012893850256117965,
+      "loss": 3.9312,
+      "step": 35925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012890302413146674,
+      "loss": 3.9719,
+      "step": 35930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012886754690576036,
+      "loss": 3.9719,
+      "step": 35935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012883207088608506,
+      "loss": 3.8828,
+      "step": 35940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012879659607446556,
+      "loss": 3.9844,
+      "step": 35945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012876112247292633,
+      "loss": 3.9406,
+      "step": 35950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001287256500834919,
+      "loss": 4.0219,
+      "step": 35955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001286901789081867,
+      "loss": 3.9641,
+      "step": 35960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000128654708949035,
+      "loss": 3.9656,
+      "step": 35965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012861924020806116,
+      "loss": 3.9609,
+      "step": 35970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012858377268728935,
+      "loss": 3.9734,
+      "step": 35975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001285483063887437,
+      "loss": 4.0281,
+      "step": 35980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001285128413144482,
+      "loss": 3.9578,
+      "step": 35985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012847737746642704,
+      "loss": 3.9438,
+      "step": 35990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012844191484670395,
+      "loss": 4.0141,
+      "step": 35995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012840645345730287,
+      "loss": 4.0094,
+      "step": 36000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012837099330024755,
+      "loss": 3.9328,
+      "step": 36005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001283355343775618,
+      "loss": 4.0094,
+      "step": 36010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012830007669126914,
+      "loss": 3.8422,
+      "step": 36015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001282646202433932,
+      "loss": 3.9656,
+      "step": 36020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012822916503595752,
+      "loss": 3.9859,
+      "step": 36025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012819371107098543,
+      "loss": 3.9656,
+      "step": 36030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012815825835050043,
+      "loss": 3.8672,
+      "step": 36035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012812280687652568,
+      "loss": 3.8922,
+      "step": 36040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012808735665108446,
+      "loss": 3.9453,
+      "step": 36045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012805190767619995,
+      "loss": 4.0109,
+      "step": 36050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012801645995389512,
+      "loss": 3.9719,
+      "step": 36055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012798101348619309,
+      "loss": 3.8672,
+      "step": 36060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012794556827511674,
+      "loss": 3.9484,
+      "step": 36065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012791012432268896,
+      "loss": 3.9531,
+      "step": 36070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012787468163093243,
+      "loss": 3.9609,
+      "step": 36075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012783924020187007,
+      "loss": 3.9656,
+      "step": 36080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012780380003752436,
+      "loss": 3.9594,
+      "step": 36085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012776836113991794,
+      "loss": 3.9125,
+      "step": 36090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012773292351107323,
+      "loss": 3.9438,
+      "step": 36095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001276974871530128,
+      "loss": 3.9188,
+      "step": 36100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001276620520677589,
+      "loss": 3.9203,
+      "step": 36105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001276266182573338,
+      "loss": 3.9141,
+      "step": 36110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012759118572375984,
+      "loss": 4.0047,
+      "step": 36115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012755575446905907,
+      "loss": 3.9859,
+      "step": 36120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012752032449525353,
+      "loss": 3.9562,
+      "step": 36125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012748489580436518,
+      "loss": 4.0031,
+      "step": 36130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001274494683984161,
+      "loss": 3.8281,
+      "step": 36135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000127414042279428,
+      "loss": 3.9453,
+      "step": 36140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012737861744942263,
+      "loss": 3.9328,
+      "step": 36145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012734319391042182,
+      "loss": 3.9828,
+      "step": 36150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001273077716644471,
+      "loss": 3.975,
+      "step": 36155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012727235071352007,
+      "loss": 4.0,
+      "step": 36160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001272369310596621,
+      "loss": 3.9359,
+      "step": 36165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012720151270489472,
+      "loss": 3.9391,
+      "step": 36170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012716609565123924,
+      "loss": 3.9766,
+      "step": 36175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012713067990071687,
+      "loss": 3.925,
+      "step": 36180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012709526545534876,
+      "loss": 3.9734,
+      "step": 36185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001270598523171561,
+      "loss": 3.9641,
+      "step": 36190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012702444048815989,
+      "loss": 4.0109,
+      "step": 36195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012698902997038104,
+      "loss": 4.0656,
+      "step": 36200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001269536207658405,
+      "loss": 3.9609,
+      "step": 36205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012691821287655908,
+      "loss": 4.0563,
+      "step": 36210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012688280630455745,
+      "loss": 3.9953,
+      "step": 36215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012684740105185625,
+      "loss": 3.8813,
+      "step": 36220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012681199712047614,
+      "loss": 4.0297,
+      "step": 36225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012677659451243765,
+      "loss": 4.0047,
+      "step": 36230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012674119322976103,
+      "loss": 4.0219,
+      "step": 36235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012670579327446683,
+      "loss": 4.075,
+      "step": 36240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012667039464857527,
+      "loss": 3.9922,
+      "step": 36245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012663499735410654,
+      "loss": 3.9734,
+      "step": 36250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012659960139308073,
+      "loss": 3.8609,
+      "step": 36255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012656420676751797,
+      "loss": 3.9859,
+      "step": 36260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012652881347943813,
+      "loss": 3.8391,
+      "step": 36265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012649342153086125,
+      "loss": 4.0187,
+      "step": 36270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012645803092380703,
+      "loss": 3.9031,
+      "step": 36275
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012642264166029527,
+      "loss": 3.9047,
+      "step": 36280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001263872537423456,
+      "loss": 3.9562,
+      "step": 36285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012635186717197764,
+      "loss": 3.9781,
+      "step": 36290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012631648195121094,
+      "loss": 3.9453,
+      "step": 36295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012628109808206486,
+      "loss": 3.9672,
+      "step": 36300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012624571556655884,
+      "loss": 4.0359,
+      "step": 36305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001262103344067121,
+      "loss": 3.9266,
+      "step": 36310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012617495460454385,
+      "loss": 4.0016,
+      "step": 36315
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012613957616207329,
+      "loss": 3.9734,
+      "step": 36320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001261041990813194,
+      "loss": 3.9672,
+      "step": 36325
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012606882336430115,
+      "loss": 3.8969,
+      "step": 36330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012603344901303753,
+      "loss": 3.9438,
+      "step": 36335
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012599807602954724,
+      "loss": 4.0531,
+      "step": 36340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012596270441584903,
+      "loss": 4.0438,
+      "step": 36345
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012592733417396168,
+      "loss": 3.9734,
+      "step": 36350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012589196530590367,
+      "loss": 3.9625,
+      "step": 36355
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012585659781369355,
+      "loss": 3.9609,
+      "step": 36360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012582123169934962,
+      "loss": 3.975,
+      "step": 36365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012578586696489042,
+      "loss": 3.9266,
+      "step": 36370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012575050361233412,
+      "loss": 3.9516,
+      "step": 36375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012571514164369887,
+      "loss": 3.9813,
+      "step": 36380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012567978106100286,
+      "loss": 4.0266,
+      "step": 36385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012564442186626415,
+      "loss": 3.9844,
+      "step": 36390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001256090640615006,
+      "loss": 3.9625,
+      "step": 36395
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012557370764873006,
+      "loss": 3.8937,
+      "step": 36400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012553835262997043,
+      "loss": 3.9703,
+      "step": 36405
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001255029990072394,
+      "loss": 3.9484,
+      "step": 36410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001254676467825545,
+      "loss": 3.9359,
+      "step": 36415
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012543229595793343,
+      "loss": 4.0141,
+      "step": 36420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012539694653539363,
+      "loss": 4.0016,
+      "step": 36425
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012536159851695244,
+      "loss": 3.9781,
+      "step": 36430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012532625190462713,
+      "loss": 3.9422,
+      "step": 36435
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012529090670043507,
+      "loss": 3.9094,
+      "step": 36440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012525556290639338,
+      "loss": 3.9875,
+      "step": 36445
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012522022052451908,
+      "loss": 3.8922,
+      "step": 36450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012518487955682914,
+      "loss": 3.925,
+      "step": 36455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012514954000534056,
+      "loss": 3.9156,
+      "step": 36460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012511420187207015,
+      "loss": 3.9594,
+      "step": 36465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012507886515903457,
+      "loss": 3.9281,
+      "step": 36470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012504352986825062,
+      "loss": 3.8766,
+      "step": 36475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012500819600173483,
+      "loss": 3.9094,
+      "step": 36480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012497286356150371,
+      "loss": 3.9422,
+      "step": 36485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012493753254957363,
+      "loss": 3.9266,
+      "step": 36490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012490220296796105,
+      "loss": 3.8969,
+      "step": 36495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012486687481868216,
+      "loss": 3.9609,
+      "step": 36500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012483154810375308,
+      "loss": 3.9719,
+      "step": 36505
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012479622282519004,
+      "loss": 4.0312,
+      "step": 36510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000124760898985009,
+      "loss": 3.9594,
+      "step": 36515
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012472557658522586,
+      "loss": 4.0016,
+      "step": 36520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012469025562785653,
+      "loss": 4.0281,
+      "step": 36525
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012465493611491676,
+      "loss": 3.9391,
+      "step": 36530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001246196180484222,
+      "loss": 4.0266,
+      "step": 36535
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012458430143038853,
+      "loss": 4.0266,
+      "step": 36540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001245489862628312,
+      "loss": 3.9469,
+      "step": 36545
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012451367254776567,
+      "loss": 3.9406,
+      "step": 36550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012447836028720735,
+      "loss": 3.9781,
+      "step": 36555
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012444304948317143,
+      "loss": 4.0109,
+      "step": 36560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012440774013767314,
+      "loss": 3.9859,
+      "step": 36565
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012437243225272765,
+      "loss": 3.9281,
+      "step": 36570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001243371258303499,
+      "loss": 3.9875,
+      "step": 36575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001243018208725548,
+      "loss": 3.9969,
+      "step": 36580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001242665173813573,
+      "loss": 3.9859,
+      "step": 36585
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001242312153587722,
+      "loss": 3.9281,
+      "step": 36590
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012419591480681403,
+      "loss": 3.9578,
+      "step": 36595
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012416061572749753,
+      "loss": 3.9047,
+      "step": 36600
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001241253181228372,
+      "loss": 3.9516,
+      "step": 36605
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001240900219948475,
+      "loss": 3.9484,
+      "step": 36610
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012405472734554264,
+      "loss": 3.9562,
+      "step": 36615
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012401943417693706,
+      "loss": 3.9219,
+      "step": 36620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001239841424910449,
+      "loss": 4.0094,
+      "step": 36625
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012394885228988025,
+      "loss": 3.8906,
+      "step": 36630
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000123913563575457,
+      "loss": 3.9562,
+      "step": 36635
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001238782763497893,
+      "loss": 3.9656,
+      "step": 36640
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012384299061489088,
+      "loss": 3.9438,
+      "step": 36645
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012380770637277543,
+      "loss": 4.0312,
+      "step": 36650
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001237724236254568,
+      "loss": 3.8844,
+      "step": 36655
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012373714237494849,
+      "loss": 3.9906,
+      "step": 36660
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012370186262326396,
+      "loss": 4.1203,
+      "step": 36665
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012366658437241662,
+      "loss": 3.9719,
+      "step": 36670
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012363130762441994,
+      "loss": 3.9656,
+      "step": 36675
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012359603238128706,
+      "loss": 3.8656,
+      "step": 36680
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001235607586450311,
+      "loss": 3.8922,
+      "step": 36685
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012352548641766524,
+      "loss": 4.0328,
+      "step": 36690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012349021570120245,
+      "loss": 4.0141,
+      "step": 36695
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012345494649765562,
+      "loss": 4.0062,
+      "step": 36700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012341967880903745,
+      "loss": 3.9797,
+      "step": 36705
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012338441263736087,
+      "loss": 3.9312,
+      "step": 36710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012334914798463843,
+      "loss": 4.0047,
+      "step": 36715
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012331388485288267,
+      "loss": 3.8312,
+      "step": 36720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012327862324410606,
+      "loss": 3.9891,
+      "step": 36725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000123243363160321,
+      "loss": 3.9203,
+      "step": 36730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012320810460353984,
+      "loss": 4.0062,
+      "step": 36735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012317284757577465,
+      "loss": 3.9484,
+      "step": 36740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001231375920790377,
+      "loss": 3.9453,
+      "step": 36745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012310233811534097,
+      "loss": 3.9484,
+      "step": 36750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012306708568669635,
+      "loss": 3.9656,
+      "step": 36755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012303183479511577,
+      "loss": 3.9719,
+      "step": 36760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000122996585442611,
+      "loss": 3.9969,
+      "step": 36765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001229613376311937,
+      "loss": 3.8375,
+      "step": 36770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012292609136287544,
+      "loss": 4.0016,
+      "step": 36775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001228908466396678,
+      "loss": 3.9609,
+      "step": 36780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012285560346358214,
+      "loss": 3.9203,
+      "step": 36785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012282036183662982,
+      "loss": 3.9875,
+      "step": 36790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012278512176082206,
+      "loss": 3.9078,
+      "step": 36795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012274988323817003,
+      "loss": 3.9969,
+      "step": 36800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001227146462706848,
+      "loss": 3.9859,
+      "step": 36805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012267941086037737,
+      "loss": 3.9219,
+      "step": 36810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012264417700925853,
+      "loss": 3.9531,
+      "step": 36815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012260894471933923,
+      "loss": 3.9594,
+      "step": 36820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012257371399263008,
+      "loss": 3.9344,
+      "step": 36825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012253848483114165,
+      "loss": 3.9969,
+      "step": 36830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001225032572368846,
+      "loss": 4.0016,
+      "step": 36835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012246803121186933,
+      "loss": 3.9125,
+      "step": 36840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012243280675810616,
+      "loss": 3.975,
+      "step": 36845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001223975838776053,
+      "loss": 3.9438,
+      "step": 36850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001223623625723771,
+      "loss": 3.9609,
+      "step": 36855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001223271428444315,
+      "loss": 3.9469,
+      "step": 36860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012229192469577846,
+      "loss": 3.9969,
+      "step": 36865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012225670812842805,
+      "loss": 4.0297,
+      "step": 36870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012222149314438996,
+      "loss": 3.9953,
+      "step": 36875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012218627974567393,
+      "loss": 3.9906,
+      "step": 36880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012215106793428957,
+      "loss": 3.975,
+      "step": 36885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001221158577122465,
+      "loss": 3.9844,
+      "step": 36890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001220806490815541,
+      "loss": 3.9109,
+      "step": 36895
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012204544204422176,
+      "loss": 3.9375,
+      "step": 36900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012201023660225869,
+      "loss": 3.9188,
+      "step": 36905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012197503275767419,
+      "loss": 3.9547,
+      "step": 36910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012193983051247724,
+      "loss": 3.9422,
+      "step": 36915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012190462986867685,
+      "loss": 3.9688,
+      "step": 36920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012186943082828198,
+      "loss": 3.9359,
+      "step": 36925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012183423339330142,
+      "loss": 3.9703,
+      "step": 36930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012179903756574387,
+      "loss": 3.9266,
+      "step": 36935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012176384334761793,
+      "loss": 3.9453,
+      "step": 36940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012172865074093222,
+      "loss": 4.025,
+      "step": 36945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012169345974769514,
+      "loss": 3.9328,
+      "step": 36950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000121658270369915,
+      "loss": 3.9359,
+      "step": 36955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012162308260960013,
+      "loss": 3.9844,
+      "step": 36960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001215878964687587,
+      "loss": 4.0031,
+      "step": 36965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012155271194939877,
+      "loss": 3.9438,
+      "step": 36970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012151752905352822,
+      "loss": 3.9531,
+      "step": 36975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012148234778315514,
+      "loss": 4.0609,
+      "step": 36980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012144716814028721,
+      "loss": 4.0172,
+      "step": 36985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012141199012693211,
+      "loss": 3.9984,
+      "step": 36990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001213768137450975,
+      "loss": 3.8891,
+      "step": 36995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012134163899679092,
+      "loss": 4.0438,
+      "step": 37000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012130646588401974,
+      "loss": 4.0234,
+      "step": 37005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001212712944087913,
+      "loss": 3.9703,
+      "step": 37010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012123612457311292,
+      "loss": 4.0438,
+      "step": 37015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012120095637899166,
+      "loss": 3.9359,
+      "step": 37020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012116578982843463,
+      "loss": 3.9234,
+      "step": 37025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012113062492344871,
+      "loss": 3.9781,
+      "step": 37030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012109546166604082,
+      "loss": 3.9781,
+      "step": 37035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012106030005821776,
+      "loss": 3.9422,
+      "step": 37040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012102514010198612,
+      "loss": 3.9844,
+      "step": 37045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012098998179935258,
+      "loss": 4.0219,
+      "step": 37050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012095482515232354,
+      "loss": 3.8422,
+      "step": 37055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012091967016290546,
+      "loss": 4.0094,
+      "step": 37060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012088451683310458,
+      "loss": 3.9234,
+      "step": 37065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012084936516492715,
+      "loss": 3.9266,
+      "step": 37070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012081421516037929,
+      "loss": 3.9781,
+      "step": 37075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012077906682146698,
+      "loss": 3.9469,
+      "step": 37080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012074392015019609,
+      "loss": 3.9875,
+      "step": 37085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012070877514857256,
+      "loss": 4.025,
+      "step": 37090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012067363181860206,
+      "loss": 3.9516,
+      "step": 37095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012063849016229017,
+      "loss": 3.8766,
+      "step": 37100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012060335018164254,
+      "loss": 3.9406,
+      "step": 37105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012056821187866453,
+      "loss": 3.8953,
+      "step": 37110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012053307525536153,
+      "loss": 3.9328,
+      "step": 37115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001204979403137387,
+      "loss": 3.95,
+      "step": 37120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012046280705580135,
+      "loss": 3.9594,
+      "step": 37125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012042767548355442,
+      "loss": 3.9828,
+      "step": 37130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012039254559900286,
+      "loss": 3.9813,
+      "step": 37135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012035741740415165,
+      "loss": 4.0438,
+      "step": 37140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012032229090100548,
+      "loss": 3.9391,
+      "step": 37145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012028716609156903,
+      "loss": 3.9156,
+      "step": 37150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012025204297784681,
+      "loss": 3.975,
+      "step": 37155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012021692156184345,
+      "loss": 3.975,
+      "step": 37160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012018180184556325,
+      "loss": 4.0125,
+      "step": 37165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012014668383101049,
+      "loss": 4.0266,
+      "step": 37170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001201115675201893,
+      "loss": 3.9828,
+      "step": 37175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012007645291510389,
+      "loss": 3.8922,
+      "step": 37180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012004134001775821,
+      "loss": 3.9125,
+      "step": 37185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001200062288301561,
+      "loss": 4.0391,
+      "step": 37190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00011997111935430145,
+      "loss": 4.0016,
+      "step": 37195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00011993601159219793,
+      "loss": 3.9797,
+      "step": 37200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00011990090554584914,
+      "loss": 3.9859,
+      "step": 37205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011986580121725848,
+      "loss": 3.9797,
+      "step": 37210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011983069860842954,
+      "loss": 4.0234,
+      "step": 37215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011979559772136553,
+      "loss": 4.0266,
+      "step": 37220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011976049855806962,
+      "loss": 3.9453,
+      "step": 37225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011972540112054503,
+      "loss": 3.9828,
+      "step": 37230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011969030541079473,
+      "loss": 3.9312,
+      "step": 37235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011965521143082159,
+      "loss": 4.0062,
+      "step": 37240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011962011918262847,
+      "loss": 3.8766,
+      "step": 37245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011958502866821811,
+      "loss": 3.9719,
+      "step": 37250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011954993988959309,
+      "loss": 3.9328,
+      "step": 37255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011951485284875591,
+      "loss": 4.0344,
+      "step": 37260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011947976754770904,
+      "loss": 3.95,
+      "step": 37265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011944468398845479,
+      "loss": 3.9031,
+      "step": 37270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011940960217299535,
+      "loss": 3.9562,
+      "step": 37275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011937452210333286,
+      "loss": 3.9594,
+      "step": 37280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011933944378146941,
+      "loss": 3.9297,
+      "step": 37285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001193043672094068,
+      "loss": 4.0609,
+      "step": 37290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011926929238914696,
+      "loss": 3.9156,
+      "step": 37295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011923421932269155,
+      "loss": 3.9625,
+      "step": 37300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001191991480120422,
+      "loss": 3.9328,
+      "step": 37305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011916407845920047,
+      "loss": 3.8844,
+      "step": 37310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011912901066616773,
+      "loss": 3.9266,
+      "step": 37315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011909394463494535,
+      "loss": 4.025,
+      "step": 37320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011905888036753454,
+      "loss": 3.9297,
+      "step": 37325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011902381786593644,
+      "loss": 3.8937,
+      "step": 37330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011898875713215197,
+      "loss": 4.0047,
+      "step": 37335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011895369816818219,
+      "loss": 3.9203,
+      "step": 37340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011891864097602785,
+      "loss": 3.9125,
+      "step": 37345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011888358555768966,
+      "loss": 4.0141,
+      "step": 37350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011884853191516819,
+      "loss": 3.9109,
+      "step": 37355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011881348005046406,
+      "loss": 3.9516,
+      "step": 37360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011877842996557766,
+      "loss": 4.0203,
+      "step": 37365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001187433816625092,
+      "loss": 3.9172,
+      "step": 37370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011870833514325904,
+      "loss": 3.9797,
+      "step": 37375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011867329040982719,
+      "loss": 3.9703,
+      "step": 37380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011863824746421368,
+      "loss": 4.0266,
+      "step": 37385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011860320630841839,
+      "loss": 3.9422,
+      "step": 37390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011856816694444118,
+      "loss": 3.9906,
+      "step": 37395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011853312937428171,
+      "loss": 3.9969,
+      "step": 37400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011849809359993954,
+      "loss": 3.9828,
+      "step": 37405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011846305962341426,
+      "loss": 3.9766,
+      "step": 37410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011842802744670523,
+      "loss": 3.9828,
+      "step": 37415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001183929970718117,
+      "loss": 4.0141,
+      "step": 37420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011835796850073284,
+      "loss": 3.9328,
+      "step": 37425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001183229417354678,
+      "loss": 3.8609,
+      "step": 37430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011828791677801556,
+      "loss": 3.9359,
+      "step": 37435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011825289363037497,
+      "loss": 4.0422,
+      "step": 37440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011821787229454473,
+      "loss": 3.9562,
+      "step": 37445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011818285277252365,
+      "loss": 3.8859,
+      "step": 37450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011814783506631023,
+      "loss": 3.9531,
+      "step": 37455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011811281917790289,
+      "loss": 4.0109,
+      "step": 37460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011807780510930008,
+      "loss": 3.9562,
+      "step": 37465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011804279286250005,
+      "loss": 3.9672,
+      "step": 37470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011800778243950088,
+      "loss": 3.9859,
+      "step": 37475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011797277384230061,
+      "loss": 3.8578,
+      "step": 37480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011793776707289727,
+      "loss": 3.9234,
+      "step": 37485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011790276213328869,
+      "loss": 3.9656,
+      "step": 37490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011786775902547248,
+      "loss": 3.9688,
+      "step": 37495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011783275775144643,
+      "loss": 3.9859,
+      "step": 37500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011779775831320803,
+      "loss": 3.9813,
+      "step": 37505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011776276071275462,
+      "loss": 3.9109,
+      "step": 37510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011772776495208358,
+      "loss": 3.9562,
+      "step": 37515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011769277103319212,
+      "loss": 3.9562,
+      "step": 37520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011765777895807731,
+      "loss": 4.0047,
+      "step": 37525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011762278872873623,
+      "loss": 3.8469,
+      "step": 37530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011758780034716567,
+      "loss": 3.9172,
+      "step": 37535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011755281381536247,
+      "loss": 3.9906,
+      "step": 37540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011751782913532335,
+      "loss": 4.0109,
+      "step": 37545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011748284630904483,
+      "loss": 3.9562,
+      "step": 37550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011744786533852343,
+      "loss": 3.8672,
+      "step": 37555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011741288622575551,
+      "loss": 3.9703,
+      "step": 37560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011737790897273731,
+      "loss": 3.9766,
+      "step": 37565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011734293358146493,
+      "loss": 4.0109,
+      "step": 37570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011730796005393458,
+      "loss": 3.8922,
+      "step": 37575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011727298839214207,
+      "loss": 3.9469,
+      "step": 37580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011723801859808323,
+      "loss": 3.9359,
+      "step": 37585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011720305067375389,
+      "loss": 3.8875,
+      "step": 37590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001171680846211496,
+      "loss": 3.9094,
+      "step": 37595
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001171331204422659,
+      "loss": 3.9828,
+      "step": 37600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011709815813909813,
+      "loss": 4.0047,
+      "step": 37605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001170631977136417,
+      "loss": 3.9047,
+      "step": 37610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011702823916789177,
+      "loss": 3.9984,
+      "step": 37615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001169932825038434,
+      "loss": 3.9766,
+      "step": 37620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011695832772349153,
+      "loss": 3.8578,
+      "step": 37625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011692337482883116,
+      "loss": 3.8859,
+      "step": 37630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011688842382185697,
+      "loss": 3.9625,
+      "step": 37635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011685347470456355,
+      "loss": 3.9813,
+      "step": 37640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011681852747894562,
+      "loss": 3.9406,
+      "step": 37645
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011678358214699751,
+      "loss": 4.0234,
+      "step": 37650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011674863871071359,
+      "loss": 3.9062,
+      "step": 37655
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011671369717208798,
+      "loss": 3.9641,
+      "step": 37660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011667875753311498,
+      "loss": 4.0203,
+      "step": 37665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011664381979578849,
+      "loss": 3.8844,
+      "step": 37670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011660888396210236,
+      "loss": 4.0516,
+      "step": 37675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001165739500340505,
+      "loss": 4.0031,
+      "step": 37680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011653901801362655,
+      "loss": 3.9719,
+      "step": 37685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011650408790282407,
+      "loss": 4.0125,
+      "step": 37690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011646915970363646,
+      "loss": 3.9453,
+      "step": 37695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011643423341805721,
+      "loss": 3.9937,
+      "step": 37700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001163993090480795,
+      "loss": 3.9937,
+      "step": 37705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011636438659569647,
+      "loss": 3.9766,
+      "step": 37710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011632946606290107,
+      "loss": 3.9531,
+      "step": 37715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011629454745168635,
+      "loss": 3.9688,
+      "step": 37720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011625963076404507,
+      "loss": 3.9625,
+      "step": 37725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011622471600196985,
+      "loss": 4.0109,
+      "step": 37730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011618980316745342,
+      "loss": 3.9547,
+      "step": 37735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011615489226248819,
+      "loss": 3.9719,
+      "step": 37740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011611998328906649,
+      "loss": 3.9969,
+      "step": 37745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011608507624918062,
+      "loss": 3.9937,
+      "step": 37750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011605017114482275,
+      "loss": 3.9516,
+      "step": 37755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011601526797798483,
+      "loss": 4.0266,
+      "step": 37760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011598036675065888,
+      "loss": 3.975,
+      "step": 37765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011594546746483671,
+      "loss": 3.9766,
+      "step": 37770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011591057012250997,
+      "loss": 3.9312,
+      "step": 37775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001158756747256703,
+      "loss": 3.9594,
+      "step": 37780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011584078127630915,
+      "loss": 4.0609,
+      "step": 37785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001158058897764179,
+      "loss": 3.9234,
+      "step": 37790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011577100022798786,
+      "loss": 3.9609,
+      "step": 37795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011573611263301014,
+      "loss": 3.9547,
+      "step": 37800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011570122699347573,
+      "loss": 3.9156,
+      "step": 37805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011566634331137564,
+      "loss": 3.9813,
+      "step": 37810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011563146158870068,
+      "loss": 3.9359,
+      "step": 37815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011559658182744149,
+      "loss": 4.0219,
+      "step": 37820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011556170402958867,
+      "loss": 3.9078,
+      "step": 37825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011552682819713277,
+      "loss": 4.0578,
+      "step": 37830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011549195433206412,
+      "loss": 3.9516,
+      "step": 37835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001154570824363729,
+      "loss": 4.05,
+      "step": 37840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011542221251204939,
+      "loss": 3.9484,
+      "step": 37845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011538734456108355,
+      "loss": 3.9281,
+      "step": 37850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011535247858546522,
+      "loss": 3.9828,
+      "step": 37855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011531761458718435,
+      "loss": 3.95,
+      "step": 37860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011528275256823056,
+      "loss": 3.9953,
+      "step": 37865
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011524789253059344,
+      "loss": 3.9875,
+      "step": 37870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011521303447626239,
+      "loss": 3.9516,
+      "step": 37875
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011517817840722687,
+      "loss": 3.9937,
+      "step": 37880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011514332432547608,
+      "loss": 3.9266,
+      "step": 37885
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011510847223299913,
+      "loss": 3.9328,
+      "step": 37890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011507362213178499,
+      "loss": 3.9625,
+      "step": 37895
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011503877402382263,
+      "loss": 3.9922,
+      "step": 37900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011500392791110086,
+      "loss": 3.8734,
+      "step": 37905
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011496908379560819,
+      "loss": 4.0016,
+      "step": 37910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001149342416793334,
+      "loss": 4.0297,
+      "step": 37915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011489940156426479,
+      "loss": 3.9531,
+      "step": 37920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011486456345239073,
+      "loss": 3.9484,
+      "step": 37925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011482972734569936,
+      "loss": 3.8875,
+      "step": 37930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011479489324617893,
+      "loss": 3.925,
+      "step": 37935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011476006115581731,
+      "loss": 3.9281,
+      "step": 37940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011472523107660236,
+      "loss": 4.0172,
+      "step": 37945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011469040301052194,
+      "loss": 4.0719,
+      "step": 37950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011465557695956361,
+      "loss": 4.0109,
+      "step": 37955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011462075292571494,
+      "loss": 4.0094,
+      "step": 37960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011458593091096323,
+      "loss": 4.0047,
+      "step": 37965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011455111091729593,
+      "loss": 3.9469,
+      "step": 37970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011451629294670015,
+      "loss": 3.9375,
+      "step": 37975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011448147700116295,
+      "loss": 3.8391,
+      "step": 37980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011444666308267126,
+      "loss": 3.9609,
+      "step": 37985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011441185119321197,
+      "loss": 3.85,
+      "step": 37990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011437704133477175,
+      "loss": 3.9672,
+      "step": 37995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011434223350933722,
+      "loss": 4.0609,
+      "step": 38000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011430742771889489,
+      "loss": 3.9375,
+      "step": 38005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011427262396543106,
+      "loss": 3.9969,
+      "step": 38010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001142378222509321,
+      "loss": 3.9875,
+      "step": 38015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000114203022577384,
+      "loss": 3.9719,
+      "step": 38020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011416822494677291,
+      "loss": 3.9562,
+      "step": 38025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011413342936108465,
+      "loss": 3.9484,
+      "step": 38030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011409863582230503,
+      "loss": 3.9781,
+      "step": 38035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011406384433241979,
+      "loss": 3.9859,
+      "step": 38040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011402905489341436,
+      "loss": 3.9844,
+      "step": 38045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011399426750727429,
+      "loss": 3.9234,
+      "step": 38050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011395948217598479,
+      "loss": 3.9766,
+      "step": 38055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011392469890153116,
+      "loss": 3.9953,
+      "step": 38060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011388991768589846,
+      "loss": 3.9359,
+      "step": 38065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011385513853107164,
+      "loss": 3.8984,
+      "step": 38070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011382036143903548,
+      "loss": 3.9656,
+      "step": 38075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011378558641177485,
+      "loss": 3.8703,
+      "step": 38080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011375081345127433,
+      "loss": 3.9422,
+      "step": 38085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001137160425595183,
+      "loss": 3.9875,
+      "step": 38090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011368127373849128,
+      "loss": 3.9016,
+      "step": 38095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011364650699017748,
+      "loss": 3.9422,
+      "step": 38100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011361174231656105,
+      "loss": 3.9391,
+      "step": 38105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001135769797196259,
+      "loss": 3.8969,
+      "step": 38110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011354221920135613,
+      "loss": 4.0234,
+      "step": 38115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011350746076373543,
+      "loss": 3.9281,
+      "step": 38120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011347270440874738,
+      "loss": 3.9359,
+      "step": 38125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001134379501383757,
+      "loss": 4.0203,
+      "step": 38130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011340319795460372,
+      "loss": 3.9375,
+      "step": 38135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011336844785941477,
+      "loss": 3.9234,
+      "step": 38140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011333369985479196,
+      "loss": 3.9203,
+      "step": 38145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011329895394271854,
+      "loss": 3.9703,
+      "step": 38150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011326421012517733,
+      "loss": 3.9859,
+      "step": 38155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001132294684041512,
+      "loss": 3.9844,
+      "step": 38160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011319472878162279,
+      "loss": 3.9094,
+      "step": 38165
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011315999125957484,
+      "loss": 4.0422,
+      "step": 38170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011312525583998974,
+      "loss": 3.975,
+      "step": 38175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011309052252484978,
+      "loss": 3.9438,
+      "step": 38180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011305579131613733,
+      "loss": 3.9109,
+      "step": 38185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011302106221583443,
+      "loss": 4.0219,
+      "step": 38190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011298633522592308,
+      "loss": 3.9766,
+      "step": 38195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001129516103483851,
+      "loss": 3.9484,
+      "step": 38200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011291688758520233,
+      "loss": 3.9797,
+      "step": 38205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001128821669383564,
+      "loss": 3.9828,
+      "step": 38210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001128474484098287,
+      "loss": 3.9594,
+      "step": 38215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011281273200160078,
+      "loss": 4.0172,
+      "step": 38220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011277801771565382,
+      "loss": 3.9609,
+      "step": 38225
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000112743305553969,
+      "loss": 3.9562,
+      "step": 38230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011270859551852725,
+      "loss": 4.05,
+      "step": 38235
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011267388761130963,
+      "loss": 3.9625,
+      "step": 38240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011263918183429684,
+      "loss": 4.0125,
+      "step": 38245
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011260447818946952,
+      "loss": 3.9781,
+      "step": 38250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011256977667880822,
+      "loss": 3.9172,
+      "step": 38255
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011253507730429344,
+      "loss": 3.9609,
+      "step": 38260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011250038006790537,
+      "loss": 3.9672,
+      "step": 38265
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011246568497162421,
+      "loss": 3.9359,
+      "step": 38270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001124309920174301,
+      "loss": 3.9672,
+      "step": 38275
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011239630120730285,
+      "loss": 3.9797,
+      "step": 38280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011236161254322235,
+      "loss": 3.9438,
+      "step": 38285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011232692602716823,
+      "loss": 3.9719,
+      "step": 38290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011229224166112007,
+      "loss": 4.0266,
+      "step": 38295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011225755944705737,
+      "loss": 3.9281,
+      "step": 38300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011222287938695935,
+      "loss": 3.925,
+      "step": 38305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011218820148280526,
+      "loss": 3.9453,
+      "step": 38310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011215352573657418,
+      "loss": 3.9469,
+      "step": 38315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011211885215024505,
+      "loss": 3.9797,
+      "step": 38320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011208418072579661,
+      "loss": 4.0047,
+      "step": 38325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011204951146520771,
+      "loss": 3.9422,
+      "step": 38330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011201484437045684,
+      "loss": 3.9719,
+      "step": 38335
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011198017944352248,
+      "loss": 3.9531,
+      "step": 38340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001119455166863829,
+      "loss": 3.9734,
+      "step": 38345
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001119108561010164,
+      "loss": 3.9469,
+      "step": 38350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011187619768940102,
+      "loss": 4.0062,
+      "step": 38355
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011184154145351465,
+      "loss": 3.975,
+      "step": 38360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011180688739533525,
+      "loss": 3.9359,
+      "step": 38365
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001117722355168405,
+      "loss": 3.9422,
+      "step": 38370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011173758582000795,
+      "loss": 4.0125,
+      "step": 38375
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000111702938306815,
+      "loss": 3.8937,
+      "step": 38380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011166829297923912,
+      "loss": 3.9984,
+      "step": 38385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011163364983925746,
+      "loss": 3.975,
+      "step": 38390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011159900888884704,
+      "loss": 4.0375,
+      "step": 38395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011156437012998497,
+      "loss": 3.9469,
+      "step": 38400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000111529733564648,
+      "loss": 3.9328,
+      "step": 38405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011149509919481285,
+      "loss": 3.9625,
+      "step": 38410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011146046702245605,
+      "loss": 3.9641,
+      "step": 38415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011142583704955414,
+      "loss": 3.9438,
+      "step": 38420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011139120927808349,
+      "loss": 3.9516,
+      "step": 38425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011135658371002022,
+      "loss": 3.9266,
+      "step": 38430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011132196034734037,
+      "loss": 3.9969,
+      "step": 38435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011128733919202007,
+      "loss": 3.9953,
+      "step": 38440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011125272024603505,
+      "loss": 4.0109,
+      "step": 38445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011121810351136097,
+      "loss": 3.9797,
+      "step": 38450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011118348898997352,
+      "loss": 3.9609,
+      "step": 38455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001111488766838481,
+      "loss": 4.0016,
+      "step": 38460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011111426659496005,
+      "loss": 3.9078,
+      "step": 38465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001110796587252845,
+      "loss": 3.9438,
+      "step": 38470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011104505307679661,
+      "loss": 3.9672,
+      "step": 38475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011101044965147133,
+      "loss": 3.9406,
+      "step": 38480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011097584845128339,
+      "loss": 3.9156,
+      "step": 38485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011094124947820759,
+      "loss": 4.0531,
+      "step": 38490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011090665273421848,
+      "loss": 4.0422,
+      "step": 38495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011087205822129042,
+      "loss": 4.0203,
+      "step": 38500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011083746594139778,
+      "loss": 4.0312,
+      "step": 38505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011080287589651477,
+      "loss": 3.9922,
+      "step": 38510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001107682880886154,
+      "loss": 3.8422,
+      "step": 38515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011073370251967362,
+      "loss": 3.9641,
+      "step": 38520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011069911919166322,
+      "loss": 3.8891,
+      "step": 38525
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011066453810655787,
+      "loss": 3.9375,
+      "step": 38530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011062995926633116,
+      "loss": 3.9281,
+      "step": 38535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011059538267295644,
+      "loss": 3.9406,
+      "step": 38540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011056080832840704,
+      "loss": 3.9891,
+      "step": 38545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011052623623465615,
+      "loss": 3.9219,
+      "step": 38550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011049166639367675,
+      "loss": 3.9359,
+      "step": 38555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011045709880744172,
+      "loss": 3.9391,
+      "step": 38560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011042253347792394,
+      "loss": 3.8859,
+      "step": 38565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011038797040709599,
+      "loss": 3.9937,
+      "step": 38570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011035340959693033,
+      "loss": 3.7734,
+      "step": 38575
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011031885104939947,
+      "loss": 3.9922,
+      "step": 38580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011028429476647562,
+      "loss": 3.9531,
+      "step": 38585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011024974075013089,
+      "loss": 3.8687,
+      "step": 38590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011021518900233725,
+      "loss": 3.9719,
+      "step": 38595
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011018063952506664,
+      "loss": 3.9094,
+      "step": 38600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001101460923202908,
+      "loss": 3.9484,
+      "step": 38605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011011154738998134,
+      "loss": 3.9078,
+      "step": 38610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011007700473610963,
+      "loss": 3.9531,
+      "step": 38615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011004246436064717,
+      "loss": 3.9578,
+      "step": 38620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011000792626556515,
+      "loss": 4.0359,
+      "step": 38625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010997339045283456,
+      "loss": 3.9922,
+      "step": 38630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010993885692442651,
+      "loss": 4.0859,
+      "step": 38635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010990432568231175,
+      "loss": 3.9937,
+      "step": 38640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010986979672846101,
+      "loss": 3.9312,
+      "step": 38645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010983527006484478,
+      "loss": 3.9797,
+      "step": 38650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010980074569343362,
+      "loss": 3.9672,
+      "step": 38655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010976622361619779,
+      "loss": 4.0109,
+      "step": 38660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001097317038351074,
+      "loss": 3.9937,
+      "step": 38665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010969718635213261,
+      "loss": 3.9766,
+      "step": 38670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001096626711692433,
+      "loss": 3.9484,
+      "step": 38675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010962815828840924,
+      "loss": 4.0078,
+      "step": 38680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001095936477116,
+      "loss": 3.9703,
+      "step": 38685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010955913944078527,
+      "loss": 3.9969,
+      "step": 38690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010952463347793435,
+      "loss": 3.9656,
+      "step": 38695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001094901298250165,
+      "loss": 4.0031,
+      "step": 38700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001094556284840008,
+      "loss": 4.0125,
+      "step": 38705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010942112945685634,
+      "loss": 3.8906,
+      "step": 38710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010938663274555196,
+      "loss": 4.0047,
+      "step": 38715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001093521383520563,
+      "loss": 4.0281,
+      "step": 38720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010931764627833807,
+      "loss": 4.0344,
+      "step": 38725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010928315652636574,
+      "loss": 3.9578,
+      "step": 38730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010924866909810752,
+      "loss": 3.9562,
+      "step": 38735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001092141839955317,
+      "loss": 4.0219,
+      "step": 38740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010917970122060638,
+      "loss": 3.8953,
+      "step": 38745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010914522077529941,
+      "loss": 3.9547,
+      "step": 38750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010911074266157863,
+      "loss": 4.0047,
+      "step": 38755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010907626688141178,
+      "loss": 4.075,
+      "step": 38760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010904179343676627,
+      "loss": 3.9844,
+      "step": 38765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010900732232960959,
+      "loss": 4.0047,
+      "step": 38770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010897285356190895,
+      "loss": 3.9234,
+      "step": 38775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010893838713563153,
+      "loss": 3.9453,
+      "step": 38780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010890392305274434,
+      "loss": 3.9656,
+      "step": 38785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010886946131521422,
+      "loss": 3.9094,
+      "step": 38790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010883500192500792,
+      "loss": 4.0484,
+      "step": 38795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010880054488409199,
+      "loss": 3.9781,
+      "step": 38800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000108766090194433,
+      "loss": 4.0328,
+      "step": 38805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010873163785799716,
+      "loss": 3.9328,
+      "step": 38810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010869718787675078,
+      "loss": 3.9531,
+      "step": 38815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010866274025265987,
+      "loss": 3.9688,
+      "step": 38820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010862829498769038,
+      "loss": 4.0563,
+      "step": 38825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010859385208380802,
+      "loss": 4.0563,
+      "step": 38830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010855941154297858,
+      "loss": 3.9031,
+      "step": 38835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010852497336716753,
+      "loss": 4.0078,
+      "step": 38840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010849053755834018,
+      "loss": 3.8625,
+      "step": 38845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010845610411846193,
+      "loss": 3.8937,
+      "step": 38850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010842167304949782,
+      "loss": 3.9375,
+      "step": 38855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010838724435341285,
+      "loss": 3.7625,
+      "step": 38860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001083528180321718,
+      "loss": 4.0047,
+      "step": 38865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001083183940877395,
+      "loss": 3.9453,
+      "step": 38870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010828397252208048,
+      "loss": 3.9016,
+      "step": 38875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010824955333715917,
+      "loss": 3.9406,
+      "step": 38880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010821513653493981,
+      "loss": 4.0219,
+      "step": 38885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010818072211738672,
+      "loss": 3.9031,
+      "step": 38890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010814631008646385,
+      "loss": 4.0469,
+      "step": 38895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010811190044413505,
+      "loss": 4.0563,
+      "step": 38900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010807749319236419,
+      "loss": 3.9922,
+      "step": 38905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010804308833311484,
+      "loss": 3.9984,
+      "step": 38910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010800868586835049,
+      "loss": 4.0016,
+      "step": 38915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010797428580003442,
+      "loss": 3.9328,
+      "step": 38920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010793988813013,
+      "loss": 3.9797,
+      "step": 38925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001079054928606002,
+      "loss": 3.9594,
+      "step": 38930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010787109999340793,
+      "loss": 3.9219,
+      "step": 38935
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010783670953051612,
+      "loss": 3.9641,
+      "step": 38940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010780232147388737,
+      "loss": 3.9656,
+      "step": 38945
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001077679358254842,
+      "loss": 4.0156,
+      "step": 38950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010773355258726895,
+      "loss": 3.95,
+      "step": 38955
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010769917176120399,
+      "loss": 3.9922,
+      "step": 38960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010766479334925138,
+      "loss": 3.9359,
+      "step": 38965
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010763041735337309,
+      "loss": 3.9688,
+      "step": 38970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010759604377553093,
+      "loss": 3.9469,
+      "step": 38975
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010756167261768672,
+      "loss": 3.9344,
+      "step": 38980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010752730388180194,
+      "loss": 3.9406,
+      "step": 38985
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010749293756983795,
+      "loss": 3.9813,
+      "step": 38990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010745857368375618,
+      "loss": 4.0547,
+      "step": 38995
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010742421222551774,
+      "loss": 3.9562,
+      "step": 39000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010738985319708357,
+      "loss": 3.9016,
+      "step": 39005
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010735549660041458,
+      "loss": 3.9937,
+      "step": 39010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010732114243747155,
+      "loss": 4.0,
+      "step": 39015
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010728679071021503,
+      "loss": 4.0438,
+      "step": 39020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010725244142060547,
+      "loss": 3.9062,
+      "step": 39025
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010721809457060324,
+      "loss": 3.9047,
+      "step": 39030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010718375016216845,
+      "loss": 3.8797,
+      "step": 39035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001071494081972612,
+      "loss": 3.8656,
+      "step": 39040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010711506867784133,
+      "loss": 3.9141,
+      "step": 39045
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010708073160586864,
+      "loss": 4.0125,
+      "step": 39050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010704639698330276,
+      "loss": 3.9016,
+      "step": 39055
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010701206481210316,
+      "loss": 3.9594,
+      "step": 39060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001069777350942291,
+      "loss": 3.9922,
+      "step": 39065
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010694340783163991,
+      "loss": 3.8547,
+      "step": 39070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010690908302629461,
+      "loss": 3.9781,
+      "step": 39075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010687476068015204,
+      "loss": 3.9047,
+      "step": 39080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010684044079517108,
+      "loss": 3.9656,
+      "step": 39085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010680612337331035,
+      "loss": 4.0297,
+      "step": 39090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010677180841652831,
+      "loss": 4.0359,
+      "step": 39095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001067374959267833,
+      "loss": 3.9328,
+      "step": 39100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010670318590603362,
+      "loss": 3.95,
+      "step": 39105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001066688783562373,
+      "loss": 4.0438,
+      "step": 39110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010663457327935222,
+      "loss": 3.9219,
+      "step": 39115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010660027067733627,
+      "loss": 3.9578,
+      "step": 39120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010656597055214708,
+      "loss": 3.9656,
+      "step": 39125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010653167290574215,
+      "loss": 3.9453,
+      "step": 39130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010649737774007875,
+      "loss": 3.9781,
+      "step": 39135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010646308505711429,
+      "loss": 3.9312,
+      "step": 39140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010642879485880577,
+      "loss": 3.9547,
+      "step": 39145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010639450714711013,
+      "loss": 4.0047,
+      "step": 39150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001063602219239841,
+      "loss": 3.9547,
+      "step": 39155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001063259391913845,
+      "loss": 3.8891,
+      "step": 39160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010629165895126777,
+      "loss": 4.0438,
+      "step": 39165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010625738120559019,
+      "loss": 3.9219,
+      "step": 39170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010622310595630817,
+      "loss": 3.9594,
+      "step": 39175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010618883320537772,
+      "loss": 3.9797,
+      "step": 39180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010615456295475478,
+      "loss": 4.0578,
+      "step": 39185
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001061202952063951,
+      "loss": 3.9359,
+      "step": 39190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010608602996225448,
+      "loss": 4.0781,
+      "step": 39195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010605176722428838,
+      "loss": 4.0219,
+      "step": 39200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010601750699445207,
+      "loss": 3.9406,
+      "step": 39205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010598324927470097,
+      "loss": 3.9078,
+      "step": 39210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001059489940669901,
+      "loss": 3.9703,
+      "step": 39215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010591474137327435,
+      "loss": 4.0734,
+      "step": 39220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010588049119550853,
+      "loss": 3.9125,
+      "step": 39225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010584624353564738,
+      "loss": 3.9719,
+      "step": 39230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001058119983956454,
+      "loss": 3.9234,
+      "step": 39235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010577775577745688,
+      "loss": 3.9641,
+      "step": 39240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010574351568303612,
+      "loss": 3.9641,
+      "step": 39245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010570927811433722,
+      "loss": 4.0219,
+      "step": 39250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010567504307331406,
+      "loss": 4.0391,
+      "step": 39255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010564081056192047,
+      "loss": 3.9844,
+      "step": 39260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010560658058211014,
+      "loss": 3.8766,
+      "step": 39265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001055723531358365,
+      "loss": 3.9516,
+      "step": 39270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010553812822505299,
+      "loss": 3.9297,
+      "step": 39275
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010550390585171277,
+      "loss": 3.7984,
+      "step": 39280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010546968601776895,
+      "loss": 3.8875,
+      "step": 39285
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010543546872517445,
+      "loss": 4.0172,
+      "step": 39290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010540125397588204,
+      "loss": 3.9141,
+      "step": 39295
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010536704177184438,
+      "loss": 3.9188,
+      "step": 39300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010533283211501395,
+      "loss": 3.9312,
+      "step": 39305
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010529862500734312,
+      "loss": 3.8922,
+      "step": 39310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010526442045078403,
+      "loss": 4.0516,
+      "step": 39315
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010523021844728882,
+      "loss": 4.0094,
+      "step": 39320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010519601899880939,
+      "loss": 3.9844,
+      "step": 39325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010516182210729747,
+      "loss": 3.9062,
+      "step": 39330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010512762777470461,
+      "loss": 3.9922,
+      "step": 39335
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010509343600298245,
+      "loss": 3.9219,
+      "step": 39340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010505924679408224,
+      "loss": 4.0187,
+      "step": 39345
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010502506014995506,
+      "loss": 4.0156,
+      "step": 39350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010499087607255212,
+      "loss": 3.8875,
+      "step": 39355
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010495669456382424,
+      "loss": 3.9672,
+      "step": 39360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010492251562572213,
+      "loss": 3.9625,
+      "step": 39365
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010488833926019635,
+      "loss": 3.9203,
+      "step": 39370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010485416546919748,
+      "loss": 3.9859,
+      "step": 39375
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010481999425467572,
+      "loss": 3.9406,
+      "step": 39380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010478582561858117,
+      "loss": 3.95,
+      "step": 39385
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010475165956286402,
+      "loss": 3.9469,
+      "step": 39390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010471749608947399,
+      "loss": 3.8937,
+      "step": 39395
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010468333520036088,
+      "loss": 3.9312,
+      "step": 39400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001046491768974741,
+      "loss": 4.0234,
+      "step": 39405
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010461502118276324,
+      "loss": 3.9797,
+      "step": 39410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010458086805817751,
+      "loss": 3.9734,
+      "step": 39415
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010454671752566602,
+      "loss": 3.8828,
+      "step": 39420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001045125695871777,
+      "loss": 3.9594,
+      "step": 39425
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010447842424466147,
+      "loss": 3.9875,
+      "step": 39430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010444428150006597,
+      "loss": 4.0187,
+      "step": 39435
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010441014135533965,
+      "loss": 3.9359,
+      "step": 39440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010437600381243103,
+      "loss": 3.8875,
+      "step": 39445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001043418688732883,
+      "loss": 4.0109,
+      "step": 39450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010430773653985949,
+      "loss": 3.9906,
+      "step": 39455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010427360681409253,
+      "loss": 3.8813,
+      "step": 39460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010423947969793528,
+      "loss": 3.9937,
+      "step": 39465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010420535519333535,
+      "loss": 3.9344,
+      "step": 39470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010417123330224017,
+      "loss": 4.0187,
+      "step": 39475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010413711402659719,
+      "loss": 3.875,
+      "step": 39480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010410299736835351,
+      "loss": 4.0328,
+      "step": 39485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001040688833294562,
+      "loss": 3.9125,
+      "step": 39490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010403477191185216,
+      "loss": 4.0422,
+      "step": 39495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010400066311748812,
+      "loss": 3.9016,
+      "step": 39500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010396655694831067,
+      "loss": 3.8641,
+      "step": 39505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010393245340626627,
+      "loss": 3.9344,
+      "step": 39510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010389835249330119,
+      "loss": 3.9672,
+      "step": 39515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010386425421136158,
+      "loss": 3.9422,
+      "step": 39520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010383015856239347,
+      "loss": 3.9688,
+      "step": 39525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010379606554834261,
+      "loss": 4.0078,
+      "step": 39530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010376197517115479,
+      "loss": 3.8859,
+      "step": 39535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001037278874327755,
+      "loss": 4.0797,
+      "step": 39540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010369380233515017,
+      "loss": 3.9297,
+      "step": 39545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010365971988022396,
+      "loss": 3.9453,
+      "step": 39550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010362564006994203,
+      "loss": 3.9188,
+      "step": 39555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010359156290624933,
+      "loss": 3.9188,
+      "step": 39560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010355748839109056,
+      "loss": 4.0,
+      "step": 39565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010352341652641043,
+      "loss": 4.0531,
+      "step": 39570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010348934731415344,
+      "loss": 4.0312,
+      "step": 39575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010345528075626388,
+      "loss": 4.0078,
+      "step": 39580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010342121685468587,
+      "loss": 3.9172,
+      "step": 39585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010338715561136359,
+      "loss": 3.8781,
+      "step": 39590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010335309702824083,
+      "loss": 3.9813,
+      "step": 39595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010331904110726133,
+      "loss": 3.9813,
+      "step": 39600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010328498785036857,
+      "loss": 3.9172,
+      "step": 39605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010325093725950615,
+      "loss": 4.0031,
+      "step": 39610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010321688933661727,
+      "loss": 3.9656,
+      "step": 39615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010318284408364494,
+      "loss": 3.9344,
+      "step": 39620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010314880150253228,
+      "loss": 4.0391,
+      "step": 39625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010311476159522207,
+      "loss": 3.9719,
+      "step": 39630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010308072436365692,
+      "loss": 3.9781,
+      "step": 39635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001030466898097793,
+      "loss": 4.0406,
+      "step": 39640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010301265793553167,
+      "loss": 4.0344,
+      "step": 39645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001029786287428562,
+      "loss": 3.9312,
+      "step": 39650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010294460223369489,
+      "loss": 3.9891,
+      "step": 39655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010291057840998972,
+      "loss": 3.9344,
+      "step": 39660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010287655727368236,
+      "loss": 3.8703,
+      "step": 39665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010284253882671445,
+      "loss": 3.9891,
+      "step": 39670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010280852307102734,
+      "loss": 4.0219,
+      "step": 39675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010277451000856244,
+      "loss": 4.0219,
+      "step": 39680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001027404996412608,
+      "loss": 4.0031,
+      "step": 39685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001027064919710634,
+      "loss": 3.9125,
+      "step": 39690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010267248699991102,
+      "loss": 3.9984,
+      "step": 39695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010263848472974444,
+      "loss": 4.0094,
+      "step": 39700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010260448516250409,
+      "loss": 3.9047,
+      "step": 39705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010257048830013031,
+      "loss": 3.9422,
+      "step": 39710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001025364941445634,
+      "loss": 3.8703,
+      "step": 39715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010250250269774334,
+      "loss": 3.9297,
+      "step": 39720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010246851396160999,
+      "loss": 3.9531,
+      "step": 39725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010243452793810316,
+      "loss": 3.9203,
+      "step": 39730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010240054462916244,
+      "loss": 3.9109,
+      "step": 39735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010236656403672721,
+      "loss": 3.9406,
+      "step": 39740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010233258616273674,
+      "loss": 3.8656,
+      "step": 39745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010229861100913023,
+      "loss": 3.9734,
+      "step": 39750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010226463857784658,
+      "loss": 3.9734,
+      "step": 39755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010223066887082461,
+      "loss": 3.9125,
+      "step": 39760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010219670189000296,
+      "loss": 3.9688,
+      "step": 39765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010216273763732019,
+      "loss": 3.8937,
+      "step": 39770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010212877611471455,
+      "loss": 3.9797,
+      "step": 39775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010209481732412433,
+      "loss": 3.9406,
+      "step": 39780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010206086126748747,
+      "loss": 3.9594,
+      "step": 39785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010202690794674188,
+      "loss": 3.9672,
+      "step": 39790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010199295736382531,
+      "loss": 3.9625,
+      "step": 39795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001019590095206753,
+      "loss": 3.9391,
+      "step": 39800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010192506441922921,
+      "loss": 3.9078,
+      "step": 39805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010189112206142441,
+      "loss": 4.0391,
+      "step": 39810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010185718244919788,
+      "loss": 3.9469,
+      "step": 39815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010182324558448653,
+      "loss": 3.9984,
+      "step": 39820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001017893114692273,
+      "loss": 4.0,
+      "step": 39825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001017553801053567,
+      "loss": 4.0109,
+      "step": 39830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010172145149481115,
+      "loss": 4.0125,
+      "step": 39835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010168752563952711,
+      "loss": 4.0375,
+      "step": 39840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010165360254144062,
+      "loss": 3.9844,
+      "step": 39845
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001016196822024877,
+      "loss": 3.9953,
+      "step": 39850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010158576462460412,
+      "loss": 3.9531,
+      "step": 39855
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010155184980972571,
+      "loss": 3.9625,
+      "step": 39860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010151793775978791,
+      "loss": 3.9859,
+      "step": 39865
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010148402847672606,
+      "loss": 3.9688,
+      "step": 39870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010145012196247533,
+      "loss": 3.9297,
+      "step": 39875
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010141621821897091,
+      "loss": 3.9922,
+      "step": 39880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010138231724814758,
+      "loss": 3.9734,
+      "step": 39885
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010134841905194006,
+      "loss": 4.0516,
+      "step": 39890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010131452363228299,
+      "loss": 3.9531,
+      "step": 39895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001012806309911108,
+      "loss": 4.0078,
+      "step": 39900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010124674113035766,
+      "loss": 4.0078,
+      "step": 39905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010121285405195767,
+      "loss": 3.9937,
+      "step": 39910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010117896975784485,
+      "loss": 3.9438,
+      "step": 39915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010114508824995295,
+      "loss": 4.0094,
+      "step": 39920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010111120953021553,
+      "loss": 3.8953,
+      "step": 39925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010107733360056617,
+      "loss": 3.9297,
+      "step": 39930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010104346046293808,
+      "loss": 4.0047,
+      "step": 39935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010100959011926444,
+      "loss": 3.9688,
+      "step": 39940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010097572257147814,
+      "loss": 3.9766,
+      "step": 39945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010094185782151218,
+      "loss": 3.9516,
+      "step": 39950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010090799587129911,
+      "loss": 4.0094,
+      "step": 39955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010087413672277147,
+      "loss": 4.0281,
+      "step": 39960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001008402803778615,
+      "loss": 3.9172,
+      "step": 39965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010080642683850157,
+      "loss": 4.0141,
+      "step": 39970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010077257610662356,
+      "loss": 4.0125,
+      "step": 39975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010073872818415938,
+      "loss": 3.9937,
+      "step": 39980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010070488307304076,
+      "loss": 3.9203,
+      "step": 39985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010067104077519923,
+      "loss": 3.9719,
+      "step": 39990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010063720129256614,
+      "loss": 4.0125,
+      "step": 39995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010060336462707274,
+      "loss": 3.9391,
+      "step": 40000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010056953078065011,
+      "loss": 4.0266,
+      "step": 40005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001005356997552291,
+      "loss": 3.9109,
+      "step": 40010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010050187155274049,
+      "loss": 3.9859,
+      "step": 40015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010046804617511487,
+      "loss": 4.0141,
+      "step": 40020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001004342236242826,
+      "loss": 3.9016,
+      "step": 40025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010040040390217404,
+      "loss": 3.9578,
+      "step": 40030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010036658701071913,
+      "loss": 3.9797,
+      "step": 40035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010033277295184794,
+      "loss": 3.9297,
+      "step": 40040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010029896172749019,
+      "loss": 3.9891,
+      "step": 40045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010026515333957551,
+      "loss": 3.9656,
+      "step": 40050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010023134779003326,
+      "loss": 3.9547,
+      "step": 40055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010019754508079286,
+      "loss": 3.9359,
+      "step": 40060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010016374521378335,
+      "loss": 3.9859,
+      "step": 40065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010012994819093369,
+      "loss": 3.9859,
+      "step": 40070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010009615401417272,
+      "loss": 4.0062,
+      "step": 40075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001000623626854291,
+      "loss": 4.0016,
+      "step": 40080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010002857420663123,
+      "loss": 3.9266,
+      "step": 40085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.99947885797074e-05,
+      "loss": 3.9203,
+      "step": 40090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.99610058065859e-05,
+      "loss": 3.8875,
+      "step": 40095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.992722588919459e-05,
+      "loss": 3.9391,
+      "step": 40100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.989344882946128e-05,
+      "loss": 3.9547,
+      "step": 40105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.985967462931375e-05,
+      "loss": 3.8641,
+      "step": 40110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.982590329067944e-05,
+      "loss": 3.9344,
+      "step": 40115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.979213481548565e-05,
+      "loss": 4.0047,
+      "step": 40120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.975836920565952e-05,
+      "loss": 3.9578,
+      "step": 40125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.972460646312818e-05,
+      "loss": 3.8609,
+      "step": 40130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.96908465898184e-05,
+      "loss": 3.9109,
+      "step": 40135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.965708958765686e-05,
+      "loss": 3.9641,
+      "step": 40140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.962333545857003e-05,
+      "loss": 3.8422,
+      "step": 40145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.958958420448433e-05,
+      "loss": 3.9531,
+      "step": 40150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.955583582732595e-05,
+      "loss": 3.9359,
+      "step": 40155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.952209032902083e-05,
+      "loss": 4.0187,
+      "step": 40160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.948834771149494e-05,
+      "loss": 4.0516,
+      "step": 40165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.945460797667393e-05,
+      "loss": 3.9109,
+      "step": 40170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.942087112648332e-05,
+      "loss": 3.9578,
+      "step": 40175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.938713716284842e-05,
+      "loss": 3.9813,
+      "step": 40180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.935340608769453e-05,
+      "loss": 3.9891,
+      "step": 40185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.931967790294668e-05,
+      "loss": 3.9375,
+      "step": 40190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.928595261052961e-05,
+      "loss": 3.9156,
+      "step": 40195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.92522302123682e-05,
+      "loss": 3.8609,
+      "step": 40200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.921851071038689e-05,
+      "loss": 3.9438,
+      "step": 40205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.91847941065101e-05,
+      "loss": 3.8859,
+      "step": 40210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.915108040266193e-05,
+      "loss": 4.0094,
+      "step": 40215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.911736960076661e-05,
+      "loss": 4.0609,
+      "step": 40220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.908366170274788e-05,
+      "loss": 3.9844,
+      "step": 40225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.904995671052946e-05,
+      "loss": 3.9906,
+      "step": 40230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.901625462603495e-05,
+      "loss": 4.0047,
+      "step": 40235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.898255545118774e-05,
+      "loss": 3.9734,
+      "step": 40240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.894885918791098e-05,
+      "loss": 3.9953,
+      "step": 40245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.891516583812775e-05,
+      "loss": 3.8312,
+      "step": 40250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.888147540376096e-05,
+      "loss": 3.9562,
+      "step": 40255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.88477878867333e-05,
+      "loss": 4.0109,
+      "step": 40260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.881410328896734e-05,
+      "loss": 3.9031,
+      "step": 40265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.878042161238541e-05,
+      "loss": 4.0438,
+      "step": 40270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.874674285890977e-05,
+      "loss": 3.9578,
+      "step": 40275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.871306703046248e-05,
+      "loss": 3.9609,
+      "step": 40280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.867939412896537e-05,
+      "loss": 4.0281,
+      "step": 40285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.864572415634021e-05,
+      "loss": 3.9688,
+      "step": 40290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.861205711450855e-05,
+      "loss": 3.9094,
+      "step": 40295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.857839300539176e-05,
+      "loss": 3.9797,
+      "step": 40300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.854473183091097e-05,
+      "loss": 4.0,
+      "step": 40305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.851107359298738e-05,
+      "loss": 3.9453,
+      "step": 40310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.847741829354178e-05,
+      "loss": 4.0016,
+      "step": 40315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.844376593449488e-05,
+      "loss": 3.8781,
+      "step": 40320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.841011651776721e-05,
+      "loss": 3.9609,
+      "step": 40325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.837647004527921e-05,
+      "loss": 4.0766,
+      "step": 40330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.834282651895105e-05,
+      "loss": 3.9609,
+      "step": 40335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.830918594070268e-05,
+      "loss": 3.9062,
+      "step": 40340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.827554831245415e-05,
+      "loss": 4.0156,
+      "step": 40345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.824191363612506e-05,
+      "loss": 4.0219,
+      "step": 40350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.820828191363495e-05,
+      "loss": 4.0234,
+      "step": 40355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.817465314690317e-05,
+      "loss": 3.8953,
+      "step": 40360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.814102733784895e-05,
+      "loss": 3.9188,
+      "step": 40365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.810740448839132e-05,
+      "loss": 4.0203,
+      "step": 40370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.807378460044906e-05,
+      "loss": 3.9625,
+      "step": 40375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.8040167675941e-05,
+      "loss": 3.9297,
+      "step": 40380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.800655371678558e-05,
+      "loss": 3.8406,
+      "step": 40385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.797294272490114e-05,
+      "loss": 3.9875,
+      "step": 40390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.793933470220586e-05,
+      "loss": 4.0016,
+      "step": 40395
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.79057296506178e-05,
+      "loss": 3.9719,
+      "step": 40400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.787212757205481e-05,
+      "loss": 3.8797,
+      "step": 40405
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.783852846843451e-05,
+      "loss": 4.0062,
+      "step": 40410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.780493234167435e-05,
+      "loss": 3.9141,
+      "step": 40415
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.777133919369182e-05,
+      "loss": 3.9531,
+      "step": 40420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.7737749026404e-05,
+      "loss": 3.9328,
+      "step": 40425
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.770416184172783e-05,
+      "loss": 3.9766,
+      "step": 40430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.767057764158026e-05,
+      "loss": 3.9406,
+      "step": 40435
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.763699642787786e-05,
+      "loss": 3.9703,
+      "step": 40440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.760341820253711e-05,
+      "loss": 3.9094,
+      "step": 40445
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.756984296747432e-05,
+      "loss": 3.9188,
+      "step": 40450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.753627072460572e-05,
+      "loss": 3.9906,
+      "step": 40455
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.750270147584717e-05,
+      "loss": 3.9453,
+      "step": 40460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.746913522311448e-05,
+      "loss": 3.8828,
+      "step": 40465
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.743557196832338e-05,
+      "loss": 3.9891,
+      "step": 40470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.740201171338926e-05,
+      "loss": 3.9734,
+      "step": 40475
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.736845446022736e-05,
+      "loss": 4.0875,
+      "step": 40480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.733490021075287e-05,
+      "loss": 3.9844,
+      "step": 40485
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.730134896688071e-05,
+      "loss": 3.9797,
+      "step": 40490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.726780073052563e-05,
+      "loss": 4.0563,
+      "step": 40495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.723425550360231e-05,
+      "loss": 4.0312,
+      "step": 40500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.720071328802506e-05,
+      "loss": 3.8516,
+      "step": 40505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.716717408570821e-05,
+      "loss": 3.9797,
+      "step": 40510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.713363789856587e-05,
+      "loss": 3.9172,
+      "step": 40515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.710010472851189e-05,
+      "loss": 3.9891,
+      "step": 40520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.706657457746004e-05,
+      "loss": 3.9234,
+      "step": 40525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.703304744732389e-05,
+      "loss": 4.0047,
+      "step": 40530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.699952334001687e-05,
+      "loss": 4.0766,
+      "step": 40535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.696600225745213e-05,
+      "loss": 3.9141,
+      "step": 40540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.693248420154275e-05,
+      "loss": 3.9188,
+      "step": 40545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.689896917420166e-05,
+      "loss": 3.9688,
+      "step": 40550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.686545717734148e-05,
+      "loss": 3.9562,
+      "step": 40555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.68319482128748e-05,
+      "loss": 4.0109,
+      "step": 40560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.6798442282714e-05,
+      "loss": 3.9703,
+      "step": 40565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.676493938877122e-05,
+      "loss": 4.0078,
+      "step": 40570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.673143953295844e-05,
+      "loss": 3.9578,
+      "step": 40575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.669794271718762e-05,
+      "loss": 3.9656,
+      "step": 40580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.666444894337035e-05,
+      "loss": 3.925,
+      "step": 40585
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.663095821341812e-05,
+      "loss": 3.9719,
+      "step": 40590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.65974705292422e-05,
+      "loss": 4.0031,
+      "step": 40595
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.656398589275383e-05,
+      "loss": 3.9469,
+      "step": 40600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.653050430586397e-05,
+      "loss": 3.9859,
+      "step": 40605
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.649702577048332e-05,
+      "loss": 3.9016,
+      "step": 40610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.646355028852262e-05,
+      "loss": 3.8875,
+      "step": 40615
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.643007786189232e-05,
+      "loss": 3.9453,
+      "step": 40620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.63966084925026e-05,
+      "loss": 3.9672,
+      "step": 40625
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.636314218226357e-05,
+      "loss": 3.95,
+      "step": 40630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.632967893308525e-05,
+      "loss": 3.9641,
+      "step": 40635
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.629621874687733e-05,
+      "loss": 3.9641,
+      "step": 40640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.626276162554932e-05,
+      "loss": 4.0125,
+      "step": 40645
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.622930757101078e-05,
+      "loss": 3.9844,
+      "step": 40650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.619585658517083e-05,
+      "loss": 3.9172,
+      "step": 40655
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.616240866993853e-05,
+      "loss": 3.9422,
+      "step": 40660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.612896382722272e-05,
+      "loss": 4.0281,
+      "step": 40665
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.60955220589322e-05,
+      "loss": 3.9969,
+      "step": 40670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.606208336697544e-05,
+      "loss": 4.0094,
+      "step": 40675
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.602864775326079e-05,
+      "loss": 3.9828,
+      "step": 40680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.599521521969636e-05,
+      "loss": 3.9469,
+      "step": 40685
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.596178576819027e-05,
+      "loss": 3.9172,
+      "step": 40690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.592835940065028e-05,
+      "loss": 4.0047,
+      "step": 40695
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.589493611898398e-05,
+      "loss": 3.9109,
+      "step": 40700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.586151592509897e-05,
+      "loss": 3.9297,
+      "step": 40705
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.582809882090246e-05,
+      "loss": 3.9703,
+      "step": 40710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.579468480830159e-05,
+      "loss": 3.9453,
+      "step": 40715
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.576127388920321e-05,
+      "loss": 3.9078,
+      "step": 40720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.572786606551424e-05,
+      "loss": 4.0391,
+      "step": 40725
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.569446133914119e-05,
+      "loss": 4.0312,
+      "step": 40730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.56610597119904e-05,
+      "loss": 3.9937,
+      "step": 40735
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.562766118596828e-05,
+      "loss": 3.9656,
+      "step": 40740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.559426576298073e-05,
+      "loss": 3.8781,
+      "step": 40745
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.556087344493369e-05,
+      "loss": 3.9484,
+      "step": 40750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.552748423373286e-05,
+      "loss": 3.9641,
+      "step": 40755
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.549409813128377e-05,
+      "loss": 4.0469,
+      "step": 40760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.546071513949174e-05,
+      "loss": 4.0031,
+      "step": 40765
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.542733526026201e-05,
+      "loss": 3.9594,
+      "step": 40770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.539395849549947e-05,
+      "loss": 3.9406,
+      "step": 40775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.536058484710898e-05,
+      "loss": 3.9438,
+      "step": 40780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.532721431699525e-05,
+      "loss": 3.9203,
+      "step": 40785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.529384690706264e-05,
+      "loss": 3.9813,
+      "step": 40790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.526048261921544e-05,
+      "loss": 3.9516,
+      "step": 40795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.522712145535784e-05,
+      "loss": 3.8672,
+      "step": 40800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.519376341739371e-05,
+      "loss": 3.9688,
+      "step": 40805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.516040850722672e-05,
+      "loss": 4.0,
+      "step": 40810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.512705672676058e-05,
+      "loss": 3.9562,
+      "step": 40815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.509370807789858e-05,
+      "loss": 3.9656,
+      "step": 40820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.506036256254394e-05,
+      "loss": 3.9562,
+      "step": 40825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.502702018259974e-05,
+      "loss": 3.9906,
+      "step": 40830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.499368093996882e-05,
+      "loss": 4.0234,
+      "step": 40835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.496034483655384e-05,
+      "loss": 4.0734,
+      "step": 40840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.492701187425721e-05,
+      "loss": 3.9734,
+      "step": 40845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.489368205498141e-05,
+      "loss": 4.0641,
+      "step": 40850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.48603553806285e-05,
+      "loss": 3.9422,
+      "step": 40855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.482703185310043e-05,
+      "loss": 3.9359,
+      "step": 40860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.479371147429891e-05,
+      "loss": 3.9141,
+      "step": 40865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.476039424612566e-05,
+      "loss": 3.9344,
+      "step": 40870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.472708017048205e-05,
+      "loss": 4.0062,
+      "step": 40875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.469376924926926e-05,
+      "loss": 3.9703,
+      "step": 40880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.466046148438843e-05,
+      "loss": 3.9828,
+      "step": 40885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.462715687774044e-05,
+      "loss": 3.9813,
+      "step": 40890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.459385543122593e-05,
+      "loss": 3.9531,
+      "step": 40895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.456055714674538e-05,
+      "loss": 3.95,
+      "step": 40900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.452726202619924e-05,
+      "loss": 3.9062,
+      "step": 40905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.449397007148763e-05,
+      "loss": 4.0594,
+      "step": 40910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.446068128451041e-05,
+      "loss": 4.0391,
+      "step": 40915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.442739566716758e-05,
+      "loss": 3.9734,
+      "step": 40920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.439411322135862e-05,
+      "loss": 3.9875,
+      "step": 40925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.436083394898299e-05,
+      "loss": 4.0281,
+      "step": 40930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.432755785193986e-05,
+      "loss": 3.9297,
+      "step": 40935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.429428493212844e-05,
+      "loss": 3.9016,
+      "step": 40940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.426101519144757e-05,
+      "loss": 3.9875,
+      "step": 40945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.422774863179596e-05,
+      "loss": 4.0719,
+      "step": 40950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.419448525507203e-05,
+      "loss": 3.9547,
+      "step": 40955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.416122506317428e-05,
+      "loss": 3.8109,
+      "step": 40960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.412796805800083e-05,
+      "loss": 3.9781,
+      "step": 40965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.409471424144955e-05,
+      "loss": 3.9391,
+      "step": 40970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.406146361541839e-05,
+      "loss": 3.9922,
+      "step": 40975
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.402821618180491e-05,
+      "loss": 4.0359,
+      "step": 40980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.399497194250648e-05,
+      "loss": 4.0328,
+      "step": 40985
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.396173089942043e-05,
+      "loss": 3.9312,
+      "step": 40990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.392849305444383e-05,
+      "loss": 3.9828,
+      "step": 40995
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.389525840947353e-05,
+      "loss": 3.9375,
+      "step": 41000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.386202696640623e-05,
+      "loss": 3.9062,
+      "step": 41005
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.38287987271385e-05,
+      "loss": 3.8766,
+      "step": 41010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.379557369356662e-05,
+      "loss": 4.0094,
+      "step": 41015
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.37623518675868e-05,
+      "loss": 3.9813,
+      "step": 41020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.372913325109498e-05,
+      "loss": 3.9641,
+      "step": 41025
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.369591784598695e-05,
+      "loss": 3.9625,
+      "step": 41030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.366270565415834e-05,
+      "loss": 3.9844,
+      "step": 41035
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.362949667750455e-05,
+      "loss": 3.9719,
+      "step": 41040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.35962909179208e-05,
+      "loss": 3.9203,
+      "step": 41045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.35630883773022e-05,
+      "loss": 4.0656,
+      "step": 41050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.352988905754361e-05,
+      "loss": 3.9531,
+      "step": 41055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.349669296053965e-05,
+      "loss": 3.95,
+      "step": 41060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.346350008818493e-05,
+      "loss": 3.9594,
+      "step": 41065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.34303104423737e-05,
+      "loss": 3.9422,
+      "step": 41070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.339712402500014e-05,
+      "loss": 3.9094,
+      "step": 41075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.33639408379581e-05,
+      "loss": 3.9391,
+      "step": 41080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.333076088314148e-05,
+      "loss": 3.8906,
+      "step": 41085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.329758416244381e-05,
+      "loss": 3.9906,
+      "step": 41090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.326441067775844e-05,
+      "loss": 3.9625,
+      "step": 41095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.323124043097868e-05,
+      "loss": 3.9766,
+      "step": 41100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.319807342399748e-05,
+      "loss": 3.8719,
+      "step": 41105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.316490965870775e-05,
+      "loss": 3.9734,
+      "step": 41110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.313174913700205e-05,
+      "loss": 4.0375,
+      "step": 41115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.309859186077298e-05,
+      "loss": 4.0,
+      "step": 41120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.306543783191274e-05,
+      "loss": 3.9438,
+      "step": 41125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.303228705231346e-05,
+      "loss": 3.9844,
+      "step": 41130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.299913952386702e-05,
+      "loss": 4.0,
+      "step": 41135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.296599524846525e-05,
+      "loss": 3.9625,
+      "step": 41140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.293285422799963e-05,
+      "loss": 3.9016,
+      "step": 41145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.289971646436145e-05,
+      "loss": 3.9328,
+      "step": 41150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.286658195944204e-05,
+      "loss": 3.9516,
+      "step": 41155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.283345071513231e-05,
+      "loss": 4.0297,
+      "step": 41160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.280032273332306e-05,
+      "loss": 4.0359,
+      "step": 41165
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.276719801590488e-05,
+      "loss": 4.0938,
+      "step": 41170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.273407656476826e-05,
+      "loss": 3.9906,
+      "step": 41175
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.270095838180342e-05,
+      "loss": 3.8125,
+      "step": 41180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.266784346890038e-05,
+      "loss": 3.9016,
+      "step": 41185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.263473182794909e-05,
+      "loss": 3.875,
+      "step": 41190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.260162346083922e-05,
+      "loss": 3.9484,
+      "step": 41195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.25685183694602e-05,
+      "loss": 3.9625,
+      "step": 41200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.253541655570134e-05,
+      "loss": 3.8906,
+      "step": 41205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.250231802145187e-05,
+      "loss": 3.95,
+      "step": 41210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.246922276860066e-05,
+      "loss": 4.0547,
+      "step": 41215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.243613079903642e-05,
+      "loss": 3.9234,
+      "step": 41220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.240304211464775e-05,
+      "loss": 3.8969,
+      "step": 41225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.236995671732306e-05,
+      "loss": 3.9344,
+      "step": 41230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.233687460895048e-05,
+      "loss": 3.9328,
+      "step": 41235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.230379579141803e-05,
+      "loss": 4.0594,
+      "step": 41240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.227072026661355e-05,
+      "loss": 3.9625,
+      "step": 41245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.223764803642459e-05,
+      "loss": 4.0563,
+      "step": 41250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.22045791027387e-05,
+      "loss": 3.9141,
+      "step": 41255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.2171513467443e-05,
+      "loss": 4.0125,
+      "step": 41260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.213845113242463e-05,
+      "loss": 3.9609,
+      "step": 41265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.210539209957045e-05,
+      "loss": 3.9547,
+      "step": 41270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.207233637076712e-05,
+      "loss": 4.0359,
+      "step": 41275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.203928394790114e-05,
+      "loss": 4.0281,
+      "step": 41280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.200623483285885e-05,
+      "loss": 3.9953,
+      "step": 41285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.197318902752632e-05,
+      "loss": 3.9703,
+      "step": 41290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.194014653378947e-05,
+      "loss": 3.9984,
+      "step": 41295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.190710735353407e-05,
+      "loss": 3.9641,
+      "step": 41300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18740714886457e-05,
+      "loss": 3.9422,
+      "step": 41305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.184103894100965e-05,
+      "loss": 3.9547,
+      "step": 41310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.180800971251108e-05,
+      "loss": 4.0812,
+      "step": 41315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17749838050351e-05,
+      "loss": 3.9875,
+      "step": 41320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.174196122046637e-05,
+      "loss": 3.875,
+      "step": 41325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17089419606895e-05,
+      "loss": 3.9594,
+      "step": 41330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.1675926027589e-05,
+      "loss": 3.9516,
+      "step": 41335
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.164291342304903e-05,
+      "loss": 3.9531,
+      "step": 41340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160990414895363e-05,
+      "loss": 3.9828,
+      "step": 41345
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.157689820718657e-05,
+      "loss": 4.0422,
+      "step": 41350
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.154389559963161e-05,
+      "loss": 3.9234,
+      "step": 41355
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.151089632817219e-05,
+      "loss": 3.9688,
+      "step": 41360
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.14779003946915e-05,
+      "loss": 3.9,
+      "step": 41365
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.144490780107276e-05,
+      "loss": 3.9188,
+      "step": 41370
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.141191854919876e-05,
+      "loss": 3.8906,
+      "step": 41375
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.137893264095224e-05,
+      "loss": 4.0203,
+      "step": 41380
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.134595007821562e-05,
+      "loss": 4.0312,
+      "step": 41385
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.131297086287135e-05,
+      "loss": 3.9453,
+      "step": 41390
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.127999499680151e-05,
+      "loss": 4.0266,
+      "step": 41395
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.124702248188802e-05,
+      "loss": 4.0109,
+      "step": 41400
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.121405332001257e-05,
+      "loss": 3.9125,
+      "step": 41405
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.118108751305681e-05,
+      "loss": 4.0125,
+      "step": 41410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.11481250629021e-05,
+      "loss": 3.9156,
+      "step": 41415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.111516597142948e-05,
+      "loss": 4.0938,
+      "step": 41420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.108221024052011e-05,
+      "loss": 4.0156,
+      "step": 41425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.104925787205467e-05,
+      "loss": 3.9609,
+      "step": 41430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.101630886791377e-05,
+      "loss": 3.8969,
+      "step": 41435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.098336322997779e-05,
+      "loss": 3.9781,
+      "step": 41440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.095042096012699e-05,
+      "loss": 3.8531,
+      "step": 41445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.09174820602414e-05,
+      "loss": 3.8969,
+      "step": 41450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.088454653220073e-05,
+      "loss": 3.8531,
+      "step": 41455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.085161437788475e-05,
+      "loss": 3.9031,
+      "step": 41460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.081868559917286e-05,
+      "loss": 4.0844,
+      "step": 41465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.078576019794426e-05,
+      "loss": 3.8813,
+      "step": 41470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.075283817607805e-05,
+      "loss": 3.9375,
+      "step": 41475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.07199195354531e-05,
+      "loss": 3.8719,
+      "step": 41480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.068700427794804e-05,
+      "loss": 3.9453,
+      "step": 41485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.065409240544143e-05,
+      "loss": 3.9312,
+      "step": 41490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.062118391981143e-05,
+      "loss": 3.9641,
+      "step": 41495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.058827882293624e-05,
+      "loss": 4.0156,
+      "step": 41500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.055537711669367e-05,
+      "loss": 4.0328,
+      "step": 41505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.052247880296147e-05,
+      "loss": 3.9813,
+      "step": 41510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.048958388361719e-05,
+      "loss": 3.9813,
+      "step": 41515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.045669236053806e-05,
+      "loss": 4.0047,
+      "step": 41520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.042380423560126e-05,
+      "loss": 3.9594,
+      "step": 41525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.039091951068369e-05,
+      "loss": 3.9969,
+      "step": 41530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.035803818766212e-05,
+      "loss": 3.9688,
+      "step": 41535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.032516026841308e-05,
+      "loss": 4.0578,
+      "step": 41540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.029228575481286e-05,
+      "loss": 4.0078,
+      "step": 41545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.025941464873765e-05,
+      "loss": 3.9719,
+      "step": 41550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.022654695206346e-05,
+      "loss": 3.9734,
+      "step": 41555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.019368266666602e-05,
+      "loss": 3.9703,
+      "step": 41560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.016082179442081e-05,
+      "loss": 3.9891,
+      "step": 41565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.012796433720332e-05,
+      "loss": 3.8687,
+      "step": 41570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.009511029688871e-05,
+      "loss": 3.9609,
+      "step": 41575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.006225967535193e-05,
+      "loss": 3.9562,
+      "step": 41580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.002941247446771e-05,
+      "loss": 4.0375,
+      "step": 41585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.999656869611076e-05,
+      "loss": 3.8547,
+      "step": 41590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.996372834215545e-05,
+      "loss": 3.9422,
+      "step": 41595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.993089141447589e-05,
+      "loss": 3.9406,
+      "step": 41600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.989805791494621e-05,
+      "loss": 3.9703,
+      "step": 41605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.986522784544017e-05,
+      "loss": 4.0516,
+      "step": 41610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.983240120783138e-05,
+      "loss": 3.9438,
+      "step": 41615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.97995780039932e-05,
+      "loss": 3.9297,
+      "step": 41620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.976675823579898e-05,
+      "loss": 3.9047,
+      "step": 41625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.973394190512169e-05,
+      "loss": 3.9297,
+      "step": 41630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.970112901383407e-05,
+      "loss": 4.0453,
+      "step": 41635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.966831956380891e-05,
+      "loss": 3.8922,
+      "step": 41640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.963551355691857e-05,
+      "loss": 3.8125,
+      "step": 41645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.960271099503531e-05,
+      "loss": 3.9266,
+      "step": 41650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.95699118800311e-05,
+      "loss": 3.9953,
+      "step": 41655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.95371162137779e-05,
+      "loss": 3.9547,
+      "step": 41660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.950432399814734e-05,
+      "loss": 3.8453,
+      "step": 41665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.947153523501085e-05,
+      "loss": 3.9719,
+      "step": 41670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.943874992623959e-05,
+      "loss": 3.9391,
+      "step": 41675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.94059680737048e-05,
+      "loss": 3.9859,
+      "step": 41680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.937318967927728e-05,
+      "loss": 3.8875,
+      "step": 41685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.934041474482759e-05,
+      "loss": 3.9891,
+      "step": 41690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.930764327222634e-05,
+      "loss": 4.0297,
+      "step": 41695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.927487526334378e-05,
+      "loss": 3.8859,
+      "step": 41700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.924211072004993e-05,
+      "loss": 3.9078,
+      "step": 41705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.920934964421463e-05,
+      "loss": 3.9047,
+      "step": 41710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.917659203770766e-05,
+      "loss": 3.9391,
+      "step": 41715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.914383790239849e-05,
+      "loss": 3.9078,
+      "step": 41720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.911108724015628e-05,
+      "loss": 3.9344,
+      "step": 41725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.907834005285023e-05,
+      "loss": 3.9469,
+      "step": 41730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.904559634234925e-05,
+      "loss": 3.9016,
+      "step": 41735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.90128561105219e-05,
+      "loss": 3.9766,
+      "step": 41740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.898011935923676e-05,
+      "loss": 3.9531,
+      "step": 41745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.894738609036212e-05,
+      "loss": 3.9734,
+      "step": 41750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.891465630576602e-05,
+      "loss": 3.9953,
+      "step": 41755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.888193000731643e-05,
+      "loss": 3.9016,
+      "step": 41760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.884920719688094e-05,
+      "loss": 3.9188,
+      "step": 41765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.881648787632712e-05,
+      "loss": 3.9781,
+      "step": 41770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.878377204752223e-05,
+      "loss": 4.0625,
+      "step": 41775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.875105971233339e-05,
+      "loss": 3.975,
+      "step": 41780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.871835087262747e-05,
+      "loss": 3.9047,
+      "step": 41785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.868564553027123e-05,
+      "loss": 3.9656,
+      "step": 41790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.865294368713109e-05,
+      "loss": 3.8641,
+      "step": 41795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.862024534507332e-05,
+      "loss": 3.9422,
+      "step": 41800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.858755050596416e-05,
+      "loss": 3.9937,
+      "step": 41805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.85548591716694e-05,
+      "loss": 3.9406,
+      "step": 41810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.85221713440547e-05,
+      "loss": 3.9688,
+      "step": 41815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.848948702498567e-05,
+      "loss": 3.8594,
+      "step": 41820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.845680621632755e-05,
+      "loss": 3.9562,
+      "step": 41825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.842412891994546e-05,
+      "loss": 3.9031,
+      "step": 41830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.839145513770419e-05,
+      "loss": 3.9656,
+      "step": 41835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.835878487146861e-05,
+      "loss": 3.9562,
+      "step": 41840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.832611812310312e-05,
+      "loss": 3.9984,
+      "step": 41845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.8293454894472e-05,
+      "loss": 3.9922,
+      "step": 41850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.826079518743934e-05,
+      "loss": 3.9406,
+      "step": 41855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.822813900386911e-05,
+      "loss": 3.9859,
+      "step": 41860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.819548634562494e-05,
+      "loss": 3.9813,
+      "step": 41865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.816283721457025e-05,
+      "loss": 4.05,
+      "step": 41870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.81301916125685e-05,
+      "loss": 4.0312,
+      "step": 41875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.809754954148268e-05,
+      "loss": 4.0531,
+      "step": 41880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.806491100317569e-05,
+      "loss": 3.8828,
+      "step": 41885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.803227599951015e-05,
+      "loss": 3.9984,
+      "step": 41890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.799964453234864e-05,
+      "loss": 3.9219,
+      "step": 41895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.79670166035534e-05,
+      "loss": 3.9156,
+      "step": 41900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.793439221498648e-05,
+      "loss": 3.9547,
+      "step": 41905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.790177136850982e-05,
+      "loss": 3.8672,
+      "step": 41910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.786915406598509e-05,
+      "loss": 3.9844,
+      "step": 41915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.783654030927371e-05,
+      "loss": 3.8766,
+      "step": 41920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.780393010023693e-05,
+      "loss": 3.9125,
+      "step": 41925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.777132344073592e-05,
+      "loss": 3.8922,
+      "step": 41930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.773872033263148e-05,
+      "loss": 3.9688,
+      "step": 41935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.770612077778428e-05,
+      "loss": 3.9438,
+      "step": 41940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.767352477805472e-05,
+      "loss": 4.0312,
+      "step": 41945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.76409323353032e-05,
+      "loss": 4.0422,
+      "step": 41950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.760834345138966e-05,
+      "loss": 3.9391,
+      "step": 41955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.757575812817394e-05,
+      "loss": 3.9422,
+      "step": 41960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.754317636751579e-05,
+      "loss": 3.925,
+      "step": 41965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.751059817127459e-05,
+      "loss": 3.8578,
+      "step": 41970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.747802354130959e-05,
+      "loss": 4.0047,
+      "step": 41975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.744545247947976e-05,
+      "loss": 3.9719,
+      "step": 41980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.741288498764404e-05,
+      "loss": 3.9813,
+      "step": 41985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.738032106766102e-05,
+      "loss": 3.975,
+      "step": 41990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.734776072138908e-05,
+      "loss": 3.9266,
+      "step": 41995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.731520395068654e-05,
+      "loss": 3.9516,
+      "step": 42000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.728265075741135e-05,
+      "loss": 3.9219,
+      "step": 42005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.725010114342135e-05,
+      "loss": 3.9391,
+      "step": 42010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.721755511057405e-05,
+      "loss": 4.0016,
+      "step": 42015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.718501266072701e-05,
+      "loss": 3.8859,
+      "step": 42020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.715247379573738e-05,
+      "loss": 3.9156,
+      "step": 42025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.711993851746211e-05,
+      "loss": 4.0062,
+      "step": 42030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.708740682775799e-05,
+      "loss": 3.9703,
+      "step": 42035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.705487872848167e-05,
+      "loss": 3.9703,
+      "step": 42040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.702235422148948e-05,
+      "loss": 4.0141,
+      "step": 42045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.698983330863759e-05,
+      "loss": 3.9141,
+      "step": 42050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.695731599178206e-05,
+      "loss": 4.0328,
+      "step": 42055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.692480227277856e-05,
+      "loss": 3.9688,
+      "step": 42060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.68922921534827e-05,
+      "loss": 3.9391,
+      "step": 42065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.685978563574976e-05,
+      "loss": 3.9094,
+      "step": 42070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.6827282721435e-05,
+      "loss": 4.0078,
+      "step": 42075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.679478341239333e-05,
+      "loss": 4.0328,
+      "step": 42080
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.67622877104794e-05,
+      "loss": 3.8734,
+      "step": 42085
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.672979561754787e-05,
+      "loss": 3.85,
+      "step": 42090
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.669730713545302e-05,
+      "loss": 4.0187,
+      "step": 42095
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.666482226604892e-05,
+      "loss": 3.9312,
+      "step": 42100
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.663234101118955e-05,
+      "loss": 3.9813,
+      "step": 42105
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.659986337272862e-05,
+      "loss": 3.9594,
+      "step": 42110
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.656738935251954e-05,
+      "loss": 3.9062,
+      "step": 42115
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.653491895241573e-05,
+      "loss": 3.9562,
+      "step": 42120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.650245217427023e-05,
+      "loss": 3.9516,
+      "step": 42125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.646998901993584e-05,
+      "loss": 3.9391,
+      "step": 42130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.643752949126538e-05,
+      "loss": 3.9438,
+      "step": 42135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.640507359011124e-05,
+      "loss": 3.8953,
+      "step": 42140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.637262131832564e-05,
+      "loss": 3.9625,
+      "step": 42145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.634017267776074e-05,
+      "loss": 4.0031,
+      "step": 42150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.630772767026833e-05,
+      "loss": 3.9188,
+      "step": 42155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.62752862977e-05,
+      "loss": 3.9312,
+      "step": 42160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.624284856190727e-05,
+      "loss": 3.9984,
+      "step": 42165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.621041446474135e-05,
+      "loss": 3.9562,
+      "step": 42170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.617798400805322e-05,
+      "loss": 3.9984,
+      "step": 42175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.614555719369365e-05,
+      "loss": 4.0297,
+      "step": 42180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.611313402351334e-05,
+      "loss": 4.0062,
+      "step": 42185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.608071449936265e-05,
+      "loss": 3.8906,
+      "step": 42190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.60482986230917e-05,
+      "loss": 3.9828,
+      "step": 42195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.601588639655057e-05,
+      "loss": 4.0031,
+      "step": 42200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.5983477821589e-05,
+      "loss": 3.9672,
+      "step": 42205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.59510729000565e-05,
+      "loss": 3.9484,
+      "step": 42210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.591867163380242e-05,
+      "loss": 3.9469,
+      "step": 42215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.588627402467598e-05,
+      "loss": 3.9922,
+      "step": 42220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.585388007452607e-05,
+      "loss": 3.9672,
+      "step": 42225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.582148978520135e-05,
+      "loss": 3.9984,
+      "step": 42230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.578910315855048e-05,
+      "loss": 3.9641,
+      "step": 42235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.575672019642168e-05,
+      "loss": 3.9062,
+      "step": 42240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.572434090066307e-05,
+      "loss": 3.9203,
+      "step": 42245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.569196527312245e-05,
+      "loss": 4.0641,
+      "step": 42250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.565959331564766e-05,
+      "loss": 3.9688,
+      "step": 42255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.562722503008609e-05,
+      "loss": 3.9469,
+      "step": 42260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.559486041828494e-05,
+      "loss": 4.0641,
+      "step": 42265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.556249948209137e-05,
+      "loss": 4.0563,
+      "step": 42270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.55301422233522e-05,
+      "loss": 3.9422,
+      "step": 42275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.549778864391399e-05,
+      "loss": 3.9734,
+      "step": 42280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.546543874562319e-05,
+      "loss": 3.9844,
+      "step": 42285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.543309253032609e-05,
+      "loss": 4.075,
+      "step": 42290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.540074999986861e-05,
+      "loss": 3.9937,
+      "step": 42295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.536841115609657e-05,
+      "loss": 3.9547,
+      "step": 42300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.533607600085548e-05,
+      "loss": 3.95,
+      "step": 42305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.530374453599085e-05,
+      "loss": 3.9469,
+      "step": 42310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.527141676334777e-05,
+      "loss": 3.9594,
+      "step": 42315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.52390926847711e-05,
+      "loss": 3.8828,
+      "step": 42320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.520677230210574e-05,
+      "loss": 3.9797,
+      "step": 42325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.517445561719613e-05,
+      "loss": 3.95,
+      "step": 42330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.514214263188659e-05,
+      "loss": 3.8531,
+      "step": 42335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.510983334802119e-05,
+      "loss": 3.9312,
+      "step": 42340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.507752776744391e-05,
+      "loss": 4.0078,
+      "step": 42345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.50452258919984e-05,
+      "loss": 4.0141,
+      "step": 42350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.501292772352804e-05,
+      "loss": 3.8625,
+      "step": 42355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.498063326387624e-05,
+      "loss": 4.0844,
+      "step": 42360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.494834251488597e-05,
+      "loss": 3.925,
+      "step": 42365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.491605547840008e-05,
+      "loss": 3.925,
+      "step": 42370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.488377215626112e-05,
+      "loss": 3.9719,
+      "step": 42375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.485149255031162e-05,
+      "loss": 3.8766,
+      "step": 42380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.481921666239374e-05,
+      "loss": 3.9172,
+      "step": 42385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.478694449434947e-05,
+      "loss": 3.9813,
+      "step": 42390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.47546760480205e-05,
+      "loss": 3.8937,
+      "step": 42395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.472241132524853e-05,
+      "loss": 3.9,
+      "step": 42400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.469015032787484e-05,
+      "loss": 4.0031,
+      "step": 42405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.465789305774051e-05,
+      "loss": 4.0234,
+      "step": 42410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.46256395166866e-05,
+      "loss": 3.9266,
+      "step": 42415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.459338970655375e-05,
+      "loss": 4.0031,
+      "step": 42420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.456114362918247e-05,
+      "loss": 3.9266,
+      "step": 42425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.452890128641298e-05,
+      "loss": 3.8906,
+      "step": 42430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.449666268008546e-05,
+      "loss": 4.0359,
+      "step": 42435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.446442781203974e-05,
+      "loss": 3.9016,
+      "step": 42440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.44321966841154e-05,
+      "loss": 3.9672,
+      "step": 42445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.439996929815196e-05,
+      "loss": 3.9781,
+      "step": 42450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.436774565598861e-05,
+      "loss": 3.8578,
+      "step": 42455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.433552575946436e-05,
+      "loss": 3.9766,
+      "step": 42460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.430330961041792e-05,
+      "loss": 3.9859,
+      "step": 42465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.4271097210688e-05,
+      "loss": 4.0656,
+      "step": 42470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.423888856211293e-05,
+      "loss": 3.9906,
+      "step": 42475
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.420668366653084e-05,
+      "loss": 3.9547,
+      "step": 42480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.417448252577957e-05,
+      "loss": 3.9219,
+      "step": 42485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.414228514169702e-05,
+      "loss": 3.9359,
+      "step": 42490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.41100915161206e-05,
+      "loss": 3.9562,
+      "step": 42495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.40779016508876e-05,
+      "loss": 4.0078,
+      "step": 42500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.404571554783514e-05,
+      "loss": 3.9094,
+      "step": 42505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.401353320880007e-05,
+      "loss": 3.8734,
+      "step": 42510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.398135463561902e-05,
+      "loss": 3.9656,
+      "step": 42515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.394917983012836e-05,
+      "loss": 3.9828,
+      "step": 42520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.391700879416447e-05,
+      "loss": 3.9734,
+      "step": 42525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.388484152956326e-05,
+      "loss": 3.9438,
+      "step": 42530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.385267803816046e-05,
+      "loss": 3.9203,
+      "step": 42535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.382051832179177e-05,
+      "loss": 3.9813,
+      "step": 42540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.37883623822925e-05,
+      "loss": 3.8781,
+      "step": 42545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.375621022149778e-05,
+      "loss": 3.9813,
+      "step": 42550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.372406184124245e-05,
+      "loss": 3.9641,
+      "step": 42555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.36919172433614e-05,
+      "loss": 3.95,
+      "step": 42560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.365977642968902e-05,
+      "loss": 3.9375,
+      "step": 42565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.362763940205961e-05,
+      "loss": 4.025,
+      "step": 42570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.359550616230717e-05,
+      "loss": 3.9609,
+      "step": 42575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.356337671226566e-05,
+      "loss": 3.9937,
+      "step": 42580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.353125105376863e-05,
+      "loss": 3.8813,
+      "step": 42585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.349912918864949e-05,
+      "loss": 3.9094,
+      "step": 42590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.34670111187415e-05,
+      "loss": 3.9625,
+      "step": 42595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.343489684587762e-05,
+      "loss": 3.9813,
+      "step": 42600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.340278637189055e-05,
+      "loss": 3.9203,
+      "step": 42605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.337067969861297e-05,
+      "loss": 4.0516,
+      "step": 42610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.33385768278771e-05,
+      "loss": 4.0266,
+      "step": 42615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.330647776151503e-05,
+      "loss": 3.9234,
+      "step": 42620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.327438250135875e-05,
+      "loss": 3.975,
+      "step": 42625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.324229104923994e-05,
+      "loss": 3.9281,
+      "step": 42630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.321020340698995e-05,
+      "loss": 3.8016,
+      "step": 42635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.317811957644016e-05,
+      "loss": 4.0094,
+      "step": 42640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.314603955942154e-05,
+      "loss": 4.0391,
+      "step": 42645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.311396335776483e-05,
+      "loss": 3.9141,
+      "step": 42650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.308189097330075e-05,
+      "loss": 3.9203,
+      "step": 42655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.304982240785961e-05,
+      "loss": 3.9172,
+      "step": 42660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.301775766327151e-05,
+      "loss": 3.8656,
+      "step": 42665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.298569674136651e-05,
+      "loss": 3.9266,
+      "step": 42670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.295363964397425e-05,
+      "loss": 3.9594,
+      "step": 42675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.29215863729242e-05,
+      "loss": 3.8734,
+      "step": 42680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.288953693004575e-05,
+      "loss": 3.9,
+      "step": 42685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.285749131716792e-05,
+      "loss": 4.0453,
+      "step": 42690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.282544953611953e-05,
+      "loss": 3.8953,
+      "step": 42695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.279341158872915e-05,
+      "loss": 3.8547,
+      "step": 42700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.276137747682533e-05,
+      "loss": 3.9797,
+      "step": 42705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.272934720223616e-05,
+      "loss": 3.9375,
+      "step": 42710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.269732076678962e-05,
+      "loss": 4.0375,
+      "step": 42715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.266529817231352e-05,
+      "loss": 3.9672,
+      "step": 42720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.263327942063534e-05,
+      "loss": 4.0031,
+      "step": 42725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.260126451358243e-05,
+      "loss": 3.9062,
+      "step": 42730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.256925345298179e-05,
+      "loss": 3.9875,
+      "step": 42735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.253724624066044e-05,
+      "loss": 4.0563,
+      "step": 42740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.250524287844496e-05,
+      "loss": 3.9422,
+      "step": 42745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.247324336816175e-05,
+      "loss": 3.8813,
+      "step": 42750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.244124771163702e-05,
+      "loss": 3.9609,
+      "step": 42755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.240925591069687e-05,
+      "loss": 3.9844,
+      "step": 42760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.237726796716701e-05,
+      "loss": 3.9891,
+      "step": 42765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.234528388287292e-05,
+      "loss": 3.8766,
+      "step": 42770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.231330365964007e-05,
+      "loss": 3.9641,
+      "step": 42775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.228132729929354e-05,
+      "loss": 3.9391,
+      "step": 42780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.224935480365818e-05,
+      "loss": 4.0328,
+      "step": 42785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.221738617455861e-05,
+      "loss": 3.9438,
+      "step": 42790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.218542141381944e-05,
+      "loss": 3.8984,
+      "step": 42795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.21534605232648e-05,
+      "loss": 3.9719,
+      "step": 42800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.212150350471867e-05,
+      "loss": 3.9734,
+      "step": 42805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.208955036000494e-05,
+      "loss": 3.8828,
+      "step": 42810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.205760109094711e-05,
+      "loss": 4.0172,
+      "step": 42815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.202565569936856e-05,
+      "loss": 3.875,
+      "step": 42820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.199371418709234e-05,
+      "loss": 3.9094,
+      "step": 42825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.196177655594148e-05,
+      "loss": 4.0047,
+      "step": 42830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.192984280773859e-05,
+      "loss": 3.9656,
+      "step": 42835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.189791294430615e-05,
+      "loss": 3.9016,
+      "step": 42840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.186598696746632e-05,
+      "loss": 4.0094,
+      "step": 42845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.183406487904127e-05,
+      "loss": 4.0234,
+      "step": 42850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.18021466808527e-05,
+      "loss": 3.9266,
+      "step": 42855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.177023237472216e-05,
+      "loss": 3.9359,
+      "step": 42860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.173832196247108e-05,
+      "loss": 3.9156,
+      "step": 42865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.170641544592059e-05,
+      "loss": 3.9688,
+      "step": 42870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.167451282689154e-05,
+      "loss": 3.9234,
+      "step": 42875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.164261410720458e-05,
+      "loss": 3.9453,
+      "step": 42880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.161071928868028e-05,
+      "loss": 4.0,
+      "step": 42885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.157882837313885e-05,
+      "loss": 3.9234,
+      "step": 42890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.154694136240023e-05,
+      "loss": 3.9469,
+      "step": 42895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.151505825828434e-05,
+      "loss": 3.9406,
+      "step": 42900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.148317906261066e-05,
+      "loss": 3.9547,
+      "step": 42905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.145130377719859e-05,
+      "loss": 3.9188,
+      "step": 42910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.141943240386716e-05,
+      "loss": 3.9531,
+      "step": 42915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.13875649444354e-05,
+      "loss": 3.9672,
+      "step": 42920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.135570140072192e-05,
+      "loss": 3.9344,
+      "step": 42925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.132384177454522e-05,
+      "loss": 3.8656,
+      "step": 42930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.129198606772341e-05,
+      "loss": 3.9203,
+      "step": 42935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.126013428207467e-05,
+      "loss": 3.9203,
+      "step": 42940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.122828641941668e-05,
+      "loss": 4.0219,
+      "step": 42945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.119644248156697e-05,
+      "loss": 3.9797,
+      "step": 42950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.116460247034301e-05,
+      "loss": 3.9031,
+      "step": 42955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.11327663875618e-05,
+      "loss": 3.8891,
+      "step": 42960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.110093423504029e-05,
+      "loss": 3.9031,
+      "step": 42965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.106910601459503e-05,
+      "loss": 3.9344,
+      "step": 42970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.103728172804262e-05,
+      "loss": 3.9719,
+      "step": 42975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.100546137719918e-05,
+      "loss": 3.9219,
+      "step": 42980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.09736449638807e-05,
+      "loss": 3.9906,
+      "step": 42985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.094183248990298e-05,
+      "loss": 3.9469,
+      "step": 42990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.091002395708155e-05,
+      "loss": 3.9766,
+      "step": 42995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.087821936723174e-05,
+      "loss": 4.0047,
+      "step": 43000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.084641872216856e-05,
+      "loss": 3.9141,
+      "step": 43005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.081462202370699e-05,
+      "loss": 3.9422,
+      "step": 43010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.078282927366165e-05,
+      "loss": 4.0187,
+      "step": 43015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.075104047384689e-05,
+      "loss": 3.8859,
+      "step": 43020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.071925562607689e-05,
+      "loss": 3.9031,
+      "step": 43025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.068747473216571e-05,
+      "loss": 3.9531,
+      "step": 43030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.065569779392704e-05,
+      "loss": 3.8703,
+      "step": 43035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.062392481317433e-05,
+      "loss": 3.9391,
+      "step": 43040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.059215579172101e-05,
+      "loss": 3.8719,
+      "step": 43045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.056039073138003e-05,
+      "loss": 3.9031,
+      "step": 43050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.052862963396427e-05,
+      "loss": 3.9469,
+      "step": 43055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.049687250128625e-05,
+      "loss": 4.0109,
+      "step": 43060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.046511933515852e-05,
+      "loss": 3.9078,
+      "step": 43065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.04333701373931e-05,
+      "loss": 3.9813,
+      "step": 43070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.040162490980192e-05,
+      "loss": 3.9469,
+      "step": 43075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.036988365419677e-05,
+      "loss": 3.9406,
+      "step": 43080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.03381463723891e-05,
+      "loss": 3.8734,
+      "step": 43085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.030641306619007e-05,
+      "loss": 3.9031,
+      "step": 43090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.027468373741084e-05,
+      "loss": 3.9672,
+      "step": 43095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.024295838786214e-05,
+      "loss": 3.9937,
+      "step": 43100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.021123701935447e-05,
+      "loss": 3.9438,
+      "step": 43105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.017951963369831e-05,
+      "loss": 3.8828,
+      "step": 43110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.01478062327037e-05,
+      "loss": 3.9703,
+      "step": 43115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.011609681818047e-05,
+      "loss": 3.9344,
+      "step": 43120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.008439139193842e-05,
+      "loss": 3.9516,
+      "step": 43125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.005268995578689e-05,
+      "loss": 3.9359,
+      "step": 43130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.002099251153509e-05,
+      "loss": 4.0266,
+      "step": 43135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.998929906099194e-05,
+      "loss": 3.9078,
+      "step": 43140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.995760960596631e-05,
+      "loss": 3.9203,
+      "step": 43145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.992592414826667e-05,
+      "loss": 3.9562,
+      "step": 43150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.989424268970125e-05,
+      "loss": 4.0,
+      "step": 43155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.986256523207821e-05,
+      "loss": 3.9672,
+      "step": 43160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.983089177720534e-05,
+      "loss": 4.0312,
+      "step": 43165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.97992223268902e-05,
+      "loss": 3.8578,
+      "step": 43170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.976755688294027e-05,
+      "loss": 3.9125,
+      "step": 43175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.973589544716268e-05,
+      "loss": 4.0391,
+      "step": 43180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.970423802136422e-05,
+      "loss": 3.9359,
+      "step": 43185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.967258460735176e-05,
+      "loss": 4.0219,
+      "step": 43190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.964093520693167e-05,
+      "loss": 3.9672,
+      "step": 43195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.960928982191023e-05,
+      "loss": 3.9188,
+      "step": 43200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.957764845409335e-05,
+      "loss": 3.8703,
+      "step": 43205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.954601110528691e-05,
+      "loss": 3.9656,
+      "step": 43210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.951437777729642e-05,
+      "loss": 3.95,
+      "step": 43215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.948274847192715e-05,
+      "loss": 4.0156,
+      "step": 43220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.945112319098428e-05,
+      "loss": 3.9047,
+      "step": 43225
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.941950193627261e-05,
+      "loss": 3.9641,
+      "step": 43230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.938788470959679e-05,
+      "loss": 4.0,
+      "step": 43235
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.935627151276116e-05,
+      "loss": 4.0859,
+      "step": 43240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.932466234756997e-05,
+      "loss": 3.9766,
+      "step": 43245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.929305721582712e-05,
+      "loss": 3.9578,
+      "step": 43250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.926145611933628e-05,
+      "loss": 3.9078,
+      "step": 43255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.922985905990102e-05,
+      "loss": 3.8281,
+      "step": 43260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.919826603932452e-05,
+      "loss": 3.975,
+      "step": 43265
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.916667705940985e-05,
+      "loss": 3.9312,
+      "step": 43270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.913509212195966e-05,
+      "loss": 3.9062,
+      "step": 43275
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.910351122877667e-05,
+      "loss": 4.0031,
+      "step": 43280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.907193438166316e-05,
+      "loss": 3.9672,
+      "step": 43285
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.904036158242118e-05,
+      "loss": 3.9844,
+      "step": 43290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.900879283285258e-05,
+      "loss": 3.8922,
+      "step": 43295
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.897722813475905e-05,
+      "loss": 3.9594,
+      "step": 43300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.894566748994198e-05,
+      "loss": 3.9828,
+      "step": 43305
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.891411090020247e-05,
+      "loss": 3.9031,
+      "step": 43310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.888255836734157e-05,
+      "loss": 4.0047,
+      "step": 43315
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.885100989315993e-05,
+      "loss": 3.9359,
+      "step": 43320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.881946547945803e-05,
+      "loss": 3.9953,
+      "step": 43325
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.878792512803603e-05,
+      "loss": 3.8859,
+      "step": 43330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.875638884069409e-05,
+      "loss": 3.9984,
+      "step": 43335
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.872485661923193e-05,
+      "loss": 3.8844,
+      "step": 43340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.8693328465449e-05,
+      "loss": 3.825,
+      "step": 43345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.866180438114475e-05,
+      "loss": 3.9547,
+      "step": 43350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.863028436811823e-05,
+      "loss": 3.9922,
+      "step": 43355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.859876842816827e-05,
+      "loss": 3.9125,
+      "step": 43360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.856725656309341e-05,
+      "loss": 3.9438,
+      "step": 43365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.85357487746922e-05,
+      "loss": 3.9734,
+      "step": 43370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.850424506476268e-05,
+      "loss": 3.9937,
+      "step": 43375
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.847274543510282e-05,
+      "loss": 3.9391,
+      "step": 43380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.844124988751021e-05,
+      "loss": 3.9344,
+      "step": 43385
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.840975842378244e-05,
+      "loss": 3.9609,
+      "step": 43390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.837827104571666e-05,
+      "loss": 3.925,
+      "step": 43395
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.834678775510981e-05,
+      "loss": 3.8984,
+      "step": 43400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.831530855375877e-05,
+      "loss": 3.9266,
+      "step": 43405
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.828383344345998e-05,
+      "loss": 4.0172,
+      "step": 43410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.825236242600973e-05,
+      "loss": 4.0078,
+      "step": 43415
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.822089550320403e-05,
+      "loss": 3.9562,
+      "step": 43420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.818943267683882e-05,
+      "loss": 4.0078,
+      "step": 43425
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.815797394870961e-05,
+      "loss": 4.0172,
+      "step": 43430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.81265193206117e-05,
+      "loss": 3.9906,
+      "step": 43435
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.809506879434035e-05,
+      "loss": 4.0016,
+      "step": 43440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.806362237169035e-05,
+      "loss": 3.8813,
+      "step": 43445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.803218005445634e-05,
+      "loss": 3.9766,
+      "step": 43450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.800074184443275e-05,
+      "loss": 3.9344,
+      "step": 43455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.796930774341381e-05,
+      "loss": 3.9797,
+      "step": 43460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.793787775319343e-05,
+      "loss": 3.8297,
+      "step": 43465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.790645187556531e-05,
+      "loss": 3.8984,
+      "step": 43470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.787503011232291e-05,
+      "loss": 3.9,
+      "step": 43475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.784361246525955e-05,
+      "loss": 3.8937,
+      "step": 43480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.781219893616819e-05,
+      "loss": 3.9547,
+      "step": 43485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.778078952684155e-05,
+      "loss": 3.8875,
+      "step": 43490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.774938423907227e-05,
+      "loss": 3.9375,
+      "step": 43495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.771798307465262e-05,
+      "loss": 3.925,
+      "step": 43500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.768658603537463e-05,
+      "loss": 3.9062,
+      "step": 43505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.765519312303013e-05,
+      "loss": 3.9641,
+      "step": 43510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.762380433941079e-05,
+      "loss": 3.9781,
+      "step": 43515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.759241968630791e-05,
+      "loss": 3.9734,
+      "step": 43520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.756103916551259e-05,
+      "loss": 3.9484,
+      "step": 43525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.75296627788158e-05,
+      "loss": 3.9641,
+      "step": 43530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.749829052800815e-05,
+      "loss": 3.9484,
+      "step": 43535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.746692241488005e-05,
+      "loss": 3.85,
+      "step": 43540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.743555844122166e-05,
+      "loss": 4.0531,
+      "step": 43545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.740419860882297e-05,
+      "loss": 3.9969,
+      "step": 43550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.737284291947371e-05,
+      "loss": 3.8859,
+      "step": 43555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.73414913749633e-05,
+      "loss": 3.9656,
+      "step": 43560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.731014397708097e-05,
+      "loss": 3.9156,
+      "step": 43565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.727880072761574e-05,
+      "loss": 3.9375,
+      "step": 43570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.72474616283564e-05,
+      "loss": 4.0109,
+      "step": 43575
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.721612668109141e-05,
+      "loss": 3.9875,
+      "step": 43580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.718479588760911e-05,
+      "loss": 4.0375,
+      "step": 43585
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.715346924969757e-05,
+      "loss": 3.8984,
+      "step": 43590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.712214676914451e-05,
+      "loss": 4.0516,
+      "step": 43595
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.709082844773761e-05,
+      "loss": 3.9953,
+      "step": 43600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.705951428726418e-05,
+      "loss": 3.9562,
+      "step": 43605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.702820428951126e-05,
+      "loss": 3.9172,
+      "step": 43610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.69968984562658e-05,
+      "loss": 3.9094,
+      "step": 43615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.696559678931441e-05,
+      "loss": 3.9266,
+      "step": 43620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.69342992904434e-05,
+      "loss": 4.0172,
+      "step": 43625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.690300596143904e-05,
+      "loss": 3.9859,
+      "step": 43630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.68717168040872e-05,
+      "loss": 3.8984,
+      "step": 43635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.684043182017348e-05,
+      "loss": 3.9125,
+      "step": 43640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.68091510114834e-05,
+      "loss": 3.8766,
+      "step": 43645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.677787437980218e-05,
+      "loss": 4.0125,
+      "step": 43650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.674660192691466e-05,
+      "loss": 3.9813,
+      "step": 43655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.671533365460572e-05,
+      "loss": 3.9453,
+      "step": 43660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.668406956465974e-05,
+      "loss": 4.0047,
+      "step": 43665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.665280965886091e-05,
+      "loss": 3.9406,
+      "step": 43670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.66215539389934e-05,
+      "loss": 3.9406,
+      "step": 43675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.659030240684088e-05,
+      "loss": 3.875,
+      "step": 43680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.655905506418689e-05,
+      "loss": 3.9813,
+      "step": 43685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.652781191281467e-05,
+      "loss": 4.0125,
+      "step": 43690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.649657295450737e-05,
+      "loss": 4.0266,
+      "step": 43695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.646533819104774e-05,
+      "loss": 3.8875,
+      "step": 43700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.64341076242183e-05,
+      "loss": 3.9469,
+      "step": 43705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.64028812558015e-05,
+      "loss": 3.9562,
+      "step": 43710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.637165908757936e-05,
+      "loss": 4.0094,
+      "step": 43715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.634044112133377e-05,
+      "loss": 3.9438,
+      "step": 43720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.630922735884623e-05,
+      "loss": 3.9766,
+      "step": 43725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.627801780189827e-05,
+      "loss": 3.9281,
+      "step": 43730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.624681245227096e-05,
+      "loss": 3.8734,
+      "step": 43735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.621561131174515e-05,
+      "loss": 3.9781,
+      "step": 43740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.618441438210148e-05,
+      "loss": 3.9094,
+      "step": 43745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.615322166512049e-05,
+      "loss": 4.0125,
+      "step": 43750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.612203316258226e-05,
+      "loss": 3.9188,
+      "step": 43755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.609084887626665e-05,
+      "loss": 3.9297,
+      "step": 43760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.60596688079535e-05,
+      "loss": 3.8922,
+      "step": 43765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.602849295942218e-05,
+      "loss": 3.9109,
+      "step": 43770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.599732133245192e-05,
+      "loss": 3.9641,
+      "step": 43775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.596615392882162e-05,
+      "loss": 3.9547,
+      "step": 43780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.593499075031012e-05,
+      "loss": 3.9156,
+      "step": 43785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.590383179869583e-05,
+      "loss": 3.8906,
+      "step": 43790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.587267707575697e-05,
+      "loss": 3.9297,
+      "step": 43795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.584152658327162e-05,
+      "loss": 4.0125,
+      "step": 43800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.581038032301754e-05,
+      "loss": 3.8438,
+      "step": 43805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.577923829677219e-05,
+      "loss": 3.9297,
+      "step": 43810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.574810050631283e-05,
+      "loss": 3.9562,
+      "step": 43815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.571696695341659e-05,
+      "loss": 3.9641,
+      "step": 43820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.568583763986023e-05,
+      "loss": 3.9422,
+      "step": 43825
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.565471256742029e-05,
+      "loss": 4.0578,
+      "step": 43830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.562359173787301e-05,
+      "loss": 3.9906,
+      "step": 43835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.559247515299459e-05,
+      "loss": 4.0062,
+      "step": 43840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.55613628145608e-05,
+      "loss": 3.8859,
+      "step": 43845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.553025472434716e-05,
+      "loss": 3.9297,
+      "step": 43850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.549915088412913e-05,
+      "loss": 3.9688,
+      "step": 43855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.546805129568175e-05,
+      "loss": 4.0078,
+      "step": 43860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.543695596077987e-05,
+      "loss": 3.9688,
+      "step": 43865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.540586488119807e-05,
+      "loss": 3.9656,
+      "step": 43870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.53747780587108e-05,
+      "loss": 3.9297,
+      "step": 43875
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.534369549509215e-05,
+      "loss": 4.0047,
+      "step": 43880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.531261719211596e-05,
+      "loss": 3.9781,
+      "step": 43885
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.528154315155599e-05,
+      "loss": 4.0391,
+      "step": 43890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.525047337518555e-05,
+      "loss": 3.8875,
+      "step": 43895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.52194078647778e-05,
+      "loss": 3.95,
+      "step": 43900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.518834662210561e-05,
+      "loss": 3.9047,
+      "step": 43905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.515728964894175e-05,
+      "loss": 3.9641,
+      "step": 43910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.512623694705861e-05,
+      "loss": 3.9078,
+      "step": 43915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.509518851822834e-05,
+      "loss": 3.9594,
+      "step": 43920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.506414436422286e-05,
+      "loss": 4.0016,
+      "step": 43925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.50331044868139e-05,
+      "loss": 3.9844,
+      "step": 43930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.500206888777297e-05,
+      "loss": 3.9875,
+      "step": 43935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.497103756887112e-05,
+      "loss": 4.0125,
+      "step": 43940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.494001053187946e-05,
+      "loss": 3.9109,
+      "step": 43945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.490898777856862e-05,
+      "loss": 3.9359,
+      "step": 43950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.487796931070911e-05,
+      "loss": 3.9328,
+      "step": 43955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.48469551300711e-05,
+      "loss": 3.9312,
+      "step": 43960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.481594523842466e-05,
+      "loss": 3.9359,
+      "step": 43965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.478493963753948e-05,
+      "loss": 4.0,
+      "step": 43970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.4753938329185e-05,
+      "loss": 3.9453,
+      "step": 43975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.47229413151306e-05,
+      "loss": 4.0047,
+      "step": 43980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.46919485971452e-05,
+      "loss": 3.9453,
+      "step": 43985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.466096017699754e-05,
+      "loss": 3.9219,
+      "step": 43990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.462997605645612e-05,
+      "loss": 4.0016,
+      "step": 43995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.459899623728928e-05,
+      "loss": 4.0016,
+      "step": 44000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.456802072126503e-05,
+      "loss": 4.0016,
+      "step": 44005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.45370495101511e-05,
+      "loss": 3.9594,
+      "step": 44010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.450608260571499e-05,
+      "loss": 3.9484,
+      "step": 44015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.447512000972409e-05,
+      "loss": 3.9828,
+      "step": 44020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.444416172394537e-05,
+      "loss": 3.9266,
+      "step": 44025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.441320775014559e-05,
+      "loss": 3.8875,
+      "step": 44030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.438225809009139e-05,
+      "loss": 3.8797,
+      "step": 44035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.435131274554903e-05,
+      "loss": 3.875,
+      "step": 44040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.432037171828454e-05,
+      "loss": 3.9547,
+      "step": 44045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.428943501006369e-05,
+      "loss": 3.875,
+      "step": 44050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.425850262265215e-05,
+      "loss": 3.9969,
+      "step": 44055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.422757455781519e-05,
+      "loss": 3.9375,
+      "step": 44060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.419665081731778e-05,
+      "loss": 3.9125,
+      "step": 44065
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.41657314029249e-05,
+      "loss": 3.9797,
+      "step": 44070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.413481631640104e-05,
+      "loss": 3.9906,
+      "step": 44075
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.41039055595105e-05,
+      "loss": 3.9422,
+      "step": 44080
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.407299913401743e-05,
+      "loss": 4.0062,
+      "step": 44085
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.404209704168563e-05,
+      "loss": 4.0563,
+      "step": 44090
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.40111992842787e-05,
+      "loss": 4.0203,
+      "step": 44095
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.39803058635599e-05,
+      "loss": 3.9484,
+      "step": 44100
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.394941678129243e-05,
+      "loss": 4.0516,
+      "step": 44105
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.39185320392391e-05,
+      "loss": 3.9109,
+      "step": 44110
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.388765163916242e-05,
+      "loss": 3.9391,
+      "step": 44115
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.385677558282487e-05,
+      "loss": 3.9406,
+      "step": 44120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.382590387198849e-05,
+      "loss": 3.9578,
+      "step": 44125
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.379503650841507e-05,
+      "loss": 3.975,
+      "step": 44130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.376417349386633e-05,
+      "loss": 3.9953,
+      "step": 44135
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.373331483010358e-05,
+      "loss": 3.9094,
+      "step": 44140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.370246051888784e-05,
+      "loss": 3.9625,
+      "step": 44145
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.36716105619801e-05,
+      "loss": 3.8922,
+      "step": 44150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.364076496114092e-05,
+      "loss": 3.9125,
+      "step": 44155
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.36099237181306e-05,
+      "loss": 3.9172,
+      "step": 44160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.357908683470937e-05,
+      "loss": 3.8906,
+      "step": 44165
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.354825431263702e-05,
+      "loss": 3.9672,
+      "step": 44170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.351742615367312e-05,
+      "loss": 3.9875,
+      "step": 44175
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.348660235957716e-05,
+      "loss": 3.9438,
+      "step": 44180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.345578293210818e-05,
+      "loss": 3.8969,
+      "step": 44185
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.342496787302506e-05,
+      "loss": 3.8813,
+      "step": 44190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.339415718408634e-05,
+      "loss": 3.9375,
+      "step": 44195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.336335086705054e-05,
+      "loss": 3.95,
+      "step": 44200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.33325489236757e-05,
+      "loss": 3.8891,
+      "step": 44205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.330175135571965e-05,
+      "loss": 3.9922,
+      "step": 44210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.327095816494009e-05,
+      "loss": 3.9453,
+      "step": 44215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.324016935309437e-05,
+      "loss": 3.9516,
+      "step": 44220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.320938492193959e-05,
+      "loss": 3.9547,
+      "step": 44225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.317860487323257e-05,
+      "loss": 3.8922,
+      "step": 44230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.314782920873004e-05,
+      "loss": 3.9562,
+      "step": 44235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.311705793018831e-05,
+      "loss": 3.9625,
+      "step": 44240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.308629103936345e-05,
+      "loss": 3.9344,
+      "step": 44245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.305552853801145e-05,
+      "loss": 3.9188,
+      "step": 44250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.302477042788785e-05,
+      "loss": 3.8813,
+      "step": 44255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.299401671074806e-05,
+      "loss": 3.9922,
+      "step": 44260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.296326738834709e-05,
+      "loss": 4.0125,
+      "step": 44265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.293252246243994e-05,
+      "loss": 3.9391,
+      "step": 44270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.290178193478119e-05,
+      "loss": 3.9234,
+      "step": 44275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.287104580712518e-05,
+      "loss": 3.9359,
+      "step": 44280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.284031408122598e-05,
+      "loss": 3.9531,
+      "step": 44285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.280958675883755e-05,
+      "loss": 3.8359,
+      "step": 44290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.277886384171345e-05,
+      "loss": 3.9609,
+      "step": 44295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.2748145331607e-05,
+      "loss": 4.0125,
+      "step": 44300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.27174312302714e-05,
+      "loss": 3.8937,
+      "step": 44305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.268672153945948e-05,
+      "loss": 3.8953,
+      "step": 44310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.26560162609238e-05,
+      "loss": 3.9953,
+      "step": 44315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.262531539641668e-05,
+      "loss": 3.9531,
+      "step": 44320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.259461894769036e-05,
+      "loss": 3.9719,
+      "step": 44325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.256392691649659e-05,
+      "loss": 3.8797,
+      "step": 44330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.253323930458695e-05,
+      "loss": 3.9953,
+      "step": 44335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.250255611371286e-05,
+      "loss": 3.9734,
+      "step": 44340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.247187734562538e-05,
+      "loss": 3.9234,
+      "step": 44345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.244120300207533e-05,
+      "loss": 3.8813,
+      "step": 44350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.241053308481328e-05,
+      "loss": 3.8687,
+      "step": 44355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.237986759558963e-05,
+      "loss": 3.9859,
+      "step": 44360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.234920653615444e-05,
+      "loss": 3.8844,
+      "step": 44365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.231854990825753e-05,
+      "loss": 4.0344,
+      "step": 44370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.228789771364841e-05,
+      "loss": 3.9578,
+      "step": 44375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.225724995407653e-05,
+      "loss": 3.9859,
+      "step": 44380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.22266066312909e-05,
+      "loss": 3.9859,
+      "step": 44385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.21959677470403e-05,
+      "loss": 3.9969,
+      "step": 44390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.216533330307336e-05,
+      "loss": 3.9469,
+      "step": 44395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.213470330113838e-05,
+      "loss": 3.9031,
+      "step": 44400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.21040777429834e-05,
+      "loss": 3.9391,
+      "step": 44405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.207345663035615e-05,
+      "loss": 3.9797,
+      "step": 44410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.20428399650043e-05,
+      "loss": 3.9438,
+      "step": 44415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.201222774867512e-05,
+      "loss": 3.925,
+      "step": 44420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.198161998311557e-05,
+      "loss": 3.9813,
+      "step": 44425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.195101667007257e-05,
+      "loss": 3.9188,
+      "step": 44430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.192041781129255e-05,
+      "loss": 3.9937,
+      "step": 44435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.188982340852184e-05,
+      "loss": 3.8547,
+      "step": 44440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.18592334635064e-05,
+      "loss": 4.0,
+      "step": 44445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.18286479779921e-05,
+      "loss": 3.9234,
+      "step": 44450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.17980669537244e-05,
+      "loss": 3.9469,
+      "step": 44455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.176749039244857e-05,
+      "loss": 3.8813,
+      "step": 44460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.173691829590956e-05,
+      "loss": 3.9484,
+      "step": 44465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.170635066585224e-05,
+      "loss": 3.9375,
+      "step": 44470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.167578750402103e-05,
+      "loss": 4.0078,
+      "step": 44475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.164522881216013e-05,
+      "loss": 3.9047,
+      "step": 44480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.161467459201366e-05,
+      "loss": 4.0,
+      "step": 44485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.158412484532528e-05,
+      "loss": 3.9656,
+      "step": 44490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.155357957383845e-05,
+      "loss": 3.9562,
+      "step": 44495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.152303877929636e-05,
+      "loss": 3.9953,
+      "step": 44500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.149250246344207e-05,
+      "loss": 3.975,
+      "step": 44505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.146197062801824e-05,
+      "loss": 3.9766,
+      "step": 44510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.143144327476729e-05,
+      "loss": 3.9547,
+      "step": 44515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.140092040543149e-05,
+      "loss": 4.0047,
+      "step": 44520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.137040202175276e-05,
+      "loss": 4.0078,
+      "step": 44525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.133988812547277e-05,
+      "loss": 3.9875,
+      "step": 44530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.130937871833292e-05,
+      "loss": 3.9531,
+      "step": 44535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.127887380207447e-05,
+      "loss": 3.9719,
+      "step": 44540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.124837337843828e-05,
+      "loss": 3.8984,
+      "step": 44545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.121787744916504e-05,
+      "loss": 3.9422,
+      "step": 44550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.118738601599507e-05,
+      "loss": 4.075,
+      "step": 44555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.115689908066866e-05,
+      "loss": 3.8828,
+      "step": 44560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.11264166449256e-05,
+      "loss": 3.9797,
+      "step": 44565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.109593871050554e-05,
+      "loss": 3.9484,
+      "step": 44570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.10654652791479e-05,
+      "loss": 4.0109,
+      "step": 44575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.103499635259179e-05,
+      "loss": 3.9625,
+      "step": 44580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.1004531932576e-05,
+      "loss": 3.8109,
+      "step": 44585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.097407202083926e-05,
+      "loss": 3.9094,
+      "step": 44590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.094361661911985e-05,
+      "loss": 3.9344,
+      "step": 44595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.091316572915581e-05,
+      "loss": 3.9266,
+      "step": 44600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.08827193526851e-05,
+      "loss": 3.95,
+      "step": 44605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.085227749144523e-05,
+      "loss": 3.8969,
+      "step": 44610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.082184014717348e-05,
+      "loss": 4.0,
+      "step": 44615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.0791407321607e-05,
+      "loss": 3.9531,
+      "step": 44620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.076097901648255e-05,
+      "loss": 4.0547,
+      "step": 44625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.073055523353663e-05,
+      "loss": 3.8547,
+      "step": 44630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.070013597450562e-05,
+      "loss": 3.9703,
+      "step": 44635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.066972124112552e-05,
+      "loss": 3.9547,
+      "step": 44640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.063931103513202e-05,
+      "loss": 3.9984,
+      "step": 44645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.060890535826076e-05,
+      "loss": 3.9516,
+      "step": 44650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.057850421224692e-05,
+      "loss": 3.9484,
+      "step": 44655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.054810759882547e-05,
+      "loss": 3.9328,
+      "step": 44660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.051771551973125e-05,
+      "loss": 3.925,
+      "step": 44665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.048732797669867e-05,
+      "loss": 3.9328,
+      "step": 44670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.045694497146197e-05,
+      "loss": 3.9547,
+      "step": 44675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.0426566505755e-05,
+      "loss": 4.0078,
+      "step": 44680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.039619258131168e-05,
+      "loss": 3.9594,
+      "step": 44685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.03658231998653e-05,
+      "loss": 3.9797,
+      "step": 44690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.033545836314901e-05,
+      "loss": 3.9672,
+      "step": 44695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.030509807289587e-05,
+      "loss": 4.0141,
+      "step": 44700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.027474233083848e-05,
+      "loss": 3.9547,
+      "step": 44705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.024439113870923e-05,
+      "loss": 4.0,
+      "step": 44710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.021404449824025e-05,
+      "loss": 3.9422,
+      "step": 44715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.018370241116346e-05,
+      "loss": 3.9406,
+      "step": 44720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.015336487921051e-05,
+      "loss": 3.9203,
+      "step": 44725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.012303190411274e-05,
+      "loss": 4.0281,
+      "step": 44730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.009270348760117e-05,
+      "loss": 3.9266,
+      "step": 44735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.006237963140679e-05,
+      "loss": 3.8875,
+      "step": 44740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.003206033726013e-05,
+      "loss": 3.9656,
+      "step": 44745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.000174560689145e-05,
+      "loss": 3.9094,
+      "step": 44750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.997143544203091e-05,
+      "loss": 3.9031,
+      "step": 44755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.99411298444083e-05,
+      "loss": 3.9813,
+      "step": 44760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.991082881575312e-05,
+      "loss": 3.95,
+      "step": 44765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.988053235779462e-05,
+      "loss": 3.8766,
+      "step": 44770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.985024047226193e-05,
+      "loss": 3.9953,
+      "step": 44775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.981995316088375e-05,
+      "loss": 4.0469,
+      "step": 44780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.978967042538854e-05,
+      "loss": 3.9328,
+      "step": 44785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.97593922675046e-05,
+      "loss": 3.9656,
+      "step": 44790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.972911868895993e-05,
+      "loss": 3.8375,
+      "step": 44795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.96988496914822e-05,
+      "loss": 3.9203,
+      "step": 44800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.96685852767988e-05,
+      "loss": 3.9656,
+      "step": 44805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.963832544663705e-05,
+      "loss": 3.9469,
+      "step": 44810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.96080702027238e-05,
+      "loss": 3.9078,
+      "step": 44815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.95778195467858e-05,
+      "loss": 3.8984,
+      "step": 44820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.954757348054931e-05,
+      "loss": 4.0141,
+      "step": 44825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.951733200574063e-05,
+      "loss": 4.0141,
+      "step": 44830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.948709512408558e-05,
+      "loss": 3.9344,
+      "step": 44835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.945686283730972e-05,
+      "loss": 3.9516,
+      "step": 44840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.942663514713853e-05,
+      "loss": 3.8969,
+      "step": 44845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.939641205529707e-05,
+      "loss": 3.9453,
+      "step": 44850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.936619356351014e-05,
+      "loss": 3.8797,
+      "step": 44855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.933597967350227e-05,
+      "loss": 4.0047,
+      "step": 44860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.930577038699789e-05,
+      "loss": 4.0266,
+      "step": 44865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.927556570572097e-05,
+      "loss": 4.0531,
+      "step": 44870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.924536563139526e-05,
+      "loss": 3.9188,
+      "step": 44875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.921517016574439e-05,
+      "loss": 3.9594,
+      "step": 44880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.918497931049154e-05,
+      "loss": 4.0375,
+      "step": 44885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.915479306735974e-05,
+      "loss": 3.9328,
+      "step": 44890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.912461143807164e-05,
+      "loss": 3.9516,
+      "step": 44895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.909443442434983e-05,
+      "loss": 3.9328,
+      "step": 44900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.906426202791647e-05,
+      "loss": 3.9266,
+      "step": 44905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.903409425049347e-05,
+      "loss": 3.9703,
+      "step": 44910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.900393109380249e-05,
+      "loss": 4.0187,
+      "step": 44915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.897377255956502e-05,
+      "loss": 3.9375,
+      "step": 44920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.894361864950217e-05,
+      "loss": 3.9234,
+      "step": 44925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.891346936533479e-05,
+      "loss": 3.9016,
+      "step": 44930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.888332470878358e-05,
+      "loss": 3.9344,
+      "step": 44935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.885318468156887e-05,
+      "loss": 3.9438,
+      "step": 44940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.882304928541073e-05,
+      "loss": 3.9766,
+      "step": 44945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.879291852202895e-05,
+      "loss": 3.9813,
+      "step": 44950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.876279239314321e-05,
+      "loss": 3.9,
+      "step": 44955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.873267090047276e-05,
+      "loss": 3.9375,
+      "step": 44960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.870255404573656e-05,
+      "loss": 3.9797,
+      "step": 44965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.867244183065351e-05,
+      "loss": 4.0219,
+      "step": 44970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.864233425694203e-05,
+      "loss": 3.9875,
+      "step": 44975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.861223132632042e-05,
+      "loss": 3.9891,
+      "step": 44980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.858213304050656e-05,
+      "loss": 3.9406,
+      "step": 44985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.855203940121827e-05,
+      "loss": 3.9016,
+      "step": 44990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.852195041017296e-05,
+      "loss": 3.9062,
+      "step": 44995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.849186606908782e-05,
+      "loss": 4.0563,
+      "step": 45000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.846178637967968e-05,
+      "loss": 3.9797,
+      "step": 45005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.843171134366532e-05,
+      "loss": 3.9062,
+      "step": 45010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.840164096276108e-05,
+      "loss": 3.8469,
+      "step": 45015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.837157523868301e-05,
+      "loss": 3.9125,
+      "step": 45020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.834151417314705e-05,
+      "loss": 3.9453,
+      "step": 45025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.831145776786878e-05,
+      "loss": 3.9844,
+      "step": 45030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.828140602456349e-05,
+      "loss": 3.9531,
+      "step": 45035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.825135894494622e-05,
+      "loss": 3.9719,
+      "step": 45040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.822131653073181e-05,
+      "loss": 3.9828,
+      "step": 45045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.819127878363477e-05,
+      "loss": 3.9031,
+      "step": 45050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.816124570536929e-05,
+      "loss": 3.9281,
+      "step": 45055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.813121729764947e-05,
+      "loss": 3.8828,
+      "step": 45060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.810119356218899e-05,
+      "loss": 3.8922,
+      "step": 45065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.807117450070129e-05,
+      "loss": 3.9766,
+      "step": 45070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.804116011489952e-05,
+      "loss": 3.8875,
+      "step": 45075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.801115040649672e-05,
+      "loss": 3.9594,
+      "step": 45080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.798114537720547e-05,
+      "loss": 3.975,
+      "step": 45085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.795114502873811e-05,
+      "loss": 3.8266,
+      "step": 45090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.792114936280689e-05,
+      "loss": 3.975,
+      "step": 45095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.789115838112358e-05,
+      "loss": 3.8734,
+      "step": 45100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.786117208539976e-05,
+      "loss": 3.8781,
+      "step": 45105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.783119047734682e-05,
+      "loss": 3.9438,
+      "step": 45110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.780121355867576e-05,
+      "loss": 3.9125,
+      "step": 45115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.777124133109733e-05,
+      "loss": 3.9141,
+      "step": 45120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.774127379632215e-05,
+      "loss": 3.9297,
+      "step": 45125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.771131095606042e-05,
+      "loss": 3.9406,
+      "step": 45130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.768135281202206e-05,
+      "loss": 3.9391,
+      "step": 45135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.76513993659169e-05,
+      "loss": 3.9922,
+      "step": 45140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.762145061945432e-05,
+      "loss": 4.0031,
+      "step": 45145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.759150657434345e-05,
+      "loss": 3.9484,
+      "step": 45150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.756156723229332e-05,
+      "loss": 3.9984,
+      "step": 45155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75316325950125e-05,
+      "loss": 3.9578,
+      "step": 45160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75017026642093e-05,
+      "loss": 3.9031,
+      "step": 45165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.747177744159197e-05,
+      "loss": 3.9156,
+      "step": 45170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.744185692886827e-05,
+      "loss": 3.9672,
+      "step": 45175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.741194112774574e-05,
+      "loss": 3.9797,
+      "step": 45180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.738203003993166e-05,
+      "loss": 3.8734,
+      "step": 45185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.735212366713315e-05,
+      "loss": 3.9359,
+      "step": 45190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.732222201105692e-05,
+      "loss": 4.0125,
+      "step": 45195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.729232507340943e-05,
+      "loss": 4.0141,
+      "step": 45200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.726243285589697e-05,
+      "loss": 4.0359,
+      "step": 45205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.723254536022544e-05,
+      "loss": 3.9172,
+      "step": 45210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.720266258810055e-05,
+      "loss": 4.0469,
+      "step": 45215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.717278454122765e-05,
+      "loss": 3.9734,
+      "step": 45220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.714291122131198e-05,
+      "loss": 3.9547,
+      "step": 45225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.711304263005837e-05,
+      "loss": 3.8844,
+      "step": 45230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.708317876917135e-05,
+      "loss": 3.9172,
+      "step": 45235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.705331964035539e-05,
+      "loss": 4.0672,
+      "step": 45240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.702346524531447e-05,
+      "loss": 3.9141,
+      "step": 45245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.699361558575242e-05,
+      "loss": 3.9641,
+      "step": 45250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.696377066337267e-05,
+      "loss": 3.9406,
+      "step": 45255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.69339304798786e-05,
+      "loss": 3.9922,
+      "step": 45260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.690409503697313e-05,
+      "loss": 3.8937,
+      "step": 45265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.6874264336359e-05,
+      "loss": 3.9859,
+      "step": 45270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.684443837973854e-05,
+      "loss": 3.9969,
+      "step": 45275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.681461716881409e-05,
+      "loss": 3.9734,
+      "step": 45280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.678480070528746e-05,
+      "loss": 3.9656,
+      "step": 45285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.675498899086024e-05,
+      "loss": 3.9016,
+      "step": 45290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.67251820272339e-05,
+      "loss": 4.0438,
+      "step": 45295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.669537981610943e-05,
+      "loss": 4.0016,
+      "step": 45300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.666558235918769e-05,
+      "loss": 3.9484,
+      "step": 45305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.663578965816915e-05,
+      "loss": 3.95,
+      "step": 45310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.66060017147542e-05,
+      "loss": 4.0641,
+      "step": 45315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.657621853064278e-05,
+      "loss": 3.8594,
+      "step": 45320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.654644010753456e-05,
+      "loss": 4.0516,
+      "step": 45325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.651666644712913e-05,
+      "loss": 3.9094,
+      "step": 45330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.64868975511256e-05,
+      "loss": 3.8422,
+      "step": 45335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.645713342122289e-05,
+      "loss": 3.9859,
+      "step": 45340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.642737405911957e-05,
+      "loss": 3.8453,
+      "step": 45345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.639761946651414e-05,
+      "loss": 3.95,
+      "step": 45350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.636786964510465e-05,
+      "loss": 3.9469,
+      "step": 45355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.633812459658889e-05,
+      "loss": 4.0297,
+      "step": 45360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.63083843226644e-05,
+      "loss": 3.9422,
+      "step": 45365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.627864882502854e-05,
+      "loss": 3.9859,
+      "step": 45370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.624891810537828e-05,
+      "loss": 3.9203,
+      "step": 45375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.62191921654103e-05,
+      "loss": 3.9688,
+      "step": 45380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.618947100682115e-05,
+      "loss": 3.9562,
+      "step": 45385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.6159754631307e-05,
+      "loss": 3.9531,
+      "step": 45390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.613004304056374e-05,
+      "loss": 3.9656,
+      "step": 45395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.610033623628695e-05,
+      "loss": 3.9953,
+      "step": 45400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.607063422017214e-05,
+      "loss": 3.925,
+      "step": 45405
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.604093699391433e-05,
+      "loss": 3.7891,
+      "step": 45410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.60112445592083e-05,
+      "loss": 3.8563,
+      "step": 45415
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.59815569177487e-05,
+      "loss": 3.8734,
+      "step": 45420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.595187407122976e-05,
+      "loss": 3.9594,
+      "step": 45425
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.592219602134547e-05,
+      "loss": 3.9516,
+      "step": 45430
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.589252276978953e-05,
+      "loss": 3.8797,
+      "step": 45435
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.586285431825549e-05,
+      "loss": 4.0266,
+      "step": 45440
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.583319066843646e-05,
+      "loss": 4.0141,
+      "step": 45445
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.580353182202538e-05,
+      "loss": 3.9062,
+      "step": 45450
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.57738777807148e-05,
+      "loss": 3.9797,
+      "step": 45455
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.57442285461972e-05,
+      "loss": 4.0078,
+      "step": 45460
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.571458412016462e-05,
+      "loss": 3.9813,
+      "step": 45465
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.568494450430879e-05,
+      "loss": 4.0031,
+      "step": 45470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.565530970032139e-05,
+      "loss": 3.9578,
+      "step": 45475
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.56256797098936e-05,
+      "loss": 3.9062,
+      "step": 45480
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.55960545347164e-05,
+      "loss": 3.9656,
+      "step": 45485
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.556643417648047e-05,
+      "loss": 4.0297,
+      "step": 45490
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.553681863687636e-05,
+      "loss": 3.9656,
+      "step": 45495
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.550720791759415e-05,
+      "loss": 3.9422,
+      "step": 45500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.547760202032368e-05,
+      "loss": 4.025,
+      "step": 45505
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.544800094675469e-05,
+      "loss": 4.0266,
+      "step": 45510
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.541840469857645e-05,
+      "loss": 4.0203,
+      "step": 45515
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.5388813277478e-05,
+      "loss": 3.9766,
+      "step": 45520
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.53592266851481e-05,
+      "loss": 3.9922,
+      "step": 45525
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.532964492327537e-05,
+      "loss": 3.9969,
+      "step": 45530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.530006799354797e-05,
+      "loss": 3.8391,
+      "step": 45535
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.527049589765386e-05,
+      "loss": 4.0453,
+      "step": 45540
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.524092863728067e-05,
+      "loss": 3.9578,
+      "step": 45545
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.521136621411593e-05,
+      "loss": 3.9516,
+      "step": 45550
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.51818086298467e-05,
+      "loss": 4.0406,
+      "step": 45555
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.515225588615978e-05,
+      "loss": 3.9406,
+      "step": 45560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.512270798474188e-05,
+      "loss": 3.9984,
+      "step": 45565
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.509316492727921e-05,
+      "loss": 3.8969,
+      "step": 45570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.506362671545778e-05,
+      "loss": 4.025,
+      "step": 45575
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.50340933509634e-05,
+      "loss": 3.9516,
+      "step": 45580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.500456483548153e-05,
+      "loss": 3.8625,
+      "step": 45585
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.497504117069728e-05,
+      "loss": 3.9281,
+      "step": 45590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.49455223582957e-05,
+      "loss": 3.9156,
+      "step": 45595
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.491600839996138e-05,
+      "loss": 3.9703,
+      "step": 45600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.48864992973786e-05,
+      "loss": 3.9609,
+      "step": 45605
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.48569950522316e-05,
+      "loss": 4.0187,
+      "step": 45610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.48274956662041e-05,
+      "loss": 3.9484,
+      "step": 45615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.479800114097959e-05,
+      "loss": 3.8469,
+      "step": 45620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.476851147824143e-05,
+      "loss": 3.9484,
+      "step": 45625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.473902667967255e-05,
+      "loss": 3.9828,
+      "step": 45630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.470954674695565e-05,
+      "loss": 3.9406,
+      "step": 45635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.46800716817731e-05,
+      "loss": 3.9578,
+      "step": 45640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.465060148580715e-05,
+      "loss": 3.9609,
+      "step": 45645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.46211361607396e-05,
+      "loss": 3.9109,
+      "step": 45650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.4591675708252e-05,
+      "loss": 3.9922,
+      "step": 45655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.456222013002576e-05,
+      "loss": 4.0062,
+      "step": 45660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.453276942774188e-05,
+      "loss": 3.95,
+      "step": 45665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.450332360308103e-05,
+      "loss": 4.0422,
+      "step": 45670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.447388265772382e-05,
+      "loss": 3.8078,
+      "step": 45675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.444444659335039e-05,
+      "loss": 3.9562,
+      "step": 45680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.441501541164059e-05,
+      "loss": 3.9906,
+      "step": 45685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.438558911427418e-05,
+      "loss": 3.8969,
+      "step": 45690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.435616770293045e-05,
+      "loss": 3.9469,
+      "step": 45695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.43267511792885e-05,
+      "loss": 4.0141,
+      "step": 45700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.429733954502709e-05,
+      "loss": 4.0406,
+      "step": 45705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.426793280182482e-05,
+      "loss": 4.0234,
+      "step": 45710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.423853095135991e-05,
+      "loss": 3.9016,
+      "step": 45715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.420913399531033e-05,
+      "loss": 3.9203,
+      "step": 45720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.417974193535369e-05,
+      "loss": 4.0219,
+      "step": 45725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.41503547731675e-05,
+      "loss": 4.0438,
+      "step": 45730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.412097251042887e-05,
+      "loss": 3.9438,
+      "step": 45735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.409159514881457e-05,
+      "loss": 3.9109,
+      "step": 45740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.406222269000128e-05,
+      "loss": 3.8656,
+      "step": 45745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.403285513566524e-05,
+      "loss": 3.9375,
+      "step": 45750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.400349248748245e-05,
+      "loss": 3.9141,
+      "step": 45755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.397413474712861e-05,
+      "loss": 3.9922,
+      "step": 45760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.394478191627923e-05,
+      "loss": 4.0281,
+      "step": 45765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.391543399660947e-05,
+      "loss": 3.9609,
+      "step": 45770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.388609098979414e-05,
+      "loss": 3.9922,
+      "step": 45775
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.385675289750796e-05,
+      "loss": 3.9547,
+      "step": 45780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.382741972142524e-05,
+      "loss": 3.9766,
+      "step": 45785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.379809146321996e-05,
+      "loss": 4.0156,
+      "step": 45790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.376876812456587e-05,
+      "loss": 4.0031,
+      "step": 45795
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.373944970713658e-05,
+      "loss": 3.9641,
+      "step": 45800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.371013621260521e-05,
+      "loss": 3.9453,
+      "step": 45805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.36808276426447e-05,
+      "loss": 3.8344,
+      "step": 45810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.365152399892764e-05,
+      "loss": 4.0375,
+      "step": 45815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.362222528312649e-05,
+      "loss": 3.9766,
+      "step": 45820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.359293149691326e-05,
+      "loss": 3.9422,
+      "step": 45825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.356364264195974e-05,
+      "loss": 3.8844,
+      "step": 45830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.353435871993752e-05,
+      "loss": 3.9609,
+      "step": 45835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.350507973251781e-05,
+      "loss": 3.9453,
+      "step": 45840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.347580568137155e-05,
+      "loss": 4.0656,
+      "step": 45845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.344653656816933e-05,
+      "loss": 3.9344,
+      "step": 45850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.341727239458171e-05,
+      "loss": 3.9641,
+      "step": 45855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.338801316227869e-05,
+      "loss": 3.9859,
+      "step": 45860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.335875887293007e-05,
+      "loss": 3.9562,
+      "step": 45865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.332950952820552e-05,
+      "loss": 3.9609,
+      "step": 45870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.330026512977421e-05,
+      "loss": 3.95,
+      "step": 45875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.327102567930513e-05,
+      "loss": 3.8891,
+      "step": 45880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.324179117846694e-05,
+      "loss": 4.0203,
+      "step": 45885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.321256162892817e-05,
+      "loss": 3.9219,
+      "step": 45890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.318333703235688e-05,
+      "loss": 4.0078,
+      "step": 45895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.315411739042091e-05,
+      "loss": 3.9641,
+      "step": 45900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.31249027047878e-05,
+      "loss": 3.8687,
+      "step": 45905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.309569297712492e-05,
+      "loss": 4.0297,
+      "step": 45910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.306648820909924e-05,
+      "loss": 3.925,
+      "step": 45915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.303728840237742e-05,
+      "loss": 3.9125,
+      "step": 45920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.300809355862602e-05,
+      "loss": 3.9937,
+      "step": 45925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.297890367951108e-05,
+      "loss": 3.9328,
+      "step": 45930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.294971876669853e-05,
+      "loss": 3.9844,
+      "step": 45935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.292053882185388e-05,
+      "loss": 4.0,
+      "step": 45940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.289136384664252e-05,
+      "loss": 3.9625,
+      "step": 45945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.286219384272945e-05,
+      "loss": 3.9328,
+      "step": 45950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.283302881177934e-05,
+      "loss": 3.8922,
+      "step": 45955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.280386875545674e-05,
+      "loss": 3.875,
+      "step": 45960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.277471367542576e-05,
+      "loss": 4.0297,
+      "step": 45965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.274556357335028e-05,
+      "loss": 3.9672,
+      "step": 45970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.271641845089388e-05,
+      "loss": 3.8641,
+      "step": 45975
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.268727830971993e-05,
+      "loss": 3.9641,
+      "step": 45980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.265814315149147e-05,
+      "loss": 3.9937,
+      "step": 45985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.262901297787118e-05,
+      "loss": 3.9406,
+      "step": 45990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.259988779052152e-05,
+      "loss": 3.9359,
+      "step": 45995
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.257076759110474e-05,
+      "loss": 4.0328,
+      "step": 46000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.25416523812827e-05,
+      "loss": 3.9875,
+      "step": 46005
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.251254216271696e-05,
+      "loss": 3.9719,
+      "step": 46010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.248343693706891e-05,
+      "loss": 3.9719,
+      "step": 46015
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.24543367059996e-05,
+      "loss": 4.0078,
+      "step": 46020
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.242524147116972e-05,
+      "loss": 4.0187,
+      "step": 46025
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.239615123423971e-05,
+      "loss": 3.9547,
+      "step": 46030
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.236706599686987e-05,
+      "loss": 3.9078,
+      "step": 46035
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.233798576072002e-05,
+      "loss": 3.9422,
+      "step": 46040
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.230891052744973e-05,
+      "loss": 3.9438,
+      "step": 46045
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.227984029871844e-05,
+      "loss": 4.0359,
+      "step": 46050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.225077507618515e-05,
+      "loss": 3.9266,
+      "step": 46055
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.222171486150857e-05,
+      "loss": 3.9297,
+      "step": 46060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.219265965634715e-05,
+      "loss": 3.9656,
+      "step": 46065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.216360946235918e-05,
+      "loss": 3.9203,
+      "step": 46070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.213456428120249e-05,
+      "loss": 4.0438,
+      "step": 46075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.210552411453465e-05,
+      "loss": 3.9656,
+      "step": 46080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.207648896401308e-05,
+      "loss": 3.9297,
+      "step": 46085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.204745883129477e-05,
+      "loss": 4.0,
+      "step": 46090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.201843371803645e-05,
+      "loss": 4.0203,
+      "step": 46095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.198941362589464e-05,
+      "loss": 3.9609,
+      "step": 46100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.196039855652548e-05,
+      "loss": 3.8438,
+      "step": 46105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.193138851158484e-05,
+      "loss": 4.0062,
+      "step": 46110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.190238349272841e-05,
+      "loss": 3.9953,
+      "step": 46115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.187338350161144e-05,
+      "loss": 3.9719,
+      "step": 46120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.184438853988896e-05,
+      "loss": 3.9375,
+      "step": 46125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.181539860921575e-05,
+      "loss": 3.975,
+      "step": 46130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.178641371124629e-05,
+      "loss": 3.9375,
+      "step": 46135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.175743384763464e-05,
+      "loss": 3.9969,
+      "step": 46140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.172845902003481e-05,
+      "loss": 3.9219,
+      "step": 46145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.169948923010035e-05,
+      "loss": 4.0312,
+      "step": 46150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.167052447948451e-05,
+      "loss": 3.9328,
+      "step": 46155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.164156476984042e-05,
+      "loss": 3.9578,
+      "step": 46160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.161261010282077e-05,
+      "loss": 3.975,
+      "step": 46165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.158366048007797e-05,
+      "loss": 3.9391,
+      "step": 46170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.155471590326416e-05,
+      "loss": 3.8547,
+      "step": 46175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.15257763740313e-05,
+      "loss": 4.0078,
+      "step": 46180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.149684189403093e-05,
+      "loss": 3.8531,
+      "step": 46185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.146791246491428e-05,
+      "loss": 4.0484,
+      "step": 46190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.143898808833246e-05,
+      "loss": 3.8656,
+      "step": 46195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.141006876593614e-05,
+      "loss": 4.0344,
+      "step": 46200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.138115449937576e-05,
+      "loss": 4.0391,
+      "step": 46205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.135224529030138e-05,
+      "loss": 4.0547,
+      "step": 46210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.132334114036298e-05,
+      "loss": 3.9937,
+      "step": 46215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.129444205121005e-05,
+      "loss": 3.9219,
+      "step": 46220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.126554802449182e-05,
+      "loss": 4.0062,
+      "step": 46225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.123665906185741e-05,
+      "loss": 3.9172,
+      "step": 46230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.120777516495542e-05,
+      "loss": 3.9844,
+      "step": 46235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.117889633543426e-05,
+      "loss": 3.9344,
+      "step": 46240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.115002257494202e-05,
+      "loss": 3.9703,
+      "step": 46245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.112115388512662e-05,
+      "loss": 3.9609,
+      "step": 46250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.109229026763555e-05,
+      "loss": 3.9016,
+      "step": 46255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.106343172411606e-05,
+      "loss": 3.9703,
+      "step": 46260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.103457825621504e-05,
+      "loss": 3.9609,
+      "step": 46265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.100572986557928e-05,
+      "loss": 3.9891,
+      "step": 46270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.097688655385512e-05,
+      "loss": 3.9781,
+      "step": 46275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.094804832268858e-05,
+      "loss": 3.9734,
+      "step": 46280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.091921517372556e-05,
+      "loss": 3.9547,
+      "step": 46285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.089038710861155e-05,
+      "loss": 3.9641,
+      "step": 46290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.086156412899175e-05,
+      "loss": 3.875,
+      "step": 46295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.083274623651102e-05,
+      "loss": 3.9219,
+      "step": 46300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.080393343281412e-05,
+      "loss": 4.0,
+      "step": 46305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.077512571954537e-05,
+      "loss": 3.9016,
+      "step": 46310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.074632309834873e-05,
+      "loss": 4.0125,
+      "step": 46315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0717525570868125e-05,
+      "loss": 3.975,
+      "step": 46320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0688733138746935e-05,
+      "loss": 3.9469,
+      "step": 46325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0659945803628365e-05,
+      "loss": 4.0484,
+      "step": 46330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.063116356715527e-05,
+      "loss": 3.9859,
+      "step": 46335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.060238643097035e-05,
+      "loss": 3.85,
+      "step": 46340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0573614396715874e-05,
+      "loss": 3.9156,
+      "step": 46345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.054484746603385e-05,
+      "loss": 3.9281,
+      "step": 46350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0516085640565956e-05,
+      "loss": 3.8891,
+      "step": 46355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0487328921953746e-05,
+      "loss": 4.0328,
+      "step": 46360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.045857731183832e-05,
+      "loss": 3.9859,
+      "step": 46365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.042983081186047e-05,
+      "loss": 3.9312,
+      "step": 46370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0401089423660886e-05,
+      "loss": 3.8906,
+      "step": 46375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.037235314887978e-05,
+      "loss": 3.9953,
+      "step": 46380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.034362198915713e-05,
+      "loss": 3.9531,
+      "step": 46385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.031489594613256e-05,
+      "loss": 3.9016,
+      "step": 46390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.028617502144559e-05,
+      "loss": 3.9172,
+      "step": 46395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0257459216735285e-05,
+      "loss": 3.8813,
+      "step": 46400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.02287485336404e-05,
+      "loss": 3.8969,
+      "step": 46405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.020004297379953e-05,
+      "loss": 3.9984,
+      "step": 46410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0171342538850885e-05,
+      "loss": 3.9922,
+      "step": 46415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.01426472304324e-05,
+      "loss": 3.9734,
+      "step": 46420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.011395705018166e-05,
+      "loss": 3.9453,
+      "step": 46425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.008527199973611e-05,
+      "loss": 3.9969,
+      "step": 46430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.005659208073277e-05,
+      "loss": 4.0016,
+      "step": 46435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.00279172948084e-05,
+      "loss": 3.9203,
+      "step": 46440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.999924764359943e-05,
+      "loss": 3.9344,
+      "step": 46445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.997058312874214e-05,
+      "loss": 3.9984,
+      "step": 46450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.994192375187236e-05,
+      "loss": 3.9062,
+      "step": 46455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.991326951462563e-05,
+      "loss": 4.0297,
+      "step": 46460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9884620418637354e-05,
+      "loss": 3.9281,
+      "step": 46465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9855976465542486e-05,
+      "loss": 4.0109,
+      "step": 46470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9827337656975746e-05,
+      "loss": 3.9734,
+      "step": 46475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.979870399457151e-05,
+      "loss": 3.8984,
+      "step": 46480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9770075479963986e-05,
+      "loss": 3.8656,
+      "step": 46485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9741452114786957e-05,
+      "loss": 3.9297,
+      "step": 46490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.971283390067392e-05,
+      "loss": 3.9406,
+      "step": 46495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.968422083925824e-05,
+      "loss": 3.9078,
+      "step": 46500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.965561293217277e-05,
+      "loss": 3.9141,
+      "step": 46505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.962701018105021e-05,
+      "loss": 3.8563,
+      "step": 46510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.959841258752284e-05,
+      "loss": 4.0234,
+      "step": 46515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.956982015322284e-05,
+      "loss": 4.0031,
+      "step": 46520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.954123287978193e-05,
+      "loss": 3.9188,
+      "step": 46525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9512650768831606e-05,
+      "loss": 3.9656,
+      "step": 46530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9484073822002976e-05,
+      "loss": 3.9641,
+      "step": 46535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.945550204092705e-05,
+      "loss": 3.9016,
+      "step": 46540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.942693542723436e-05,
+      "loss": 4.0078,
+      "step": 46545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.939837398255514e-05,
+      "loss": 3.9406,
+      "step": 46550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.936981770851954e-05,
+      "loss": 3.8781,
+      "step": 46555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.934126660675718e-05,
+      "loss": 3.9531,
+      "step": 46560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.931272067889743e-05,
+      "loss": 3.8359,
+      "step": 46565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.928417992656953e-05,
+      "loss": 3.9516,
+      "step": 46570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.925564435140223e-05,
+      "loss": 3.9609,
+      "step": 46575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.922711395502401e-05,
+      "loss": 4.0016,
+      "step": 46580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.919858873906322e-05,
+      "loss": 3.8672,
+      "step": 46585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.917006870514773e-05,
+      "loss": 3.9719,
+      "step": 46590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.914155385490514e-05,
+      "loss": 3.9734,
+      "step": 46595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.911304418996289e-05,
+      "loss": 3.8625,
+      "step": 46600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.908453971194799e-05,
+      "loss": 3.9375,
+      "step": 46605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.905604042248717e-05,
+      "loss": 3.9875,
+      "step": 46610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.902754632320686e-05,
+      "loss": 3.9047,
+      "step": 46615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8999057415733316e-05,
+      "loss": 3.9234,
+      "step": 46620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8970573701692346e-05,
+      "loss": 3.8875,
+      "step": 46625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.894209518270948e-05,
+      "loss": 3.9703,
+      "step": 46630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.891362186041007e-05,
+      "loss": 3.8563,
+      "step": 46635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.888515373641905e-05,
+      "loss": 3.9922,
+      "step": 46640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8856690812361064e-05,
+      "loss": 3.9688,
+      "step": 46645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.882823308986058e-05,
+      "loss": 3.9828,
+      "step": 46650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8799780570541625e-05,
+      "loss": 3.9984,
+      "step": 46655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.877133325602795e-05,
+      "loss": 3.9547,
+      "step": 46660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.874289114794314e-05,
+      "loss": 3.9312,
+      "step": 46665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.871445424791035e-05,
+      "loss": 3.9188,
+      "step": 46670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.868602255755241e-05,
+      "loss": 3.9688,
+      "step": 46675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8657596078492044e-05,
+      "loss": 3.9328,
+      "step": 46680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.862917481235147e-05,
+      "loss": 3.9344,
+      "step": 46685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8600758760752716e-05,
+      "loss": 3.9562,
+      "step": 46690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.857234792531742e-05,
+      "loss": 3.9016,
+      "step": 46695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.854394230766712e-05,
+      "loss": 3.9875,
+      "step": 46700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8515541909422864e-05,
+      "loss": 4.0047,
+      "step": 46705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8487146732205444e-05,
+      "loss": 3.9797,
+      "step": 46710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8458756777635365e-05,
+      "loss": 3.8766,
+      "step": 46715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.843037204733291e-05,
+      "loss": 3.85,
+      "step": 46720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.840199254291795e-05,
+      "loss": 4.0578,
+      "step": 46725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.837361826601008e-05,
+      "loss": 3.9641,
+      "step": 46730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.834524921822872e-05,
+      "loss": 4.0078,
+      "step": 46735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8316885401192835e-05,
+      "loss": 3.9,
+      "step": 46740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8288526816521134e-05,
+      "loss": 3.9781,
+      "step": 46745
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8260173465832025e-05,
+      "loss": 3.8687,
+      "step": 46750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8231825350743715e-05,
+      "loss": 3.8781,
+      "step": 46755
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.820348247287399e-05,
+      "loss": 3.9344,
+      "step": 46760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.817514483384034e-05,
+      "loss": 3.9266,
+      "step": 46765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8146812435260074e-05,
+      "loss": 3.9406,
+      "step": 46770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8118485278750095e-05,
+      "loss": 4.0016,
+      "step": 46775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.809016336592703e-05,
+      "loss": 3.9156,
+      "step": 46780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.806184669840717e-05,
+      "loss": 3.8469,
+      "step": 46785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.803353527780663e-05,
+      "loss": 3.9156,
+      "step": 46790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.800522910574113e-05,
+      "loss": 3.9469,
+      "step": 46795
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.797692818382608e-05,
+      "loss": 3.8984,
+      "step": 46800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.794863251367656e-05,
+      "loss": 3.9344,
+      "step": 46805
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.792034209690753e-05,
+      "loss": 3.8703,
+      "step": 46810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7892056935133455e-05,
+      "loss": 3.9937,
+      "step": 46815
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.786377702996854e-05,
+      "loss": 3.9188,
+      "step": 46820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.78355023830268e-05,
+      "loss": 4.0719,
+      "step": 46825
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.780723299592184e-05,
+      "loss": 3.8734,
+      "step": 46830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.777896887026701e-05,
+      "loss": 3.9562,
+      "step": 46835
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.775071000767527e-05,
+      "loss": 3.925,
+      "step": 46840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.772245640975947e-05,
+      "loss": 3.9094,
+      "step": 46845
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.769420807813197e-05,
+      "loss": 3.8359,
+      "step": 46850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.766596501440489e-05,
+      "loss": 4.0078,
+      "step": 46855
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.763772722019016e-05,
+      "loss": 3.9703,
+      "step": 46860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.760949469709922e-05,
+      "loss": 3.9469,
+      "step": 46865
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.758126744674337e-05,
+      "loss": 3.9094,
+      "step": 46870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.755304547073345e-05,
+      "loss": 3.9453,
+      "step": 46875
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.752482877068018e-05,
+      "loss": 4.0141,
+      "step": 46880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.749661734819388e-05,
+      "loss": 3.8656,
+      "step": 46885
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7468411204884555e-05,
+      "loss": 3.9969,
+      "step": 46890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7440210342361874e-05,
+      "loss": 3.8937,
+      "step": 46895
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.741201476223538e-05,
+      "loss": 4.0,
+      "step": 46900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.738382446611415e-05,
+      "loss": 3.9641,
+      "step": 46905
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.735563945560694e-05,
+      "loss": 3.9219,
+      "step": 46910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.732745973232237e-05,
+      "loss": 3.8594,
+      "step": 46915
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.729928529786864e-05,
+      "loss": 3.9422,
+      "step": 46920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7271116153853626e-05,
+      "loss": 3.9359,
+      "step": 46925
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.724295230188493e-05,
+      "loss": 3.8813,
+      "step": 46930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7214793743569933e-05,
+      "loss": 3.9547,
+      "step": 46935
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.718664048051564e-05,
+      "loss": 3.9031,
+      "step": 46940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7158492514328654e-05,
+      "loss": 3.8875,
+      "step": 46945
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.713034984661553e-05,
+      "loss": 3.8594,
+      "step": 46950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.710221247898231e-05,
+      "loss": 3.9766,
+      "step": 46955
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.707408041303478e-05,
+      "loss": 3.9828,
+      "step": 46960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7045953650378406e-05,
+      "loss": 3.9609,
+      "step": 46965
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.701783219261847e-05,
+      "loss": 3.9125,
+      "step": 46970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6989716041359844e-05,
+      "loss": 3.9781,
+      "step": 46975
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6961605198207086e-05,
+      "loss": 4.0172,
+      "step": 46980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.693349966476445e-05,
+      "loss": 4.0047,
+      "step": 46985
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.690539944263604e-05,
+      "loss": 3.9125,
+      "step": 46990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.687730453342545e-05,
+      "loss": 3.9656,
+      "step": 46995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.684921493873605e-05,
+      "loss": 3.8953,
+      "step": 47000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.682113066017097e-05,
+      "loss": 3.9734,
+      "step": 47005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.679305169933299e-05,
+      "loss": 3.8781,
+      "step": 47010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.676497805782453e-05,
+      "loss": 4.0234,
+      "step": 47015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.673690973724771e-05,
+      "loss": 3.9484,
+      "step": 47020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6708846739204515e-05,
+      "loss": 4.0062,
+      "step": 47025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.668078906529645e-05,
+      "loss": 3.9016,
+      "step": 47030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.66527367171247e-05,
+      "loss": 3.9734,
+      "step": 47035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6624689696290326e-05,
+      "loss": 3.8734,
+      "step": 47040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6596648004393944e-05,
+      "loss": 3.9859,
+      "step": 47045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.656861164303585e-05,
+      "loss": 3.85,
+      "step": 47050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.654058061381607e-05,
+      "loss": 3.8656,
+      "step": 47055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.651255491833443e-05,
+      "loss": 3.9719,
+      "step": 47060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.648453455819031e-05,
+      "loss": 3.9969,
+      "step": 47065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6456519534982766e-05,
+      "loss": 3.8219,
+      "step": 47070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.642850985031073e-05,
+      "loss": 3.9375,
+      "step": 47075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.640050550577269e-05,
+      "loss": 3.8828,
+      "step": 47080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.637250650296676e-05,
+      "loss": 4.0094,
+      "step": 47085
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6344512843491e-05,
+      "loss": 3.9641,
+      "step": 47090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6316524528942905e-05,
+      "loss": 3.9078,
+      "step": 47095
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.628854156091977e-05,
+      "loss": 3.9656,
+      "step": 47100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.626056394101866e-05,
+      "loss": 3.975,
+      "step": 47105
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.62325916708362e-05,
+      "loss": 3.9219,
+      "step": 47110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.620462475196875e-05,
+      "loss": 3.9141,
+      "step": 47115
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6176663186012466e-05,
+      "loss": 3.9328,
+      "step": 47120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.614870697456307e-05,
+      "loss": 3.9516,
+      "step": 47125
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6120756119215984e-05,
+      "loss": 3.9266,
+      "step": 47130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6092810621566444e-05,
+      "loss": 4.0172,
+      "step": 47135
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6064870483209275e-05,
+      "loss": 3.9641,
+      "step": 47140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.603693570573897e-05,
+      "loss": 3.8719,
+      "step": 47145
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6009006290749855e-05,
+      "loss": 3.9297,
+      "step": 47150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5981082239835837e-05,
+      "loss": 3.8828,
+      "step": 47155
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5953163554590516e-05,
+      "loss": 3.9297,
+      "step": 47160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.592525023660718e-05,
+      "loss": 4.0281,
+      "step": 47165
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5897342287478956e-05,
+      "loss": 3.8656,
+      "step": 47170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.586943970879849e-05,
+      "loss": 3.9891,
+      "step": 47175
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.584154250215813e-05,
+      "loss": 3.8391,
+      "step": 47180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.581365066915009e-05,
+      "loss": 3.9312,
+      "step": 47185
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5785764211366086e-05,
+      "loss": 3.9953,
+      "step": 47190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.575788313039761e-05,
+      "loss": 4.0328,
+      "step": 47195
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.57300074278358e-05,
+      "loss": 3.9047,
+      "step": 47200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5702137105271604e-05,
+      "loss": 3.8563,
+      "step": 47205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.567427216429557e-05,
+      "loss": 3.9453,
+      "step": 47210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.564641260649785e-05,
+      "loss": 3.9984,
+      "step": 47215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.561855843346855e-05,
+      "loss": 4.0094,
+      "step": 47220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.559070964679721e-05,
+      "loss": 3.9969,
+      "step": 47225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.556286624807319e-05,
+      "loss": 3.9703,
+      "step": 47230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5535028238885456e-05,
+      "loss": 4.0125,
+      "step": 47235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.550719562082283e-05,
+      "loss": 3.9453,
+      "step": 47240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.547936839547368e-05,
+      "loss": 4.0922,
+      "step": 47245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.545154656442609e-05,
+      "loss": 4.0094,
+      "step": 47250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.542373012926782e-05,
+      "loss": 3.9344,
+      "step": 47255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.539591909158646e-05,
+      "loss": 3.9641,
+      "step": 47260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5368113452969115e-05,
+      "loss": 3.8719,
+      "step": 47265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.534031321500263e-05,
+      "loss": 3.9266,
+      "step": 47270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.531251837927365e-05,
+      "loss": 4.0328,
+      "step": 47275
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.528472894736838e-05,
+      "loss": 3.9344,
+      "step": 47280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.525694492087279e-05,
+      "loss": 3.9156,
+      "step": 47285
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.522916630137244e-05,
+      "loss": 3.9266,
+      "step": 47290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5201393090452755e-05,
+      "loss": 3.9578,
+      "step": 47295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.517362528969874e-05,
+      "loss": 3.8937,
+      "step": 47300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.514586290069502e-05,
+      "loss": 4.0328,
+      "step": 47305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5118105925026105e-05,
+      "loss": 3.9562,
+      "step": 47310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5090354364276064e-05,
+      "loss": 3.975,
+      "step": 47315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5062608220028645e-05,
+      "loss": 3.9062,
+      "step": 47320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.50348674938673e-05,
+      "loss": 3.9188,
+      "step": 47325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5007132187375273e-05,
+      "loss": 3.9156,
+      "step": 47330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.49794023021354e-05,
+      "loss": 3.9156,
+      "step": 47335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.495167783973019e-05,
+      "loss": 3.9172,
+      "step": 47340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.492395880174186e-05,
+      "loss": 4.0016,
+      "step": 47345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.489624518975242e-05,
+      "loss": 3.9203,
+      "step": 47350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.486853700534346e-05,
+      "loss": 3.8922,
+      "step": 47355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.484083425009622e-05,
+      "loss": 3.95,
+      "step": 47360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.481313692559181e-05,
+      "loss": 3.9266,
+      "step": 47365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4785445033410864e-05,
+      "loss": 3.9141,
+      "step": 47370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.475775857513375e-05,
+      "loss": 3.9609,
+      "step": 47375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.473007755234051e-05,
+      "loss": 3.9562,
+      "step": 47380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.470240196661099e-05,
+      "loss": 4.0297,
+      "step": 47385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4674731819524575e-05,
+      "loss": 3.8656,
+      "step": 47390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.464706711266038e-05,
+      "loss": 3.9844,
+      "step": 47395
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4619407847597315e-05,
+      "loss": 3.9531,
+      "step": 47400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.459175402591386e-05,
+      "loss": 3.9344,
+      "step": 47405
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4564105649188204e-05,
+      "loss": 3.8203,
+      "step": 47410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.453646271899821e-05,
+      "loss": 3.9844,
+      "step": 47415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4508825236921537e-05,
+      "loss": 3.9969,
+      "step": 47420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4481193204535435e-05,
+      "loss": 3.9,
+      "step": 47425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.445356662341687e-05,
+      "loss": 3.9328,
+      "step": 47430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4425945495142406e-05,
+      "loss": 3.9078,
+      "step": 47435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4398329821288536e-05,
+      "loss": 3.8875,
+      "step": 47440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4370719603431194e-05,
+      "loss": 3.9094,
+      "step": 47445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4343114843146065e-05,
+      "loss": 3.8984,
+      "step": 47450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.431551554200866e-05,
+      "loss": 4.0484,
+      "step": 47455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4287921701594026e-05,
+      "loss": 3.9844,
+      "step": 47460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.426033332347692e-05,
+      "loss": 3.9172,
+      "step": 47465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.42327504092318e-05,
+      "loss": 3.9391,
+      "step": 47470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4205172960432896e-05,
+      "loss": 4.0172,
+      "step": 47475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.417760097865403e-05,
+      "loss": 3.9641,
+      "step": 47480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4150034465468675e-05,
+      "loss": 3.9047,
+      "step": 47485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.412247342245015e-05,
+      "loss": 3.9906,
+      "step": 47490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.409491785117133e-05,
+      "loss": 3.975,
+      "step": 47495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4067367753204804e-05,
+      "loss": 3.9797,
+      "step": 47500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4039823130122814e-05,
+      "loss": 3.9969,
+      "step": 47505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.401228398349742e-05,
+      "loss": 3.9641,
+      "step": 47510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.398475031490025e-05,
+      "loss": 3.9656,
+      "step": 47515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.395722212590265e-05,
+      "loss": 3.9922,
+      "step": 47520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.39296994180756e-05,
+      "loss": 4.0047,
+      "step": 47525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.390218219298993e-05,
+      "loss": 3.9844,
+      "step": 47530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.387467045221601e-05,
+      "loss": 3.9344,
+      "step": 47535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3847164197323864e-05,
+      "loss": 3.9703,
+      "step": 47540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3819663429883384e-05,
+      "loss": 3.9484,
+      "step": 47545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3792168151464015e-05,
+      "loss": 4.0125,
+      "step": 47550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3764678363634826e-05,
+      "loss": 3.9656,
+      "step": 47555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.37371940679648e-05,
+      "loss": 3.8672,
+      "step": 47560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.37097152660224e-05,
+      "loss": 3.9156,
+      "step": 47565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3682241959375826e-05,
+      "loss": 3.975,
+      "step": 47570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.365477414959295e-05,
+      "loss": 3.8578,
+      "step": 47575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3627311838241495e-05,
+      "loss": 3.9953,
+      "step": 47580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.359985502688864e-05,
+      "loss": 3.95,
+      "step": 47585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.35724037171013e-05,
+      "loss": 3.95,
+      "step": 47590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.354495791044625e-05,
+      "loss": 3.9781,
+      "step": 47595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3517517608489764e-05,
+      "loss": 3.9078,
+      "step": 47600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3490082812797816e-05,
+      "loss": 3.8703,
+      "step": 47605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.34626535249362e-05,
+      "loss": 3.9312,
+      "step": 47610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.343522974647027e-05,
+      "loss": 4.0125,
+      "step": 47615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.340781147896504e-05,
+      "loss": 4.0156,
+      "step": 47620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.338039872398541e-05,
+      "loss": 3.9453,
+      "step": 47625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.335299148309574e-05,
+      "loss": 4.0578,
+      "step": 47630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.332558975786012e-05,
+      "loss": 4.0062,
+      "step": 47635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.329819354984246e-05,
+      "loss": 3.9516,
+      "step": 47640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3270802860606254e-05,
+      "loss": 3.8188,
+      "step": 47645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3243417691714595e-05,
+      "loss": 3.9813,
+      "step": 47650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3216038044730486e-05,
+      "loss": 4.0484,
+      "step": 47655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.318866392121644e-05,
+      "loss": 3.9688,
+      "step": 47660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3161295322734625e-05,
+      "loss": 3.8594,
+      "step": 47665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.31339322508471e-05,
+      "loss": 3.9797,
+      "step": 47670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3106574707115396e-05,
+      "loss": 3.9547,
+      "step": 47675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.307922269310083e-05,
+      "loss": 3.8953,
+      "step": 47680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.305187621036433e-05,
+      "loss": 3.775,
+      "step": 47685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.302453526046668e-05,
+      "loss": 4.0219,
+      "step": 47690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.2997199844968144e-05,
+      "loss": 3.9953,
+      "step": 47695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.296986996542878e-05,
+      "loss": 3.925,
+      "step": 47700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.294254562340826e-05,
+      "loss": 3.9984,
+      "step": 47705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.291522682046607e-05,
+      "loss": 3.9109,
+      "step": 47710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.2887913558161266e-05,
+      "loss": 4.0031,
+      "step": 47715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.286060583805256e-05,
+      "loss": 3.9031,
+      "step": 47720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.283330366169851e-05,
+      "loss": 3.8859,
+      "step": 47725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.280600703065719e-05,
+      "loss": 3.9078,
+      "step": 47730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.2778715946486424e-05,
+      "loss": 4.0266,
+      "step": 47735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.2751430410743684e-05,
+      "loss": 3.9125,
+      "step": 47740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.272415042498626e-05,
+      "loss": 3.9188,
+      "step": 47745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.269687599077095e-05,
+      "loss": 4.0109,
+      "step": 47750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2669607109654263e-05,
+      "loss": 3.925,
+      "step": 47755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2642343783192555e-05,
+      "loss": 3.9984,
+      "step": 47760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.261508601294168e-05,
+      "loss": 3.8859,
+      "step": 47765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.258783380045724e-05,
+      "loss": 4.0359,
+      "step": 47770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.256058714729449e-05,
+      "loss": 3.925,
+      "step": 47775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.253334605500849e-05,
+      "loss": 3.975,
+      "step": 47780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.250611052515383e-05,
+      "loss": 3.9641,
+      "step": 47785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.247888055928485e-05,
+      "loss": 3.9047,
+      "step": 47790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.245165615895552e-05,
+      "loss": 3.9719,
+      "step": 47795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.242443732571966e-05,
+      "loss": 3.9516,
+      "step": 47800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2397224061130564e-05,
+      "loss": 4.0,
+      "step": 47805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.237001636674127e-05,
+      "loss": 3.9656,
+      "step": 47810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.234281424410462e-05,
+      "loss": 3.9,
+      "step": 47815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.231561769477298e-05,
+      "loss": 3.9062,
+      "step": 47820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2288426720298465e-05,
+      "loss": 3.9984,
+      "step": 47825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.226124132223282e-05,
+      "loss": 3.9828,
+      "step": 47830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2234061502127625e-05,
+      "loss": 3.9953,
+      "step": 47835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.220688726153396e-05,
+      "loss": 3.9422,
+      "step": 47840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.217971860200264e-05,
+      "loss": 3.9922,
+      "step": 47845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.215255552508428e-05,
+      "loss": 3.8875,
+      "step": 47850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.212539803232901e-05,
+      "loss": 3.8922,
+      "step": 47855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2098246125286725e-05,
+      "loss": 3.9438,
+      "step": 47860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.207109980550694e-05,
+      "loss": 3.9312,
+      "step": 47865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2043959074539e-05,
+      "loss": 3.9453,
+      "step": 47870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.201682393393175e-05,
+      "loss": 3.9937,
+      "step": 47875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.198969438523382e-05,
+      "loss": 3.9891,
+      "step": 47880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.196257042999346e-05,
+      "loss": 3.9984,
+      "step": 47885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.193545206975871e-05,
+      "loss": 4.0406,
+      "step": 47890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.190833930607718e-05,
+      "loss": 3.9203,
+      "step": 47895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.188123214049612e-05,
+      "loss": 3.9547,
+      "step": 47900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.18541305745627e-05,
+      "loss": 3.9078,
+      "step": 47905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.18270346098235e-05,
+      "loss": 3.9344,
+      "step": 47910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1799944247824926e-05,
+      "loss": 3.9312,
+      "step": 47915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.177285949011295e-05,
+      "loss": 3.8781,
+      "step": 47920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.174578033823342e-05,
+      "loss": 3.9469,
+      "step": 47925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.171870679373169e-05,
+      "loss": 3.9688,
+      "step": 47930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.16916388581528e-05,
+      "loss": 3.9906,
+      "step": 47935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1664576533041616e-05,
+      "loss": 4.0109,
+      "step": 47940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.163751981994256e-05,
+      "loss": 3.8906,
+      "step": 47945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1610468720399715e-05,
+      "loss": 3.9016,
+      "step": 47950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.158342323595689e-05,
+      "loss": 3.9406,
+      "step": 47955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.155638336815764e-05,
+      "loss": 3.9719,
+      "step": 47960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.15293491185451e-05,
+      "loss": 3.9359,
+      "step": 47965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.150232048866211e-05,
+      "loss": 3.9203,
+      "step": 47970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.147529748005115e-05,
+      "loss": 3.9578,
+      "step": 47975
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.144828009425452e-05,
+      "loss": 3.8844,
+      "step": 47980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.142126833281405e-05,
+      "loss": 3.8859,
+      "step": 47985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.139426219727128e-05,
+      "loss": 3.8922,
+      "step": 47990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.136726168916753e-05,
+      "loss": 3.9375,
+      "step": 47995
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.134026681004366e-05,
+      "loss": 3.9438,
+      "step": 48000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.131327756144031e-05,
+      "loss": 3.7828,
+      "step": 48005
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1286293944897675e-05,
+      "loss": 3.9672,
+      "step": 48010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.125931596195582e-05,
+      "loss": 3.9672,
+      "step": 48015
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.123234361415434e-05,
+      "loss": 3.9328,
+      "step": 48020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.12053769030325e-05,
+      "loss": 3.9766,
+      "step": 48025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1178415830129374e-05,
+      "loss": 3.9219,
+      "step": 48030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1151460396983604e-05,
+      "loss": 3.9297,
+      "step": 48035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1124510605133526e-05,
+      "loss": 3.9328,
+      "step": 48040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.109756645611712e-05,
+      "loss": 3.9797,
+      "step": 48045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1070627951472204e-05,
+      "loss": 3.975,
+      "step": 48050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1043695092736095e-05,
+      "loss": 3.9484,
+      "step": 48055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.101676788144581e-05,
+      "loss": 4.0109,
+      "step": 48060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0989846319138186e-05,
+      "loss": 3.9172,
+      "step": 48065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0962930407349605e-05,
+      "loss": 3.975,
+      "step": 48070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.093602014761608e-05,
+      "loss": 3.9562,
+      "step": 48075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.09091155414735e-05,
+      "loss": 3.9047,
+      "step": 48080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.088221659045725e-05,
+      "loss": 3.9609,
+      "step": 48085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.085532329610243e-05,
+      "loss": 3.7906,
+      "step": 48090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0828435659943934e-05,
+      "loss": 3.9641,
+      "step": 48095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.080155368351619e-05,
+      "loss": 3.9516,
+      "step": 48100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.077467736835328e-05,
+      "loss": 3.925,
+      "step": 48105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.074780671598919e-05,
+      "loss": 3.9688,
+      "step": 48110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0720941727957336e-05,
+      "loss": 3.9609,
+      "step": 48115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0694082405790876e-05,
+      "loss": 3.9125,
+      "step": 48120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0667228751022764e-05,
+      "loss": 3.9937,
+      "step": 48125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.06403807651855e-05,
+      "loss": 3.8719,
+      "step": 48130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.061353844981124e-05,
+      "loss": 3.9656,
+      "step": 48135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.058670180643199e-05,
+      "loss": 3.9516,
+      "step": 48140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.055987083657928e-05,
+      "loss": 3.9297,
+      "step": 48145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0533045541784325e-05,
+      "loss": 3.9359,
+      "step": 48150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0506225923578016e-05,
+      "loss": 3.9391,
+      "step": 48155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.047941198349103e-05,
+      "loss": 3.9813,
+      "step": 48160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.045260372305363e-05,
+      "loss": 3.9391,
+      "step": 48165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.042580114379569e-05,
+      "loss": 3.9578,
+      "step": 48170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.039900424724694e-05,
+      "loss": 3.9109,
+      "step": 48175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0372213034936634e-05,
+      "loss": 3.8516,
+      "step": 48180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.034542750839375e-05,
+      "loss": 3.9047,
+      "step": 48185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0318647669146885e-05,
+      "loss": 3.9047,
+      "step": 48190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0291873518724466e-05,
+      "loss": 3.9562,
+      "step": 48195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.026510505865445e-05,
+      "loss": 3.9719,
+      "step": 48200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.023834229046447e-05,
+      "loss": 3.9375,
+      "step": 48205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.021158521568196e-05,
+      "loss": 3.9703,
+      "step": 48210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0184833835833936e-05,
+      "loss": 3.9656,
+      "step": 48215
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0158088152447076e-05,
+      "loss": 4.0109,
+      "step": 48220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.01313481670477e-05,
+      "loss": 3.9797,
+      "step": 48225
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.010461388116199e-05,
+      "loss": 3.9234,
+      "step": 48230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0077885296315614e-05,
+      "loss": 3.9781,
+      "step": 48235
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0051162414033956e-05,
+      "loss": 3.9172,
+      "step": 48240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.002444523584205e-05,
+      "loss": 3.9438,
+      "step": 48245
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9997733763264764e-05,
+      "loss": 3.9766,
+      "step": 48250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.997102799782647e-05,
+      "loss": 3.9484,
+      "step": 48255
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9944327941051215e-05,
+      "loss": 3.9438,
+      "step": 48260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.991763359446287e-05,
+      "loss": 3.9875,
+      "step": 48265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.989094495958483e-05,
+      "loss": 4.0375,
+      "step": 48270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.986426203794024e-05,
+      "loss": 3.9391,
+      "step": 48275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9837584831051816e-05,
+      "loss": 3.9703,
+      "step": 48280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.981091334044215e-05,
+      "loss": 4.0125,
+      "step": 48285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.978424756763333e-05,
+      "loss": 3.9625,
+      "step": 48290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.975758751414714e-05,
+      "loss": 3.9516,
+      "step": 48295
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.973093318150513e-05,
+      "loss": 3.9359,
+      "step": 48300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9704284571228456e-05,
+      "loss": 3.9438,
+      "step": 48305
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.967764168483795e-05,
+      "loss": 4.025,
+      "step": 48310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.965100452385407e-05,
+      "loss": 4.0203,
+      "step": 48315
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.962437308979708e-05,
+      "loss": 3.9484,
+      "step": 48320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.959774738418681e-05,
+      "loss": 3.8937,
+      "step": 48325
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9571127408542806e-05,
+      "loss": 4.0141,
+      "step": 48330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9544513164384194e-05,
+      "loss": 3.9344,
+      "step": 48335
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.951790465322996e-05,
+      "loss": 4.0094,
+      "step": 48340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.949130187659861e-05,
+      "loss": 3.9375,
+      "step": 48345
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.946470483600833e-05,
+      "loss": 3.9328,
+      "step": 48350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9438113532977084e-05,
+      "loss": 3.8937,
+      "step": 48355
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.941152796902241e-05,
+      "loss": 3.9531,
+      "step": 48360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.938494814566153e-05,
+      "loss": 3.9578,
+      "step": 48365
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.935837406441134e-05,
+      "loss": 3.975,
+      "step": 48370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9331805726788504e-05,
+      "loss": 3.9562,
+      "step": 48375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.930524313430923e-05,
+      "loss": 3.9,
+      "step": 48380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.927868628848941e-05,
+      "loss": 3.9328,
+      "step": 48385
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9252135190844714e-05,
+      "loss": 3.9688,
+      "step": 48390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.92255898428904e-05,
+      "loss": 4.0609,
+      "step": 48395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9199050246141416e-05,
+      "loss": 3.9375,
+      "step": 48400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.917251640211229e-05,
+      "loss": 3.9922,
+      "step": 48405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.914598831231745e-05,
+      "loss": 3.9609,
+      "step": 48410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9119465978270796e-05,
+      "loss": 3.9297,
+      "step": 48415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9092949401485946e-05,
+      "loss": 4.0453,
+      "step": 48420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9066438583476165e-05,
+      "loss": 3.9062,
+      "step": 48425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.903993352575453e-05,
+      "loss": 4.0109,
+      "step": 48430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9013434229833636e-05,
+      "loss": 3.9547,
+      "step": 48435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.898694069722575e-05,
+      "loss": 3.9906,
+      "step": 48440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8960452929442967e-05,
+      "loss": 3.9547,
+      "step": 48445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.893397092799688e-05,
+      "loss": 3.9453,
+      "step": 48450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.890749469439883e-05,
+      "loss": 3.9734,
+      "step": 48455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.888102423015976e-05,
+      "loss": 3.8359,
+      "step": 48460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.885455953679045e-05,
+      "loss": 3.9578,
+      "step": 48465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8828100615801194e-05,
+      "loss": 3.9969,
+      "step": 48470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.880164746870196e-05,
+      "loss": 3.9031,
+      "step": 48475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.877520009700252e-05,
+      "loss": 3.9078,
+      "step": 48480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8748758502212165e-05,
+      "loss": 3.9672,
+      "step": 48485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8722322685839955e-05,
+      "loss": 3.9406,
+      "step": 48490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.869589264939451e-05,
+      "loss": 3.9656,
+      "step": 48495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8669468394384305e-05,
+      "loss": 3.9688,
+      "step": 48500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.864304992231732e-05,
+      "loss": 3.9281,
+      "step": 48505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.861663723470125e-05,
+      "loss": 3.9531,
+      "step": 48510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.859023033304343e-05,
+      "loss": 3.925,
+      "step": 48515
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8563829218851016e-05,
+      "loss": 3.9688,
+      "step": 48520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.853743389363065e-05,
+      "loss": 3.9547,
+      "step": 48525
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8511044358888684e-05,
+      "loss": 3.9359,
+      "step": 48530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.848466061613126e-05,
+      "loss": 3.9844,
+      "step": 48535
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.845828266686406e-05,
+      "loss": 3.9172,
+      "step": 48540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.843191051259245e-05,
+      "loss": 3.8875,
+      "step": 48545
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.840554415482148e-05,
+      "loss": 3.8125,
+      "step": 48550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8379183595055966e-05,
+      "loss": 3.8828,
+      "step": 48555
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.835282883480024e-05,
+      "loss": 3.9641,
+      "step": 48560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8326479875558335e-05,
+      "loss": 3.9688,
+      "step": 48565
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.830013671883409e-05,
+      "loss": 3.9359,
+      "step": 48570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.827379936613087e-05,
+      "loss": 4.0172,
+      "step": 48575
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.824746781895168e-05,
+      "loss": 3.9688,
+      "step": 48580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.822114207879938e-05,
+      "loss": 3.9562,
+      "step": 48585
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.819482214717631e-05,
+      "loss": 3.875,
+      "step": 48590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.816850802558454e-05,
+      "loss": 3.9359,
+      "step": 48595
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.814219971552588e-05,
+      "loss": 4.0281,
+      "step": 48600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8115897218501733e-05,
+      "loss": 4.0187,
+      "step": 48605
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.808960053601311e-05,
+      "loss": 3.9531,
+      "step": 48610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8063309669560865e-05,
+      "loss": 3.9391,
+      "step": 48615
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.803702462064539e-05,
+      "loss": 3.9172,
+      "step": 48620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.80107453907667e-05,
+      "loss": 3.9312,
+      "step": 48625
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.798447198142466e-05,
+      "loss": 3.9781,
+      "step": 48630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.795820439411866e-05,
+      "loss": 3.9625,
+      "step": 48635
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.793194263034773e-05,
+      "loss": 3.8297,
+      "step": 48640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7905686691610756e-05,
+      "loss": 3.9516,
+      "step": 48645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.787943657940607e-05,
+      "loss": 3.9578,
+      "step": 48650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7853192295231764e-05,
+      "loss": 3.9359,
+      "step": 48655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.782695384058568e-05,
+      "loss": 4.0547,
+      "step": 48660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.780072121696521e-05,
+      "loss": 3.9172,
+      "step": 48665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7774494425867436e-05,
+      "loss": 3.9969,
+      "step": 48670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.774827346878908e-05,
+      "loss": 3.8844,
+      "step": 48675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7722058347226696e-05,
+      "loss": 3.9375,
+      "step": 48680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7695849062676304e-05,
+      "loss": 3.9484,
+      "step": 48685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7669645616633675e-05,
+      "loss": 3.9344,
+      "step": 48690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7643448010594203e-05,
+      "loss": 3.9937,
+      "step": 48695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7617256246053085e-05,
+      "loss": 4.0031,
+      "step": 48700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.759107032450504e-05,
+      "loss": 3.8672,
+      "step": 48705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.756489024744444e-05,
+      "loss": 3.8719,
+      "step": 48710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7538716016365506e-05,
+      "loss": 3.8859,
+      "step": 48715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.751254763276193e-05,
+      "loss": 4.0266,
+      "step": 48720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.748638509812715e-05,
+      "loss": 3.9188,
+      "step": 48725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.746022841395423e-05,
+      "loss": 3.9922,
+      "step": 48730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.743407758173601e-05,
+      "loss": 3.8109,
+      "step": 48735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.740793260296489e-05,
+      "loss": 3.8875,
+      "step": 48740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.738179347913291e-05,
+      "loss": 3.8625,
+      "step": 48745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7355660211731926e-05,
+      "loss": 3.9578,
+      "step": 48750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.732953280225332e-05,
+      "loss": 3.9172,
+      "step": 48755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.730341125218819e-05,
+      "loss": 4.0672,
+      "step": 48760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7277295563027254e-05,
+      "loss": 3.9062,
+      "step": 48765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.725118573626102e-05,
+      "loss": 3.9937,
+      "step": 48770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.722508177337953e-05,
+      "loss": 3.9953,
+      "step": 48775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.719898367587255e-05,
+      "loss": 3.9781,
+      "step": 48780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7172891445229444e-05,
+      "loss": 3.9156,
+      "step": 48785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.714680508293941e-05,
+      "loss": 3.8937,
+      "step": 48790
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.712072459049111e-05,
+      "loss": 3.9047,
+      "step": 48795
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7094649969372954e-05,
+      "loss": 3.8953,
+      "step": 48800
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.706858122107311e-05,
+      "loss": 3.9172,
+      "step": 48805
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7042518347079284e-05,
+      "loss": 3.9719,
+      "step": 48810
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.701646134887885e-05,
+      "loss": 3.9875,
+      "step": 48815
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.699041022795887e-05,
+      "loss": 4.0219,
+      "step": 48820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.696436498580617e-05,
+      "loss": 3.9125,
+      "step": 48825
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6938325623907114e-05,
+      "loss": 3.9109,
+      "step": 48830
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.691229214374772e-05,
+      "loss": 3.9234,
+      "step": 48835
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.688626454681379e-05,
+      "loss": 3.9766,
+      "step": 48840
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6860242834590715e-05,
+      "loss": 3.8797,
+      "step": 48845
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6834227008563524e-05,
+      "loss": 3.9734,
+      "step": 48850
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.680821707021691e-05,
+      "loss": 3.9469,
+      "step": 48855
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.678221302103536e-05,
+      "loss": 3.9438,
+      "step": 48860
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.675621486250289e-05,
+      "loss": 3.8984,
+      "step": 48865
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.673022259610319e-05,
+      "loss": 3.925,
+      "step": 48870
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.670423622331961e-05,
+      "loss": 3.9391,
+      "step": 48875
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.667825574563529e-05,
+      "loss": 3.8813,
+      "step": 48880
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.665228116453289e-05,
+      "loss": 3.9266,
+      "step": 48885
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6626312481494724e-05,
+      "loss": 3.9312,
+      "step": 48890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.660034969800293e-05,
+      "loss": 3.8609,
+      "step": 48895
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.657439281553917e-05,
+      "loss": 4.0172,
+      "step": 48900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.654844183558479e-05,
+      "loss": 3.9562,
+      "step": 48905
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.652249675962077e-05,
+      "loss": 3.9172,
+      "step": 48910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6496557589127915e-05,
+      "loss": 3.8734,
+      "step": 48915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.647062432558649e-05,
+      "loss": 3.8766,
+      "step": 48920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.644469697047649e-05,
+      "loss": 3.85,
+      "step": 48925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6418775525277665e-05,
+      "loss": 3.9406,
+      "step": 48930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.639285999146933e-05,
+      "loss": 3.9141,
+      "step": 48935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.636695037053048e-05,
+      "loss": 3.9156,
+      "step": 48940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.634104666393972e-05,
+      "loss": 3.8813,
+      "step": 48945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.631514887317549e-05,
+      "loss": 3.9219,
+      "step": 48950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.628925699971573e-05,
+      "loss": 3.9391,
+      "step": 48955
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.626337104503806e-05,
+      "loss": 4.0109,
+      "step": 48960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6237491010619795e-05,
+      "loss": 3.8594,
+      "step": 48965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.621161689793796e-05,
+      "loss": 3.9078,
+      "step": 48970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.618574870846918e-05,
+      "loss": 3.9781,
+      "step": 48975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.615988644368969e-05,
+      "loss": 3.9953,
+      "step": 48980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.613403010507555e-05,
+      "loss": 4.0141,
+      "step": 48985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.610817969410235e-05,
+      "loss": 3.9141,
+      "step": 48990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6082335212245355e-05,
+      "loss": 3.9422,
+      "step": 48995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.605649666097949e-05,
+      "loss": 3.9484,
+      "step": 49000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.603066404177942e-05,
+      "loss": 3.9703,
+      "step": 49005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6004837356119395e-05,
+      "loss": 3.9656,
+      "step": 49010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.5979016605473284e-05,
+      "loss": 3.9703,
+      "step": 49015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.595320179131481e-05,
+      "loss": 3.8734,
+      "step": 49020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.592739291511714e-05,
+      "loss": 3.8687,
+      "step": 49025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.59015899783532e-05,
+      "loss": 3.9281,
+      "step": 49030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.587579298249552e-05,
+      "loss": 3.9656,
+      "step": 49035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.585000192901644e-05,
+      "loss": 3.9375,
+      "step": 49040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.58242168193878e-05,
+      "loss": 3.9734,
+      "step": 49045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.579843765508111e-05,
+      "loss": 3.8906,
+      "step": 49050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.57726644375677e-05,
+      "loss": 3.9156,
+      "step": 49055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.574689716831838e-05,
+      "loss": 3.9688,
+      "step": 49060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5721135848803653e-05,
+      "loss": 3.9391,
+      "step": 49065
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.569538048049382e-05,
+      "loss": 3.9594,
+      "step": 49070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.566963106485871e-05,
+      "loss": 3.8438,
+      "step": 49075
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5643887603367755e-05,
+      "loss": 3.8984,
+      "step": 49080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.561815009749027e-05,
+      "loss": 4.0234,
+      "step": 49085
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.559241854869502e-05,
+      "loss": 3.9688,
+      "step": 49090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5566692958450486e-05,
+      "loss": 4.0469,
+      "step": 49095
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.554097332822491e-05,
+      "loss": 3.9297,
+      "step": 49100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.551525965948607e-05,
+      "loss": 3.9766,
+      "step": 49105
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.548955195370145e-05,
+      "loss": 3.9234,
+      "step": 49110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5463850212338156e-05,
+      "loss": 3.9922,
+      "step": 49115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.543815443686304e-05,
+      "loss": 3.9594,
+      "step": 49120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.541246462874257e-05,
+      "loss": 3.8828,
+      "step": 49125
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5386780789442766e-05,
+      "loss": 3.9031,
+      "step": 49130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5361102920429545e-05,
+      "loss": 4.0312,
+      "step": 49135
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.533543102316829e-05,
+      "loss": 3.95,
+      "step": 49140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.530976509912402e-05,
+      "loss": 3.9203,
+      "step": 49145
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.528410514976163e-05,
+      "loss": 3.9266,
+      "step": 49150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.525845117654545e-05,
+      "loss": 3.9203,
+      "step": 49155
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.523280318093953e-05,
+      "loss": 3.8875,
+      "step": 49160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5207161164407696e-05,
+      "loss": 3.9547,
+      "step": 49165
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5181525128413294e-05,
+      "loss": 3.95,
+      "step": 49170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5155895074419365e-05,
+      "loss": 3.8828,
+      "step": 49175
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.513027100388857e-05,
+      "loss": 3.9594,
+      "step": 49180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.510465291828339e-05,
+      "loss": 3.9438,
+      "step": 49185
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.50790408190658e-05,
+      "loss": 3.9062,
+      "step": 49190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.505343470769742e-05,
+      "loss": 3.9141,
+      "step": 49195
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5027834585639694e-05,
+      "loss": 4.0422,
+      "step": 49200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.500224045435359e-05,
+      "loss": 3.9266,
+      "step": 49205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.497665231529977e-05,
+      "loss": 3.9016,
+      "step": 49210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4951070169938474e-05,
+      "loss": 4.0266,
+      "step": 49215
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.49254940197298e-05,
+      "loss": 3.8875,
+      "step": 49220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.489992386613332e-05,
+      "loss": 3.8453,
+      "step": 49225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.487435971060834e-05,
+      "loss": 3.9469,
+      "step": 49230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4848801554613754e-05,
+      "loss": 3.9797,
+      "step": 49235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.482324939960824e-05,
+      "loss": 4.0047,
+      "step": 49240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.479770324705006e-05,
+      "loss": 3.8969,
+      "step": 49245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.477216309839704e-05,
+      "loss": 3.8781,
+      "step": 49250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.47466289551069e-05,
+      "loss": 4.0016,
+      "step": 49255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.47211008186368e-05,
+      "loss": 3.9906,
+      "step": 49260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.469557869044363e-05,
+      "loss": 3.8734,
+      "step": 49265
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4670062571983914e-05,
+      "loss": 3.8766,
+      "step": 49270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.464455246471393e-05,
+      "loss": 3.9297,
+      "step": 49275
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4619048370089503e-05,
+      "loss": 3.8844,
+      "step": 49280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.459355028956613e-05,
+      "loss": 4.0219,
+      "step": 49285
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.456805822459905e-05,
+      "loss": 4.0531,
+      "step": 49290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.454257217664306e-05,
+      "loss": 3.9844,
+      "step": 49295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4517092147152675e-05,
+      "loss": 3.9281,
+      "step": 49300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.449161813758197e-05,
+      "loss": 3.8672,
+      "step": 49305
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4466150149384846e-05,
+      "loss": 3.9438,
+      "step": 49310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.444068818401472e-05,
+      "loss": 3.9406,
+      "step": 49315
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.441523224292471e-05,
+      "loss": 3.9688,
+      "step": 49320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.438978232756754e-05,
+      "loss": 3.8438,
+      "step": 49325
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.436433843939573e-05,
+      "loss": 3.9875,
+      "step": 49330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.433890057986133e-05,
+      "loss": 3.8484,
+      "step": 49335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4313468750416033e-05,
+      "loss": 3.9641,
+      "step": 49340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.42880429525113e-05,
+      "loss": 4.0125,
+      "step": 49345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4262623187598175e-05,
+      "loss": 3.9078,
+      "step": 49350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.423720945712736e-05,
+      "loss": 3.975,
+      "step": 49355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.421180176254915e-05,
+      "loss": 3.9813,
+      "step": 49360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.418640010531368e-05,
+      "loss": 3.9172,
+      "step": 49365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.416100448687056e-05,
+      "loss": 3.8859,
+      "step": 49370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.41356149086691e-05,
+      "loss": 3.9094,
+      "step": 49375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.411023137215835e-05,
+      "loss": 3.9328,
+      "step": 49380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4084853878786925e-05,
+      "loss": 3.9703,
+      "step": 49385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.405948243000311e-05,
+      "loss": 3.9547,
+      "step": 49390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.403411702725481e-05,
+      "loss": 3.9562,
+      "step": 49395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.400875767198973e-05,
+      "loss": 3.8906,
+      "step": 49400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.398340436565508e-05,
+      "loss": 3.9234,
+      "step": 49405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.395805710969778e-05,
+      "loss": 3.9047,
+      "step": 49410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.393271590556435e-05,
+      "loss": 3.9141,
+      "step": 49415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.39073807547011e-05,
+      "loss": 3.9531,
+      "step": 49420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.388205165855389e-05,
+      "loss": 3.925,
+      "step": 49425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3856728618568164e-05,
+      "loss": 4.0594,
+      "step": 49430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3831411636189254e-05,
+      "loss": 4.0719,
+      "step": 49435
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3806100712861906e-05,
+      "loss": 4.0219,
+      "step": 49440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3780795850030664e-05,
+      "loss": 3.9234,
+      "step": 49445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.37554970491396e-05,
+      "loss": 3.9781,
+      "step": 49450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3730204311632625e-05,
+      "loss": 3.975,
+      "step": 49455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3704917638953146e-05,
+      "loss": 3.8766,
+      "step": 49460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.367963703254422e-05,
+      "loss": 3.8984,
+      "step": 49465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.365436249384873e-05,
+      "loss": 3.9297,
+      "step": 49470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3629094024309035e-05,
+      "loss": 3.8937,
+      "step": 49475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.360383162536723e-05,
+      "loss": 3.9219,
+      "step": 49480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.357857529846496e-05,
+      "loss": 3.9609,
+      "step": 49485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.355332504504371e-05,
+      "loss": 3.9516,
+      "step": 49490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.352808086654449e-05,
+      "loss": 3.9969,
+      "step": 49495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3502842764407974e-05,
+      "loss": 4.0109,
+      "step": 49500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.347761074007445e-05,
+      "loss": 3.9469,
+      "step": 49505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3452384794984e-05,
+      "loss": 3.975,
+      "step": 49510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.342716493057625e-05,
+      "loss": 3.9797,
+      "step": 49515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.340195114829043e-05,
+      "loss": 3.9453,
+      "step": 49520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.33767434495656e-05,
+      "loss": 3.9484,
+      "step": 49525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.33515418358403e-05,
+      "loss": 3.9359,
+      "step": 49530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.332634630855283e-05,
+      "loss": 3.9266,
+      "step": 49535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3301156869141006e-05,
+      "loss": 3.9406,
+      "step": 49540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.327597351904251e-05,
+      "loss": 3.8672,
+      "step": 49545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3250796259694506e-05,
+      "loss": 3.9125,
+      "step": 49550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.322562509253382e-05,
+      "loss": 4.0016,
+      "step": 49555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.320046001899706e-05,
+      "loss": 3.9141,
+      "step": 49560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3175301040520365e-05,
+      "loss": 3.9922,
+      "step": 49565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.315014815853949e-05,
+      "loss": 3.9422,
+      "step": 49570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3125001374490045e-05,
+      "loss": 4.0094,
+      "step": 49575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3099860689807065e-05,
+      "loss": 3.9703,
+      "step": 49580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3074726105925326e-05,
+      "loss": 4.0172,
+      "step": 49585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.304959762427932e-05,
+      "loss": 3.9375,
+      "step": 49590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3024475246303106e-05,
+      "loss": 3.9234,
+      "step": 49595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.299935897343038e-05,
+      "loss": 4.0062,
+      "step": 49600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2974248807094596e-05,
+      "loss": 3.9438,
+      "step": 49605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2949144748728776e-05,
+      "loss": 3.9,
+      "step": 49610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.292404679976556e-05,
+      "loss": 3.8922,
+      "step": 49615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.289895496163736e-05,
+      "loss": 3.9609,
+      "step": 49620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2873869235776146e-05,
+      "loss": 3.8797,
+      "step": 49625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2848789623613505e-05,
+      "loss": 4.0156,
+      "step": 49630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.282371612658083e-05,
+      "loss": 3.9625,
+      "step": 49635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.279864874610903e-05,
+      "loss": 3.9516,
+      "step": 49640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2773587483628644e-05,
+      "loss": 3.9594,
+      "step": 49645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.274853234057001e-05,
+      "loss": 3.9719,
+      "step": 49650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.272348331836299e-05,
+      "loss": 3.9688,
+      "step": 49655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.269844041843715e-05,
+      "loss": 3.9969,
+      "step": 49660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.267340364222161e-05,
+      "loss": 3.9609,
+      "step": 49665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.264837299114534e-05,
+      "loss": 3.8844,
+      "step": 49670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.262334846663678e-05,
+      "loss": 3.9234,
+      "step": 49675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2598330070124086e-05,
+      "loss": 3.8766,
+      "step": 49680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2573317803035014e-05,
+      "loss": 3.9312,
+      "step": 49685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.254831166679711e-05,
+      "loss": 3.9766,
+      "step": 49690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.252331166283745e-05,
+      "loss": 3.9094,
+      "step": 49695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2498317792582714e-05,
+      "loss": 3.9688,
+      "step": 49700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2473330057459396e-05,
+      "loss": 3.9047,
+      "step": 49705
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.244834845889351e-05,
+      "loss": 3.9,
+      "step": 49710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2423372998310776e-05,
+      "loss": 4.0359,
+      "step": 49715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.239840367713648e-05,
+      "loss": 4.0172,
+      "step": 49720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.23734404967957e-05,
+      "loss": 3.9562,
+      "step": 49725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.234848345871308e-05,
+      "loss": 3.9906,
+      "step": 49730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.232353256431285e-05,
+      "loss": 3.9422,
+      "step": 49735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2298587815019045e-05,
+      "loss": 3.9234,
+      "step": 49740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2273649212255244e-05,
+      "loss": 3.9656,
+      "step": 49745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2248716757444676e-05,
+      "loss": 3.9062,
+      "step": 49750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.222379045201021e-05,
+      "loss": 3.8516,
+      "step": 49755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.219887029737445e-05,
+      "loss": 3.9359,
+      "step": 49760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2173956294959586e-05,
+      "loss": 3.9359,
+      "step": 49765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2149048446187436e-05,
+      "loss": 3.9766,
+      "step": 49770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2124146752479464e-05,
+      "loss": 3.9078,
+      "step": 49775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2099251215256894e-05,
+      "loss": 3.9609,
+      "step": 49780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.207436183594046e-05,
+      "loss": 4.0016,
+      "step": 49785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.204947861595059e-05,
+      "loss": 3.9547,
+      "step": 49790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2024601556707424e-05,
+      "loss": 3.8344,
+      "step": 49795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1999730659630674e-05,
+      "loss": 3.9609,
+      "step": 49800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.197486592613972e-05,
+      "loss": 4.0609,
+      "step": 49805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.195000735765356e-05,
+      "loss": 3.9375,
+      "step": 49810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.192515495559095e-05,
+      "loss": 3.9672,
+      "step": 49815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.190030872137018e-05,
+      "loss": 3.9547,
+      "step": 49820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.187546865640918e-05,
+      "loss": 3.9547,
+      "step": 49825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.185063476212566e-05,
+      "loss": 4.0047,
+      "step": 49830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1825807039936866e-05,
+      "loss": 3.9641,
+      "step": 49835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.18009854912597e-05,
+      "loss": 3.9234,
+      "step": 49840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1776170117510696e-05,
+      "loss": 3.9516,
+      "step": 49845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.175136092010615e-05,
+      "loss": 3.9688,
+      "step": 49850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1726557900461885e-05,
+      "loss": 3.8641,
+      "step": 49855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1701761059993414e-05,
+      "loss": 3.9125,
+      "step": 49860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.167697040011585e-05,
+      "loss": 3.9531,
+      "step": 49865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1652185922244086e-05,
+      "loss": 3.9516,
+      "step": 49870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.162740762779254e-05,
+      "loss": 3.9406,
+      "step": 49875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.160263551817524e-05,
+      "loss": 3.95,
+      "step": 49880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1577869594806054e-05,
+      "loss": 3.9062,
+      "step": 49885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1553109859098306e-05,
+      "loss": 3.9484,
+      "step": 49890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1528356312465054e-05,
+      "loss": 3.8984,
+      "step": 49895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.150360895631893e-05,
+      "loss": 3.9859,
+      "step": 49900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1478867792072364e-05,
+      "loss": 3.9109,
+      "step": 49905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1454132821137274e-05,
+      "loss": 3.9609,
+      "step": 49910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.142940404492528e-05,
+      "loss": 3.9766,
+      "step": 49915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.140468146484771e-05,
+      "loss": 3.9828,
+      "step": 49920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.137996508231544e-05,
+      "loss": 3.8875,
+      "step": 49925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.135525489873907e-05,
+      "loss": 3.8484,
+      "step": 49930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.133055091552874e-05,
+      "loss": 3.9641,
+      "step": 49935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1305853134094394e-05,
+      "loss": 3.8937,
+      "step": 49940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1281161555845504e-05,
+      "loss": 4.0,
+      "step": 49945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.125647618219123e-05,
+      "loss": 3.8734,
+      "step": 49950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.123179701454031e-05,
+      "loss": 3.9453,
+      "step": 49955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.120712405430127e-05,
+      "loss": 3.9813,
+      "step": 49960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1182457302882167e-05,
+      "loss": 3.9203,
+      "step": 49965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1157796761690695e-05,
+      "loss": 3.9891,
+      "step": 49970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1133142432134305e-05,
+      "loss": 3.8797,
+      "step": 49975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.110849431561997e-05,
+      "loss": 3.9781,
+      "step": 49980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1083852413554394e-05,
+      "loss": 3.9734,
+      "step": 49985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1059216727343836e-05,
+      "loss": 3.9422,
+      "step": 49990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.103458725839433e-05,
+      "loss": 4.0,
+      "step": 49995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.100996400811144e-05,
+      "loss": 4.0391,
+      "step": 50000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.098534697790039e-05,
+      "loss": 3.9766,
+      "step": 50005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.096073616916616e-05,
+      "loss": 4.0266,
+      "step": 50010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.093613158331324e-05,
+      "loss": 3.9406,
+      "step": 50015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.091153322174582e-05,
+      "loss": 4.0109,
+      "step": 50020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.088694108586766e-05,
+      "loss": 3.8813,
+      "step": 50025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.086235517708238e-05,
+      "loss": 3.9344,
+      "step": 50030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0837775496793025e-05,
+      "loss": 3.8578,
+      "step": 50035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.081320204640229e-05,
+      "loss": 3.9984,
+      "step": 50040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.078863482731272e-05,
+      "loss": 4.0125,
+      "step": 50045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0764073840926305e-05,
+      "loss": 3.9266,
+      "step": 50050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0739519088644685e-05,
+      "loss": 3.9672,
+      "step": 50055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.071497057186931e-05,
+      "loss": 3.9422,
+      "step": 50060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0690428292001106e-05,
+      "loss": 3.9359,
+      "step": 50065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0665892250440664e-05,
+      "loss": 3.9625,
+      "step": 50070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.064136244858837e-05,
+      "loss": 3.8906,
+      "step": 50075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.061683888784405e-05,
+      "loss": 4.0172,
+      "step": 50080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0592321569607314e-05,
+      "loss": 3.95,
+      "step": 50085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.05678104952773e-05,
+      "loss": 4.0,
+      "step": 50090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.054330566625294e-05,
+      "loss": 3.9094,
+      "step": 50095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.05188070839327e-05,
+      "loss": 3.9172,
+      "step": 50100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.049431474971467e-05,
+      "loss": 3.9984,
+      "step": 50105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0469828664996703e-05,
+      "loss": 3.8297,
+      "step": 50110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.044534883117618e-05,
+      "loss": 3.8844,
+      "step": 50115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.042087524965013e-05,
+      "loss": 3.8984,
+      "step": 50120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.039640792181536e-05,
+      "loss": 3.8969,
+      "step": 50125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.037194684906817e-05,
+      "loss": 3.9375,
+      "step": 50130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.034749203280451e-05,
+      "loss": 3.9219,
+      "step": 50135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.032304347442012e-05,
+      "loss": 3.875,
+      "step": 50140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.029860117531022e-05,
+      "loss": 3.8266,
+      "step": 50145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0274165136869695e-05,
+      "loss": 3.9641,
+      "step": 50150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.02497353604932e-05,
+      "loss": 3.9422,
+      "step": 50155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.022531184757489e-05,
+      "loss": 3.9922,
+      "step": 50160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0200894599508656e-05,
+      "loss": 3.9328,
+      "step": 50165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0176483617687905e-05,
+      "loss": 3.9891,
+      "step": 50170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.015207890350588e-05,
+      "loss": 3.8219,
+      "step": 50175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0127680458355315e-05,
+      "loss": 3.9219,
+      "step": 50180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0103288283628596e-05,
+      "loss": 3.8797,
+      "step": 50185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.007890238071786e-05,
+      "loss": 3.9781,
+      "step": 50190
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0054522751014784e-05,
+      "loss": 3.9859,
+      "step": 50195
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0030149395910696e-05,
+      "loss": 3.8906,
+      "step": 50200
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.000578231679656e-05,
+      "loss": 3.9609,
+      "step": 50205
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9981421515063084e-05,
+      "loss": 3.9562,
+      "step": 50210
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.99570669921005e-05,
+      "loss": 4.0062,
+      "step": 50215
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9932718749298756e-05,
+      "loss": 3.9547,
+      "step": 50220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9908376788047294e-05,
+      "loss": 4.0078,
+      "step": 50225
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.988404110973546e-05,
+      "loss": 3.9688,
+      "step": 50230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9859711715752026e-05,
+      "loss": 3.9297,
+      "step": 50235
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.983538860748543e-05,
+      "loss": 3.9156,
+      "step": 50240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9811071786323904e-05,
+      "loss": 3.9469,
+      "step": 50245
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.978676125365511e-05,
+      "loss": 4.0719,
+      "step": 50250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9762457010866525e-05,
+      "loss": 3.9031,
+      "step": 50255
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.97381590593451e-05,
+      "loss": 3.9016,
+      "step": 50260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9713867400477634e-05,
+      "loss": 4.0563,
+      "step": 50265
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.968958203565039e-05,
+      "loss": 4.0078,
+      "step": 50270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.966530296624931e-05,
+      "loss": 3.9875,
+      "step": 50275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9641030193660074e-05,
+      "loss": 3.9516,
+      "step": 50280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9616763719267916e-05,
+      "loss": 3.8813,
+      "step": 50285
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9592503544457694e-05,
+      "loss": 3.9484,
+      "step": 50290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.956824967061392e-05,
+      "loss": 3.9641,
+      "step": 50295
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9544002099120824e-05,
+      "loss": 3.85,
+      "step": 50300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.951976083136221e-05,
+      "loss": 4.0312,
+      "step": 50305
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9495525868721493e-05,
+      "loss": 3.9875,
+      "step": 50310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.947129721258175e-05,
+      "loss": 3.9937,
+      "step": 50315
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.944707486432578e-05,
+      "loss": 3.9328,
+      "step": 50320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.942285882533593e-05,
+      "loss": 3.9578,
+      "step": 50325
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.939864909699416e-05,
+      "loss": 3.8438,
+      "step": 50330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.937444568068219e-05,
+      "loss": 3.925,
+      "step": 50335
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.93502485777813e-05,
+      "loss": 3.8672,
+      "step": 50340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9326057789672396e-05,
+      "loss": 3.9422,
+      "step": 50345
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9301873317736014e-05,
+      "loss": 3.8719,
+      "step": 50350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.927769516335244e-05,
+      "loss": 3.9172,
+      "step": 50355
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.925352332790152e-05,
+      "loss": 3.9469,
+      "step": 50360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9229357812762655e-05,
+      "loss": 3.9031,
+      "step": 50365
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.920519861931507e-05,
+      "loss": 3.9312,
+      "step": 50370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9181045748937506e-05,
+      "loss": 3.9953,
+      "step": 50375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.915689920300834e-05,
+      "loss": 3.9188,
+      "step": 50380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.91327589829056e-05,
+      "loss": 3.9312,
+      "step": 50385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9108625090007065e-05,
+      "loss": 3.9766,
+      "step": 50390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.908449752568999e-05,
+      "loss": 3.9,
+      "step": 50395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9060376291331355e-05,
+      "loss": 3.9438,
+      "step": 50400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.90362613883077e-05,
+      "loss": 3.8984,
+      "step": 50405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.901215281799537e-05,
+      "loss": 3.9484,
+      "step": 50410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.898805058177019e-05,
+      "loss": 3.9219,
+      "step": 50415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8963954681007635e-05,
+      "loss": 4.0203,
+      "step": 50420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.893986511708295e-05,
+      "loss": 3.8672,
+      "step": 50425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.891578189137089e-05,
+      "loss": 3.95,
+      "step": 50430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8891705005245864e-05,
+      "loss": 4.0219,
+      "step": 50435
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.886763446008193e-05,
+      "loss": 3.9859,
+      "step": 50440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.884357025725287e-05,
+      "loss": 3.9672,
+      "step": 50445
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.881951239813197e-05,
+      "loss": 3.9312,
+      "step": 50450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.879546088409219e-05,
+      "loss": 3.9359,
+      "step": 50455
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.877141571650625e-05,
+      "loss": 3.9781,
+      "step": 50460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.874737689674635e-05,
+      "loss": 3.9516,
+      "step": 50465
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8723344426184386e-05,
+      "loss": 3.9688,
+      "step": 50470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8699318306191846e-05,
+      "loss": 3.9219,
+      "step": 50475
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.867529853814001e-05,
+      "loss": 4.0156,
+      "step": 50480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8651285123399615e-05,
+      "loss": 3.9062,
+      "step": 50485
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.862727806334114e-05,
+      "loss": 4.0,
+      "step": 50490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.860327735933461e-05,
+      "loss": 3.9219,
+      "step": 50495
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.857928301274984e-05,
+      "loss": 3.9109,
+      "step": 50500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8555295024956145e-05,
+      "loss": 3.925,
+      "step": 50505
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8531313397322465e-05,
+      "loss": 3.9125,
+      "step": 50510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.850733813121756e-05,
+      "loss": 3.9,
+      "step": 50515
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.848336922800961e-05,
+      "loss": 3.9656,
+      "step": 50520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8459406689066544e-05,
+      "loss": 3.9156,
+      "step": 50525
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.843545051575586e-05,
+      "loss": 3.9875,
+      "step": 50530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.841150070944484e-05,
+      "loss": 3.9703,
+      "step": 50535
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8387557271500244e-05,
+      "loss": 3.9609,
+      "step": 50540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8363620203288464e-05,
+      "loss": 3.95,
+      "step": 50545
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.833968950617572e-05,
+      "loss": 3.95,
+      "step": 50550
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.831576518152768e-05,
+      "loss": 3.9453,
+      "step": 50555
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.829184723070964e-05,
+      "loss": 3.8656,
+      "step": 50560
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.826793565508672e-05,
+      "loss": 3.9453,
+      "step": 50565
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8244030456023495e-05,
+      "loss": 3.9734,
+      "step": 50570
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8220131634884196e-05,
+      "loss": 3.9828,
+      "step": 50575
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.819623919303281e-05,
+      "loss": 4.0391,
+      "step": 50580
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8172353131832844e-05,
+      "loss": 3.8922,
+      "step": 50585
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.814847345264745e-05,
+      "loss": 3.9,
+      "step": 50590
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.812460015683951e-05,
+      "loss": 3.9188,
+      "step": 50595
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.810073324577143e-05,
+      "loss": 3.9375,
+      "step": 50600
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8076872720805277e-05,
+      "loss": 3.8813,
+      "step": 50605
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8053018583302835e-05,
+      "loss": 3.8578,
+      "step": 50610
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.802917083462544e-05,
+      "loss": 3.9781,
+      "step": 50615
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.800532947613402e-05,
+      "loss": 3.9609,
+      "step": 50620
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7981494509189315e-05,
+      "loss": 3.8687,
+      "step": 50625
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7957665935151514e-05,
+      "loss": 3.9828,
+      "step": 50630
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.79338437553805e-05,
+      "loss": 3.9188,
+      "step": 50635
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.791002797123588e-05,
+      "loss": 3.9641,
+      "step": 50640
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7886218584076786e-05,
+      "loss": 4.0375,
+      "step": 50645
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7862415595262026e-05,
+      "loss": 3.9937,
+      "step": 50650
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.783861900614999e-05,
+      "loss": 3.9203,
+      "step": 50655
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.781482881809882e-05,
+      "loss": 3.9109,
+      "step": 50660
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.779104503246621e-05,
+      "loss": 3.9969,
+      "step": 50665
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7767267650609486e-05,
+      "loss": 3.9172,
+      "step": 50670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7743496673885594e-05,
+      "loss": 3.9641,
+      "step": 50675
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.771973210365124e-05,
+      "loss": 3.9969,
+      "step": 50680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.76959739412626e-05,
+      "loss": 3.9281,
+      "step": 50685
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7672222188075516e-05,
+      "loss": 4.0172,
+      "step": 50690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.764847684544561e-05,
+      "loss": 3.8578,
+      "step": 50695
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.762473791472796e-05,
+      "loss": 4.0078,
+      "step": 50700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.760100539727739e-05,
+      "loss": 3.9578,
+      "step": 50705
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7577279294448235e-05,
+      "loss": 3.9125,
+      "step": 50710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.755355960759464e-05,
+      "loss": 3.9359,
+      "step": 50715
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7529846338070263e-05,
+      "loss": 3.9922,
+      "step": 50720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.750613948722837e-05,
+      "loss": 3.9203,
+      "step": 50725
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7482439056421995e-05,
+      "loss": 4.0156,
+      "step": 50730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.745874504700368e-05,
+      "loss": 3.9969,
+      "step": 50735
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.743505746032565e-05,
+      "loss": 4.0187,
+      "step": 50740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.741137629773973e-05,
+      "loss": 3.9594,
+      "step": 50745
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.738770156059746e-05,
+      "loss": 3.8047,
+      "step": 50750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.736403325024994e-05,
+      "loss": 3.9922,
+      "step": 50755
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.734037136804791e-05,
+      "loss": 3.8672,
+      "step": 50760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.731671591534172e-05,
+      "loss": 3.9844,
+      "step": 50765
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.729306689348148e-05,
+      "loss": 3.9453,
+      "step": 50770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.726942430381678e-05,
+      "loss": 3.8641,
+      "step": 50775
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.724578814769688e-05,
+      "loss": 4.0062,
+      "step": 50780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7222158426470765e-05,
+      "loss": 3.9484,
+      "step": 50785
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7198535141486976e-05,
+      "loss": 4.0047,
+      "step": 50790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.717491829409366e-05,
+      "loss": 3.9688,
+      "step": 50795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.715130788563861e-05,
+      "loss": 4.0078,
+      "step": 50800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.712770391746936e-05,
+      "loss": 3.9359,
+      "step": 50805
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7104106390932945e-05,
+      "loss": 3.9828,
+      "step": 50810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.708051530737603e-05,
+      "loss": 3.9141,
+      "step": 50815
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7056930668145053e-05,
+      "loss": 3.9188,
+      "step": 50820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.703335247458594e-05,
+      "loss": 3.9813,
+      "step": 50825
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.700978072804429e-05,
+      "loss": 4.0156,
+      "step": 50830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.698621542986534e-05,
+      "loss": 3.9531,
+      "step": 50835
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.696265658139402e-05,
+      "loss": 3.9547,
+      "step": 50840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6939104183974815e-05,
+      "loss": 3.9469,
+      "step": 50845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6915558238951824e-05,
+      "loss": 4.0328,
+      "step": 50850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6892018747668814e-05,
+      "loss": 3.95,
+      "step": 50855
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.686848571146925e-05,
+      "loss": 3.8922,
+      "step": 50860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6844959131696125e-05,
+      "loss": 3.9953,
+      "step": 50865
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6821439009692076e-05,
+      "loss": 3.9547,
+      "step": 50870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.679792534679946e-05,
+      "loss": 3.9312,
+      "step": 50875
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.677441814436017e-05,
+      "loss": 3.95,
+      "step": 50880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.675091740371577e-05,
+      "loss": 4.0406,
+      "step": 50885
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.67274231262074e-05,
+      "loss": 3.9266,
+      "step": 50890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.670393531317598e-05,
+      "loss": 3.9125,
+      "step": 50895
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.668045396596189e-05,
+      "loss": 3.9578,
+      "step": 50900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6656979085905195e-05,
+      "loss": 3.9562,
+      "step": 50905
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6633510674345706e-05,
+      "loss": 3.9266,
+      "step": 50910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6610048732622684e-05,
+      "loss": 3.9734,
+      "step": 50915
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.658659326207514e-05,
+      "loss": 3.9609,
+      "step": 50920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.656314426404162e-05,
+      "loss": 3.9547,
+      "step": 50925
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.653970173986043e-05,
+      "loss": 3.9281,
+      "step": 50930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.651626569086943e-05,
+      "loss": 3.9203,
+      "step": 50935
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6492836118406115e-05,
+      "loss": 3.9594,
+      "step": 50940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6469413023807544e-05,
+      "loss": 3.8859,
+      "step": 50945
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6445996408410575e-05,
+      "loss": 3.9891,
+      "step": 50950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6422586273551555e-05,
+      "loss": 3.9422,
+      "step": 50955
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.639918262056645e-05,
+      "loss": 3.9469,
+      "step": 50960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.637578545079101e-05,
+      "loss": 3.9766,
+      "step": 50965
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.635239476556046e-05,
+      "loss": 4.0766,
+      "step": 50970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.632901056620971e-05,
+      "loss": 4.0078,
+      "step": 50975
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6305632854073256e-05,
+      "loss": 3.9266,
+      "step": 50980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.628226163048535e-05,
+      "loss": 4.025,
+      "step": 50985
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.625889689677976e-05,
+      "loss": 3.9688,
+      "step": 50990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.623553865428986e-05,
+      "loss": 3.9703,
+      "step": 50995
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.621218690434878e-05,
+      "loss": 3.9844,
+      "step": 51000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.618884164828919e-05,
+      "loss": 3.9281,
+      "step": 51005
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6165502887443395e-05,
+      "loss": 3.9,
+      "step": 51010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6142170623143293e-05,
+      "loss": 3.9578,
+      "step": 51015
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6118844856720535e-05,
+      "loss": 3.8813,
+      "step": 51020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.609552558950632e-05,
+      "loss": 4.0422,
+      "step": 51025
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.607221282283139e-05,
+      "loss": 3.8875,
+      "step": 51030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6048906558026325e-05,
+      "loss": 3.9937,
+      "step": 51035
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6025606796421156e-05,
+      "loss": 3.9969,
+      "step": 51040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6002313539345615e-05,
+      "loss": 3.8687,
+      "step": 51045
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.597902678812899e-05,
+      "loss": 3.9641,
+      "step": 51050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.595574654410037e-05,
+      "loss": 3.9578,
+      "step": 51055
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5932472808588296e-05,
+      "loss": 3.8563,
+      "step": 51060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.590920558292097e-05,
+      "loss": 3.975,
+      "step": 51065
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5885944868426345e-05,
+      "loss": 3.8719,
+      "step": 51070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.586269066643185e-05,
+      "loss": 3.9516,
+      "step": 51075
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.583944297826458e-05,
+      "loss": 3.8781,
+      "step": 51080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.581620180525135e-05,
+      "loss": 3.9547,
+      "step": 51085
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5792967148718525e-05,
+      "loss": 3.9766,
+      "step": 51090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.576973900999203e-05,
+      "loss": 3.9391,
+      "step": 51095
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.574651739039761e-05,
+      "loss": 3.9922,
+      "step": 51100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.572330229126047e-05,
+      "loss": 3.9359,
+      "step": 51105
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5700093713905444e-05,
+      "loss": 3.9125,
+      "step": 51110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.567689165965716e-05,
+      "loss": 3.9594,
+      "step": 51115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.565369612983971e-05,
+      "loss": 3.9641,
+      "step": 51120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5630507125776806e-05,
+      "loss": 3.9562,
+      "step": 51125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.560732464879195e-05,
+      "loss": 3.9375,
+      "step": 51130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5584148700208116e-05,
+      "loss": 4.0469,
+      "step": 51135
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5560979281347905e-05,
+      "loss": 3.9969,
+      "step": 51140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.553781639353372e-05,
+      "loss": 3.8797,
+      "step": 51145
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5514660038087376e-05,
+      "loss": 3.9813,
+      "step": 51150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.549151021633046e-05,
+      "loss": 4.0375,
+      "step": 51155
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.546836692958406e-05,
+      "loss": 3.9766,
+      "step": 51160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.544523017916905e-05,
+      "loss": 3.9844,
+      "step": 51165
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.542209996640581e-05,
+      "loss": 3.9484,
+      "step": 51170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.539897629261435e-05,
+      "loss": 3.9219,
+      "step": 51175
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5375859159114417e-05,
+      "loss": 3.9656,
+      "step": 51180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.535274856722527e-05,
+      "loss": 3.8891,
+      "step": 51185
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.532964451826582e-05,
+      "loss": 4.0141,
+      "step": 51190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.53065470135546e-05,
+      "loss": 3.95,
+      "step": 51195
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.528345605440985e-05,
+      "loss": 4.05,
+      "step": 51200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.526037164214932e-05,
+      "loss": 3.9,
+      "step": 51205
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5237293778090465e-05,
+      "loss": 3.9484,
+      "step": 51210
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5214222463550306e-05,
+      "loss": 3.9891,
+      "step": 51215
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5191157699845575e-05,
+      "loss": 4.0203,
+      "step": 51220
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.516809948829256e-05,
+      "loss": 4.0234,
+      "step": 51225
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.514504783020715e-05,
+      "loss": 3.9203,
+      "step": 51230
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5122002726905006e-05,
+      "loss": 4.0203,
+      "step": 51235
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5098964179701244e-05,
+      "loss": 3.8766,
+      "step": 51240
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.507593218991068e-05,
+      "loss": 3.9406,
+      "step": 51245
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.505290675884772e-05,
+      "loss": 3.7984,
+      "step": 51250
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.502988788782653e-05,
+      "loss": 3.9297,
+      "step": 51255
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5006875578160705e-05,
+      "loss": 4.0344,
+      "step": 51260
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.498386983116357e-05,
+      "loss": 4.05,
+      "step": 51265
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.496087064814812e-05,
+      "loss": 3.9797,
+      "step": 51270
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.493787803042689e-05,
+      "loss": 3.925,
+      "step": 51275
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.491489197931207e-05,
+      "loss": 3.8297,
+      "step": 51280
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.489191249611543e-05,
+      "loss": 3.8797,
+      "step": 51285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.486893958214849e-05,
+      "loss": 3.9375,
+      "step": 51290
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.48459732387223e-05,
+      "loss": 4.0297,
+      "step": 51295
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.482301346714751e-05,
+      "loss": 3.9234,
+      "step": 51300
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.480006026873443e-05,
+      "loss": 3.8797,
+      "step": 51305
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.477711364479309e-05,
+      "loss": 3.8766,
+      "step": 51310
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.475417359663296e-05,
+      "loss": 3.9719,
+      "step": 51315
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4731240125563256e-05,
+      "loss": 3.9828,
+      "step": 51320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.470831323289283e-05,
+      "loss": 3.925,
+      "step": 51325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.468539291993012e-05,
+      "loss": 3.9891,
+      "step": 51330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.466247918798313e-05,
+      "loss": 3.9438,
+      "step": 51335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.463957203835957e-05,
+      "loss": 3.9172,
+      "step": 51340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.461667147236681e-05,
+      "loss": 3.9781,
+      "step": 51345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4593777491311755e-05,
+      "loss": 3.9016,
+      "step": 51350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.457089009650089e-05,
+      "loss": 3.8703,
+      "step": 51355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4548009289240535e-05,
+      "loss": 3.9203,
+      "step": 51360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4525135070836425e-05,
+      "loss": 3.9766,
+      "step": 51365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4502267442594013e-05,
+      "loss": 3.9062,
+      "step": 51370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.447940640581831e-05,
+      "loss": 3.9562,
+      "step": 51375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.445655196181406e-05,
+      "loss": 3.8875,
+      "step": 51380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.443370411188554e-05,
+      "loss": 3.9391,
+      "step": 51385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4410862857336685e-05,
+      "loss": 3.8719,
+      "step": 51390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.438802819947102e-05,
+      "loss": 4.0344,
+      "step": 51395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4365200139591776e-05,
+      "loss": 3.9547,
+      "step": 51400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4342378679001714e-05,
+      "loss": 3.9344,
+      "step": 51405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4319563819003246e-05,
+      "loss": 3.9516,
+      "step": 51410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.429675556089845e-05,
+      "loss": 3.9328,
+      "step": 51415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4273953905989004e-05,
+      "loss": 4.0172,
+      "step": 51420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.425115885557617e-05,
+      "loss": 3.9094,
+      "step": 51425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.422837041096085e-05,
+      "loss": 3.9625,
+      "step": 51430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4205588573443627e-05,
+      "loss": 3.8734,
+      "step": 51435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.418281334432465e-05,
+      "loss": 3.95,
+      "step": 51440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.416004472490364e-05,
+      "loss": 3.8937,
+      "step": 51445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4137282716480125e-05,
+      "loss": 3.9094,
+      "step": 51450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4114527320353065e-05,
+      "loss": 3.9734,
+      "step": 51455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.409177853782111e-05,
+      "loss": 3.8719,
+      "step": 51460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.40690363701825e-05,
+      "loss": 3.8937,
+      "step": 51465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.404630081873522e-05,
+      "loss": 3.9766,
+      "step": 51470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4023571884776754e-05,
+      "loss": 3.9672,
+      "step": 51475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4000849569604214e-05,
+      "loss": 3.8625,
+      "step": 51480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3978133874514365e-05,
+      "loss": 3.9688,
+      "step": 51485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.395542480080365e-05,
+      "loss": 3.9875,
+      "step": 51490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.393272234976804e-05,
+      "loss": 3.9094,
+      "step": 51495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.391002652270314e-05,
+      "loss": 4.0297,
+      "step": 51500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3887337320904264e-05,
+      "loss": 3.875,
+      "step": 51505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.386465474566626e-05,
+      "loss": 3.8078,
+      "step": 51510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.384197879828361e-05,
+      "loss": 3.9891,
+      "step": 51515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3819309480050415e-05,
+      "loss": 3.95,
+      "step": 51520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.379664679226048e-05,
+      "loss": 3.9281,
+      "step": 51525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3773990736207134e-05,
+      "loss": 3.9547,
+      "step": 51530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.37513413131833e-05,
+      "loss": 3.9141,
+      "step": 51535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.372869852448169e-05,
+      "loss": 3.9234,
+      "step": 51540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3706062371394486e-05,
+      "loss": 3.9266,
+      "step": 51545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3683432855213474e-05,
+      "loss": 3.9562,
+      "step": 51550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3660809977230225e-05,
+      "loss": 4.0375,
+      "step": 51555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.363819373873578e-05,
+      "loss": 3.9688,
+      "step": 51560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3615584141020815e-05,
+      "loss": 3.9406,
+      "step": 51565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.359298118537574e-05,
+      "loss": 3.9453,
+      "step": 51570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.357038487309046e-05,
+      "loss": 4.0344,
+      "step": 51575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.354779520545451e-05,
+      "loss": 3.9406,
+      "step": 51580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3525212183757166e-05,
+      "loss": 3.9219,
+      "step": 51585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3502635809287224e-05,
+      "loss": 3.8797,
+      "step": 51590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3480066083333045e-05,
+      "loss": 3.9688,
+      "step": 51595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3457503007182796e-05,
+      "loss": 4.0484,
+      "step": 51600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.343494658212409e-05,
+      "loss": 3.8438,
+      "step": 51605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3412396809444194e-05,
+      "loss": 3.8516,
+      "step": 51610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.338985369043011e-05,
+      "loss": 4.0484,
+      "step": 51615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.336731722636834e-05,
+      "loss": 3.9594,
+      "step": 51620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.334478741854503e-05,
+      "loss": 4.0828,
+      "step": 51625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.332226426824591e-05,
+      "loss": 3.9312,
+      "step": 51630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3299747776756485e-05,
+      "loss": 3.9859,
+      "step": 51635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3277237945361724e-05,
+      "loss": 3.9844,
+      "step": 51640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.325473477534621e-05,
+      "loss": 3.8594,
+      "step": 51645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.32322382679943e-05,
+      "loss": 3.8891,
+      "step": 51650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.320974842458981e-05,
+      "loss": 3.9266,
+      "step": 51655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.318726524641627e-05,
+      "loss": 3.9562,
+      "step": 51660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.316478873475672e-05,
+      "loss": 3.9875,
+      "step": 51665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.314231889089401e-05,
+      "loss": 3.9094,
+      "step": 51670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.311985571611043e-05,
+      "loss": 3.9469,
+      "step": 51675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3097399211687936e-05,
+      "loss": 3.9578,
+      "step": 51680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3074949378908186e-05,
+      "loss": 3.9312,
+      "step": 51685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3052506219052375e-05,
+      "loss": 3.9562,
+      "step": 51690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.303006973340131e-05,
+      "loss": 4.0141,
+      "step": 51695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3007639923235425e-05,
+      "loss": 3.9922,
+      "step": 51700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.298521678983486e-05,
+      "loss": 4.0328,
+      "step": 51705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2962800334479266e-05,
+      "loss": 3.9547,
+      "step": 51710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2940390558447914e-05,
+      "loss": 3.9328,
+      "step": 51715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.291798746301982e-05,
+      "loss": 3.9156,
+      "step": 51720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2895591049473466e-05,
+      "loss": 3.8469,
+      "step": 51725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.287320131908706e-05,
+      "loss": 3.8875,
+      "step": 51730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2850818273138306e-05,
+      "loss": 3.85,
+      "step": 51735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.28284419129047e-05,
+      "loss": 3.9031,
+      "step": 51740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.280607223966324e-05,
+      "loss": 3.925,
+      "step": 51745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.278370925469053e-05,
+      "loss": 3.9953,
+      "step": 51750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2761352959262824e-05,
+      "loss": 3.9875,
+      "step": 51755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.273900335465605e-05,
+      "loss": 3.9297,
+      "step": 51760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2716660442145685e-05,
+      "loss": 3.9516,
+      "step": 51765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.269432422300678e-05,
+      "loss": 3.9562,
+      "step": 51770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.267199469851416e-05,
+      "loss": 3.9781,
+      "step": 51775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2649671869942126e-05,
+      "loss": 3.9594,
+      "step": 51780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2627355738564655e-05,
+      "loss": 4.0109,
+      "step": 51785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2605046305655264e-05,
+      "loss": 3.9844,
+      "step": 51790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.258274357248727e-05,
+      "loss": 3.9016,
+      "step": 51795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.256044754033343e-05,
+      "loss": 3.9656,
+      "step": 51800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2538158210466144e-05,
+      "loss": 3.9047,
+      "step": 51805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2515875584157556e-05,
+      "loss": 3.9469,
+      "step": 51810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.249359966267928e-05,
+      "loss": 3.9406,
+      "step": 51815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.247133044730261e-05,
+      "loss": 3.8859,
+      "step": 51820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.244906793929842e-05,
+      "loss": 3.9203,
+      "step": 51825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2426812139937316e-05,
+      "loss": 3.9703,
+      "step": 51830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.240456305048939e-05,
+      "loss": 4.0062,
+      "step": 51835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.23823206722244e-05,
+      "loss": 3.9328,
+      "step": 51840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.236008500641169e-05,
+      "loss": 3.9172,
+      "step": 51845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.233785605432031e-05,
+      "loss": 3.925,
+      "step": 51850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.231563381721886e-05,
+      "loss": 3.8,
+      "step": 51855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2293418296375494e-05,
+      "loss": 3.9156,
+      "step": 51860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2271209493058165e-05,
+      "loss": 3.9984,
+      "step": 51865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.224900740853426e-05,
+      "loss": 3.8609,
+      "step": 51870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2226812044070874e-05,
+      "loss": 3.9828,
+      "step": 51875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2204623400934646e-05,
+      "loss": 3.8906,
+      "step": 51880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.218244148039199e-05,
+      "loss": 3.9469,
+      "step": 51885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.216026628370875e-05,
+      "loss": 3.875,
+      "step": 51890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.213809781215047e-05,
+      "loss": 3.9484,
+      "step": 51895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.211593606698235e-05,
+      "loss": 3.925,
+      "step": 51900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2093781049469146e-05,
+      "loss": 4.0344,
+      "step": 51905
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.207163276087523e-05,
+      "loss": 3.9234,
+      "step": 51910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.204949120246459e-05,
+      "loss": 4.0359,
+      "step": 51915
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2027356375500904e-05,
+      "loss": 3.9719,
+      "step": 51920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2005228281247376e-05,
+      "loss": 4.0094,
+      "step": 51925
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1983106920966874e-05,
+      "loss": 3.9922,
+      "step": 51930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.196099229592179e-05,
+      "loss": 3.9328,
+      "step": 51935
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.193888440737431e-05,
+      "loss": 3.9891,
+      "step": 51940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.191678325658613e-05,
+      "loss": 3.9594,
+      "step": 51945
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.189468884481844e-05,
+      "loss": 4.0016,
+      "step": 51950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.187260117333233e-05,
+      "loss": 4.0141,
+      "step": 51955
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1850520243388274e-05,
+      "loss": 3.9953,
+      "step": 51960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.182844605624642e-05,
+      "loss": 3.9359,
+      "step": 51965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1806378613166514e-05,
+      "loss": 4.0125,
+      "step": 51970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1784317915408035e-05,
+      "loss": 4.0234,
+      "step": 51975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.176226396422995e-05,
+      "loss": 3.9516,
+      "step": 51980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.174021676089084e-05,
+      "loss": 3.9656,
+      "step": 51985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1718176306649e-05,
+      "loss": 3.95,
+      "step": 51990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.169614260276226e-05,
+      "loss": 3.9578,
+      "step": 51995
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.167411565048808e-05,
+      "loss": 4.0094,
+      "step": 52000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.165209545108351e-05,
+      "loss": 3.9562,
+      "step": 52005
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1630082005805325e-05,
+      "loss": 4.0203,
+      "step": 52010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.160807531590976e-05,
+      "loss": 4.0203,
+      "step": 52015
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.158607538265279e-05,
+      "loss": 3.9078,
+      "step": 52020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.156408220728987e-05,
+      "loss": 3.9453,
+      "step": 52025
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1542095791076254e-05,
+      "loss": 3.9688,
+      "step": 52030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1520116135266676e-05,
+      "loss": 3.9109,
+      "step": 52035
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1498143241115464e-05,
+      "loss": 3.9984,
+      "step": 52040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.14761771098767e-05,
+      "loss": 3.85,
+      "step": 52045
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1454217742803945e-05,
+      "loss": 3.8984,
+      "step": 52050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.143226514115038e-05,
+      "loss": 3.9141,
+      "step": 52055
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.141031930616894e-05,
+      "loss": 3.9625,
+      "step": 52060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.138838023911202e-05,
+      "loss": 3.9844,
+      "step": 52065
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.136644794123165e-05,
+      "loss": 3.8969,
+      "step": 52070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.134452241377959e-05,
+      "loss": 3.9953,
+      "step": 52075
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.132260365800709e-05,
+      "loss": 3.9766,
+      "step": 52080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.130069167516501e-05,
+      "loss": 3.9484,
+      "step": 52085
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1278786466503956e-05,
+      "loss": 3.9469,
+      "step": 52090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1256888033274025e-05,
+      "loss": 3.9484,
+      "step": 52095
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.123499637672491e-05,
+      "loss": 3.9375,
+      "step": 52100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.121311149810606e-05,
+      "loss": 3.9562,
+      "step": 52105
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1191233398666395e-05,
+      "loss": 4.0125,
+      "step": 52110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.116936207965448e-05,
+      "loss": 3.8531,
+      "step": 52115
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1147497542318575e-05,
+      "loss": 3.9625,
+      "step": 52120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1125639787906455e-05,
+      "loss": 3.9312,
+      "step": 52125
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.110378881766551e-05,
+      "loss": 3.9219,
+      "step": 52130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.108194463284286e-05,
+      "loss": 4.0438,
+      "step": 52135
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1060107234685105e-05,
+      "loss": 3.9688,
+      "step": 52140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.103827662443852e-05,
+      "loss": 3.9406,
+      "step": 52145
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1016452803348916e-05,
+      "loss": 4.0328,
+      "step": 52150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.099463577266189e-05,
+      "loss": 3.9344,
+      "step": 52155
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0972825533622474e-05,
+      "loss": 3.8922,
+      "step": 52160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.095102208747537e-05,
+      "loss": 3.9797,
+      "step": 52165
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.092922543546496e-05,
+      "loss": 3.8859,
+      "step": 52170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0907435578835154e-05,
+      "loss": 3.9469,
+      "step": 52175
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0885652518829496e-05,
+      "loss": 4.0906,
+      "step": 52180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.086387625669109e-05,
+      "loss": 3.9297,
+      "step": 52185
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.084210679366282e-05,
+      "loss": 3.8609,
+      "step": 52190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0820344130987026e-05,
+      "loss": 3.9719,
+      "step": 52195
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.079858826990568e-05,
+      "loss": 3.9813,
+      "step": 52200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0776839211660374e-05,
+      "loss": 3.9391,
+      "step": 52205
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.075509695749241e-05,
+      "loss": 3.9734,
+      "step": 52210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0733361508642566e-05,
+      "loss": 3.9047,
+      "step": 52215
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.071163286635126e-05,
+      "loss": 3.9391,
+      "step": 52220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.068991103185861e-05,
+      "loss": 3.9516,
+      "step": 52225
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.066819600640426e-05,
+      "loss": 3.8594,
+      "step": 52230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0646487791227486e-05,
+      "loss": 3.9484,
+      "step": 52235
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.062478638756713e-05,
+      "loss": 3.9328,
+      "step": 52240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.060309179666178e-05,
+      "loss": 3.9703,
+      "step": 52245
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0581404019749524e-05,
+      "loss": 4.0563,
+      "step": 52250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0559723058068006e-05,
+      "loss": 3.9641,
+      "step": 52255
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.053804891285467e-05,
+      "loss": 3.9453,
+      "step": 52260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.051638158534641e-05,
+      "loss": 3.9875,
+      "step": 52265
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0494721076779798e-05,
+      "loss": 3.9359,
+      "step": 52270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0473067388390943e-05,
+      "loss": 3.9516,
+      "step": 52275
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.045142052141572e-05,
+      "loss": 3.9391,
+      "step": 52280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0429780477089444e-05,
+      "loss": 4.0062,
+      "step": 52285
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0408147256647158e-05,
+      "loss": 4.0891,
+      "step": 52290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.038652086132341e-05,
+      "loss": 3.9734,
+      "step": 52295
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0364901292352494e-05,
+      "loss": 3.8547,
+      "step": 52300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0343288550968225e-05,
+      "loss": 4.0281,
+      "step": 52305
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.032168263840398e-05,
+      "loss": 3.8937,
+      "step": 52310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0300083555892895e-05,
+      "loss": 3.9391,
+      "step": 52315
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0278491304667614e-05,
+      "loss": 3.9266,
+      "step": 52320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.025690588596038e-05,
+      "loss": 3.9453,
+      "step": 52325
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0235327301003033e-05,
+      "loss": 3.9188,
+      "step": 52330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.021375555102718e-05,
+      "loss": 3.9875,
+      "step": 52335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.019219063726384e-05,
+      "loss": 3.9125,
+      "step": 52340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0170632560943713e-05,
+      "loss": 3.8906,
+      "step": 52345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0149081323297196e-05,
+      "loss": 3.9047,
+      "step": 52350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.012753692555419e-05,
+      "loss": 3.775,
+      "step": 52355
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.010599936894421e-05,
+      "loss": 3.9625,
+      "step": 52360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0084468654696375e-05,
+      "loss": 3.9453,
+      "step": 52365
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0062944784039534e-05,
+      "loss": 3.8594,
+      "step": 52370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0041427758202018e-05,
+      "loss": 3.9375,
+      "step": 52375
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.00199175784118e-05,
+      "loss": 3.9406,
+      "step": 52380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.999841424589641e-05,
+      "loss": 3.9359,
+      "step": 52385
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.997691776188316e-05,
+      "loss": 3.9109,
+      "step": 52390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9955428127598803e-05,
+      "loss": 3.9234,
+      "step": 52395
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.99339453442697e-05,
+      "loss": 3.9969,
+      "step": 52400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.991246941312197e-05,
+      "loss": 4.0203,
+      "step": 52405
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.98910003353812e-05,
+      "loss": 3.8875,
+      "step": 52410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9869538112272625e-05,
+      "loss": 3.9734,
+      "step": 52415
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.984808274502108e-05,
+      "loss": 3.9844,
+      "step": 52420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.982663423485109e-05,
+      "loss": 3.9016,
+      "step": 52425
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9805192582986688e-05,
+      "loss": 3.8687,
+      "step": 52430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9783757790651497e-05,
+      "loss": 3.9531,
+      "step": 52435
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9762329859068907e-05,
+      "loss": 3.9641,
+      "step": 52440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9740908789461755e-05,
+      "loss": 3.9547,
+      "step": 52445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9719494583052543e-05,
+      "loss": 3.9625,
+      "step": 52450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.969808724106335e-05,
+      "loss": 3.9609,
+      "step": 52455
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9676686764715963e-05,
+      "loss": 3.9562,
+      "step": 52460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9655293155231684e-05,
+      "loss": 3.9828,
+      "step": 52465
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9633906413831442e-05,
+      "loss": 3.9172,
+      "step": 52470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9612526541735726e-05,
+      "loss": 3.9719,
+      "step": 52475
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9591153540164806e-05,
+      "loss": 3.9828,
+      "step": 52480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9569787410338363e-05,
+      "loss": 3.9516,
+      "step": 52485
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9548428153475732e-05,
+      "loss": 3.9281,
+      "step": 52490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9527075770795976e-05,
+      "loss": 3.8531,
+      "step": 52495
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9505730263517652e-05,
+      "loss": 3.9516,
+      "step": 52500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9484391632858915e-05,
+      "loss": 3.9937,
+      "step": 52505
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.946305988003755e-05,
+      "loss": 3.9609,
+      "step": 52510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9441735006271038e-05,
+      "loss": 3.8859,
+      "step": 52515
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.942041701277633e-05,
+      "loss": 3.9984,
+      "step": 52520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9399105900770037e-05,
+      "loss": 4.0312,
+      "step": 52525
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.937780167146845e-05,
+      "loss": 4.0453,
+      "step": 52530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.935650432608736e-05,
+      "loss": 3.9656,
+      "step": 52535
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9335213865842183e-05,
+      "loss": 3.9391,
+      "step": 52540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9313930291948028e-05,
+      "loss": 4.0297,
+      "step": 52545
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9292653605619528e-05,
+      "loss": 3.925,
+      "step": 52550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9271383808070888e-05,
+      "loss": 3.9703,
+      "step": 52555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.925012090051609e-05,
+      "loss": 3.9391,
+      "step": 52560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.922886488416853e-05,
+      "loss": 3.8734,
+      "step": 52565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.920761576024127e-05,
+      "loss": 3.9734,
+      "step": 52570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9186373529947094e-05,
+      "loss": 3.9688,
+      "step": 52575
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9165138194498226e-05,
+      "loss": 3.9562,
+      "step": 52580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9143909755106588e-05,
+      "loss": 3.9922,
+      "step": 52585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.912268821298365e-05,
+      "loss": 3.9891,
+      "step": 52590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9101473569340604e-05,
+      "loss": 3.9297,
+      "step": 52595
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9080265825388143e-05,
+      "loss": 3.9953,
+      "step": 52600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9059064982336533e-05,
+      "loss": 3.8781,
+      "step": 52605
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9037871041395787e-05,
+      "loss": 3.9281,
+      "step": 52610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9016684003775442e-05,
+      "loss": 3.9375,
+      "step": 52615
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8995503870684562e-05,
+      "loss": 3.9234,
+      "step": 52620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.897433064333201e-05,
+      "loss": 3.9344,
+      "step": 52625
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8953164322926077e-05,
+      "loss": 3.9703,
+      "step": 52630
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8932004910674704e-05,
+      "loss": 3.9109,
+      "step": 52635
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.891085240778555e-05,
+      "loss": 3.9109,
+      "step": 52640
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.888970681546573e-05,
+      "loss": 4.0016,
+      "step": 52645
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8868568134922027e-05,
+      "loss": 3.8766,
+      "step": 52650
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8847436367360794e-05,
+      "loss": 3.9203,
+      "step": 52655
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8826311513988116e-05,
+      "loss": 3.9359,
+      "step": 52660
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8805193576009527e-05,
+      "loss": 3.9094,
+      "step": 52665
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.87840825546302e-05,
+      "loss": 3.8672,
+      "step": 52670
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8762978451055018e-05,
+      "loss": 3.9734,
+      "step": 52675
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.874188126648837e-05,
+      "loss": 3.9453,
+      "step": 52680
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.872079100213426e-05,
+      "loss": 3.8406,
+      "step": 52685
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.869970765919627e-05,
+      "loss": 3.9375,
+      "step": 52690
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8678631238877726e-05,
+      "loss": 3.9641,
+      "step": 52695
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8657561742381398e-05,
+      "loss": 3.975,
+      "step": 52700
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8636499170909687e-05,
+      "loss": 3.9516,
+      "step": 52705
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.861544352566474e-05,
+      "loss": 4.0484,
+      "step": 52710
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8594394807848138e-05,
+      "loss": 3.9359,
+      "step": 52715
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8573353018661137e-05,
+      "loss": 4.0281,
+      "step": 52720
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8552318159304583e-05,
+      "loss": 3.9328,
+      "step": 52725
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8531290230978975e-05,
+      "loss": 3.9953,
+      "step": 52730
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8510269234884377e-05,
+      "loss": 3.9359,
+      "step": 52735
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.848925517222042e-05,
+      "loss": 3.9453,
+      "step": 52740
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.846824804418637e-05,
+      "loss": 3.975,
+      "step": 52745
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.844724785198118e-05,
+      "loss": 3.8844,
+      "step": 52750
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8426254596803277e-05,
+      "loss": 4.0094,
+      "step": 52755
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8405268279850734e-05,
+      "loss": 3.8953,
+      "step": 52760
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.838428890232131e-05,
+      "loss": 3.9844,
+      "step": 52765
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.836331646541225e-05,
+      "loss": 3.9078,
+      "step": 52770
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.834235097032046e-05,
+      "loss": 3.8672,
+      "step": 52775
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.83213924182424e-05,
+      "loss": 3.9047,
+      "step": 52780
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8300440810374263e-05,
+      "loss": 3.9719,
+      "step": 52785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8279496147911727e-05,
+      "loss": 3.9391,
+      "step": 52790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.825855843205004e-05,
+      "loss": 3.8859,
+      "step": 52795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8237627663984234e-05,
+      "loss": 4.0484,
+      "step": 52800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.821670384490875e-05,
+      "loss": 4.0547,
+      "step": 52805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8195786976017742e-05,
+      "loss": 4.0062,
+      "step": 52810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.817487705850489e-05,
+      "loss": 3.9828,
+      "step": 52815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.815397409356361e-05,
+      "loss": 3.8734,
+      "step": 52820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8133078082386788e-05,
+      "loss": 3.95,
+      "step": 52825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8112189026166963e-05,
+      "loss": 3.8969,
+      "step": 52830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.809130692609623e-05,
+      "loss": 3.8922,
+      "step": 52835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.807043178336643e-05,
+      "loss": 3.9172,
+      "step": 52840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.804956359916885e-05,
+      "loss": 3.9016,
+      "step": 52845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8028702374694416e-05,
+      "loss": 3.9734,
+      "step": 52850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8007848111133746e-05,
+      "loss": 3.9438,
+      "step": 52855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7987000809676953e-05,
+      "loss": 3.9813,
+      "step": 52860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7966160471513816e-05,
+      "loss": 3.9984,
+      "step": 52865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7945327097833636e-05,
+      "loss": 3.9969,
+      "step": 52870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7924500689825445e-05,
+      "loss": 3.9547,
+      "step": 52875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7903681248677794e-05,
+      "loss": 3.8516,
+      "step": 52880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7882868775578816e-05,
+      "loss": 3.9266,
+      "step": 52885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7862063271716322e-05,
+      "loss": 4.0109,
+      "step": 52890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7841264738277687e-05,
+      "loss": 3.9688,
+      "step": 52895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7820473176449853e-05,
+      "loss": 4.0187,
+      "step": 52900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7799688587419376e-05,
+      "loss": 3.9312,
+      "step": 52905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7778910972372492e-05,
+      "loss": 3.9641,
+      "step": 52910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.775814033249498e-05,
+      "loss": 3.9312,
+      "step": 52915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7737376668972182e-05,
+      "loss": 3.8797,
+      "step": 52920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7716619982989064e-05,
+      "loss": 3.8578,
+      "step": 52925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7695870275730293e-05,
+      "loss": 3.8656,
+      "step": 52930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7675127548380005e-05,
+      "loss": 3.9875,
+      "step": 52935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7654391802121955e-05,
+      "loss": 3.9375,
+      "step": 52940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7633663038139632e-05,
+      "loss": 3.9531,
+      "step": 52945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.761294125761595e-05,
+      "loss": 4.0156,
+      "step": 52950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7592226461733534e-05,
+      "loss": 3.8828,
+      "step": 52955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.757151865167454e-05,
+      "loss": 3.9375,
+      "step": 52960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.755081782862082e-05,
+      "loss": 3.9141,
+      "step": 52965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7530123993753743e-05,
+      "loss": 3.8969,
+      "step": 52970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.750943714825429e-05,
+      "loss": 3.9344,
+      "step": 52975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7488757293303104e-05,
+      "loss": 3.8984,
+      "step": 52980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7468084430080384e-05,
+      "loss": 3.8734,
+      "step": 52985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7447418559765897e-05,
+      "loss": 3.9891,
+      "step": 52990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7426759683539042e-05,
+      "loss": 3.9688,
+      "step": 52995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7406107802578863e-05,
+      "loss": 3.9094,
+      "step": 53000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.738546291806396e-05,
+      "loss": 3.9172,
+      "step": 53005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.736482503117252e-05,
+      "loss": 3.9516,
+      "step": 53010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7344194143082305e-05,
+      "loss": 3.9469,
+      "step": 53015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7323570254970817e-05,
+      "loss": 3.9156,
+      "step": 53020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7302953368015017e-05,
+      "loss": 4.0172,
+      "step": 53025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7282343483391483e-05,
+      "loss": 3.7938,
+      "step": 53030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.726174060227649e-05,
+      "loss": 3.8875,
+      "step": 53035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.724114472584579e-05,
+      "loss": 3.9312,
+      "step": 53040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.722055585527479e-05,
+      "loss": 3.8641,
+      "step": 53045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7199973991738555e-05,
+      "loss": 3.875,
+      "step": 53050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.717939913641165e-05,
+      "loss": 3.9359,
+      "step": 53055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.715883129046825e-05,
+      "loss": 3.9125,
+      "step": 53060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7138270455082256e-05,
+      "loss": 4.0156,
+      "step": 53065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.711771663142704e-05,
+      "loss": 3.9203,
+      "step": 53070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.709716982067553e-05,
+      "loss": 3.95,
+      "step": 53075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7076630024000458e-05,
+      "loss": 3.9406,
+      "step": 53080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.705609724257398e-05,
+      "loss": 3.8969,
+      "step": 53085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.703557147756785e-05,
+      "loss": 3.8656,
+      "step": 53090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.701505273015358e-05,
+      "loss": 3.8109,
+      "step": 53095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6994541001502114e-05,
+      "loss": 3.9828,
+      "step": 53100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6974036292784014e-05,
+      "loss": 3.9547,
+      "step": 53105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6953538605169593e-05,
+      "loss": 4.0031,
+      "step": 53110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6933047939828605e-05,
+      "loss": 3.9516,
+      "step": 53115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.691256429793041e-05,
+      "loss": 3.925,
+      "step": 53120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6892087680644076e-05,
+      "loss": 3.9641,
+      "step": 53125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6871618089138198e-05,
+      "loss": 3.9062,
+      "step": 53130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.685115552458096e-05,
+      "loss": 3.8484,
+      "step": 53135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6830699988140126e-05,
+      "loss": 3.9672,
+      "step": 53140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6810251480983162e-05,
+      "loss": 3.9234,
+      "step": 53145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.678981000427704e-05,
+      "loss": 3.8203,
+      "step": 53150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6769375559188322e-05,
+      "loss": 3.9922,
+      "step": 53155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6748948146883277e-05,
+      "loss": 4.0031,
+      "step": 53160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.672852776852765e-05,
+      "loss": 3.8578,
+      "step": 53165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.670811442528685e-05,
+      "loss": 3.9531,
+      "step": 53170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.668770811832581e-05,
+      "loss": 3.9,
+      "step": 53175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6667308848809233e-05,
+      "loss": 3.8141,
+      "step": 53180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6646916617901232e-05,
+      "loss": 3.9734,
+      "step": 53185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6626531426765595e-05,
+      "loss": 3.9266,
+      "step": 53190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6606153276565694e-05,
+      "loss": 3.8719,
+      "step": 53195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.658578216846458e-05,
+      "loss": 3.9781,
+      "step": 53200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6565418103624792e-05,
+      "loss": 3.9312,
+      "step": 53205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6545061083208462e-05,
+      "loss": 4.0312,
+      "step": 53210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.652471110837745e-05,
+      "loss": 3.9391,
+      "step": 53215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.650436818029309e-05,
+      "loss": 3.9469,
+      "step": 53220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.648403230011637e-05,
+      "loss": 3.9578,
+      "step": 53225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.646370346900781e-05,
+      "loss": 3.8891,
+      "step": 53230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.644338168812766e-05,
+      "loss": 3.9203,
+      "step": 53235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6423066958635625e-05,
+      "loss": 3.9344,
+      "step": 53240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.640275928169107e-05,
+      "loss": 3.9562,
+      "step": 53245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6382458658453003e-05,
+      "loss": 3.7922,
+      "step": 53250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6362165090079962e-05,
+      "loss": 3.9469,
+      "step": 53255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6341878577730086e-05,
+      "loss": 4.0078,
+      "step": 53260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.63215991225611e-05,
+      "loss": 3.9016,
+      "step": 53265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6301326725730426e-05,
+      "loss": 3.8969,
+      "step": 53270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6281061388394976e-05,
+      "loss": 4.0266,
+      "step": 53275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6260803111711305e-05,
+      "loss": 3.9703,
+      "step": 53280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6240551896835503e-05,
+      "loss": 4.0172,
+      "step": 53285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6220307744923386e-05,
+      "loss": 3.8984,
+      "step": 53290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6200070657130256e-05,
+      "loss": 3.9391,
+      "step": 53295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6179840634611012e-05,
+      "loss": 3.9094,
+      "step": 53300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.615961767852024e-05,
+      "loss": 3.9062,
+      "step": 53305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6139401790012053e-05,
+      "loss": 3.9719,
+      "step": 53310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6119192970240148e-05,
+      "loss": 3.9203,
+      "step": 53315
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6098991220357834e-05,
+      "loss": 3.9375,
+      "step": 53320
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6078796541518088e-05,
+      "loss": 3.9844,
+      "step": 53325
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6058608934873388e-05,
+      "loss": 3.9484,
+      "step": 53330
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6038428401575794e-05,
+      "loss": 3.9953,
+      "step": 53335
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6018254942777105e-05,
+      "loss": 3.9484,
+      "step": 53340
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5998088559628564e-05,
+      "loss": 3.8781,
+      "step": 53345
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5977929253281086e-05,
+      "loss": 3.9531,
+      "step": 53350
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5957777024885118e-05,
+      "loss": 3.9844,
+      "step": 53355
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.593763187559082e-05,
+      "loss": 3.8563,
+      "step": 53360
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5917493806547857e-05,
+      "loss": 3.9234,
+      "step": 53365
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5897362818905493e-05,
+      "loss": 3.9297,
+      "step": 53370
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5877238913812592e-05,
+      "loss": 3.9297,
+      "step": 53375
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.585712209241767e-05,
+      "loss": 3.9281,
+      "step": 53380
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5837012355868773e-05,
+      "loss": 3.8578,
+      "step": 53385
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.581690970531354e-05,
+      "loss": 3.925,
+      "step": 53390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5796814141899274e-05,
+      "loss": 3.975,
+      "step": 53395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5776725666772835e-05,
+      "loss": 3.9328,
+      "step": 53400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.575664428108065e-05,
+      "loss": 3.9453,
+      "step": 53405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5736569985968725e-05,
+      "loss": 4.025,
+      "step": 53410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.571650278258279e-05,
+      "loss": 3.8312,
+      "step": 53415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.569644267206804e-05,
+      "loss": 3.8875,
+      "step": 53420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.567638965556925e-05,
+      "loss": 3.9062,
+      "step": 53425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5656343734230972e-05,
+      "loss": 3.9797,
+      "step": 53430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5636304909197137e-05,
+      "loss": 3.8734,
+      "step": 53435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5616273181611396e-05,
+      "loss": 3.9953,
+      "step": 53440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5596248552616904e-05,
+      "loss": 3.8297,
+      "step": 53445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5576231023356563e-05,
+      "loss": 3.9406,
+      "step": 53450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5556220594972726e-05,
+      "loss": 3.8984,
+      "step": 53455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5536217268607396e-05,
+      "loss": 3.9656,
+      "step": 53460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5516221045402107e-05,
+      "loss": 3.9359,
+      "step": 53465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.549623192649815e-05,
+      "loss": 3.9703,
+      "step": 53470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.547624991303626e-05,
+      "loss": 3.9094,
+      "step": 53475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.545627500615676e-05,
+      "loss": 3.9297,
+      "step": 53480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5436307206999722e-05,
+      "loss": 3.9219,
+      "step": 53485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5416346516704635e-05,
+      "loss": 4.0016,
+      "step": 53490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5396392936410696e-05,
+      "loss": 3.9281,
+      "step": 53495
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.537644646725659e-05,
+      "loss": 3.9859,
+      "step": 53500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5356507110380762e-05,
+      "loss": 3.9797,
+      "step": 53505
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5336574866921094e-05,
+      "loss": 3.9156,
+      "step": 53510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5316649738015106e-05,
+      "loss": 3.9219,
+      "step": 53515
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5296731724799984e-05,
+      "loss": 3.9125,
+      "step": 53520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5276820828412425e-05,
+      "loss": 3.9438,
+      "step": 53525
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.52569170499887e-05,
+      "loss": 3.9328,
+      "step": 53530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5237020390664786e-05,
+      "loss": 4.0187,
+      "step": 53535
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5217130851576173e-05,
+      "loss": 3.9672,
+      "step": 53540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5197248433857954e-05,
+      "loss": 4.0391,
+      "step": 53545
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.517737313864477e-05,
+      "loss": 4.0,
+      "step": 53550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5157504967070996e-05,
+      "loss": 3.9578,
+      "step": 53555
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.513764392027046e-05,
+      "loss": 3.9797,
+      "step": 53560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5117789999376592e-05,
+      "loss": 3.8859,
+      "step": 53565
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5097943205522546e-05,
+      "loss": 3.9656,
+      "step": 53570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5078103539840956e-05,
+      "loss": 3.8359,
+      "step": 53575
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5058271003463993e-05,
+      "loss": 3.9359,
+      "step": 53580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.503844559752362e-05,
+      "loss": 4.0297,
+      "step": 53585
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5018627323151203e-05,
+      "loss": 3.9406,
+      "step": 53590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4998816181477767e-05,
+      "loss": 3.8922,
+      "step": 53595
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4979012173633987e-05,
+      "loss": 3.8953,
+      "step": 53600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.495921530075005e-05,
+      "loss": 3.9937,
+      "step": 53605
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4939425563955734e-05,
+      "loss": 3.9297,
+      "step": 53610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4919642964380515e-05,
+      "loss": 3.9484,
+      "step": 53615
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.489986750315334e-05,
+      "loss": 3.9734,
+      "step": 53620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4880099181402765e-05,
+      "loss": 3.8937,
+      "step": 53625
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.486033800025704e-05,
+      "loss": 3.9062,
+      "step": 53630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4840583960843907e-05,
+      "loss": 3.8719,
+      "step": 53635
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4820837064290745e-05,
+      "loss": 3.9828,
+      "step": 53640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.480109731172445e-05,
+      "loss": 3.9422,
+      "step": 53645
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.478136470427164e-05,
+      "loss": 3.9031,
+      "step": 53650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4761639243058446e-05,
+      "loss": 4.0203,
+      "step": 53655
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.474192092921055e-05,
+      "loss": 3.9625,
+      "step": 53660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4722209763853356e-05,
+      "loss": 3.9219,
+      "step": 53665
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4702505748111743e-05,
+      "loss": 4.0234,
+      "step": 53670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4682808883110234e-05,
+      "loss": 3.9328,
+      "step": 53675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4663119169972854e-05,
+      "loss": 3.9094,
+      "step": 53680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.464343660982343e-05,
+      "loss": 3.9234,
+      "step": 53685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.462376120378516e-05,
+      "loss": 3.9062,
+      "step": 53690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4604092952980908e-05,
+      "loss": 3.9734,
+      "step": 53695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4584431858533215e-05,
+      "loss": 3.9594,
+      "step": 53700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.45647779215641e-05,
+      "loss": 3.9562,
+      "step": 53705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.454513114319523e-05,
+      "loss": 3.9891,
+      "step": 53710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.452549152454779e-05,
+      "loss": 3.875,
+      "step": 53715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.450585906674269e-05,
+      "loss": 3.9562,
+      "step": 53720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4486233770900328e-05,
+      "loss": 3.9984,
+      "step": 53725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4466615638140723e-05,
+      "loss": 3.8734,
+      "step": 53730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4447004669583454e-05,
+      "loss": 3.9156,
+      "step": 53735
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4427400866347785e-05,
+      "loss": 3.9422,
+      "step": 53740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.440780422955246e-05,
+      "loss": 3.9375,
+      "step": 53745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4388214760315845e-05,
+      "loss": 3.9953,
+      "step": 53750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4368632459755965e-05,
+      "loss": 3.8703,
+      "step": 53755
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.434905732899036e-05,
+      "loss": 3.9297,
+      "step": 53760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.432948936913619e-05,
+      "loss": 3.8672,
+      "step": 53765
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4309928581310156e-05,
+      "loss": 3.925,
+      "step": 53770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4290374966628652e-05,
+      "loss": 3.9109,
+      "step": 53775
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.42708285262076e-05,
+      "loss": 3.9297,
+      "step": 53780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4251289261162456e-05,
+      "loss": 3.9297,
+      "step": 53785
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.423175717260841e-05,
+      "loss": 4.0469,
+      "step": 53790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4212232261660132e-05,
+      "loss": 3.9688,
+      "step": 53795
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4192714529431884e-05,
+      "loss": 3.9484,
+      "step": 53800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.417320397703755e-05,
+      "loss": 3.9578,
+      "step": 53805
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4153700605590633e-05,
+      "loss": 3.9641,
+      "step": 53810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.413420441620416e-05,
+      "loss": 3.9391,
+      "step": 53815
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4114715409990796e-05,
+      "loss": 3.9359,
+      "step": 53820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.409523358806275e-05,
+      "loss": 3.9547,
+      "step": 53825
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4075758951531916e-05,
+      "loss": 3.9062,
+      "step": 53830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4056291501509666e-05,
+      "loss": 3.9047,
+      "step": 53835
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4036831239106984e-05,
+      "loss": 3.9875,
+      "step": 53840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.401737816543453e-05,
+      "loss": 3.875,
+      "step": 53845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3997932281602467e-05,
+      "loss": 3.9688,
+      "step": 53850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.397849358872059e-05,
+      "loss": 3.9859,
+      "step": 53855
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.395906208789819e-05,
+      "loss": 3.9078,
+      "step": 53860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3939637780244337e-05,
+      "loss": 3.9203,
+      "step": 53865
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3920220666867508e-05,
+      "loss": 3.9922,
+      "step": 53870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3900810748875832e-05,
+      "loss": 3.9656,
+      "step": 53875
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.388140802737709e-05,
+      "loss": 3.8828,
+      "step": 53880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3862012503478568e-05,
+      "loss": 3.9875,
+      "step": 53885
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3842624178287174e-05,
+      "loss": 3.85,
+      "step": 53890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3823243052909358e-05,
+      "loss": 3.8953,
+      "step": 53895
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.380386912845127e-05,
+      "loss": 3.9828,
+      "step": 53900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3784502406018556e-05,
+      "loss": 3.8875,
+      "step": 53905
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3765142886716487e-05,
+      "loss": 3.9266,
+      "step": 53910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3745790571649842e-05,
+      "loss": 3.9781,
+      "step": 53915
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.372644546192317e-05,
+      "loss": 3.9312,
+      "step": 53920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3707107558640426e-05,
+      "loss": 4.0125,
+      "step": 53925
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.368777686290522e-05,
+      "loss": 3.9734,
+      "step": 53930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3668453375820828e-05,
+      "loss": 4.0125,
+      "step": 53935
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3649137098489977e-05,
+      "loss": 3.8563,
+      "step": 53940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3629828032015076e-05,
+      "loss": 3.9703,
+      "step": 53945
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3610526177498053e-05,
+      "loss": 3.9891,
+      "step": 53950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3591231536040545e-05,
+      "loss": 3.9422,
+      "step": 53955
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.357194410874364e-05,
+      "loss": 3.9953,
+      "step": 53960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3552663896708073e-05,
+      "loss": 3.9031,
+      "step": 53965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3533390901034198e-05,
+      "loss": 3.875,
+      "step": 53970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3514125122821936e-05,
+      "loss": 3.9453,
+      "step": 53975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3494866563170737e-05,
+      "loss": 3.7922,
+      "step": 53980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3475615223179694e-05,
+      "loss": 4.0312,
+      "step": 53985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.345637110394754e-05,
+      "loss": 3.8719,
+      "step": 53990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.34371342065725e-05,
+      "loss": 3.9281,
+      "step": 53995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3417904532152426e-05,
+      "loss": 3.9703,
+      "step": 54000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3398682081784725e-05,
+      "loss": 3.8656,
+      "step": 54005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3379466856566502e-05,
+      "loss": 3.9906,
+      "step": 54010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3360258857594316e-05,
+      "loss": 3.9203,
+      "step": 54015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3341058085964355e-05,
+      "loss": 3.9047,
+      "step": 54020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.332186454277247e-05,
+      "loss": 3.9594,
+      "step": 54025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3302678229113996e-05,
+      "loss": 3.9219,
+      "step": 54030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3283499146083885e-05,
+      "loss": 3.9828,
+      "step": 54035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.326432729477674e-05,
+      "loss": 3.9922,
+      "step": 54040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3245162676286662e-05,
+      "loss": 3.9875,
+      "step": 54045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.322600529170735e-05,
+      "loss": 4.0094,
+      "step": 54050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3206855142132203e-05,
+      "loss": 3.9016,
+      "step": 54055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.318771222865406e-05,
+      "loss": 3.9937,
+      "step": 54060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.31685765523654e-05,
+      "loss": 4.0016,
+      "step": 54065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3149448114358325e-05,
+      "loss": 3.9922,
+      "step": 54070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3130326915724523e-05,
+      "loss": 3.9281,
+      "step": 54075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3111212957555164e-05,
+      "loss": 3.9062,
+      "step": 54080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.309210624094116e-05,
+      "loss": 3.9375,
+      "step": 54085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3073006766972908e-05,
+      "loss": 3.95,
+      "step": 54090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3053914536740382e-05,
+      "loss": 3.9422,
+      "step": 54095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3034829551333245e-05,
+      "loss": 3.8766,
+      "step": 54100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3015751811840626e-05,
+      "loss": 3.8891,
+      "step": 54105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.299668131935129e-05,
+      "loss": 3.9625,
+      "step": 54110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.297761807495363e-05,
+      "loss": 4.0,
+      "step": 54115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2958562079735575e-05,
+      "loss": 3.9688,
+      "step": 54120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2939513334784658e-05,
+      "loss": 3.9047,
+      "step": 54125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2920471841187928e-05,
+      "loss": 3.9281,
+      "step": 54130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2901437600032192e-05,
+      "loss": 4.0031,
+      "step": 54135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.288241061240367e-05,
+      "loss": 3.9906,
+      "step": 54140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2863390879388204e-05,
+      "loss": 3.9516,
+      "step": 54145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2844378402071335e-05,
+      "loss": 3.9312,
+      "step": 54150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2825373181538052e-05,
+      "loss": 3.9484,
+      "step": 54155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2806375218873013e-05,
+      "loss": 4.0281,
+      "step": 54160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2787384515160378e-05,
+      "loss": 3.9453,
+      "step": 54165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2768401071484034e-05,
+      "loss": 3.95,
+      "step": 54170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.274942488892731e-05,
+      "loss": 3.9297,
+      "step": 54175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2730455968573192e-05,
+      "loss": 4.0109,
+      "step": 54180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2711494311504215e-05,
+      "loss": 3.8109,
+      "step": 54185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.269253991880256e-05,
+      "loss": 4.0109,
+      "step": 54190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.267359279154995e-05,
+      "loss": 4.0234,
+      "step": 54195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2654652930827667e-05,
+      "loss": 3.9594,
+      "step": 54200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2635720337716663e-05,
+      "loss": 3.9203,
+      "step": 54205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.261679501329739e-05,
+      "loss": 4.0062,
+      "step": 54210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2597876958649918e-05,
+      "loss": 3.9422,
+      "step": 54215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2578966174853867e-05,
+      "loss": 3.8078,
+      "step": 54220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2560062662988575e-05,
+      "loss": 3.9578,
+      "step": 54225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2541166424132796e-05,
+      "loss": 3.9688,
+      "step": 54230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2522277459364914e-05,
+      "loss": 3.8937,
+      "step": 54235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2503395769763e-05,
+      "loss": 3.9828,
+      "step": 54240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2484521356404595e-05,
+      "loss": 3.9219,
+      "step": 54245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2465654220366868e-05,
+      "loss": 4.0625,
+      "step": 54250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.244679436272653e-05,
+      "loss": 3.9609,
+      "step": 54255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2427941784559976e-05,
+      "loss": 3.9625,
+      "step": 54260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2409096486943108e-05,
+      "loss": 3.8844,
+      "step": 54265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2390258470951423e-05,
+      "loss": 3.9234,
+      "step": 54270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2371427737659982e-05,
+      "loss": 3.9219,
+      "step": 54275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.235260428814349e-05,
+      "loss": 3.9031,
+      "step": 54280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2333788123476204e-05,
+      "loss": 4.0156,
+      "step": 54285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2314979244731916e-05,
+      "loss": 3.9891,
+      "step": 54290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.229617765298414e-05,
+      "loss": 3.8594,
+      "step": 54295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2277383349305804e-05,
+      "loss": 3.9656,
+      "step": 54300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2258596334769545e-05,
+      "loss": 3.8844,
+      "step": 54305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2239816610447504e-05,
+      "loss": 3.9266,
+      "step": 54310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2221044177411484e-05,
+      "loss": 3.9562,
+      "step": 54315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2202279036732802e-05,
+      "loss": 3.9344,
+      "step": 54320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.218352118948236e-05,
+      "loss": 3.9891,
+      "step": 54325
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.216477063673073e-05,
+      "loss": 3.8969,
+      "step": 54330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.214602737954799e-05,
+      "loss": 3.9047,
+      "step": 54335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.21272914190038e-05,
+      "loss": 3.9578,
+      "step": 54340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2108562756167413e-05,
+      "loss": 3.9562,
+      "step": 54345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2089841392107728e-05,
+      "loss": 3.9266,
+      "step": 54350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2071127327893135e-05,
+      "loss": 4.0234,
+      "step": 54355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.205242056459165e-05,
+      "loss": 3.9188,
+      "step": 54360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2033721103270853e-05,
+      "loss": 3.9,
+      "step": 54365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.201502894499797e-05,
+      "loss": 3.9328,
+      "step": 54370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1996344090839735e-05,
+      "loss": 3.9906,
+      "step": 54375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1977666541862475e-05,
+      "loss": 3.9453,
+      "step": 54380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1958996299132176e-05,
+      "loss": 3.9516,
+      "step": 54385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1940333363714312e-05,
+      "loss": 3.925,
+      "step": 54390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1921677736673988e-05,
+      "loss": 4.0047,
+      "step": 54395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1903029419075845e-05,
+      "loss": 3.9594,
+      "step": 54400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1884388411984222e-05,
+      "loss": 3.9672,
+      "step": 54405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1865754716462897e-05,
+      "loss": 4.0219,
+      "step": 54410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1847128333575303e-05,
+      "loss": 3.9797,
+      "step": 54415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1828509264384493e-05,
+      "loss": 3.9641,
+      "step": 54420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1809897509953046e-05,
+      "loss": 3.9594,
+      "step": 54425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1791293071343112e-05,
+      "loss": 4.0453,
+      "step": 54430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1772695949616415e-05,
+      "loss": 3.9562,
+      "step": 54435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1754106145834383e-05,
+      "loss": 4.0016,
+      "step": 54440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.17355236610579e-05,
+      "loss": 3.9516,
+      "step": 54445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1716948496347443e-05,
+      "loss": 3.975,
+      "step": 54450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1698380652763102e-05,
+      "loss": 3.9203,
+      "step": 54455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1679820131364588e-05,
+      "loss": 3.8594,
+      "step": 54460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1661266933211135e-05,
+      "loss": 4.0078,
+      "step": 54465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.164272105936153e-05,
+      "loss": 3.8719,
+      "step": 54470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1624182510874254e-05,
+      "loss": 3.9422,
+      "step": 54475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.160565128880727e-05,
+      "loss": 4.0125,
+      "step": 54480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1587127394218172e-05,
+      "loss": 3.9656,
+      "step": 54485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1568610828164075e-05,
+      "loss": 4.0016,
+      "step": 54490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1550101591701775e-05,
+      "loss": 3.9719,
+      "step": 54495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1531599685887578e-05,
+      "loss": 3.9078,
+      "step": 54500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1513105111777378e-05,
+      "loss": 3.9297,
+      "step": 54505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1494617870426678e-05,
+      "loss": 4.0125,
+      "step": 54510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1476137962890565e-05,
+      "loss": 3.9672,
+      "step": 54515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.145766539022365e-05,
+      "loss": 3.9797,
+      "step": 54520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.143920015348015e-05,
+      "loss": 3.9969,
+      "step": 54525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1420742253713945e-05,
+      "loss": 4.0078,
+      "step": 54530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.140229169197839e-05,
+      "loss": 3.9375,
+      "step": 54535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1383848469326426e-05,
+      "loss": 3.9641,
+      "step": 54540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1365412586810677e-05,
+      "loss": 3.95,
+      "step": 54545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.134698404548325e-05,
+      "loss": 3.9594,
+      "step": 54550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1328562846395823e-05,
+      "loss": 3.9937,
+      "step": 54555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1310148990599773e-05,
+      "loss": 4.0156,
+      "step": 54560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.129174247914596e-05,
+      "loss": 4.0109,
+      "step": 54565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1273343313084766e-05,
+      "loss": 3.9438,
+      "step": 54570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.125495149346635e-05,
+      "loss": 3.9438,
+      "step": 54575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.123656702134028e-05,
+      "loss": 4.0531,
+      "step": 54580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1218189897755716e-05,
+      "loss": 3.9656,
+      "step": 54585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1199820123761523e-05,
+      "loss": 3.9344,
+      "step": 54590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1181457700406028e-05,
+      "loss": 3.9375,
+      "step": 54595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1163102628737144e-05,
+      "loss": 3.8266,
+      "step": 54600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1144754909802465e-05,
+      "loss": 3.9266,
+      "step": 54605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.112641454464905e-05,
+      "loss": 4.0375,
+      "step": 54610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1108081534323586e-05,
+      "loss": 3.9688,
+      "step": 54615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.108975587987236e-05,
+      "loss": 3.9766,
+      "step": 54620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1071437582341227e-05,
+      "loss": 3.8984,
+      "step": 54625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.10531266427756e-05,
+      "loss": 4.0375,
+      "step": 54630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1034823062220435e-05,
+      "loss": 4.0187,
+      "step": 54635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1016526841720406e-05,
+      "loss": 3.9703,
+      "step": 54640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0998237982319636e-05,
+      "loss": 3.9703,
+      "step": 54645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0979956485061866e-05,
+      "loss": 3.8312,
+      "step": 54650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.096168235099044e-05,
+      "loss": 3.9797,
+      "step": 54655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.094341558114827e-05,
+      "loss": 3.9562,
+      "step": 54660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.092515617657783e-05,
+      "loss": 3.9406,
+      "step": 54665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0906904138321164e-05,
+      "loss": 3.9688,
+      "step": 54670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0888659467419956e-05,
+      "loss": 3.9125,
+      "step": 54675
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0870422164915424e-05,
+      "loss": 3.9594,
+      "step": 54680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0852192231848335e-05,
+      "loss": 3.9078,
+      "step": 54685
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0833969669259125e-05,
+      "loss": 3.9141,
+      "step": 54690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0815754478187747e-05,
+      "loss": 3.9547,
+      "step": 54695
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.079754665967373e-05,
+      "loss": 3.9375,
+      "step": 54700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.077934621475615e-05,
+      "loss": 4.0016,
+      "step": 54705
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0761153144473804e-05,
+      "loss": 3.9531,
+      "step": 54710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.074296744986491e-05,
+      "loss": 3.9688,
+      "step": 54715
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0724789131967345e-05,
+      "loss": 3.9391,
+      "step": 54720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0706618191818524e-05,
+      "loss": 4.0531,
+      "step": 54725
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0688454630455504e-05,
+      "loss": 3.9578,
+      "step": 54730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0670298448914867e-05,
+      "loss": 3.9984,
+      "step": 54735
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.065214964823274e-05,
+      "loss": 3.9234,
+      "step": 54740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.063400822944496e-05,
+      "loss": 4.0234,
+      "step": 54745
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.061587419358681e-05,
+      "loss": 3.975,
+      "step": 54750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0597747541693222e-05,
+      "loss": 4.0203,
+      "step": 54755
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0579628274798625e-05,
+      "loss": 3.9734,
+      "step": 54760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0561516393937182e-05,
+      "loss": 4.0047,
+      "step": 54765
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0543411900142483e-05,
+      "loss": 3.975,
+      "step": 54770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.052531479444773e-05,
+      "loss": 3.9609,
+      "step": 54775
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.05072250778858e-05,
+      "loss": 3.9516,
+      "step": 54780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.048914275148903e-05,
+      "loss": 3.825,
+      "step": 54785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.04710678162894e-05,
+      "loss": 3.9688,
+      "step": 54790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.045300027331838e-05,
+      "loss": 3.8828,
+      "step": 54795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0434940123607185e-05,
+      "loss": 4.0078,
+      "step": 54800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.041688736818645e-05,
+      "loss": 3.975,
+      "step": 54805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0398842008086475e-05,
+      "loss": 3.9516,
+      "step": 54810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0380804044337063e-05,
+      "loss": 4.0672,
+      "step": 54815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.036277347796771e-05,
+      "loss": 3.8844,
+      "step": 54820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.034475031000739e-05,
+      "loss": 3.9641,
+      "step": 54825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0326734541484652e-05,
+      "loss": 4.0391,
+      "step": 54830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.030872617342772e-05,
+      "loss": 3.9969,
+      "step": 54835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0290725206864305e-05,
+      "loss": 3.8813,
+      "step": 54840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.027273164282172e-05,
+      "loss": 3.9234,
+      "step": 54845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.025474548232683e-05,
+      "loss": 3.9641,
+      "step": 54850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.023676672640619e-05,
+      "loss": 3.8906,
+      "step": 54855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0218795376085783e-05,
+      "loss": 3.975,
+      "step": 54860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0200831432391224e-05,
+      "loss": 3.8516,
+      "step": 54865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0182874896347773e-05,
+      "loss": 3.9891,
+      "step": 54870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0164925768980178e-05,
+      "loss": 3.9516,
+      "step": 54875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.014698405131282e-05,
+      "loss": 4.0078,
+      "step": 54880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.012904974436958e-05,
+      "loss": 3.9438,
+      "step": 54885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.011112284917402e-05,
+      "loss": 3.9375,
+      "step": 54890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0093203366749245e-05,
+      "loss": 3.9703,
+      "step": 54895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.007529129811788e-05,
+      "loss": 3.9,
+      "step": 54900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0057386644302144e-05,
+      "loss": 3.95,
+      "step": 54905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.003948940632395e-05,
+      "loss": 3.9625,
+      "step": 54910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0021599585204616e-05,
+      "loss": 3.9609,
+      "step": 54915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.000371718196513e-05,
+      "loss": 3.9062,
+      "step": 54920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9985842197626067e-05,
+      "loss": 3.9109,
+      "step": 54925
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9967974633207552e-05,
+      "loss": 4.075,
+      "step": 54930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.995011448972926e-05,
+      "loss": 3.8937,
+      "step": 54935
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9932261768210472e-05,
+      "loss": 3.9141,
+      "step": 54940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9914416469670097e-05,
+      "loss": 3.8578,
+      "step": 54945
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9896578595126523e-05,
+      "loss": 3.9312,
+      "step": 54950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9878748145597735e-05,
+      "loss": 3.9031,
+      "step": 54955
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9860925122101385e-05,
+      "loss": 3.9781,
+      "step": 54960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9843109525654594e-05,
+      "loss": 3.8891,
+      "step": 54965
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9825301357274114e-05,
+      "loss": 3.8953,
+      "step": 54970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9807500617976217e-05,
+      "loss": 3.9656,
+      "step": 54975
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9789707308776876e-05,
+      "loss": 3.9484,
+      "step": 54980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.977192143069149e-05,
+      "loss": 3.9047,
+      "step": 54985
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9754142984735117e-05,
+      "loss": 3.9547,
+      "step": 54990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9736371971922345e-05,
+      "loss": 4.0125,
+      "step": 54995
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9718608393267443e-05,
+      "loss": 3.9484,
+      "step": 55000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9700852249784123e-05,
+      "loss": 3.9141,
+      "step": 55005
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9683103542485717e-05,
+      "loss": 3.8547,
+      "step": 55010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9665362272385205e-05,
+      "loss": 3.8859,
+      "step": 55015
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9647628440495057e-05,
+      "loss": 3.9516,
+      "step": 55020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9629902047827294e-05,
+      "loss": 3.9312,
+      "step": 55025
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9612183095393645e-05,
+      "loss": 3.9422,
+      "step": 55030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9594471584205286e-05,
+      "loss": 3.9781,
+      "step": 55035
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9576767515272983e-05,
+      "loss": 3.9781,
+      "step": 55040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9559070889607203e-05,
+      "loss": 3.9391,
+      "step": 55045
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9541381708217817e-05,
+      "loss": 3.8734,
+      "step": 55050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9523699972114355e-05,
+      "loss": 3.9688,
+      "step": 55055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9506025682305955e-05,
+      "loss": 3.9203,
+      "step": 55060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.948835883980126e-05,
+      "loss": 3.8937,
+      "step": 55065
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.94706994456085e-05,
+      "loss": 3.9328,
+      "step": 55070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.945304750073555e-05,
+      "loss": 3.9078,
+      "step": 55075
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9435403006189793e-05,
+      "loss": 3.8609,
+      "step": 55080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9417765962978148e-05,
+      "loss": 3.9594,
+      "step": 55085
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9400136372107235e-05,
+      "loss": 4.0578,
+      "step": 55090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.938251423458314e-05,
+      "loss": 3.8937,
+      "step": 55095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9364899551411583e-05,
+      "loss": 3.8672,
+      "step": 55100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9347292323597775e-05,
+      "loss": 3.9297,
+      "step": 55105
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9329692552146643e-05,
+      "loss": 3.9969,
+      "step": 55110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9312100238062568e-05,
+      "loss": 3.8953,
+      "step": 55115
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9294515382349528e-05,
+      "loss": 3.9703,
+      "step": 55120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9276937986011124e-05,
+      "loss": 3.9672,
+      "step": 55125
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9259368050050506e-05,
+      "loss": 3.9891,
+      "step": 55130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9241805575470332e-05,
+      "loss": 3.9047,
+      "step": 55135
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9224250563272968e-05,
+      "loss": 3.9688,
+      "step": 55140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.920670301446025e-05,
+      "loss": 3.9344,
+      "step": 55145
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9189162930033626e-05,
+      "loss": 3.9203,
+      "step": 55150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.917163031099406e-05,
+      "loss": 3.9625,
+      "step": 55155
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9154105158342215e-05,
+      "loss": 3.9266,
+      "step": 55160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9136587473078215e-05,
+      "loss": 3.9484,
+      "step": 55165
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9119077256201803e-05,
+      "loss": 3.9984,
+      "step": 55170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9101574508712253e-05,
+      "loss": 3.9266,
+      "step": 55175
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.908407923160853e-05,
+      "loss": 3.9906,
+      "step": 55180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9066591425889016e-05,
+      "loss": 4.0094,
+      "step": 55185
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.904911109255175e-05,
+      "loss": 3.9781,
+      "step": 55190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9031638232594383e-05,
+      "loss": 3.9312,
+      "step": 55195
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.901417284701408e-05,
+      "loss": 3.9641,
+      "step": 55200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8996714936807566e-05,
+      "loss": 3.925,
+      "step": 55205
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8979264502971143e-05,
+      "loss": 3.9391,
+      "step": 55210
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.896182154650076e-05,
+      "loss": 3.8625,
+      "step": 55215
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8944386068391893e-05,
+      "loss": 3.9297,
+      "step": 55220
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8926958069639524e-05,
+      "loss": 3.8687,
+      "step": 55225
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8909537551238346e-05,
+      "loss": 3.9047,
+      "step": 55230
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.889212451418251e-05,
+      "loss": 3.9453,
+      "step": 55235
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.887471895946579e-05,
+      "loss": 3.9609,
+      "step": 55240
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8857320888081473e-05,
+      "loss": 3.9359,
+      "step": 55245
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8839930301022548e-05,
+      "loss": 3.9125,
+      "step": 55250
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8822547199281472e-05,
+      "loss": 3.9594,
+      "step": 55255
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.880517158385027e-05,
+      "loss": 3.8859,
+      "step": 55260
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.878780345572058e-05,
+      "loss": 4.0219,
+      "step": 55265
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.877044281588364e-05,
+      "loss": 3.9359,
+      "step": 55270
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8753089665330196e-05,
+      "loss": 3.9125,
+      "step": 55275
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.873574400505055e-05,
+      "loss": 3.9516,
+      "step": 55280
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.87184058360347e-05,
+      "loss": 3.9203,
+      "step": 55285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8701075159272116e-05,
+      "loss": 4.0094,
+      "step": 55290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8683751975751843e-05,
+      "loss": 3.9859,
+      "step": 55295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.866643628646247e-05,
+      "loss": 3.9891,
+      "step": 55300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.864912809239231e-05,
+      "loss": 4.0406,
+      "step": 55305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8631827394529076e-05,
+      "loss": 4.0062,
+      "step": 55310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.861453419386012e-05,
+      "loss": 3.9156,
+      "step": 55315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.859724849137239e-05,
+      "loss": 4.0219,
+      "step": 55320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.857997028805238e-05,
+      "loss": 4.0016,
+      "step": 55325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8562699584886155e-05,
+      "loss": 3.8828,
+      "step": 55330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8545436382859318e-05,
+      "loss": 3.9312,
+      "step": 55335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8528180682957155e-05,
+      "loss": 3.9797,
+      "step": 55340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.851093248616441e-05,
+      "loss": 3.9906,
+      "step": 55345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8493691793465443e-05,
+      "loss": 3.9359,
+      "step": 55350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8476458605844134e-05,
+      "loss": 3.9766,
+      "step": 55355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.845923292428407e-05,
+      "loss": 3.9062,
+      "step": 55360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8442014749768285e-05,
+      "loss": 3.9125,
+      "step": 55365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8424804083279372e-05,
+      "loss": 3.9625,
+      "step": 55370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.840760092579964e-05,
+      "loss": 3.9281,
+      "step": 55375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8390405278310815e-05,
+      "loss": 3.8859,
+      "step": 55380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8373217141794273e-05,
+      "loss": 3.8969,
+      "step": 55385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8356036517230888e-05,
+      "loss": 3.9656,
+      "step": 55390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8338863405601263e-05,
+      "loss": 3.9281,
+      "step": 55395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8321697807885395e-05,
+      "loss": 3.9937,
+      "step": 55400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8304539725062917e-05,
+      "loss": 3.9219,
+      "step": 55405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8287389158113096e-05,
+      "loss": 3.9641,
+      "step": 55410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8270246108014684e-05,
+      "loss": 3.8719,
+      "step": 55415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8253110575746065e-05,
+      "loss": 3.9641,
+      "step": 55420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.823598256228509e-05,
+      "loss": 4.025,
+      "step": 55425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.821886206860936e-05,
+      "loss": 3.9203,
+      "step": 55430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8201749095695888e-05,
+      "loss": 3.9469,
+      "step": 55435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.81846436445213e-05,
+      "loss": 3.9797,
+      "step": 55440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8167545716061814e-05,
+      "loss": 3.8547,
+      "step": 55445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.815045531129325e-05,
+      "loss": 3.9297,
+      "step": 55450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8133372431190905e-05,
+      "loss": 3.9359,
+      "step": 55455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8116297076729723e-05,
+      "loss": 3.9578,
+      "step": 55460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8099229248884206e-05,
+      "loss": 3.9062,
+      "step": 55465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8082168948628424e-05,
+      "loss": 3.8828,
+      "step": 55470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8065116176935995e-05,
+      "loss": 3.8703,
+      "step": 55475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8048070934780077e-05,
+      "loss": 3.8859,
+      "step": 55480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8031033223133528e-05,
+      "loss": 3.9578,
+      "step": 55485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.801400304296865e-05,
+      "loss": 3.8453,
+      "step": 55490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7996980395257333e-05,
+      "loss": 3.8922,
+      "step": 55495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7979965280971113e-05,
+      "loss": 3.8797,
+      "step": 55500
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.796295770108102e-05,
+      "loss": 3.9406,
+      "step": 55505
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7945957656557673e-05,
+      "loss": 3.9234,
+      "step": 55510
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.792896514837125e-05,
+      "loss": 4.0438,
+      "step": 55515
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7911980177491552e-05,
+      "loss": 3.85,
+      "step": 55520
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.789500274488791e-05,
+      "loss": 4.0062,
+      "step": 55525
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.787803285152918e-05,
+      "loss": 3.925,
+      "step": 55530
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7861070498383885e-05,
+      "loss": 3.9688,
+      "step": 55535
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.784411568642007e-05,
+      "loss": 4.0703,
+      "step": 55540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.782716841660529e-05,
+      "loss": 3.9438,
+      "step": 55545
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7810228689906798e-05,
+      "loss": 3.8891,
+      "step": 55550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7793296507291326e-05,
+      "loss": 4.0172,
+      "step": 55555
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.777637186972514e-05,
+      "loss": 3.9188,
+      "step": 55560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7759454778174203e-05,
+      "loss": 3.9844,
+      "step": 55565
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7742545233603957e-05,
+      "loss": 3.9297,
+      "step": 55570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.772564323697939e-05,
+      "loss": 3.8516,
+      "step": 55575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7708748789265158e-05,
+      "loss": 3.9109,
+      "step": 55580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7691861891425418e-05,
+      "loss": 3.9172,
+      "step": 55585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.767498254442385e-05,
+      "loss": 3.8781,
+      "step": 55590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.765811074922383e-05,
+      "loss": 3.9016,
+      "step": 55595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.764124650678821e-05,
+      "loss": 3.975,
+      "step": 55600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7624389818079383e-05,
+      "loss": 3.9109,
+      "step": 55605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.760754068405946e-05,
+      "loss": 3.8391,
+      "step": 55610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7590699105689947e-05,
+      "loss": 4.0031,
+      "step": 55615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7573865083932036e-05,
+      "loss": 3.8937,
+      "step": 55620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7557038619746393e-05,
+      "loss": 3.9766,
+      "step": 55625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7540219714093366e-05,
+      "loss": 3.9516,
+      "step": 55630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7523408367932807e-05,
+      "loss": 3.9594,
+      "step": 55635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7506604582224058e-05,
+      "loss": 3.8359,
+      "step": 55640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.748980835792623e-05,
+      "loss": 3.9156,
+      "step": 55645
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7473019695997824e-05,
+      "loss": 3.9422,
+      "step": 55650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.745623859739697e-05,
+      "loss": 4.0234,
+      "step": 55655
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7439465063081348e-05,
+      "loss": 3.8797,
+      "step": 55660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.742269909400829e-05,
+      "loss": 3.9797,
+      "step": 55665
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7405940691134583e-05,
+      "loss": 3.8625,
+      "step": 55670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7389189855416614e-05,
+      "loss": 3.9156,
+      "step": 55675
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7372446587810412e-05,
+      "loss": 4.0359,
+      "step": 55680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7355710889271483e-05,
+      "loss": 3.9844,
+      "step": 55685
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7338982760754932e-05,
+      "loss": 3.9703,
+      "step": 55690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.732226220321541e-05,
+      "loss": 3.9703,
+      "step": 55695
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7305549217607233e-05,
+      "loss": 3.9094,
+      "step": 55700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.728884380488418e-05,
+      "loss": 3.9219,
+      "step": 55705
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7272145965999595e-05,
+      "loss": 3.8594,
+      "step": 55710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.725545570190644e-05,
+      "loss": 3.9547,
+      "step": 55715
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7238773013557265e-05,
+      "loss": 3.9391,
+      "step": 55720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7222097901904132e-05,
+      "loss": 3.9484,
+      "step": 55725
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.720543036789867e-05,
+      "loss": 3.9453,
+      "step": 55730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7188770412492125e-05,
+      "loss": 3.7719,
+      "step": 55735
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7172118036635287e-05,
+      "loss": 3.9094,
+      "step": 55740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.715547324127849e-05,
+      "loss": 4.0187,
+      "step": 55745
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7138836027371627e-05,
+      "loss": 3.9781,
+      "step": 55750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7122206395864256e-05,
+      "loss": 3.8578,
+      "step": 55755
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7105584347705386e-05,
+      "loss": 3.9578,
+      "step": 55760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7088969883843624e-05,
+      "loss": 3.9375,
+      "step": 55765
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7072363005227208e-05,
+      "loss": 3.9641,
+      "step": 55770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.705576371280385e-05,
+      "loss": 3.9422,
+      "step": 55775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.703917200752091e-05,
+      "loss": 3.9438,
+      "step": 55780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7022587890325224e-05,
+      "loss": 3.9719,
+      "step": 55785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.700601136216331e-05,
+      "loss": 3.9609,
+      "step": 55790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.698944242398117e-05,
+      "loss": 4.0375,
+      "step": 55795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6972881076724386e-05,
+      "loss": 4.0031,
+      "step": 55800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6956327321338086e-05,
+      "loss": 3.9312,
+      "step": 55805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6939781158767064e-05,
+      "loss": 3.9906,
+      "step": 55810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6923242589955577e-05,
+      "loss": 3.9203,
+      "step": 55815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6906711615847456e-05,
+      "loss": 3.8828,
+      "step": 55820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6890188237386164e-05,
+      "loss": 3.9453,
+      "step": 55825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6873672455514696e-05,
+      "loss": 3.9937,
+      "step": 55830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6857164271175582e-05,
+      "loss": 3.8641,
+      "step": 55835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6840663685310934e-05,
+      "loss": 3.9828,
+      "step": 55840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.68241706988625e-05,
+      "loss": 4.0047,
+      "step": 55845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6807685312771473e-05,
+      "loss": 3.9219,
+      "step": 55850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.679120752797869e-05,
+      "loss": 3.9984,
+      "step": 55855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6774737345424595e-05,
+      "loss": 3.9578,
+      "step": 55860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6758274766049083e-05,
+      "loss": 3.8594,
+      "step": 55865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6741819790791692e-05,
+      "loss": 3.9562,
+      "step": 55870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6725372420591487e-05,
+      "loss": 3.9359,
+      "step": 55875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6708932656387165e-05,
+      "loss": 3.9828,
+      "step": 55880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6692500499116922e-05,
+      "loss": 3.8969,
+      "step": 55885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6676075949718548e-05,
+      "loss": 3.9781,
+      "step": 55890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6659659009129355e-05,
+      "loss": 3.9297,
+      "step": 55895
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.664324967828633e-05,
+      "loss": 3.9641,
+      "step": 55900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6626847958125923e-05,
+      "loss": 4.0391,
+      "step": 55905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6610453849584147e-05,
+      "loss": 3.8797,
+      "step": 55910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6594067353596662e-05,
+      "loss": 3.9391,
+      "step": 55915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.657768847109865e-05,
+      "loss": 3.9641,
+      "step": 55920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6561317203024826e-05,
+      "loss": 3.9422,
+      "step": 55925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6544953550309482e-05,
+      "loss": 3.8156,
+      "step": 55930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.652859751388655e-05,
+      "loss": 3.9688,
+      "step": 55935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6512249094689435e-05,
+      "loss": 3.9828,
+      "step": 55940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6495908293651138e-05,
+      "loss": 3.9891,
+      "step": 55945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.647957511170425e-05,
+      "loss": 3.9047,
+      "step": 55950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.646324954978091e-05,
+      "loss": 3.8687,
+      "step": 55955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.644693160881282e-05,
+      "loss": 3.925,
+      "step": 55960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.64306212897312e-05,
+      "loss": 3.8781,
+      "step": 55965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6414318593466942e-05,
+      "loss": 4.0266,
+      "step": 55970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6398023520950417e-05,
+      "loss": 3.9672,
+      "step": 55975
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.638173607311158e-05,
+      "loss": 3.9703,
+      "step": 55980
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6365456250879938e-05,
+      "loss": 3.9719,
+      "step": 55985
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6349184055184632e-05,
+      "loss": 3.9781,
+      "step": 55990
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6332919486954304e-05,
+      "loss": 3.9156,
+      "step": 55995
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6316662547117143e-05,
+      "loss": 3.9422,
+      "step": 56000
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6300413236600974e-05,
+      "loss": 4.0203,
+      "step": 56005
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.628417155633312e-05,
+      "loss": 3.8828,
+      "step": 56010
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6267937507240475e-05,
+      "loss": 3.85,
+      "step": 56015
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.625171109024958e-05,
+      "loss": 3.9703,
+      "step": 56020
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6235492306286445e-05,
+      "loss": 3.9797,
+      "step": 56025
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6219281156276647e-05,
+      "loss": 3.9312,
+      "step": 56030
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.620307764114543e-05,
+      "loss": 4.0328,
+      "step": 56035
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6186881761817466e-05,
+      "loss": 3.9547,
+      "step": 56040
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6170693519217053e-05,
+      "loss": 3.8641,
+      "step": 56045
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.615451291426812e-05,
+      "loss": 3.9016,
+      "step": 56050
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.613833994789404e-05,
+      "loss": 3.9391,
+      "step": 56055
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.612217462101783e-05,
+      "loss": 3.9281,
+      "step": 56060
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6106016934562017e-05,
+      "loss": 3.9359,
+      "step": 56065
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.608986688944876e-05,
+      "loss": 3.8781,
+      "step": 56070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6073724486599727e-05,
+      "loss": 3.9719,
+      "step": 56075
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6057589726936144e-05,
+      "loss": 3.9047,
+      "step": 56080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.604146261137886e-05,
+      "loss": 3.9984,
+      "step": 56085
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6025343140848252e-05,
+      "loss": 3.9516,
+      "step": 56090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6009231316264204e-05,
+      "loss": 4.0062,
+      "step": 56095
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.599312713854628e-05,
+      "loss": 3.9828,
+      "step": 56100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5977030608613544e-05,
+      "loss": 3.9375,
+      "step": 56105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5960941727384563e-05,
+      "loss": 3.9813,
+      "step": 56110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5944860495777618e-05,
+      "loss": 3.9172,
+      "step": 56115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5928786914710423e-05,
+      "loss": 3.8578,
+      "step": 56120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.591272098510028e-05,
+      "loss": 3.9172,
+      "step": 56125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5896662707864105e-05,
+      "loss": 3.8859,
+      "step": 56130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5880612083918347e-05,
+      "loss": 3.9641,
+      "step": 56135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5864569114178993e-05,
+      "loss": 3.9125,
+      "step": 56140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.584853379956161e-05,
+      "loss": 3.9969,
+      "step": 56145
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5832506140981378e-05,
+      "loss": 4.0281,
+      "step": 56150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5816486139352968e-05,
+      "loss": 3.8609,
+      "step": 56155
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5800473795590653e-05,
+      "loss": 3.9234,
+      "step": 56160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.578446911060823e-05,
+      "loss": 3.9531,
+      "step": 56165
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5768472085319135e-05,
+      "loss": 3.9984,
+      "step": 56170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5752482720636288e-05,
+      "loss": 3.9672,
+      "step": 56175
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.573650101747218e-05,
+      "loss": 3.9531,
+      "step": 56180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.572052697673896e-05,
+      "loss": 3.9703,
+      "step": 56185
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.570456059934822e-05,
+      "loss": 3.8953,
+      "step": 56190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5688601886211176e-05,
+      "loss": 3.8937,
+      "step": 56195
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5672650838238553e-05,
+      "loss": 3.9766,
+      "step": 56200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5656707456340757e-05,
+      "loss": 3.9906,
+      "step": 56205
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5640771741427622e-05,
+      "loss": 3.8672,
+      "step": 56210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.562484369440859e-05,
+      "loss": 3.8297,
+      "step": 56215
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.560892331619272e-05,
+      "loss": 3.9328,
+      "step": 56220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5593010607688577e-05,
+      "loss": 3.825,
+      "step": 56225
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5577105569804292e-05,
+      "loss": 4.0016,
+      "step": 56230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5561208203447523e-05,
+      "loss": 3.9547,
+      "step": 56235
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5545318509525623e-05,
+      "loss": 4.0031,
+      "step": 56240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5529436488945384e-05,
+      "loss": 3.9359,
+      "step": 56245
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5513562142613168e-05,
+      "loss": 3.9016,
+      "step": 56250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.549769547143493e-05,
+      "loss": 3.9125,
+      "step": 56255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.548183647631621e-05,
+      "loss": 3.9906,
+      "step": 56260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5465985158162074e-05,
+      "loss": 3.9594,
+      "step": 56265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5450141517877128e-05,
+      "loss": 3.9109,
+      "step": 56270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.543430555636562e-05,
+      "loss": 3.9719,
+      "step": 56275
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.541847727453128e-05,
+      "loss": 3.9641,
+      "step": 56280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5402656673277437e-05,
+      "loss": 3.9953,
+      "step": 56285
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5386843753506938e-05,
+      "loss": 3.9703,
+      "step": 56290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5371038516122287e-05,
+      "loss": 3.9859,
+      "step": 56295
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5355240962025474e-05,
+      "loss": 3.9594,
+      "step": 56300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.533945109211801e-05,
+      "loss": 3.9828,
+      "step": 56305
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5323668907301113e-05,
+      "loss": 4.0016,
+      "step": 56310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5307894408475425e-05,
+      "loss": 3.8953,
+      "step": 56315
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5292127596541208e-05,
+      "loss": 3.8641,
+      "step": 56320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5276368472398248e-05,
+      "loss": 3.9734,
+      "step": 56325
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5260617036945972e-05,
+      "loss": 3.9281,
+      "step": 56330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5244873291083292e-05,
+      "loss": 3.9125,
+      "step": 56335
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5229137235708694e-05,
+      "loss": 3.8844,
+      "step": 56340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5213408871720223e-05,
+      "loss": 3.8703,
+      "step": 56345
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5197688200015545e-05,
+      "loss": 3.925,
+      "step": 56350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.518197522149181e-05,
+      "loss": 3.8672,
+      "step": 56355
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5166269937045733e-05,
+      "loss": 3.9328,
+      "step": 56360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.515057234757368e-05,
+      "loss": 4.0047,
+      "step": 56365
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.513488245397147e-05,
+      "loss": 4.0594,
+      "step": 56370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.511920025713455e-05,
+      "loss": 3.9047,
+      "step": 56375
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5103525757957857e-05,
+      "loss": 3.9578,
+      "step": 56380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5087858957336007e-05,
+      "loss": 3.9391,
+      "step": 56385
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5072199856163053e-05,
+      "loss": 3.9672,
+      "step": 56390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5056548455332663e-05,
+      "loss": 3.9094,
+      "step": 56395
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5040904755738103e-05,
+      "loss": 3.9969,
+      "step": 56400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5025268758272147e-05,
+      "loss": 3.9594,
+      "step": 56405
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5009640463827111e-05,
+      "loss": 3.9984,
+      "step": 56410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4994019873294915e-05,
+      "loss": 3.9937,
+      "step": 56415
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4978406987567065e-05,
+      "loss": 3.9469,
+      "step": 56420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4962801807534547e-05,
+      "loss": 3.9766,
+      "step": 56425
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4947204334087981e-05,
+      "loss": 3.8984,
+      "step": 56430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4931614568117455e-05,
+      "loss": 3.9391,
+      "step": 56435
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4916032510512776e-05,
+      "loss": 3.9094,
+      "step": 56440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4900458162163147e-05,
+      "loss": 3.8703,
+      "step": 56445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4884891523957377e-05,
+      "loss": 3.9328,
+      "step": 56450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4869332596783933e-05,
+      "loss": 3.8687,
+      "step": 56455
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.485378138153071e-05,
+      "loss": 3.9859,
+      "step": 56460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4838237879085229e-05,
+      "loss": 3.9781,
+      "step": 56465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4822702090334548e-05,
+      "loss": 3.9406,
+      "step": 56470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4807174016165324e-05,
+      "loss": 4.0625,
+      "step": 56475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4791653657463732e-05,
+      "loss": 3.9906,
+      "step": 56480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4776141015115479e-05,
+      "loss": 3.9359,
+      "step": 56485
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4760636090005956e-05,
+      "loss": 3.9078,
+      "step": 56490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4745138883019975e-05,
+      "loss": 3.9156,
+      "step": 56495
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4729649395041977e-05,
+      "loss": 4.0078,
+      "step": 56500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4714167626955919e-05,
+      "loss": 3.9203,
+      "step": 56505
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.46986935796454e-05,
+      "loss": 3.8953,
+      "step": 56510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4683227253993508e-05,
+      "loss": 3.9562,
+      "step": 56515
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4667768650882856e-05,
+      "loss": 3.8891,
+      "step": 56520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4652317771195753e-05,
+      "loss": 3.9531,
+      "step": 56525
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4636874615813931e-05,
+      "loss": 3.9406,
+      "step": 56530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4621439185618716e-05,
+      "loss": 3.9625,
+      "step": 56535
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.460601148149107e-05,
+      "loss": 3.9719,
+      "step": 56540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4590591504311405e-05,
+      "loss": 3.9766,
+      "step": 56545
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4575179254959734e-05,
+      "loss": 3.9281,
+      "step": 56550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4559774734315687e-05,
+      "loss": 3.8922,
+      "step": 56555
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4544377943258361e-05,
+      "loss": 3.9438,
+      "step": 56560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4528988882666454e-05,
+      "loss": 3.8859,
+      "step": 56565
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4513607553418228e-05,
+      "loss": 3.9266,
+      "step": 56570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4498233956391514e-05,
+      "loss": 3.8719,
+      "step": 56575
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4482868092463645e-05,
+      "loss": 3.9109,
+      "step": 56580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4467509962511586e-05,
+      "loss": 3.9438,
+      "step": 56585
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4452159567411837e-05,
+      "loss": 3.9406,
+      "step": 56590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4436816908040378e-05,
+      "loss": 4.0203,
+      "step": 56595
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4421481985272909e-05,
+      "loss": 3.9484,
+      "step": 56600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4406154799984548e-05,
+      "loss": 3.9641,
+      "step": 56605
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4390835353050012e-05,
+      "loss": 3.95,
+      "step": 56610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4375523645343567e-05,
+      "loss": 3.9328,
+      "step": 56615
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4360219677739115e-05,
+      "loss": 3.9375,
+      "step": 56620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4344923451110007e-05,
+      "loss": 3.9797,
+      "step": 56625
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4329634966329195e-05,
+      "loss": 3.9219,
+      "step": 56630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4314354224269248e-05,
+      "loss": 3.8609,
+      "step": 56635
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299081225802183e-05,
+      "loss": 3.9406,
+      "step": 56640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4283815971799672e-05,
+      "loss": 3.8109,
+      "step": 56645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4268558463132851e-05,
+      "loss": 3.9625,
+      "step": 56650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253308700672521e-05,
+      "loss": 3.9,
+      "step": 56655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4238066685288974e-05,
+      "loss": 3.9016,
+      "step": 56660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4222832417852025e-05,
+      "loss": 3.9625,
+      "step": 56665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4207605899231183e-05,
+      "loss": 3.9703,
+      "step": 56670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.419238713029537e-05,
+      "loss": 3.9609,
+      "step": 56675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4177176111913135e-05,
+      "loss": 4.0109,
+      "step": 56680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.416197284495254e-05,
+      "loss": 3.9969,
+      "step": 56685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4146777330281306e-05,
+      "loss": 3.9422,
+      "step": 56690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4131589568766588e-05,
+      "loss": 3.9922,
+      "step": 56695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4116409561275177e-05,
+      "loss": 4.0109,
+      "step": 56700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4101237308673362e-05,
+      "loss": 3.9937,
+      "step": 56705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4086072811827087e-05,
+      "loss": 3.9984,
+      "step": 56710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4070916071601757e-05,
+      "loss": 3.9609,
+      "step": 56715
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.405576708886233e-05,
+      "loss": 3.9125,
+      "step": 56720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4040625864473431e-05,
+      "loss": 3.8578,
+      "step": 56725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4025492399299137e-05,
+      "loss": 4.0109,
+      "step": 56730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4010366694203107e-05,
+      "loss": 3.975,
+      "step": 56735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3995248750048565e-05,
+      "loss": 3.8703,
+      "step": 56740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3980138567698323e-05,
+      "loss": 3.9609,
+      "step": 56745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3965036148014707e-05,
+      "loss": 3.9891,
+      "step": 56750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3949941491859595e-05,
+      "loss": 3.8922,
+      "step": 56755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3934854600094464e-05,
+      "loss": 3.9969,
+      "step": 56760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3919775473580324e-05,
+      "loss": 3.8797,
+      "step": 56765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3904704113177739e-05,
+      "loss": 3.9516,
+      "step": 56770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3889640519746787e-05,
+      "loss": 3.9406,
+      "step": 56775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3874584694147228e-05,
+      "loss": 3.9859,
+      "step": 56780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3859536637238245e-05,
+      "loss": 3.9547,
+      "step": 56785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3844496349878665e-05,
+      "loss": 3.9656,
+      "step": 56790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3829463832926785e-05,
+      "loss": 3.9203,
+      "step": 56795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3814439087240586e-05,
+      "loss": 4.0187,
+      "step": 56800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3799422113677484e-05,
+      "loss": 3.8797,
+      "step": 56805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.378441291309449e-05,
+      "loss": 3.9328,
+      "step": 56810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.376941148634822e-05,
+      "loss": 3.9891,
+      "step": 56815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.375441783429479e-05,
+      "loss": 3.9594,
+      "step": 56820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3739431957789899e-05,
+      "loss": 3.9344,
+      "step": 56825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3724453857688727e-05,
+      "loss": 3.9813,
+      "step": 56830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3709483534846178e-05,
+      "loss": 3.9406,
+      "step": 56835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3694520990116548e-05,
+      "loss": 4.025,
+      "step": 56840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.367956622435374e-05,
+      "loss": 3.9625,
+      "step": 56845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3664619238411267e-05,
+      "loss": 3.9344,
+      "step": 56850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3649680033142135e-05,
+      "loss": 3.9094,
+      "step": 56855
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3634748609398943e-05,
+      "loss": 3.9922,
+      "step": 56860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3619824968033778e-05,
+      "loss": 3.9219,
+      "step": 56865
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3604909109898371e-05,
+      "loss": 3.95,
+      "step": 56870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3590001035843995e-05,
+      "loss": 3.9062,
+      "step": 56875
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3575100746721419e-05,
+      "loss": 3.9484,
+      "step": 56880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3560208243380977e-05,
+      "loss": 3.8937,
+      "step": 56885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3545323526672657e-05,
+      "loss": 3.9922,
+      "step": 56890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3530446597445898e-05,
+      "loss": 3.9109,
+      "step": 56895
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3515577456549686e-05,
+      "loss": 3.9937,
+      "step": 56900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3500716104832694e-05,
+      "loss": 3.9469,
+      "step": 56905
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.348586254314299e-05,
+      "loss": 4.0187,
+      "step": 56910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3471016772328296e-05,
+      "loss": 3.8844,
+      "step": 56915
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3456178793235823e-05,
+      "loss": 3.9375,
+      "step": 56920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3441348606712438e-05,
+      "loss": 3.9422,
+      "step": 56925
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3426526213604466e-05,
+      "loss": 3.8422,
+      "step": 56930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3411711614757798e-05,
+      "loss": 3.9094,
+      "step": 56935
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3396904811017972e-05,
+      "loss": 3.9656,
+      "step": 56940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3382105803229964e-05,
+      "loss": 3.9375,
+      "step": 56945
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3367314592238366e-05,
+      "loss": 3.9953,
+      "step": 56950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3352531178887298e-05,
+      "loss": 3.9156,
+      "step": 56955
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3337755564020474e-05,
+      "loss": 3.9062,
+      "step": 56960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.332298774848115e-05,
+      "loss": 3.9859,
+      "step": 56965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3308227733112103e-05,
+      "loss": 3.9438,
+      "step": 56970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3293475518755658e-05,
+      "loss": 3.9937,
+      "step": 56975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3278731106253809e-05,
+      "loss": 3.9828,
+      "step": 56980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3263994496447966e-05,
+      "loss": 3.9828,
+      "step": 56985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3249265690179123e-05,
+      "loss": 3.9141,
+      "step": 56990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3234544688287908e-05,
+      "loss": 3.9156,
+      "step": 56995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3219831491614446e-05,
+      "loss": 3.925,
+      "step": 57000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.320512610099837e-05,
+      "loss": 3.8781,
+      "step": 57005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3190428517278972e-05,
+      "loss": 3.9109,
+      "step": 57010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3175738741295033e-05,
+      "loss": 3.8469,
+      "step": 57015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3161056773884898e-05,
+      "loss": 3.8516,
+      "step": 57020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.314638261588643e-05,
+      "loss": 3.8469,
+      "step": 57025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3131716268137143e-05,
+      "loss": 3.9047,
+      "step": 57030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3117057731474019e-05,
+      "loss": 3.9422,
+      "step": 57035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3102407006733601e-05,
+      "loss": 3.9781,
+      "step": 57040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3087764094752057e-05,
+      "loss": 3.9016,
+      "step": 57045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3073128996365017e-05,
+      "loss": 3.8859,
+      "step": 57050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3058501712407716e-05,
+      "loss": 3.8453,
+      "step": 57055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3043882243714965e-05,
+      "loss": 3.8766,
+      "step": 57060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3029270591121065e-05,
+      "loss": 3.9328,
+      "step": 57065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3014666755459902e-05,
+      "loss": 3.9156,
+      "step": 57070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3000070737564937e-05,
+      "loss": 3.9688,
+      "step": 57075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2985482538269176e-05,
+      "loss": 3.8922,
+      "step": 57080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2970902158405116e-05,
+      "loss": 4.0016,
+      "step": 57085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2956329598804944e-05,
+      "loss": 3.9,
+      "step": 57090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2941764860300241e-05,
+      "loss": 3.9672,
+      "step": 57095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2927207943722246e-05,
+      "loss": 4.0375,
+      "step": 57100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2912658849901742e-05,
+      "loss": 3.9031,
+      "step": 57105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2898117579669031e-05,
+      "loss": 4.0078,
+      "step": 57110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2883584133853953e-05,
+      "loss": 4.0312,
+      "step": 57115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2869058513286007e-05,
+      "loss": 3.8766,
+      "step": 57120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2854540718794115e-05,
+      "loss": 3.9922,
+      "step": 57125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2840030751206832e-05,
+      "loss": 3.9172,
+      "step": 57130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2825528611352192e-05,
+      "loss": 3.9641,
+      "step": 57135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2811034300057921e-05,
+      "loss": 3.8953,
+      "step": 57140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2796547818151154e-05,
+      "loss": 3.975,
+      "step": 57145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2782069166458646e-05,
+      "loss": 3.8484,
+      "step": 57150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.276759834580669e-05,
+      "loss": 3.9281,
+      "step": 57155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2753135357021154e-05,
+      "loss": 3.8969,
+      "step": 57160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.273868020092743e-05,
+      "loss": 3.9406,
+      "step": 57165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2724232878350444e-05,
+      "loss": 3.9984,
+      "step": 57170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.270979339011478e-05,
+      "loss": 3.9234,
+      "step": 57175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2695361737044452e-05,
+      "loss": 3.8516,
+      "step": 57180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2680937919963096e-05,
+      "loss": 3.9562,
+      "step": 57185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2666521939693824e-05,
+      "loss": 3.9859,
+      "step": 57190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2652113797059443e-05,
+      "loss": 3.9844,
+      "step": 57195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2637713492882178e-05,
+      "loss": 3.9234,
+      "step": 57200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.262332102798384e-05,
+      "loss": 3.9156,
+      "step": 57205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2608936403185854e-05,
+      "loss": 3.8875,
+      "step": 57210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2594559619309147e-05,
+      "loss": 3.9594,
+      "step": 57215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2580190677174162e-05,
+      "loss": 3.9875,
+      "step": 57220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.256582957760096e-05,
+      "loss": 4.0047,
+      "step": 57225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2551476321409138e-05,
+      "loss": 3.9781,
+      "step": 57230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2537130909417836e-05,
+      "loss": 3.8844,
+      "step": 57235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2522793342445736e-05,
+      "loss": 3.8906,
+      "step": 57240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2508463621311082e-05,
+      "loss": 3.8563,
+      "step": 57245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2494141746831687e-05,
+      "loss": 3.9813,
+      "step": 57250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2479827719824913e-05,
+      "loss": 3.9203,
+      "step": 57255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.246552154110761e-05,
+      "loss": 3.9844,
+      "step": 57260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2451223211496304e-05,
+      "loss": 3.9375,
+      "step": 57265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.243693273180696e-05,
+      "loss": 3.9875,
+      "step": 57270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2422650102855142e-05,
+      "loss": 3.8984,
+      "step": 57275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2408375325455933e-05,
+      "loss": 3.9391,
+      "step": 57280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2394108400424062e-05,
+      "loss": 3.9891,
+      "step": 57285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.23798493285737e-05,
+      "loss": 3.9156,
+      "step": 57290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2365598110718606e-05,
+      "loss": 3.9625,
+      "step": 57295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2351354747672133e-05,
+      "loss": 3.9,
+      "step": 57300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2337119240247129e-05,
+      "loss": 3.9062,
+      "step": 57305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.232289158925603e-05,
+      "loss": 4.0297,
+      "step": 57310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2308671795510766e-05,
+      "loss": 4.0,
+      "step": 57315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2294459859822925e-05,
+      "loss": 3.8844,
+      "step": 57320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.228025578300354e-05,
+      "loss": 3.9656,
+      "step": 57325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2266059565863263e-05,
+      "loss": 3.9578,
+      "step": 57330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2251871209212228e-05,
+      "loss": 3.9813,
+      "step": 57335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2237690713860238e-05,
+      "loss": 3.9813,
+      "step": 57340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2223518080616528e-05,
+      "loss": 3.8937,
+      "step": 57345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2209353310289904e-05,
+      "loss": 3.8922,
+      "step": 57350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2195196403688834e-05,
+      "loss": 3.9781,
+      "step": 57355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2181047361621204e-05,
+      "loss": 3.8859,
+      "step": 57360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2166906184894504e-05,
+      "loss": 3.9734,
+      "step": 57365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2152772874315736e-05,
+      "loss": 4.0172,
+      "step": 57370
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2138647430691555e-05,
+      "loss": 3.8781,
+      "step": 57375
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2124529854828086e-05,
+      "loss": 3.9703,
+      "step": 57380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2110420147530981e-05,
+      "loss": 3.7797,
+      "step": 57385
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2096318309605529e-05,
+      "loss": 3.9969,
+      "step": 57390
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2082224341856506e-05,
+      "loss": 3.9766,
+      "step": 57395
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.206813824508825e-05,
+      "loss": 3.8969,
+      "step": 57400
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2054060020104633e-05,
+      "loss": 3.9375,
+      "step": 57405
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2039989667709149e-05,
+      "loss": 3.8641,
+      "step": 57410
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.202592718870477e-05,
+      "loss": 4.0281,
+      "step": 57415
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2011872583894055e-05,
+      "loss": 3.8859,
+      "step": 57420
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1997825854079063e-05,
+      "loss": 3.8734,
+      "step": 57425
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1983787000061501e-05,
+      "loss": 3.8891,
+      "step": 57430
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1969756022642529e-05,
+      "loss": 3.9906,
+      "step": 57435
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1955732922622875e-05,
+      "loss": 4.0156,
+      "step": 57440
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1941717700802912e-05,
+      "loss": 3.9797,
+      "step": 57445
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1927710357982434e-05,
+      "loss": 3.9062,
+      "step": 57450
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1913710894960855e-05,
+      "loss": 3.975,
+      "step": 57455
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1899719312537115e-05,
+      "loss": 3.9984,
+      "step": 57460
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1885735611509728e-05,
+      "loss": 3.9875,
+      "step": 57465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1871759792676754e-05,
+      "loss": 3.975,
+      "step": 57470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1857791856835758e-05,
+      "loss": 3.9547,
+      "step": 57475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1843831804783931e-05,
+      "loss": 3.9562,
+      "step": 57480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1829879637317973e-05,
+      "loss": 3.9062,
+      "step": 57485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1815935355234113e-05,
+      "loss": 3.9359,
+      "step": 57490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.180199895932813e-05,
+      "loss": 3.9453,
+      "step": 57495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1788070450395454e-05,
+      "loss": 3.9828,
+      "step": 57500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1774149829230917e-05,
+      "loss": 4.0234,
+      "step": 57505
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1760237096628982e-05,
+      "loss": 3.8937,
+      "step": 57510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1746332253383667e-05,
+      "loss": 3.8922,
+      "step": 57515
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1732435300288534e-05,
+      "loss": 3.8625,
+      "step": 57520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1718546238136634e-05,
+      "loss": 3.9062,
+      "step": 57525
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1704665067720681e-05,
+      "loss": 3.9625,
+      "step": 57530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1690791789832826e-05,
+      "loss": 3.9703,
+      "step": 57535
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1676926405264819e-05,
+      "loss": 3.8594,
+      "step": 57540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.166306891480801e-05,
+      "loss": 3.9375,
+      "step": 57545
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.16492193192532e-05,
+      "loss": 3.9609,
+      "step": 57550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1635377619390773e-05,
+      "loss": 3.9078,
+      "step": 57555
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1621543816010731e-05,
+      "loss": 3.8828,
+      "step": 57560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1607717909902542e-05,
+      "loss": 3.8563,
+      "step": 57565
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1593899901855225e-05,
+      "loss": 3.9719,
+      "step": 57570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1580089792657416e-05,
+      "loss": 3.9797,
+      "step": 57575
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1566287583097267e-05,
+      "loss": 3.9109,
+      "step": 57580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.15524932739624e-05,
+      "loss": 3.9469,
+      "step": 57585
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1538706866040148e-05,
+      "loss": 3.9281,
+      "step": 57590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1524928360117252e-05,
+      "loss": 3.8641,
+      "step": 57595
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1511157756980067e-05,
+      "loss": 3.8969,
+      "step": 57600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1497395057414443e-05,
+      "loss": 3.9594,
+      "step": 57605
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.148364026220589e-05,
+      "loss": 3.9969,
+      "step": 57610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.146989337213936e-05,
+      "loss": 3.9734,
+      "step": 57615
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1456154387999345e-05,
+      "loss": 3.9109,
+      "step": 57620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1442423310570014e-05,
+      "loss": 3.8922,
+      "step": 57625
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1428700140634962e-05,
+      "loss": 3.9781,
+      "step": 57630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1414984878977356e-05,
+      "loss": 3.9641,
+      "step": 57635
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1401277526379927e-05,
+      "loss": 3.9969,
+      "step": 57640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1387578083624994e-05,
+      "loss": 3.9312,
+      "step": 57645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.137388655149435e-05,
+      "loss": 3.9672,
+      "step": 57650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1360202930769368e-05,
+      "loss": 3.9531,
+      "step": 57655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1346527222230994e-05,
+      "loss": 3.8312,
+      "step": 57660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1332859426659719e-05,
+      "loss": 3.9281,
+      "step": 57665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1319199544835533e-05,
+      "loss": 3.95,
+      "step": 57670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1305547577537983e-05,
+      "loss": 3.9406,
+      "step": 57675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1291903525546259e-05,
+      "loss": 3.9984,
+      "step": 57680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.127826738963899e-05,
+      "loss": 3.975,
+      "step": 57685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1264639170594386e-05,
+      "loss": 3.9438,
+      "step": 57690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1251018869190193e-05,
+      "loss": 3.9047,
+      "step": 57695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1237406486203771e-05,
+      "loss": 3.9656,
+      "step": 57700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.122380202241195e-05,
+      "loss": 3.9531,
+      "step": 57705
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1210205478591139e-05,
+      "loss": 3.8516,
+      "step": 57710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1196616855517305e-05,
+      "loss": 4.0297,
+      "step": 57715
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1183036153965958e-05,
+      "loss": 3.9688,
+      "step": 57720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1169463374712128e-05,
+      "loss": 3.9516,
+      "step": 57725
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1155898518530398e-05,
+      "loss": 3.8875,
+      "step": 57730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1142341586194981e-05,
+      "loss": 4.0281,
+      "step": 57735
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1128792578479522e-05,
+      "loss": 3.9344,
+      "step": 57740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1115251496157256e-05,
+      "loss": 4.0,
+      "step": 57745
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1101718340001014e-05,
+      "loss": 3.9406,
+      "step": 57750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1088193110783111e-05,
+      "loss": 4.0062,
+      "step": 57755
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1074675809275446e-05,
+      "loss": 3.9625,
+      "step": 57760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1061166436249419e-05,
+      "loss": 3.95,
+      "step": 57765
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1047664992476062e-05,
+      "loss": 3.8453,
+      "step": 57770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1034171478725862e-05,
+      "loss": 3.9641,
+      "step": 57775
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1020685895768933e-05,
+      "loss": 3.8891,
+      "step": 57780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1007208244374827e-05,
+      "loss": 3.8641,
+      "step": 57785
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0993738525312812e-05,
+      "loss": 3.8656,
+      "step": 57790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.098027673935154e-05,
+      "loss": 4.0094,
+      "step": 57795
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.096682288725928e-05,
+      "loss": 3.9078,
+      "step": 57800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0953376969803868e-05,
+      "loss": 3.9188,
+      "step": 57805
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0939938987752673e-05,
+      "loss": 3.9281,
+      "step": 57810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.092650894187258e-05,
+      "loss": 3.8734,
+      "step": 57815
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0913086832930012e-05,
+      "loss": 3.9031,
+      "step": 57820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0899672661691039e-05,
+      "loss": 3.9328,
+      "step": 57825
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0886266428921164e-05,
+      "loss": 3.9016,
+      "step": 57830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0872868135385477e-05,
+      "loss": 3.9641,
+      "step": 57835
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0859477781848663e-05,
+      "loss": 3.9016,
+      "step": 57840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.084609536907488e-05,
+      "loss": 3.8453,
+      "step": 57845
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0832720897827868e-05,
+      "loss": 3.9203,
+      "step": 57850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0819354368870881e-05,
+      "loss": 3.9656,
+      "step": 57855
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0805995782966808e-05,
+      "loss": 3.9703,
+      "step": 57860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0792645140877976e-05,
+      "loss": 3.9266,
+      "step": 57865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0779302443366339e-05,
+      "loss": 3.9813,
+      "step": 57870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0765967691193323e-05,
+      "loss": 3.8422,
+      "step": 57875
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0752640885119984e-05,
+      "loss": 3.9469,
+      "step": 57880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0739322025906883e-05,
+      "loss": 3.9953,
+      "step": 57885
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0726011114314093e-05,
+      "loss": 3.9078,
+      "step": 57890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0712708151101323e-05,
+      "loss": 3.9422,
+      "step": 57895
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0699413137027735e-05,
+      "loss": 3.9984,
+      "step": 57900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0686126072852086e-05,
+      "loss": 3.9594,
+      "step": 57905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0672846959332638e-05,
+      "loss": 3.9203,
+      "step": 57910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0659575797227298e-05,
+      "loss": 3.9438,
+      "step": 57915
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0646312587293433e-05,
+      "loss": 3.9813,
+      "step": 57920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0633057330287915e-05,
+      "loss": 3.9813,
+      "step": 57925
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.061981002696729e-05,
+      "loss": 3.9234,
+      "step": 57930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0606570678087573e-05,
+      "loss": 3.9312,
+      "step": 57935
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0593339284404323e-05,
+      "loss": 3.9453,
+      "step": 57940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0580115846672621e-05,
+      "loss": 3.9469,
+      "step": 57945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0566900365647214e-05,
+      "loss": 3.9719,
+      "step": 57950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0553692842082245e-05,
+      "loss": 3.8875,
+      "step": 57955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0540493276731482e-05,
+      "loss": 3.9031,
+      "step": 57960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.052730167034822e-05,
+      "loss": 3.9937,
+      "step": 57965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0514118023685326e-05,
+      "loss": 3.9719,
+      "step": 57970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0500942337495194e-05,
+      "loss": 3.8563,
+      "step": 57975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.048777461252971e-05,
+      "loss": 3.9297,
+      "step": 57980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0474614849540435e-05,
+      "loss": 3.8219,
+      "step": 57985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0461463049278358e-05,
+      "loss": 3.8891,
+      "step": 57990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0448319212494039e-05,
+      "loss": 3.9562,
+      "step": 57995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0435183339937613e-05,
+      "loss": 3.9391,
+      "step": 58000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.042205543235875e-05,
+      "loss": 3.9688,
+      "step": 58005
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0408935490506665e-05,
+      "loss": 3.9312,
+      "step": 58010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0395823515130091e-05,
+      "loss": 4.0266,
+      "step": 58015
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0382719506977349e-05,
+      "loss": 3.9594,
+      "step": 58020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0369623466796306e-05,
+      "loss": 3.8844,
+      "step": 58025
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.035653539533428e-05,
+      "loss": 3.9609,
+      "step": 58030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.034345529333831e-05,
+      "loss": 3.8797,
+      "step": 58035
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0330383161554811e-05,
+      "loss": 3.9906,
+      "step": 58040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0317319000729808e-05,
+      "loss": 3.95,
+      "step": 58045
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0304262811608915e-05,
+      "loss": 4.0344,
+      "step": 58050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0291214594937225e-05,
+      "loss": 3.8844,
+      "step": 58055
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0278174351459389e-05,
+      "loss": 3.8859,
+      "step": 58060
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0265142081919647e-05,
+      "loss": 3.8797,
+      "step": 58065
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0252117787061737e-05,
+      "loss": 3.9875,
+      "step": 58070
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0239101467628946e-05,
+      "loss": 3.9578,
+      "step": 58075
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0226093124364131e-05,
+      "loss": 3.9219,
+      "step": 58080
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0213092758009683e-05,
+      "loss": 3.9797,
+      "step": 58085
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.020010036930749e-05,
+      "loss": 3.9766,
+      "step": 58090
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0187115958999092e-05,
+      "loss": 4.0016,
+      "step": 58095
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0174139527825498e-05,
+      "loss": 3.9031,
+      "step": 58100
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0161171076527214e-05,
+      "loss": 3.9641,
+      "step": 58105
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0148210605844432e-05,
+      "loss": 3.8719,
+      "step": 58110
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0135258116516776e-05,
+      "loss": 3.8781,
+      "step": 58115
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0122313609283422e-05,
+      "loss": 3.9953,
+      "step": 58120
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.010937708488313e-05,
+      "loss": 3.9531,
+      "step": 58125
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0096448544054192e-05,
+      "loss": 3.9203,
+      "step": 58130
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0083527987534468e-05,
+      "loss": 3.8391,
+      "step": 58135
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0070615416061283e-05,
+      "loss": 3.9172,
+      "step": 58140
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0057710830371568e-05,
+      "loss": 3.9891,
+      "step": 58145
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0044814231201831e-05,
+      "loss": 3.9453,
+      "step": 58150
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0031925619288068e-05,
+      "loss": 3.9422,
+      "step": 58155
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.001904499536577e-05,
+      "loss": 3.9656,
+      "step": 58160
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.000617236017014e-05,
+      "loss": 3.9641,
+      "step": 58165
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.99330771443575e-06,
+      "loss": 3.9141,
+      "step": 58170
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.9804510588968e-06,
+      "loss": 3.9844,
+      "step": 58175
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.967602394287016e-06,
+      "loss": 3.8328,
+      "step": 58180
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.954761721339699e-06,
+      "loss": 3.9453,
+      "step": 58185
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.941929040787643e-06,
+      "loss": 4.0469,
+      "step": 58190
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.929104353363198e-06,
+      "loss": 3.9937,
+      "step": 58195
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.916287659798311e-06,
+      "loss": 3.9141,
+      "step": 58200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.903478960824413e-06,
+      "loss": 3.9969,
+      "step": 58205
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.890678257172486e-06,
+      "loss": 4.0016,
+      "step": 58210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.877885549573062e-06,
+      "loss": 3.9516,
+      "step": 58215
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.86510083875624e-06,
+      "loss": 3.9906,
+      "step": 58220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.85232412545167e-06,
+      "loss": 3.8797,
+      "step": 58225
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.839555410388473e-06,
+      "loss": 3.9344,
+      "step": 58230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.826794694295359e-06,
+      "loss": 3.9078,
+      "step": 58235
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.814041977900617e-06,
+      "loss": 3.9234,
+      "step": 58240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.801297261932044e-06,
+      "loss": 3.9188,
+      "step": 58245
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.788560547116946e-06,
+      "loss": 3.8656,
+      "step": 58250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.775831834182257e-06,
+      "loss": 3.9406,
+      "step": 58255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.763111123854379e-06,
+      "loss": 3.9766,
+      "step": 58260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.75039841685928e-06,
+      "loss": 3.9641,
+      "step": 58265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.737693713922462e-06,
+      "loss": 3.9078,
+      "step": 58270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.724997015769032e-06,
+      "loss": 3.9625,
+      "step": 58275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.712308323123558e-06,
+      "loss": 3.9906,
+      "step": 58280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.699627636710162e-06,
+      "loss": 3.9719,
+      "step": 58285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.686954957252596e-06,
+      "loss": 3.8813,
+      "step": 58290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.674290285474051e-06,
+      "loss": 3.8422,
+      "step": 58295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.661633622097298e-06,
+      "loss": 3.8797,
+      "step": 58300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.648984967844641e-06,
+      "loss": 3.9719,
+      "step": 58305
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.63634432343799e-06,
+      "loss": 3.9594,
+      "step": 58310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.623711689598712e-06,
+      "loss": 3.9219,
+      "step": 58315
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.611087067047751e-06,
+      "loss": 3.9547,
+      "step": 58320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.598470456505597e-06,
+      "loss": 3.9375,
+      "step": 58325
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.585861858692306e-06,
+      "loss": 3.9766,
+      "step": 58330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.57326127432742e-06,
+      "loss": 3.9859,
+      "step": 58335
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.560668704130065e-06,
+      "loss": 3.8984,
+      "step": 58340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.548084148818914e-06,
+      "loss": 3.9984,
+      "step": 58345
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.535507609112158e-06,
+      "loss": 3.8797,
+      "step": 58350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.522939085727544e-06,
+      "loss": 3.9281,
+      "step": 58355
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.510378579382345e-06,
+      "loss": 3.9016,
+      "step": 58360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.497826090793403e-06,
+      "loss": 3.9562,
+      "step": 58365
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.485281620677116e-06,
+      "loss": 3.9406,
+      "step": 58370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.472745169749324e-06,
+      "loss": 3.9219,
+      "step": 58375
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.460216738725574e-06,
+      "loss": 3.8875,
+      "step": 58380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.447696328320809e-06,
+      "loss": 3.9922,
+      "step": 58385
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.435183939249574e-06,
+      "loss": 3.9297,
+      "step": 58390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.42267957222595e-06,
+      "loss": 3.9328,
+      "step": 58395
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.410183227963597e-06,
+      "loss": 3.9062,
+      "step": 58400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.397694907175646e-06,
+      "loss": 3.8422,
+      "step": 58405
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.385214610574826e-06,
+      "loss": 4.0203,
+      "step": 58410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.372742338873368e-06,
+      "loss": 3.8609,
+      "step": 58415
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.360278092783102e-06,
+      "loss": 3.9422,
+      "step": 58420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.347821873015327e-06,
+      "loss": 3.9156,
+      "step": 58425
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.335373680280922e-06,
+      "loss": 3.8813,
+      "step": 58430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.322933515290337e-06,
+      "loss": 3.9656,
+      "step": 58435
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.310501378753521e-06,
+      "loss": 3.9562,
+      "step": 58440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.298077271379972e-06,
+      "loss": 3.925,
+      "step": 58445
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.285661193878708e-06,
+      "loss": 3.9266,
+      "step": 58450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.273253146958376e-06,
+      "loss": 3.9156,
+      "step": 58455
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.260853131327062e-06,
+      "loss": 3.9031,
+      "step": 58460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.248461147692432e-06,
+      "loss": 3.9547,
+      "step": 58465
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.23607719676172e-06,
+      "loss": 3.9422,
+      "step": 58470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.22370127924168e-06,
+      "loss": 3.8734,
+      "step": 58475
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.211333395838594e-06,
+      "loss": 3.9625,
+      "step": 58480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.198973547258281e-06,
+      "loss": 3.9484,
+      "step": 58485
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.186621734206162e-06,
+      "loss": 3.9266,
+      "step": 58490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.174277957387139e-06,
+      "loss": 3.9844,
+      "step": 58495
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.161942217505631e-06,
+      "loss": 3.9531,
+      "step": 58500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.149614515265708e-06,
+      "loss": 3.9688,
+      "step": 58505
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.137294851370874e-06,
+      "loss": 3.8984,
+      "step": 58510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.124983226524218e-06,
+      "loss": 3.9625,
+      "step": 58515
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.112679641428362e-06,
+      "loss": 3.8484,
+      "step": 58520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.100384096785506e-06,
+      "loss": 3.9344,
+      "step": 58525
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.088096593297295e-06,
+      "loss": 3.9625,
+      "step": 58530
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.075817131665031e-06,
+      "loss": 3.8937,
+      "step": 58535
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.063545712589488e-06,
+      "loss": 3.9453,
+      "step": 58540
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.051282336770988e-06,
+      "loss": 3.9094,
+      "step": 58545
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.039027004909421e-06,
+      "loss": 3.8328,
+      "step": 58550
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.026779717704197e-06,
+      "loss": 3.9844,
+      "step": 58555
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.014540475854254e-06,
+      "loss": 3.9391,
+      "step": 58560
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.002309280058067e-06,
+      "loss": 3.8516,
+      "step": 58565
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.990086131013746e-06,
+      "loss": 3.9234,
+      "step": 58570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.977871029418798e-06,
+      "loss": 4.0172,
+      "step": 58575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.965663975970349e-06,
+      "loss": 3.9422,
+      "step": 58580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.95346497136511e-06,
+      "loss": 3.9828,
+      "step": 58585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.941274016299222e-06,
+      "loss": 3.8469,
+      "step": 58590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.929091111468429e-06,
+      "loss": 3.9672,
+      "step": 58595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.916916257568042e-06,
+      "loss": 4.0234,
+      "step": 58600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.904749455292854e-06,
+      "loss": 3.8828,
+      "step": 58605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.892590705337227e-06,
+      "loss": 3.8203,
+      "step": 58610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.88044000839509e-06,
+      "loss": 3.8875,
+      "step": 58615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.868297365159854e-06,
+      "loss": 3.8922,
+      "step": 58620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.856162776324515e-06,
+      "loss": 3.9828,
+      "step": 58625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.844036242581587e-06,
+      "loss": 3.8781,
+      "step": 58630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.83191776462313e-06,
+      "loss": 3.8813,
+      "step": 58635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.819807343140778e-06,
+      "loss": 3.9734,
+      "step": 58640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.807704978825609e-06,
+      "loss": 3.9922,
+      "step": 58645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.795610672368386e-06,
+      "loss": 3.9906,
+      "step": 58650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.783524424459276e-06,
+      "loss": 3.95,
+      "step": 58655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.77144623578806e-06,
+      "loss": 3.9781,
+      "step": 58660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.759376107044037e-06,
+      "loss": 3.9828,
+      "step": 58665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.747314038916053e-06,
+      "loss": 3.9312,
+      "step": 58670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.735260032092511e-06,
+      "loss": 3.95,
+      "step": 58675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.723214087261293e-06,
+      "loss": 3.9875,
+      "step": 58680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.711176205109882e-06,
+      "loss": 3.9328,
+      "step": 58685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.699146386325295e-06,
+      "loss": 3.9859,
+      "step": 58690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.687124631594067e-06,
+      "loss": 3.9266,
+      "step": 58695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.675110941602265e-06,
+      "loss": 4.0203,
+      "step": 58700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.66310531703554e-06,
+      "loss": 3.9734,
+      "step": 58705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.65110775857903e-06,
+      "loss": 3.9547,
+      "step": 58710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.63911826691745e-06,
+      "loss": 3.9469,
+      "step": 58715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.627136842735022e-06,
+      "loss": 3.9297,
+      "step": 58720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.615163486715549e-06,
+      "loss": 3.8859,
+      "step": 58725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.60319819954235e-06,
+      "loss": 3.9594,
+      "step": 58730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.591240981898263e-06,
+      "loss": 3.9906,
+      "step": 58735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.579291834465724e-06,
+      "loss": 3.9188,
+      "step": 58740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.567350757926638e-06,
+      "loss": 3.9172,
+      "step": 58745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.555417752962513e-06,
+      "loss": 3.8797,
+      "step": 58750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.543492820254316e-06,
+      "loss": 4.0094,
+      "step": 58755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.531575960482673e-06,
+      "loss": 3.8937,
+      "step": 58760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.519667174327638e-06,
+      "loss": 3.9828,
+      "step": 58765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.507766462468868e-06,
+      "loss": 3.9219,
+      "step": 58770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.495873825585486e-06,
+      "loss": 4.0516,
+      "step": 58775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.483989264356268e-06,
+      "loss": 3.9359,
+      "step": 58780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.47211277945945e-06,
+      "loss": 3.8906,
+      "step": 58785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.460244371572794e-06,
+      "loss": 3.9219,
+      "step": 58790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.448384041373673e-06,
+      "loss": 3.9,
+      "step": 58795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.436531789538947e-06,
+      "loss": 4.0047,
+      "step": 58800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.424687616745008e-06,
+      "loss": 3.9703,
+      "step": 58805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.412851523667796e-06,
+      "loss": 3.9984,
+      "step": 58810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.40102351098284e-06,
+      "loss": 3.975,
+      "step": 58815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.389203579365134e-06,
+      "loss": 3.975,
+      "step": 58820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.377391729489236e-06,
+      "loss": 3.9844,
+      "step": 58825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.365587962029274e-06,
+      "loss": 3.8969,
+      "step": 58830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.353792277658894e-06,
+      "loss": 3.9156,
+      "step": 58835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.342004677051256e-06,
+      "loss": 3.9453,
+      "step": 58840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.330225160879057e-06,
+      "loss": 3.9047,
+      "step": 58845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.318453729814605e-06,
+      "loss": 3.9281,
+      "step": 58850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.306690384529686e-06,
+      "loss": 3.8922,
+      "step": 58855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.294935125695624e-06,
+      "loss": 3.9609,
+      "step": 58860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.283187953983268e-06,
+      "loss": 4.0031,
+      "step": 58865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.271448870063063e-06,
+      "loss": 3.9328,
+      "step": 58870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.259717874604976e-06,
+      "loss": 3.9078,
+      "step": 58875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.247994968278438e-06,
+      "loss": 3.9016,
+      "step": 58880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.236280151752529e-06,
+      "loss": 4.0438,
+      "step": 58885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.224573425695797e-06,
+      "loss": 3.9359,
+      "step": 58890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.212874790776357e-06,
+      "loss": 4.0422,
+      "step": 58895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.201184247661814e-06,
+      "loss": 3.9203,
+      "step": 58900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.189501797019393e-06,
+      "loss": 3.9875,
+      "step": 58905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.177827439515783e-06,
+      "loss": 3.9828,
+      "step": 58910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.166161175817248e-06,
+      "loss": 4.0703,
+      "step": 58915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.15450300658959e-06,
+      "loss": 3.9797,
+      "step": 58920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.142852932498162e-06,
+      "loss": 4.0109,
+      "step": 58925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.131210954207796e-06,
+      "loss": 3.9,
+      "step": 58930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.119577072382894e-06,
+      "loss": 3.925,
+      "step": 58935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.107951287687443e-06,
+      "loss": 3.975,
+      "step": 58940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.09633360078491e-06,
+      "loss": 3.9203,
+      "step": 58945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.084724012338317e-06,
+      "loss": 3.9781,
+      "step": 58950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.073122523010201e-06,
+      "loss": 3.8234,
+      "step": 58955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.061529133462712e-06,
+      "loss": 3.8953,
+      "step": 58960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.04994384435746e-06,
+      "loss": 3.8687,
+      "step": 58965
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.038366656355578e-06,
+      "loss": 3.9391,
+      "step": 58970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.026797570117854e-06,
+      "loss": 3.9547,
+      "step": 58975
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.01523658630448e-06,
+      "loss": 3.8891,
+      "step": 58980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.003683705575275e-06,
+      "loss": 3.9219,
+      "step": 58985
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.992138928589514e-06,
+      "loss": 3.9953,
+      "step": 58990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.980602256006118e-06,
+      "loss": 3.9266,
+      "step": 58995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.969073688483463e-06,
+      "loss": 3.9813,
+      "step": 59000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.957553226679469e-06,
+      "loss": 4.0281,
+      "step": 59005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.94604087125163e-06,
+      "loss": 3.95,
+      "step": 59010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.934536622856969e-06,
+      "loss": 3.9109,
+      "step": 59015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.923040482151977e-06,
+      "loss": 4.0047,
+      "step": 59020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.911552449792812e-06,
+      "loss": 3.9469,
+      "step": 59025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.900072526435069e-06,
+      "loss": 3.9344,
+      "step": 59030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.888600712733872e-06,
+      "loss": 3.9078,
+      "step": 59035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.877137009343982e-06,
+      "loss": 3.8969,
+      "step": 59040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.86568141691959e-06,
+      "loss": 3.8844,
+      "step": 59045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.854233936114473e-06,
+      "loss": 3.9156,
+      "step": 59050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.842794567581946e-06,
+      "loss": 3.9344,
+      "step": 59055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.83136331197488e-06,
+      "loss": 3.9297,
+      "step": 59060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.819940169945593e-06,
+      "loss": 3.9328,
+      "step": 59065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.808525142146077e-06,
+      "loss": 3.8891,
+      "step": 59070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.79711822922776e-06,
+      "loss": 3.9188,
+      "step": 59075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.785719431841609e-06,
+      "loss": 3.9141,
+      "step": 59080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.774328750638198e-06,
+      "loss": 3.9047,
+      "step": 59085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.762946186267577e-06,
+      "loss": 4.0172,
+      "step": 59090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.751571739379325e-06,
+      "loss": 3.9391,
+      "step": 59095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.740205410622624e-06,
+      "loss": 3.9844,
+      "step": 59100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.728847200646122e-06,
+      "loss": 3.9516,
+      "step": 59105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.717497110098065e-06,
+      "loss": 3.9469,
+      "step": 59110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.706155139626152e-06,
+      "loss": 3.9156,
+      "step": 59115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.694821289877718e-06,
+      "loss": 3.9,
+      "step": 59120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.683495561499575e-06,
+      "loss": 4.0344,
+      "step": 59125
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.672177955138076e-06,
+      "loss": 3.925,
+      "step": 59130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.660868471439103e-06,
+      "loss": 3.9703,
+      "step": 59135
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.649567111048122e-06,
+      "loss": 3.9406,
+      "step": 59140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.638273874610084e-06,
+      "loss": 3.8188,
+      "step": 59145
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.626988762769493e-06,
+      "loss": 3.8984,
+      "step": 59150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.615711776170413e-06,
+      "loss": 4.0328,
+      "step": 59155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.6044429154563995e-06,
+      "loss": 3.8969,
+      "step": 59160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.593182181270569e-06,
+      "loss": 3.9891,
+      "step": 59165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.5819295742555745e-06,
+      "loss": 3.9937,
+      "step": 59170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.570685095053619e-06,
+      "loss": 3.9594,
+      "step": 59175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.559448744306406e-06,
+      "loss": 3.9031,
+      "step": 59180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.548220522655191e-06,
+      "loss": 3.9453,
+      "step": 59185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.537000430740808e-06,
+      "loss": 3.9156,
+      "step": 59190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.5257884692035475e-06,
+      "loss": 3.9484,
+      "step": 59195
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.514584638683313e-06,
+      "loss": 3.9562,
+      "step": 59200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.503388939819443e-06,
+      "loss": 3.9422,
+      "step": 59205
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.492201373250961e-06,
+      "loss": 3.9844,
+      "step": 59210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4810219396162886e-06,
+      "loss": 3.9203,
+      "step": 59215
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.469850639553465e-06,
+      "loss": 3.9438,
+      "step": 59220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.458687473699998e-06,
+      "loss": 3.9891,
+      "step": 59225
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.447532442693027e-06,
+      "loss": 3.9141,
+      "step": 59230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.436385547169127e-06,
+      "loss": 3.8406,
+      "step": 59235
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.425246787764438e-06,
+      "loss": 3.9719,
+      "step": 59240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4141161651147196e-06,
+      "loss": 3.9031,
+      "step": 59245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4029936798551456e-06,
+      "loss": 3.9203,
+      "step": 59250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3918793326204765e-06,
+      "loss": 3.8922,
+      "step": 59255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.38077312404502e-06,
+      "loss": 3.9047,
+      "step": 59260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.36967505476262e-06,
+      "loss": 3.9688,
+      "step": 59265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.358585125406635e-06,
+      "loss": 3.9937,
+      "step": 59270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.347503336609928e-06,
+      "loss": 3.9875,
+      "step": 59275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.336429689005025e-06,
+      "loss": 3.9562,
+      "step": 59280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3253641832238365e-06,
+      "loss": 3.9828,
+      "step": 59285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.314306819897891e-06,
+      "loss": 3.9422,
+      "step": 59290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.303257599658202e-06,
+      "loss": 3.9391,
+      "step": 59295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.292216523135397e-06,
+      "loss": 3.8656,
+      "step": 59300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.281183590959572e-06,
+      "loss": 3.9922,
+      "step": 59305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.270158803760373e-06,
+      "loss": 3.9438,
+      "step": 59310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.259142162166981e-06,
+      "loss": 3.9859,
+      "step": 59315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.248133666808142e-06,
+      "loss": 3.9453,
+      "step": 59320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.237133318312088e-06,
+      "loss": 3.9438,
+      "step": 59325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.2261411173066e-06,
+      "loss": 3.9891,
+      "step": 59330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.215157064419042e-06,
+      "loss": 3.9219,
+      "step": 59335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.204181160276245e-06,
+      "loss": 3.9203,
+      "step": 59340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.193213405504628e-06,
+      "loss": 3.975,
+      "step": 59345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.18225380073007e-06,
+      "loss": 3.9734,
+      "step": 59350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.171302346578106e-06,
+      "loss": 3.9625,
+      "step": 59355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.160359043673703e-06,
+      "loss": 3.9359,
+      "step": 59360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.149423892641376e-06,
+      "loss": 3.925,
+      "step": 59365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.138496894105244e-06,
+      "loss": 3.9859,
+      "step": 59370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.127578048688876e-06,
+      "loss": 3.9453,
+      "step": 59375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.116667357015421e-06,
+      "loss": 3.9625,
+      "step": 59380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.105764819707532e-06,
+      "loss": 3.8594,
+      "step": 59385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.094870437387462e-06,
+      "loss": 3.9703,
+      "step": 59390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.08398421067693e-06,
+      "loss": 3.8922,
+      "step": 59395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.073106140197221e-06,
+      "loss": 4.0391,
+      "step": 59400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.062236226569124e-06,
+      "loss": 3.7969,
+      "step": 59405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.051374470413007e-06,
+      "loss": 4.0125,
+      "step": 59410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.040520872348759e-06,
+      "loss": 3.8469,
+      "step": 59415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.029675432995768e-06,
+      "loss": 4.0328,
+      "step": 59420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.018838152973005e-06,
+      "loss": 3.8984,
+      "step": 59425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.008009032898959e-06,
+      "loss": 3.9125,
+      "step": 59430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.997188073391652e-06,
+      "loss": 3.9531,
+      "step": 59435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.98637527506859e-06,
+      "loss": 3.9141,
+      "step": 59440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.975570638546929e-06,
+      "loss": 4.0141,
+      "step": 59445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.96477416444326e-06,
+      "loss": 3.9438,
+      "step": 59450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.953985853373706e-06,
+      "loss": 3.8594,
+      "step": 59455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.943205705954008e-06,
+      "loss": 4.0078,
+      "step": 59460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.932433722799375e-06,
+      "loss": 3.9172,
+      "step": 59465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.921669904524563e-06,
+      "loss": 3.9703,
+      "step": 59470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.9109142517438325e-06,
+      "loss": 3.9844,
+      "step": 59475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.900166765071058e-06,
+      "loss": 3.9188,
+      "step": 59480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.889427445119583e-06,
+      "loss": 3.9516,
+      "step": 59485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.878696292502283e-06,
+      "loss": 3.9703,
+      "step": 59490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.86797330783162e-06,
+      "loss": 4.0344,
+      "step": 59495
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.857258491719536e-06,
+      "loss": 4.0187,
+      "step": 59500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.846551844777509e-06,
+      "loss": 3.9422,
+      "step": 59505
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8358533676166e-06,
+      "loss": 3.9641,
+      "step": 59510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.825163060847355e-06,
+      "loss": 3.8641,
+      "step": 59515
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.814480925079868e-06,
+      "loss": 3.8859,
+      "step": 59520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.803806960923802e-06,
+      "loss": 3.975,
+      "step": 59525
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.793141168988269e-06,
+      "loss": 3.8766,
+      "step": 59530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.782483549882017e-06,
+      "loss": 3.8406,
+      "step": 59535
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.7718341042132255e-06,
+      "loss": 3.9234,
+      "step": 59540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.761192832589707e-06,
+      "loss": 3.9594,
+      "step": 59545
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.7505597356187435e-06,
+      "loss": 3.9813,
+      "step": 59550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.73993481390715e-06,
+      "loss": 4.0,
+      "step": 59555
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.7293180680613235e-06,
+      "loss": 3.9312,
+      "step": 59560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.71870949868713e-06,
+      "loss": 3.9969,
+      "step": 59565
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.70810910639002e-06,
+      "loss": 3.9312,
+      "step": 59570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.697516891774957e-06,
+      "loss": 3.9141,
+      "step": 59575
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.686932855446442e-06,
+      "loss": 3.8734,
+      "step": 59580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.676356998008492e-06,
+      "loss": 3.975,
+      "step": 59585
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.6657893200646904e-06,
+      "loss": 3.9453,
+      "step": 59590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.655229822218139e-06,
+      "loss": 3.9359,
+      "step": 59595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.644678505071438e-06,
+      "loss": 3.9188,
+      "step": 59600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.634135369226789e-06,
+      "loss": 3.9734,
+      "step": 59605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.6236004152858625e-06,
+      "loss": 3.9562,
+      "step": 59610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.613073643849925e-06,
+      "loss": 3.8844,
+      "step": 59615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.602555055519681e-06,
+      "loss": 4.0062,
+      "step": 59620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.592044650895484e-06,
+      "loss": 3.9078,
+      "step": 59625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.581542430577136e-06,
+      "loss": 3.8547,
+      "step": 59630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.571048395164008e-06,
+      "loss": 3.9078,
+      "step": 59635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.560562545254988e-06,
+      "loss": 3.9219,
+      "step": 59640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.550084881448531e-06,
+      "loss": 4.0453,
+      "step": 59645
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.5396154043425575e-06,
+      "loss": 3.9078,
+      "step": 59650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.529154114534574e-06,
+      "loss": 3.9578,
+      "step": 59655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.518701012621636e-06,
+      "loss": 3.8625,
+      "step": 59660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.508256099200265e-06,
+      "loss": 3.9969,
+      "step": 59665
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.497819374866587e-06,
+      "loss": 3.9125,
+      "step": 59670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.487390840216189e-06,
+      "loss": 3.9562,
+      "step": 59675
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.476970495844263e-06,
+      "loss": 3.925,
+      "step": 59680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.466558342345484e-06,
+      "loss": 3.85,
+      "step": 59685
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4561543803140606e-06,
+      "loss": 3.9141,
+      "step": 59690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.445758610343782e-06,
+      "loss": 3.8906,
+      "step": 59695
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4353710330279255e-06,
+      "loss": 4.0422,
+      "step": 59700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.424991648959282e-06,
+      "loss": 3.9969,
+      "step": 59705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4146204587302295e-06,
+      "loss": 4.0297,
+      "step": 59710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.404257462932644e-06,
+      "loss": 4.0312,
+      "step": 59715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.393902662157968e-06,
+      "loss": 3.9594,
+      "step": 59720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.383556056997097e-06,
+      "loss": 3.9453,
+      "step": 59725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.3732176480405575e-06,
+      "loss": 3.9969,
+      "step": 59730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.362887435878361e-06,
+      "loss": 3.9188,
+      "step": 59735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.352565421100053e-06,
+      "loss": 4.0172,
+      "step": 59740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.342251604294679e-06,
+      "loss": 3.9312,
+      "step": 59745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.331945986050885e-06,
+      "loss": 3.8906,
+      "step": 59750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.3216485669568e-06,
+      "loss": 4.0359,
+      "step": 59755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.311359347600104e-06,
+      "loss": 3.8406,
+      "step": 59760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.301078328567993e-06,
+      "loss": 4.0047,
+      "step": 59765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.290805510447216e-06,
+      "loss": 3.9766,
+      "step": 59770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.280540893824054e-06,
+      "loss": 3.9828,
+      "step": 59775
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.270284479284271e-06,
+      "loss": 3.9109,
+      "step": 59780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.2600362674132485e-06,
+      "loss": 3.8719,
+      "step": 59785
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.249796258795836e-06,
+      "loss": 3.9641,
+      "step": 59790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.239564454016416e-06,
+      "loss": 3.8875,
+      "step": 59795
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.229340853658921e-06,
+      "loss": 3.9453,
+      "step": 59800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.2191254583068336e-06,
+      "loss": 3.9188,
+      "step": 59805
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.208918268543139e-06,
+      "loss": 4.0266,
+      "step": 59810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.198719284950338e-06,
+      "loss": 4.0641,
+      "step": 59815
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.18852850811053e-06,
+      "loss": 3.8687,
+      "step": 59820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.178345938605284e-06,
+      "loss": 3.9625,
+      "step": 59825
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.168171577015718e-06,
+      "loss": 3.9984,
+      "step": 59830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.158005423922469e-06,
+      "loss": 3.8656,
+      "step": 59835
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.147847479905737e-06,
+      "loss": 3.8969,
+      "step": 59840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.137697745545262e-06,
+      "loss": 3.9906,
+      "step": 59845
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.1275562214202434e-06,
+      "loss": 3.8672,
+      "step": 59850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.11742290810947e-06,
+      "loss": 3.9484,
+      "step": 59855
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.10729780619128e-06,
+      "loss": 3.9297,
+      "step": 59860
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.097180916243477e-06,
+      "loss": 3.8891,
+      "step": 59865
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.087072238843449e-06,
+      "loss": 3.8953,
+      "step": 59870
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.076971774568118e-06,
+      "loss": 3.8734,
+      "step": 59875
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.066879523993906e-06,
+      "loss": 3.9094,
+      "step": 59880
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.05679548769677e-06,
+      "loss": 3.8719,
+      "step": 59885
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.046719666252181e-06,
+      "loss": 3.9531,
+      "step": 59890
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.036652060235231e-06,
+      "loss": 3.9344,
+      "step": 59895
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.026592670220426e-06,
+      "loss": 3.8844,
+      "step": 59900
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.016541496781874e-06,
+      "loss": 3.9672,
+      "step": 59905
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.006498540493216e-06,
+      "loss": 3.8484,
+      "step": 59910
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.996463801927576e-06,
+      "loss": 3.9281,
+      "step": 59915
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9864372816576466e-06,
+      "loss": 3.9188,
+      "step": 59920
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.976418980255637e-06,
+      "loss": 3.9859,
+      "step": 59925
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.966408898293306e-06,
+      "loss": 3.9688,
+      "step": 59930
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.956407036341931e-06,
+      "loss": 4.0141,
+      "step": 59935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9464133949723045e-06,
+      "loss": 3.9406,
+      "step": 59940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9364279747547534e-06,
+      "loss": 3.9203,
+      "step": 59945
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.92645077625919e-06,
+      "loss": 3.8766,
+      "step": 59950
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.916481800054973e-06,
+      "loss": 3.9109,
+      "step": 59955
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.906521046711049e-06,
+      "loss": 3.9047,
+      "step": 59960
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.896568516795897e-06,
+      "loss": 3.8891,
+      "step": 59965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.886624210877478e-06,
+      "loss": 3.9969,
+      "step": 59970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.876688129523321e-06,
+      "loss": 3.9953,
+      "step": 59975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8667602733004744e-06,
+      "loss": 4.0047,
+      "step": 59980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.856840642775567e-06,
+      "loss": 3.8859,
+      "step": 59985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.846929238514663e-06,
+      "loss": 3.9172,
+      "step": 59990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.837026061083394e-06,
+      "loss": 3.9078,
+      "step": 59995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8271311110470065e-06,
+      "loss": 4.0016,
+      "step": 60000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.817244388970149e-06,
+      "loss": 3.9703,
+      "step": 60005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.807365895417055e-06,
+      "loss": 3.9937,
+      "step": 60010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.797495630951521e-06,
+      "loss": 3.9438,
+      "step": 60015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.787633596136848e-06,
+      "loss": 3.9297,
+      "step": 60020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7777797915358195e-06,
+      "loss": 3.9859,
+      "step": 60025
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7679342177108345e-06,
+      "loss": 3.9172,
+      "step": 60030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.758096875223761e-06,
+      "loss": 3.9203,
+      "step": 60035
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.748267764636017e-06,
+      "loss": 4.0563,
+      "step": 60040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.738446886508569e-06,
+      "loss": 3.8734,
+      "step": 60045
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.728634241401886e-06,
+      "loss": 4.0422,
+      "step": 60050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.718829829875954e-06,
+      "loss": 4.0109,
+      "step": 60055
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.709033652490358e-06,
+      "loss": 3.8547,
+      "step": 60060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.699245709804134e-06,
+      "loss": 3.9547,
+      "step": 60065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6894660023758535e-06,
+      "loss": 3.9109,
+      "step": 60070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.679694530763717e-06,
+      "loss": 3.9469,
+      "step": 60075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.669931295525332e-06,
+      "loss": 3.9391,
+      "step": 60080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6601762972178996e-06,
+      "loss": 3.9203,
+      "step": 60085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.650429536398143e-06,
+      "loss": 4.0094,
+      "step": 60090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.640691013622317e-06,
+      "loss": 3.9797,
+      "step": 60095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.630960729446176e-06,
+      "loss": 4.0172,
+      "step": 60100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.621238684425028e-06,
+      "loss": 3.9188,
+      "step": 60105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6115248791137605e-06,
+      "loss": 3.9484,
+      "step": 60110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.60181931406668e-06,
+      "loss": 3.8953,
+      "step": 60115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.592121989837728e-06,
+      "loss": 4.0109,
+      "step": 60120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.582432906980294e-06,
+      "loss": 4.0734,
+      "step": 60125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.57275206604737e-06,
+      "loss": 3.8266,
+      "step": 60130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.56307946759143e-06,
+      "loss": 3.9922,
+      "step": 60135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5534151121644665e-06,
+      "loss": 3.9297,
+      "step": 60140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.543759000318055e-06,
+      "loss": 3.9062,
+      "step": 60145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.534111132603286e-06,
+      "loss": 3.8687,
+      "step": 60150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.524471509570722e-06,
+      "loss": 3.9062,
+      "step": 60155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.514840131770504e-06,
+      "loss": 3.8766,
+      "step": 60160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.50521699975231e-06,
+      "loss": 4.0312,
+      "step": 60165
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4956021140653505e-06,
+      "loss": 3.9453,
+      "step": 60170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.485995475258304e-06,
+      "loss": 3.9281,
+      "step": 60175
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.476397083879464e-06,
+      "loss": 3.8484,
+      "step": 60180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4668069404765756e-06,
+      "loss": 3.9469,
+      "step": 60185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4572250455969856e-06,
+      "loss": 3.9703,
+      "step": 60190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.447651399787506e-06,
+      "loss": 3.9813,
+      "step": 60195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.438086003594516e-06,
+      "loss": 3.9828,
+      "step": 60200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.428528857563913e-06,
+      "loss": 3.8844,
+      "step": 60205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.418979962241127e-06,
+      "loss": 3.9484,
+      "step": 60210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.409439318171105e-06,
+      "loss": 3.9703,
+      "step": 60215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.3999069258983456e-06,
+      "loss": 4.0328,
+      "step": 60220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.3903827859668456e-06,
+      "loss": 4.0141,
+      "step": 60225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.380866898920155e-06,
+      "loss": 3.9641,
+      "step": 60230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.3713592653013725e-06,
+      "loss": 3.8406,
+      "step": 60235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.361859885653064e-06,
+      "loss": 3.9312,
+      "step": 60240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.352368760517395e-06,
+      "loss": 3.9406,
+      "step": 60245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.3428858904359666e-06,
+      "loss": 3.9406,
+      "step": 60250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.333411275950028e-06,
+      "loss": 3.9375,
+      "step": 60255
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3239449176002645e-06,
+      "loss": 3.95,
+      "step": 60260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3144868159269275e-06,
+      "loss": 3.9125,
+      "step": 60265
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3050369714698005e-06,
+      "loss": 3.9016,
+      "step": 60270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.295595384768203e-06,
+      "loss": 3.9609,
+      "step": 60275
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.286162056360921e-06,
+      "loss": 3.9516,
+      "step": 60280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.27673698678634e-06,
+      "loss": 3.9469,
+      "step": 60285
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.267320176582346e-06,
+      "loss": 4.0094,
+      "step": 60290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2579116262863765e-06,
+      "loss": 3.8359,
+      "step": 60295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2485113364353515e-06,
+      "loss": 3.9422,
+      "step": 60300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.239119307565759e-06,
+      "loss": 3.9031,
+      "step": 60305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.229735540213603e-06,
+      "loss": 4.0594,
+      "step": 60310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2203600349144226e-06,
+      "loss": 3.9062,
+      "step": 60315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2109927922032395e-06,
+      "loss": 3.9406,
+      "step": 60320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2016338126147085e-06,
+      "loss": 3.8844,
+      "step": 60325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.192283096682903e-06,
+      "loss": 4.1078,
+      "step": 60330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.182940644941497e-06,
+      "loss": 3.875,
+      "step": 60335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.173606457923612e-06,
+      "loss": 3.9109,
+      "step": 60340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.164280536162007e-06,
+      "loss": 3.9328,
+      "step": 60345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.15496288018889e-06,
+      "loss": 3.9031,
+      "step": 60350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.145653490536017e-06,
+      "loss": 3.9266,
+      "step": 60355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.136352367734697e-06,
+      "loss": 3.9484,
+      "step": 60360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.127059512315723e-06,
+      "loss": 3.9547,
+      "step": 60365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.117774924809453e-06,
+      "loss": 3.8687,
+      "step": 60370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.1084986057457475e-06,
+      "loss": 3.9797,
+      "step": 60375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0992305556540155e-06,
+      "loss": 4.0016,
+      "step": 60380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.089970775063184e-06,
+      "loss": 3.8734,
+      "step": 60385
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0807192645017145e-06,
+      "loss": 3.9594,
+      "step": 60390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.071476024497567e-06,
+      "loss": 3.9625,
+      "step": 60395
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.062241055578286e-06,
+      "loss": 3.8234,
+      "step": 60400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.053014358270902e-06,
+      "loss": 3.9422,
+      "step": 60405
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.043795933101958e-06,
+      "loss": 3.9078,
+      "step": 60410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0345857805976e-06,
+      "loss": 3.9172,
+      "step": 60415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.025383901283425e-06,
+      "loss": 3.9797,
+      "step": 60420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.01619029568458e-06,
+      "loss": 4.0297,
+      "step": 60425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0070049643257285e-06,
+      "loss": 4.0266,
+      "step": 60430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.997827907731134e-06,
+      "loss": 3.9281,
+      "step": 60435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9886591264244765e-06,
+      "loss": 3.9859,
+      "step": 60440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.97949862092904e-06,
+      "loss": 3.8594,
+      "step": 60445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.970346391767621e-06,
+      "loss": 3.9844,
+      "step": 60450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.961202439462553e-06,
+      "loss": 4.0172,
+      "step": 60455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.952066764535651e-06,
+      "loss": 3.9797,
+      "step": 60460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.942939367508281e-06,
+      "loss": 3.8797,
+      "step": 60465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.933820248901377e-06,
+      "loss": 3.9094,
+      "step": 60470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.924709409235356e-06,
+      "loss": 3.9422,
+      "step": 60475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9156068490301685e-06,
+      "loss": 3.9141,
+      "step": 60480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.906512568805315e-06,
+      "loss": 3.9734,
+      "step": 60485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.897426569079782e-06,
+      "loss": 3.9547,
+      "step": 60490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8883488503721365e-06,
+      "loss": 3.8844,
+      "step": 60495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.879279413200415e-06,
+      "loss": 3.9281,
+      "step": 60500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.870218258082237e-06,
+      "loss": 4.0312,
+      "step": 60505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.861165385534704e-06,
+      "loss": 4.0312,
+      "step": 60510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.852120796074471e-06,
+      "loss": 3.9047,
+      "step": 60515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8430844902177244e-06,
+      "loss": 3.9109,
+      "step": 60520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.834056468480152e-06,
+      "loss": 3.8906,
+      "step": 60525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.825036731376991e-06,
+      "loss": 3.9469,
+      "step": 60530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.816025279422997e-06,
+      "loss": 3.9484,
+      "step": 60535
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.807022113132475e-06,
+      "loss": 3.9328,
+      "step": 60540
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.798027233019181e-06,
+      "loss": 3.8891,
+      "step": 60545
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.7890406395965195e-06,
+      "loss": 3.8922,
+      "step": 60550
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.780062333377316e-06,
+      "loss": 3.9141,
+      "step": 60555
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.771092314873959e-06,
+      "loss": 3.9672,
+      "step": 60560
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.7621305845983895e-06,
+      "loss": 3.9109,
+      "step": 60565
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.753177143062065e-06,
+      "loss": 3.9953,
+      "step": 60570
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.744231990775927e-06,
+      "loss": 3.9766,
+      "step": 60575
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.735295128250499e-06,
+      "loss": 3.9172,
+      "step": 60580
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.726366555995792e-06,
+      "loss": 3.9594,
+      "step": 60585
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.717446274521364e-06,
+      "loss": 3.8828,
+      "step": 60590
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.708534284336324e-06,
+      "loss": 3.9328,
+      "step": 60595
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.699630585949249e-06,
+      "loss": 3.9766,
+      "step": 60600
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.690735179868282e-06,
+      "loss": 3.9391,
+      "step": 60605
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.6818480666010684e-06,
+      "loss": 3.9234,
+      "step": 60610
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.672969246654834e-06,
+      "loss": 3.9234,
+      "step": 60615
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.664098720536258e-06,
+      "loss": 3.9406,
+      "step": 60620
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.655236488751601e-06,
+      "loss": 3.9031,
+      "step": 60625
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.646382551806627e-06,
+      "loss": 3.9531,
+      "step": 60630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.637536910206646e-06,
+      "loss": 3.9312,
+      "step": 60635
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.628699564456473e-06,
+      "loss": 3.9,
+      "step": 60640
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.61987051506042e-06,
+      "loss": 3.9734,
+      "step": 60645
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.611049762522401e-06,
+      "loss": 3.8906,
+      "step": 60650
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.602237307345813e-06,
+      "loss": 3.9078,
+      "step": 60655
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5934331500335706e-06,
+      "loss": 3.9562,
+      "step": 60660
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.584637291088139e-06,
+      "loss": 3.9391,
+      "step": 60665
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5758497310115e-06,
+      "loss": 3.9547,
+      "step": 60670
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.567070470305151e-06,
+      "loss": 3.9,
+      "step": 60675
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.558299509470109e-06,
+      "loss": 3.9234,
+      "step": 60680
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.549536849006974e-06,
+      "loss": 3.9016,
+      "step": 60685
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.540782489415812e-06,
+      "loss": 3.9344,
+      "step": 60690
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.532036431196239e-06,
+      "loss": 3.9828,
+      "step": 60695
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.523298674847359e-06,
+      "loss": 3.9438,
+      "step": 60700
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.514569220867903e-06,
+      "loss": 3.9781,
+      "step": 60705
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.505848069756007e-06,
+      "loss": 3.8875,
+      "step": 60710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.497135222009407e-06,
+      "loss": 3.9219,
+      "step": 60715
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.488430678125354e-06,
+      "loss": 3.8594,
+      "step": 60720
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.479734438600602e-06,
+      "loss": 3.9422,
+      "step": 60725
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.471046503931469e-06,
+      "loss": 3.9344,
+      "step": 60730
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.462366874613727e-06,
+      "loss": 3.9578,
+      "step": 60735
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4536955511427774e-06,
+      "loss": 3.9328,
+      "step": 60740
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.445032534013477e-06,
+      "loss": 3.9297,
+      "step": 60745
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.436377823720211e-06,
+      "loss": 3.9031,
+      "step": 60750
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.427731420756919e-06,
+      "loss": 3.9422,
+      "step": 60755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.419093325617041e-06,
+      "loss": 3.9328,
+      "step": 60760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.410463538793562e-06,
+      "loss": 3.9797,
+      "step": 60765
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.401842060778976e-06,
+      "loss": 3.9562,
+      "step": 60770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.393228892065337e-06,
+      "loss": 3.9297,
+      "step": 60775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.384624033144168e-06,
+      "loss": 3.9766,
+      "step": 60780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.376027484506578e-06,
+      "loss": 3.975,
+      "step": 60785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.36743924664314e-06,
+      "loss": 3.9656,
+      "step": 60790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.358859320043995e-06,
+      "loss": 3.95,
+      "step": 60795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.350287705198835e-06,
+      "loss": 3.8953,
+      "step": 60800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.341724402596786e-06,
+      "loss": 3.9641,
+      "step": 60805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.333169412726606e-06,
+      "loss": 3.9062,
+      "step": 60810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.324622736076505e-06,
+      "loss": 4.0031,
+      "step": 60815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.316084373134243e-06,
+      "loss": 3.9422,
+      "step": 60820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3075543243870966e-06,
+      "loss": 3.9469,
+      "step": 60825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.29903259032191e-06,
+      "loss": 4.0016,
+      "step": 60830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2905191714249946e-06,
+      "loss": 3.9672,
+      "step": 60835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.28201406818221e-06,
+      "loss": 3.9312,
+      "step": 60840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.273517281078953e-06,
+      "loss": 3.9328,
+      "step": 60845
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2650288106001185e-06,
+      "loss": 3.9625,
+      "step": 60850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.256548657230185e-06,
+      "loss": 3.9797,
+      "step": 60855
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2480768214530505e-06,
+      "loss": 4.0078,
+      "step": 60860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.239613303752276e-06,
+      "loss": 3.9266,
+      "step": 60865
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.231158104610827e-06,
+      "loss": 3.9641,
+      "step": 60870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.222711224511266e-06,
+      "loss": 3.9594,
+      "step": 60875
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.214272663935625e-06,
+      "loss": 3.9547,
+      "step": 60880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.205842423365535e-06,
+      "loss": 3.9875,
+      "step": 60885
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.197420503282095e-06,
+      "loss": 3.8469,
+      "step": 60890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.18900690416592e-06,
+      "loss": 3.9734,
+      "step": 60895
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.18060162649721e-06,
+      "loss": 3.9234,
+      "step": 60900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.172204670755646e-06,
+      "loss": 3.8375,
+      "step": 60905
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.163816037420431e-06,
+      "loss": 3.9125,
+      "step": 60910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.155435726970313e-06,
+      "loss": 3.8703,
+      "step": 60915
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.147063739883561e-06,
+      "loss": 4.0391,
+      "step": 60920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.138700076637958e-06,
+      "loss": 3.9703,
+      "step": 60925
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.1303447377108225e-06,
+      "loss": 3.8656,
+      "step": 60930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.12199772357899e-06,
+      "loss": 3.8844,
+      "step": 60935
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.113659034718831e-06,
+      "loss": 3.8875,
+      "step": 60940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.105328671606245e-06,
+      "loss": 3.9719,
+      "step": 60945
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.097006634716604e-06,
+      "loss": 3.9844,
+      "step": 60950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.088692924524911e-06,
+      "loss": 3.9484,
+      "step": 60955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.080387541505587e-06,
+      "loss": 3.8109,
+      "step": 60960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.072090486132634e-06,
+      "loss": 3.975,
+      "step": 60965
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.063801758879543e-06,
+      "loss": 4.0344,
+      "step": 60970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.055521360219399e-06,
+      "loss": 3.9344,
+      "step": 60975
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.047249290624743e-06,
+      "loss": 3.975,
+      "step": 60980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.038985550567631e-06,
+      "loss": 3.9234,
+      "step": 60985
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.030730140519716e-06,
+      "loss": 3.9625,
+      "step": 60990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.022483060952142e-06,
+      "loss": 4.0078,
+      "step": 60995
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.0142443123355295e-06,
+      "loss": 4.0203,
+      "step": 61000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.006013895140087e-06,
+      "loss": 3.925,
+      "step": 61005
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.997791809835537e-06,
+      "loss": 3.9172,
+      "step": 61010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.989578056891074e-06,
+      "loss": 3.9641,
+      "step": 61015
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.9813726367755205e-06,
+      "loss": 3.9203,
+      "step": 61020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.97317554995712e-06,
+      "loss": 3.9766,
+      "step": 61025
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.964986796903663e-06,
+      "loss": 3.9656,
+      "step": 61030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.956806378082527e-06,
+      "loss": 3.8344,
+      "step": 61035
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.948634293960556e-06,
+      "loss": 3.8859,
+      "step": 61040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.940470545004093e-06,
+      "loss": 3.9688,
+      "step": 61045
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.932315131679098e-06,
+      "loss": 3.9375,
+      "step": 61050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.924168054450982e-06,
+      "loss": 3.9484,
+      "step": 61055
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.916029313784675e-06,
+      "loss": 3.9266,
+      "step": 61060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.907898910144702e-06,
+      "loss": 3.9656,
+      "step": 61065
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8997768439950285e-06,
+      "loss": 3.85,
+      "step": 61070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.891663115799215e-06,
+      "loss": 3.9328,
+      "step": 61075
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.883557726020259e-06,
+      "loss": 3.9453,
+      "step": 61080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.875460675120773e-06,
+      "loss": 3.9688,
+      "step": 61085
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.867371963562871e-06,
+      "loss": 3.9469,
+      "step": 61090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.859291591808133e-06,
+      "loss": 3.9719,
+      "step": 61095
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.851219560317759e-06,
+      "loss": 3.8609,
+      "step": 61100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.843155869552378e-06,
+      "loss": 3.9547,
+      "step": 61105
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8351005199722064e-06,
+      "loss": 3.9641,
+      "step": 61110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.827053512036943e-06,
+      "loss": 3.9188,
+      "step": 61115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.819014846205854e-06,
+      "loss": 3.875,
+      "step": 61120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8109845229377044e-06,
+      "loss": 3.9516,
+      "step": 61125
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8029625426907627e-06,
+      "loss": 3.9172,
+      "step": 61130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.794948905922879e-06,
+      "loss": 3.9188,
+      "step": 61135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7869436130913866e-06,
+      "loss": 3.9719,
+      "step": 61140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7789466646531207e-06,
+      "loss": 3.95,
+      "step": 61145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.770958061064466e-06,
+      "loss": 3.9359,
+      "step": 61150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.762977802781375e-06,
+      "loss": 3.9328,
+      "step": 61155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7550058902592493e-06,
+      "loss": 3.9266,
+      "step": 61160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7470423239530423e-06,
+      "loss": 3.9156,
+      "step": 61165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7390871043172733e-06,
+      "loss": 3.9516,
+      "step": 61170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7311402318059126e-06,
+      "loss": 3.9406,
+      "step": 61175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.723201706872497e-06,
+      "loss": 3.9531,
+      "step": 61180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.715271529970065e-06,
+      "loss": 3.8906,
+      "step": 61185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.707349701551221e-06,
+      "loss": 3.975,
+      "step": 61190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6994362220680527e-06,
+      "loss": 3.8641,
+      "step": 61195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6915310919722e-06,
+      "loss": 3.9594,
+      "step": 61200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6836343117147504e-06,
+      "loss": 3.9297,
+      "step": 61205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6757458817464445e-06,
+      "loss": 4.0109,
+      "step": 61210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.667865802517439e-06,
+      "loss": 3.9312,
+      "step": 61215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.659994074477457e-06,
+      "loss": 3.9984,
+      "step": 61220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.652130698075739e-06,
+      "loss": 3.8484,
+      "step": 61225
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6442756737610436e-06,
+      "loss": 3.9203,
+      "step": 61230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.636429001981678e-06,
+      "loss": 3.9641,
+      "step": 61235
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6285906831854173e-06,
+      "loss": 4.0375,
+      "step": 61240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.620760717819637e-06,
+      "loss": 4.0078,
+      "step": 61245
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.612939106331164e-06,
+      "loss": 3.9875,
+      "step": 61250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.605125849166374e-06,
+      "loss": 3.9391,
+      "step": 61255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5973209467712093e-06,
+      "loss": 3.8422,
+      "step": 61260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.589524399591065e-06,
+      "loss": 3.8891,
+      "step": 61265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5817362080708844e-06,
+      "loss": 3.9781,
+      "step": 61270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.573956372655146e-06,
+      "loss": 3.9375,
+      "step": 61275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5661848937878766e-06,
+      "loss": 3.9391,
+      "step": 61280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5584217719125564e-06,
+      "loss": 3.9281,
+      "step": 61285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5506670074722467e-06,
+      "loss": 3.9438,
+      "step": 61290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5429206009094933e-06,
+      "loss": 3.9984,
+      "step": 61295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5351825526664102e-06,
+      "loss": 4.0969,
+      "step": 61300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.52745286318461e-06,
+      "loss": 3.9953,
+      "step": 61305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.51973153290519e-06,
+      "loss": 3.8859,
+      "step": 61310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.512018562268848e-06,
+      "loss": 3.9062,
+      "step": 61315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5043139517157314e-06,
+      "loss": 3.9141,
+      "step": 61320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4966177016855722e-06,
+      "loss": 3.9344,
+      "step": 61325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4889298126175856e-06,
+      "loss": 3.9688,
+      "step": 61330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.481250284950504e-06,
+      "loss": 3.9844,
+      "step": 61335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.473579119122627e-06,
+      "loss": 4.0172,
+      "step": 61340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4659163155717207e-06,
+      "loss": 3.8563,
+      "step": 61345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4582618747351187e-06,
+      "loss": 3.8828,
+      "step": 61350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4506157970496718e-06,
+      "loss": 3.9688,
+      "step": 61355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.442978082951714e-06,
+      "loss": 3.9344,
+      "step": 61360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.435348732877147e-06,
+      "loss": 3.9547,
+      "step": 61365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4277277472613717e-06,
+      "loss": 3.9094,
+      "step": 61370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.420115126539341e-06,
+      "loss": 3.925,
+      "step": 61375
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.412510871145474e-06,
+      "loss": 3.8219,
+      "step": 61380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4049149815137566e-06,
+      "loss": 3.8984,
+      "step": 61385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.397327458077709e-06,
+      "loss": 3.9797,
+      "step": 61390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3897483012703175e-06,
+      "loss": 3.8609,
+      "step": 61395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3821775115241536e-06,
+      "loss": 3.9203,
+      "step": 61400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.374615089271271e-06,
+      "loss": 3.8922,
+      "step": 61405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3670610349432582e-06,
+      "loss": 4.0078,
+      "step": 61410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3595153489712367e-06,
+      "loss": 3.9813,
+      "step": 61415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3519780317858127e-06,
+      "loss": 3.9766,
+      "step": 61420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3444490838171576e-06,
+      "loss": 3.9766,
+      "step": 61425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3369285054949626e-06,
+      "loss": 3.9312,
+      "step": 61430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3294162972483996e-06,
+      "loss": 3.9188,
+      "step": 61435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.321912459506226e-06,
+      "loss": 3.9109,
+      "step": 61440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.314416992696667e-06,
+      "loss": 3.9641,
+      "step": 61445
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.306929897247479e-06,
+      "loss": 3.9438,
+      "step": 61450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2994511735859708e-06,
+      "loss": 3.9234,
+      "step": 61455
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2919808221389342e-06,
+      "loss": 3.9594,
+      "step": 61460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2845188433327284e-06,
+      "loss": 4.0203,
+      "step": 61465
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.277065237593196e-06,
+      "loss": 3.8953,
+      "step": 61470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2696200053456966e-06,
+      "loss": 3.8906,
+      "step": 61475
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.262183147015157e-06,
+      "loss": 4.0109,
+      "step": 61480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.254754663025988e-06,
+      "loss": 3.9969,
+      "step": 61485
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2473345538021335e-06,
+      "loss": 3.9047,
+      "step": 61490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2399228197670713e-06,
+      "loss": 3.9922,
+      "step": 61495
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2325194613437634e-06,
+      "loss": 3.9688,
+      "step": 61500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2251244789547547e-06,
+      "loss": 3.9734,
+      "step": 61505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.217737873022058e-06,
+      "loss": 3.9156,
+      "step": 61510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2103596439672198e-06,
+      "loss": 3.95,
+      "step": 61515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2029897922113355e-06,
+      "loss": 3.8969,
+      "step": 61520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1956283181749864e-06,
+      "loss": 3.8312,
+      "step": 61525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.188275222278319e-06,
+      "loss": 3.9656,
+      "step": 61530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1809305049409316e-06,
+      "loss": 3.9609,
+      "step": 61535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.173594166582022e-06,
+      "loss": 3.8797,
+      "step": 61540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1662662076202716e-06,
+      "loss": 4.0016,
+      "step": 61545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1589466284738795e-06,
+      "loss": 3.9734,
+      "step": 61550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1516354295605784e-06,
+      "loss": 4.0469,
+      "step": 61555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.144332611297634e-06,
+      "loss": 3.8906,
+      "step": 61560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1370381741017803e-06,
+      "loss": 3.8984,
+      "step": 61565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1297521183893504e-06,
+      "loss": 3.875,
+      "step": 61570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1224744445761283e-06,
+      "loss": 3.9734,
+      "step": 61575
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1152051530774826e-06,
+      "loss": 4.0016,
+      "step": 61580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.107944244308247e-06,
+      "loss": 3.8875,
+      "step": 61585
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.100691718682824e-06,
+      "loss": 3.9531,
+      "step": 61590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.093447576615116e-06,
+      "loss": 3.9484,
+      "step": 61595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.086211818518508e-06,
+      "loss": 3.9438,
+      "step": 61600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0789844448059874e-06,
+      "loss": 3.975,
+      "step": 61605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0717654558900073e-06,
+      "loss": 3.9625,
+      "step": 61610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0645548521825546e-06,
+      "loss": 4.0094,
+      "step": 61615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0573526340951337e-06,
+      "loss": 3.8656,
+      "step": 61620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.050158802038799e-06,
+      "loss": 3.9078,
+      "step": 61625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.042973356424072e-06,
+      "loss": 3.9281,
+      "step": 61630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0357962976610417e-06,
+      "loss": 3.9609,
+      "step": 61635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0286276261593135e-06,
+      "loss": 4.0547,
+      "step": 61640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0214673423279933e-06,
+      "loss": 3.8937,
+      "step": 61645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0143154465757215e-06,
+      "loss": 4.0297,
+      "step": 61650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0071719393106376e-06,
+      "loss": 4.0172,
+      "step": 61655
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0000368209404656e-06,
+      "loss": 4.0719,
+      "step": 61660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9929100918723635e-06,
+      "loss": 3.9344,
+      "step": 61665
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9857917525130724e-06,
+      "loss": 3.9875,
+      "step": 61670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.978681803268851e-06,
+      "loss": 3.9531,
+      "step": 61675
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9715802445454405e-06,
+      "loss": 4.0078,
+      "step": 61680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.964487076748151e-06,
+      "loss": 3.9047,
+      "step": 61685
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.957402300281758e-06,
+      "loss": 3.9484,
+      "step": 61690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.950325915550622e-06,
+      "loss": 3.9781,
+      "step": 61695
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9432579229585862e-06,
+      "loss": 3.9141,
+      "step": 61700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.936198322909011e-06,
+      "loss": 3.925,
+      "step": 61705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.929147115804792e-06,
+      "loss": 3.9641,
+      "step": 61710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.922104302048356e-06,
+      "loss": 3.9453,
+      "step": 61715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9150698820416152e-06,
+      "loss": 3.9844,
+      "step": 61720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9080438561860154e-06,
+      "loss": 3.8672,
+      "step": 61725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9010262248825856e-06,
+      "loss": 3.9141,
+      "step": 61730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8940169885317723e-06,
+      "loss": 3.8781,
+      "step": 61735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8870161475336052e-06,
+      "loss": 3.8937,
+      "step": 61740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8800237022876314e-06,
+      "loss": 3.9203,
+      "step": 61745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8730396531929156e-06,
+      "loss": 3.9578,
+      "step": 61750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.866064000648022e-06,
+      "loss": 3.9656,
+      "step": 61755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8590967450510484e-06,
+      "loss": 4.0141,
+      "step": 61760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8521378867996438e-06,
+      "loss": 3.9453,
+      "step": 61765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8451874262909234e-06,
+      "loss": 3.9438,
+      "step": 61770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8382453639215707e-06,
+      "loss": 3.9219,
+      "step": 61775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8313117000877518e-06,
+      "loss": 3.9234,
+      "step": 61780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8243864351851996e-06,
+      "loss": 3.9109,
+      "step": 61785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8174695696090986e-06,
+      "loss": 3.9875,
+      "step": 61790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.810561103754216e-06,
+      "loss": 3.9484,
+      "step": 61795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8036610380148373e-06,
+      "loss": 3.9797,
+      "step": 61800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7967693727847298e-06,
+      "loss": 3.9484,
+      "step": 61805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7898861084571955e-06,
+      "loss": 4.0469,
+      "step": 61810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7830112454250707e-06,
+      "loss": 3.9578,
+      "step": 61815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7761447840807085e-06,
+      "loss": 3.9062,
+      "step": 61820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7692867248159778e-06,
+      "loss": 3.8813,
+      "step": 61825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.762437068022266e-06,
+      "loss": 3.9375,
+      "step": 61830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7555958140904777e-06,
+      "loss": 3.9422,
+      "step": 61835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7487629634110497e-06,
+      "loss": 3.8875,
+      "step": 61840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.741938516373954e-06,
+      "loss": 3.9516,
+      "step": 61845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7351224733686118e-06,
+      "loss": 3.9531,
+      "step": 61850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7283148347840624e-06,
+      "loss": 3.8969,
+      "step": 61855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.721515601008811e-06,
+      "loss": 3.9547,
+      "step": 61860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.714724772430865e-06,
+      "loss": 3.9547,
+      "step": 61865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.707942349437797e-06,
+      "loss": 3.9578,
+      "step": 61870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.701168332416681e-06,
+      "loss": 3.9656,
+      "step": 61875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6944027217541075e-06,
+      "loss": 3.8469,
+      "step": 61880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.687645517836168e-06,
+      "loss": 3.8937,
+      "step": 61885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.680896721048553e-06,
+      "loss": 3.9328,
+      "step": 61890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6741563317763724e-06,
+      "loss": 3.9547,
+      "step": 61895
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.667424350404301e-06,
+      "loss": 3.9281,
+      "step": 61900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.660700777316549e-06,
+      "loss": 3.9344,
+      "step": 61905
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.653985612896825e-06,
+      "loss": 3.8422,
+      "step": 61910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.647278857528373e-06,
+      "loss": 3.8563,
+      "step": 61915
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.640580511593954e-06,
+      "loss": 3.9344,
+      "step": 61920
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6338905754757946e-06,
+      "loss": 3.8922,
+      "step": 61925
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6272090495557562e-06,
+      "loss": 3.9422,
+      "step": 61930
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.620535934215118e-06,
+      "loss": 3.9844,
+      "step": 61935
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6138712298347076e-06,
+      "loss": 3.9281,
+      "step": 61940
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.607214936794888e-06,
+      "loss": 3.9703,
+      "step": 61945
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.600567055475572e-06,
+      "loss": 3.9813,
+      "step": 61950
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.593927586256106e-06,
+      "loss": 3.9438,
+      "step": 61955
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5872965295154037e-06,
+      "loss": 3.9594,
+      "step": 61960
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5806738856319287e-06,
+      "loss": 3.9922,
+      "step": 61965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.574059654983646e-06,
+      "loss": 3.9547,
+      "step": 61970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.567453837947986e-06,
+      "loss": 3.9625,
+      "step": 61975
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5608564349019813e-06,
+      "loss": 3.9625,
+      "step": 61980
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5542674462221303e-06,
+      "loss": 3.8969,
+      "step": 61985
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.547686872284449e-06,
+      "loss": 3.975,
+      "step": 61990
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.541114713464537e-06,
+      "loss": 3.8656,
+      "step": 61995
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5345509701374443e-06,
+      "loss": 4.0125,
+      "step": 62000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.527995642677738e-06,
+      "loss": 3.9703,
+      "step": 62005
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5214487314595687e-06,
+      "loss": 3.9062,
+      "step": 62010
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.514910236856571e-06,
+      "loss": 3.9734,
+      "step": 62015
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5083801592418628e-06,
+      "loss": 3.8531,
+      "step": 62020
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5018584989881462e-06,
+      "loss": 3.9609,
+      "step": 62025
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.495345256467607e-06,
+      "loss": 3.8844,
+      "step": 62030
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.488840432051947e-06,
+      "loss": 3.9516,
+      "step": 62035
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.482344026112404e-06,
+      "loss": 3.9141,
+      "step": 62040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.47585603901973e-06,
+      "loss": 3.9406,
+      "step": 62045
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4693764711441967e-06,
+      "loss": 3.9672,
+      "step": 62050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4629053228555738e-06,
+      "loss": 3.8828,
+      "step": 62055
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4564425945232004e-06,
+      "loss": 3.9953,
+      "step": 62060
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4499882865158802e-06,
+      "loss": 3.9422,
+      "step": 62065
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4435423992019697e-06,
+      "loss": 4.0156,
+      "step": 62070
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4371049329493575e-06,
+      "loss": 3.9266,
+      "step": 62075
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4306758881253995e-06,
+      "loss": 3.8719,
+      "step": 62080
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4242552650970027e-06,
+      "loss": 3.9859,
+      "step": 62085
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4178430642306232e-06,
+      "loss": 3.9594,
+      "step": 62090
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4114392858921692e-06,
+      "loss": 4.0,
+      "step": 62095
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.405043930447148e-06,
+      "loss": 3.9391,
+      "step": 62100
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3986569982604843e-06,
+      "loss": 3.8703,
+      "step": 62105
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.392278489696736e-06,
+      "loss": 4.0516,
+      "step": 62110
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.385908405119896e-06,
+      "loss": 4.0344,
+      "step": 62115
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3795467448935067e-06,
+      "loss": 3.9359,
+      "step": 62120
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3731935093806444e-06,
+      "loss": 3.9984,
+      "step": 62125
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.366848698943885e-06,
+      "loss": 3.9578,
+      "step": 62130
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3605123139453063e-06,
+      "loss": 4.0031,
+      "step": 62135
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.354184354746552e-06,
+      "loss": 3.9453,
+      "step": 62140
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3478648217087503e-06,
+      "loss": 3.9281,
+      "step": 62145
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3415537151925455e-06,
+      "loss": 3.8828,
+      "step": 62150
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3352510355581166e-06,
+      "loss": 3.9953,
+      "step": 62155
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3289567831651923e-06,
+      "loss": 4.0031,
+      "step": 62160
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.322670958372952e-06,
+      "loss": 3.9047,
+      "step": 62165
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.316393561540125e-06,
+      "loss": 3.9547,
+      "step": 62170
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3101245930249755e-06,
+      "loss": 3.9734,
+      "step": 62175
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.303864053185267e-06,
+      "loss": 3.9266,
+      "step": 62180
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2976119423783134e-06,
+      "loss": 3.9328,
+      "step": 62185
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2913682609608965e-06,
+      "loss": 3.9937,
+      "step": 62190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2851330092893305e-06,
+      "loss": 3.9891,
+      "step": 62195
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2789061877195147e-06,
+      "loss": 3.9156,
+      "step": 62200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.272687796606765e-06,
+      "loss": 3.9484,
+      "step": 62205
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.26647783630598e-06,
+      "loss": 3.975,
+      "step": 62210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.260276307171577e-06,
+      "loss": 3.9937,
+      "step": 62215
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.254083209557456e-06,
+      "loss": 3.9297,
+      "step": 62220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.247898543817084e-06,
+      "loss": 3.9094,
+      "step": 62225
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2417223103033787e-06,
+      "loss": 3.9797,
+      "step": 62230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2355545093688755e-06,
+      "loss": 3.9422,
+      "step": 62235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2293951413655255e-06,
+      "loss": 3.9453,
+      "step": 62240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2232442066448477e-06,
+      "loss": 3.9,
+      "step": 62245
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.217101705557911e-06,
+      "loss": 3.95,
+      "step": 62250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2109676384552355e-06,
+      "loss": 3.9953,
+      "step": 62255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2048420056869076e-06,
+      "loss": 3.9719,
+      "step": 62260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1987248076024976e-06,
+      "loss": 3.8969,
+      "step": 62265
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1926160445511433e-06,
+      "loss": 4.0516,
+      "step": 62270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1865157168814483e-06,
+      "loss": 3.8766,
+      "step": 62275
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.180423824941585e-06,
+      "loss": 3.8719,
+      "step": 62280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1743403690791915e-06,
+      "loss": 4.0203,
+      "step": 62285
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1682653496414738e-06,
+      "loss": 3.9672,
+      "step": 62290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1621987669751206e-06,
+      "loss": 3.9797,
+      "step": 62295
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1561406214263386e-06,
+      "loss": 3.8953,
+      "step": 62300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1500909133409015e-06,
+      "loss": 3.8703,
+      "step": 62305
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1440496430640496e-06,
+      "loss": 3.925,
+      "step": 62310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.138016810940557e-06,
+      "loss": 4.0094,
+      "step": 62315
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1319924173146986e-06,
+      "loss": 3.9859,
+      "step": 62320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1259764625303155e-06,
+      "loss": 3.9562,
+      "step": 62325
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.11996894693075e-06,
+      "loss": 3.8953,
+      "step": 62330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.113969870858795e-06,
+      "loss": 4.0438,
+      "step": 62335
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1079792346568925e-06,
+      "loss": 4.0328,
+      "step": 62340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1019970386668695e-06,
+      "loss": 3.9203,
+      "step": 62345
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.096023283230169e-06,
+      "loss": 3.8578,
+      "step": 62350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0900579686876693e-06,
+      "loss": 3.8797,
+      "step": 62355
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.084101095379864e-06,
+      "loss": 3.9516,
+      "step": 62360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0781526636466815e-06,
+      "loss": 3.9516,
+      "step": 62365
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0722126738276168e-06,
+      "loss": 3.9672,
+      "step": 62370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.066281126261632e-06,
+      "loss": 4.0281,
+      "step": 62375
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0603580212872728e-06,
+      "loss": 3.8922,
+      "step": 62380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.054443359242569e-06,
+      "loss": 3.9125,
+      "step": 62385
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0485371404650507e-06,
+      "loss": 3.9594,
+      "step": 62390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0426393652917983e-06,
+      "loss": 3.9188,
+      "step": 62395
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0367500340594256e-06,
+      "loss": 4.0141,
+      "step": 62400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0308691471039973e-06,
+      "loss": 3.95,
+      "step": 62405
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.024996704761128e-06,
+      "loss": 3.8859,
+      "step": 62410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.019132707366e-06,
+      "loss": 3.9797,
+      "step": 62415
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.013277155253262e-06,
+      "loss": 4.0078,
+      "step": 62420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.007430048757064e-06,
+      "loss": 3.8813,
+      "step": 62425
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.001591388211138e-06,
+      "loss": 3.9047,
+      "step": 62430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9957611739486677e-06,
+      "loss": 4.0281,
+      "step": 62435
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9899394063024045e-06,
+      "loss": 3.9828,
+      "step": 62440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.984126085604565e-06,
+      "loss": 3.9062,
+      "step": 62445
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.978321212186951e-06,
+      "loss": 3.9203,
+      "step": 62450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.972524786380847e-06,
+      "loss": 3.9297,
+      "step": 62455
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.966736808517039e-06,
+      "loss": 3.9703,
+      "step": 62460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.960957278925829e-06,
+      "loss": 3.9641,
+      "step": 62465
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9551861979370863e-06,
+      "loss": 3.8625,
+      "step": 62470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.949423565880165e-06,
+      "loss": 3.9438,
+      "step": 62475
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9436693830839343e-06,
+      "loss": 4.0016,
+      "step": 62480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9379236498767827e-06,
+      "loss": 3.8969,
+      "step": 62485
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9321863665866133e-06,
+      "loss": 3.8875,
+      "step": 62490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.926457533540865e-06,
+      "loss": 3.9906,
+      "step": 62495
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.920737151066476e-06,
+      "loss": 3.9891,
+      "step": 62500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9150252194899183e-06,
+      "loss": 3.9844,
+      "step": 62505
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.909321739137165e-06,
+      "loss": 3.8969,
+      "step": 62510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.903626710333722e-06,
+      "loss": 4.0078,
+      "step": 62515
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8979401334045964e-06,
+      "loss": 4.0391,
+      "step": 62520
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8922620086743112e-06,
+      "loss": 4.0094,
+      "step": 62525
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8865923364669411e-06,
+      "loss": 3.9438,
+      "step": 62530
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8809311171060436e-06,
+      "loss": 3.9281,
+      "step": 62535
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.875278350914694e-06,
+      "loss": 3.9109,
+      "step": 62540
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8696340382155172e-06,
+      "loss": 3.9391,
+      "step": 62545
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8639981793306225e-06,
+      "loss": 3.9937,
+      "step": 62550
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8583707745816524e-06,
+      "loss": 3.9156,
+      "step": 62555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8527518242897666e-06,
+      "loss": 3.9344,
+      "step": 62560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8471413287756254e-06,
+      "loss": 3.9531,
+      "step": 62565
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8415392883594394e-06,
+      "loss": 3.9875,
+      "step": 62570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8359457033609027e-06,
+      "loss": 3.8687,
+      "step": 62575
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8303605740992433e-06,
+      "loss": 3.8672,
+      "step": 62580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.824783900893223e-06,
+      "loss": 4.0672,
+      "step": 62585
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8192156840610705e-06,
+      "loss": 3.9094,
+      "step": 62590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8136559239205984e-06,
+      "loss": 3.8641,
+      "step": 62595
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8081046207890859e-06,
+      "loss": 3.9781,
+      "step": 62600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8025617749833465e-06,
+      "loss": 4.0062,
+      "step": 62605
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7970273868197106e-06,
+      "loss": 3.9047,
+      "step": 62610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.791501456614025e-06,
+      "loss": 4.0156,
+      "step": 62615
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7859839846816714e-06,
+      "loss": 3.9703,
+      "step": 62620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7804749713374978e-06,
+      "loss": 3.9766,
+      "step": 62625
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7749744168959356e-06,
+      "loss": 3.9406,
+      "step": 62630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7694823216709008e-06,
+      "loss": 3.8516,
+      "step": 62635
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7639986859758092e-06,
+      "loss": 3.9937,
+      "step": 62640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7585235101236273e-06,
+      "loss": 3.8578,
+      "step": 62645
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7530567944268214e-06,
+      "loss": 3.9359,
+      "step": 62650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.747598539197359e-06,
+      "loss": 4.0141,
+      "step": 62655
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7421487447467741e-06,
+      "loss": 3.925,
+      "step": 62660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.736707411386068e-06,
+      "loss": 3.9078,
+      "step": 62665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7312745394257754e-06,
+      "loss": 3.8859,
+      "step": 62670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7258501291759653e-06,
+      "loss": 3.9844,
+      "step": 62675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7204341809461898e-06,
+      "loss": 3.9172,
+      "step": 62680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7150266950455683e-06,
+      "loss": 3.9453,
+      "step": 62685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7096276717826707e-06,
+      "loss": 4.0156,
+      "step": 62690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7042371114656339e-06,
+      "loss": 3.9672,
+      "step": 62695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6988550144021118e-06,
+      "loss": 3.925,
+      "step": 62700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6934813808992586e-06,
+      "loss": 4.0062,
+      "step": 62705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6881162112637292e-06,
+      "loss": 3.9391,
+      "step": 62710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6827595058017118e-06,
+      "loss": 4.0203,
+      "step": 62715
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6774112648189454e-06,
+      "loss": 3.9453,
+      "step": 62720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6720714886206354e-06,
+      "loss": 4.0422,
+      "step": 62725
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.666740177511522e-06,
+      "loss": 3.8984,
+      "step": 62730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6614173317958446e-06,
+      "loss": 3.9922,
+      "step": 62735
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.656102951777427e-06,
+      "loss": 3.8203,
+      "step": 62740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.650797037759527e-06,
+      "loss": 3.9578,
+      "step": 62745
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6454995900449353e-06,
+      "loss": 3.8984,
+      "step": 62750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6402106089360268e-06,
+      "loss": 3.8516,
+      "step": 62755
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6349300947346267e-06,
+      "loss": 3.9734,
+      "step": 62760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6296580477420773e-06,
+      "loss": 3.9813,
+      "step": 62765
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6243944682592714e-06,
+      "loss": 3.9688,
+      "step": 62770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.619139356586585e-06,
+      "loss": 4.0109,
+      "step": 62775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6138927130239455e-06,
+      "loss": 3.9,
+      "step": 62780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6086545378707627e-06,
+      "loss": 4.0234,
+      "step": 62785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6034248314260145e-06,
+      "loss": 3.8578,
+      "step": 62790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5982035939881122e-06,
+      "loss": 3.9875,
+      "step": 62795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5929908258550672e-06,
+      "loss": 3.7828,
+      "step": 62800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5877865273243584e-06,
+      "loss": 3.9656,
+      "step": 62805
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5825906986929981e-06,
+      "loss": 3.8781,
+      "step": 62810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5774033402575325e-06,
+      "loss": 3.925,
+      "step": 62815
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5722244523139582e-06,
+      "loss": 3.8844,
+      "step": 62820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.567054035157872e-06,
+      "loss": 3.9937,
+      "step": 62825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5618920890843378e-06,
+      "loss": 3.9016,
+      "step": 62830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.55673861438797e-06,
+      "loss": 3.9125,
+      "step": 62835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5515936113628335e-06,
+      "loss": 3.9266,
+      "step": 62840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5464570803025933e-06,
+      "loss": 3.9109,
+      "step": 62845
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5413290215003814e-06,
+      "loss": 3.9094,
+      "step": 62850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5362094352488475e-06,
+      "loss": 3.9953,
+      "step": 62855
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5310983218401574e-06,
+      "loss": 3.9359,
+      "step": 62860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5259956815660446e-06,
+      "loss": 4.0359,
+      "step": 62865
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.520901514717676e-06,
+      "loss": 3.9844,
+      "step": 62870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5158158215857862e-06,
+      "loss": 3.8484,
+      "step": 62875
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.510738602460626e-06,
+      "loss": 3.9062,
+      "step": 62880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5056698576319637e-06,
+      "loss": 3.8906,
+      "step": 62885
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5006095873890344e-06,
+      "loss": 3.9031,
+      "step": 62890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4955577920206574e-06,
+      "loss": 4.0312,
+      "step": 62895
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4905144718151352e-06,
+      "loss": 3.9359,
+      "step": 62900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4854796270602875e-06,
+      "loss": 4.0031,
+      "step": 62905
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4804532580434514e-06,
+      "loss": 3.8719,
+      "step": 62910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4754353650514805e-06,
+      "loss": 3.9312,
+      "step": 62915
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4704259483707625e-06,
+      "loss": 4.0219,
+      "step": 62920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4654250082871688e-06,
+      "loss": 3.8687,
+      "step": 62925
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4604325450861043e-06,
+      "loss": 4.05,
+      "step": 62930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4554485590525077e-06,
+      "loss": 3.8234,
+      "step": 62935
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4504730504707851e-06,
+      "loss": 3.9719,
+      "step": 62940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.445506019624909e-06,
+      "loss": 3.9094,
+      "step": 62945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4405474667983364e-06,
+      "loss": 4.025,
+      "step": 62950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4355973922740737e-06,
+      "loss": 4.0203,
+      "step": 62955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4306557963346122e-06,
+      "loss": 3.9469,
+      "step": 62960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4257226792619426e-06,
+      "loss": 3.9359,
+      "step": 62965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4207980413376563e-06,
+      "loss": 3.9234,
+      "step": 62970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4158818828427455e-06,
+      "loss": 3.9734,
+      "step": 62975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.410974204057802e-06,
+      "loss": 4.0,
+      "step": 62980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4060750052629021e-06,
+      "loss": 3.9062,
+      "step": 62985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4011842867376556e-06,
+      "loss": 3.9734,
+      "step": 62990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3963020487611553e-06,
+      "loss": 3.9609,
+      "step": 62995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3914282916120456e-06,
+      "loss": 3.9453,
+      "step": 63000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3865630155684705e-06,
+      "loss": 3.8969,
+      "step": 63005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3817062209080908e-06,
+      "loss": 3.8484,
+      "step": 63010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3768579079080689e-06,
+      "loss": 3.9703,
+      "step": 63015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3720180768451327e-06,
+      "loss": 3.9516,
+      "step": 63020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3671867279954617e-06,
+      "loss": 3.9031,
+      "step": 63025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3623638616347853e-06,
+      "loss": 3.9625,
+      "step": 63030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3575494780383666e-06,
+      "loss": 3.9328,
+      "step": 63035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3527435774809525e-06,
+      "loss": 3.9641,
+      "step": 63040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3479461602367902e-06,
+      "loss": 3.8594,
+      "step": 63045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3431572265797108e-06,
+      "loss": 3.9844,
+      "step": 63050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3383767767829956e-06,
+      "loss": 3.9375,
+      "step": 63055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.333604811119443e-06,
+      "loss": 3.9562,
+      "step": 63060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3288413298614352e-06,
+      "loss": 3.9547,
+      "step": 63065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.324086333280805e-06,
+      "loss": 3.8828,
+      "step": 63070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3193398216489182e-06,
+      "loss": 3.9937,
+      "step": 63075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3146017952366584e-06,
+      "loss": 3.9969,
+      "step": 63080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3098722543144424e-06,
+      "loss": 3.9094,
+      "step": 63085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3051511991521546e-06,
+      "loss": 3.9562,
+      "step": 63090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3004386300192294e-06,
+      "loss": 3.8984,
+      "step": 63095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2957345471846515e-06,
+      "loss": 3.8344,
+      "step": 63100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2910389509168395e-06,
+      "loss": 3.9203,
+      "step": 63105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.286351841483796e-06,
+      "loss": 4.0047,
+      "step": 63110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2816732191530066e-06,
+      "loss": 3.9312,
+      "step": 63115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2770030841914913e-06,
+      "loss": 3.9781,
+      "step": 63120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2723414368657537e-06,
+      "loss": 3.9203,
+      "step": 63125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2676882774418472e-06,
+      "loss": 4.0,
+      "step": 63130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2630436061853433e-06,
+      "loss": 3.8969,
+      "step": 63135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2584074233612795e-06,
+      "loss": 3.9609,
+      "step": 63140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2537797292342777e-06,
+      "loss": 3.8625,
+      "step": 63145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.24916052406841e-06,
+      "loss": 3.9578,
+      "step": 63150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2445498081273319e-06,
+      "loss": 3.95,
+      "step": 63155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.23994758167415e-06,
+      "loss": 3.9641,
+      "step": 63160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.235353844971504e-06,
+      "loss": 3.9016,
+      "step": 63165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.230768598281584e-06,
+      "loss": 3.9625,
+      "step": 63170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2261918418660642e-06,
+      "loss": 3.9344,
+      "step": 63175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2216235759861358e-06,
+      "loss": 3.9891,
+      "step": 63180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.217063800902507e-06,
+      "loss": 3.9656,
+      "step": 63185
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.212512516875419e-06,
+      "loss": 3.9703,
+      "step": 63190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.207969724164598e-06,
+      "loss": 3.9781,
+      "step": 63195
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2034354230293032e-06,
+      "loss": 3.9953,
+      "step": 63200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1989096137283106e-06,
+      "loss": 3.9234,
+      "step": 63205
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1943922965199304e-06,
+      "loss": 4.0141,
+      "step": 63210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1898834716619231e-06,
+      "loss": 3.9922,
+      "step": 63215
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.185383139411633e-06,
+      "loss": 3.9547,
+      "step": 63220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1808913000259046e-06,
+      "loss": 3.9547,
+      "step": 63225
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1764079537610826e-06,
+      "loss": 3.9219,
+      "step": 63230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.171933100872996e-06,
+      "loss": 3.8766,
+      "step": 63235
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1674667416170734e-06,
+      "loss": 4.0094,
+      "step": 63240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1630088762481948e-06,
+      "loss": 3.9219,
+      "step": 63245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1585595050207563e-06,
+      "loss": 3.8203,
+      "step": 63250
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1541186281887049e-06,
+      "loss": 3.8438,
+      "step": 63255
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1496862460054713e-06,
+      "loss": 4.0328,
+      "step": 63260
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1452623587240028e-06,
+      "loss": 3.9516,
+      "step": 63265
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1408469665967811e-06,
+      "loss": 4.0078,
+      "step": 63270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1364400698758048e-06,
+      "loss": 3.9813,
+      "step": 63275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1320416688125556e-06,
+      "loss": 3.8203,
+      "step": 63280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1276517636580496e-06,
+      "loss": 3.9672,
+      "step": 63285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1232703546628362e-06,
+      "loss": 3.9484,
+      "step": 63290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1188974420769659e-06,
+      "loss": 3.8328,
+      "step": 63295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1145330261499719e-06,
+      "loss": 3.9641,
+      "step": 63300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1101771071309717e-06,
+      "loss": 3.9156,
+      "step": 63305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1058296852685167e-06,
+      "loss": 3.9016,
+      "step": 63310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1014907608107415e-06,
+      "loss": 4.0141,
+      "step": 63315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.097160334005265e-06,
+      "loss": 3.9859,
+      "step": 63320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0928384050992056e-06,
+      "loss": 3.8953,
+      "step": 63325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0885249743392498e-06,
+      "loss": 3.9703,
+      "step": 63330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0842200419715507e-06,
+      "loss": 3.9969,
+      "step": 63335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0799236082417785e-06,
+      "loss": 4.0187,
+      "step": 63340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0756356733951532e-06,
+      "loss": 4.0031,
+      "step": 63345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.071356237676363e-06,
+      "loss": 3.9422,
+      "step": 63350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0670853013296621e-06,
+      "loss": 3.9078,
+      "step": 63355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0628228645987724e-06,
+      "loss": 3.9203,
+      "step": 63360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0585689277269493e-06,
+      "loss": 3.8859,
+      "step": 63365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.054323490956982e-06,
+      "loss": 3.9828,
+      "step": 63370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0500865545311598e-06,
+      "loss": 3.9734,
+      "step": 63375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0458581186912728e-06,
+      "loss": 3.9422,
+      "step": 63380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.041638183678628e-06,
+      "loss": 3.9438,
+      "step": 63385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0374267497340993e-06,
+      "loss": 3.9438,
+      "step": 63390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0332238170979778e-06,
+      "loss": 3.9219,
+      "step": 63395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.029029386010155e-06,
+      "loss": 3.9422,
+      "step": 63400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0248434567100228e-06,
+      "loss": 3.9016,
+      "step": 63405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0206660294364565e-06,
+      "loss": 3.9469,
+      "step": 63410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0164971044278325e-06,
+      "loss": 3.8578,
+      "step": 63415
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0123366819221269e-06,
+      "loss": 3.9031,
+      "step": 63420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0081847621567495e-06,
+      "loss": 3.9922,
+      "step": 63425
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0040413453686446e-06,
+      "loss": 3.925,
+      "step": 63430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.999064317942728e-07,
+      "loss": 3.9797,
+      "step": 63435
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.957800216696288e-07,
+      "loss": 3.9266,
+      "step": 63440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.916621152302073e-07,
+      "loss": 3.9609,
+      "step": 63445
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.875527127110205e-07,
+      "loss": 3.8859,
+      "step": 63450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.834518143465643e-07,
+      "loss": 3.9641,
+      "step": 63455
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.793594203709177e-07,
+      "loss": 4.0,
+      "step": 63460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.752755310175942e-07,
+      "loss": 3.8656,
+      "step": 63465
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.712001465196905e-07,
+      "loss": 3.9391,
+      "step": 63470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.671332671097876e-07,
+      "loss": 4.0047,
+      "step": 63475
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.63074893019966e-07,
+      "loss": 3.8656,
+      "step": 63480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.590250244818575e-07,
+      "loss": 3.9172,
+      "step": 63485
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.549836617265771e-07,
+      "loss": 3.9219,
+      "step": 63490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.509508049847569e-07,
+      "loss": 4.0078,
+      "step": 63495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.469264544865629e-07,
+      "loss": 3.8328,
+      "step": 63500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.429106104616613e-07,
+      "loss": 3.9984,
+      "step": 63505
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.38903273139252e-07,
+      "loss": 3.9688,
+      "step": 63510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.34904442748019e-07,
+      "loss": 3.9406,
+      "step": 63515
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.309141195161796e-07,
+      "loss": 4.0297,
+      "step": 63520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.269323036714515e-07,
+      "loss": 3.9516,
+      "step": 63525
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.229589954411032e-07,
+      "loss": 3.9188,
+      "step": 63530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.189941950518532e-07,
+      "loss": 3.9969,
+      "step": 63535
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.150379027300203e-07,
+      "loss": 3.9656,
+      "step": 63540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.110901187013408e-07,
+      "loss": 3.9125,
+      "step": 63545
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.071508431911678e-07,
+      "loss": 3.9766,
+      "step": 63550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.032200764242714e-07,
+      "loss": 3.9266,
+      "step": 63555
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.992978186249888e-07,
+      "loss": 3.9438,
+      "step": 63560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.95384070017191e-07,
+      "loss": 3.9578,
+      "step": 63565
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.91478830824216e-07,
+      "loss": 3.9281,
+      "step": 63570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.875821012689355e-07,
+      "loss": 3.9344,
+      "step": 63575
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.836938815737382e-07,
+      "loss": 3.9078,
+      "step": 63580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.798141719605134e-07,
+      "loss": 3.9344,
+      "step": 63585
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.759429726507005e-07,
+      "loss": 3.9609,
+      "step": 63590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.720802838652064e-07,
+      "loss": 3.8391,
+      "step": 63595
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.682261058245044e-07,
+      "loss": 3.9078,
+      "step": 63600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.643804387485353e-07,
+      "loss": 3.975,
+      "step": 63605
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.605432828567571e-07,
+      "loss": 3.9078,
+      "step": 63610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.567146383681611e-07,
+      "loss": 4.0031,
+      "step": 63615
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.528945055012726e-07,
+      "loss": 3.8672,
+      "step": 63620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.490828844741004e-07,
+      "loss": 3.9203,
+      "step": 63625
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.45279775504154e-07,
+      "loss": 4.0359,
+      "step": 63630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.414851788084931e-07,
+      "loss": 3.9016,
+      "step": 63635
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.37699094603661e-07,
+      "loss": 3.9234,
+      "step": 63640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.339215231057517e-07,
+      "loss": 3.8078,
+      "step": 63645
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.301524645303259e-07,
+      "loss": 3.9797,
+      "step": 63650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.263919190925117e-07,
+      "loss": 3.9531,
+      "step": 63655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.226398870069206e-07,
+      "loss": 3.9531,
+      "step": 63660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.188963684876648e-07,
+      "loss": 3.9391,
+      "step": 63665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.151613637483734e-07,
+      "loss": 3.9203,
+      "step": 63670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.114348730022424e-07,
+      "loss": 3.8672,
+      "step": 63675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.077168964619185e-07,
+      "loss": 3.9391,
+      "step": 63680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.040074343395986e-07,
+      "loss": 3.9438,
+      "step": 63685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.003064868469799e-07,
+      "loss": 3.8859,
+      "step": 63690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.966140541952603e-07,
+      "loss": 3.9156,
+      "step": 63695
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.929301365951879e-07,
+      "loss": 3.9266,
+      "step": 63700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.892547342569944e-07,
+      "loss": 3.9547,
+      "step": 63705
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.855878473904287e-07,
+      "loss": 3.9219,
+      "step": 63710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.819294762047901e-07,
+      "loss": 3.8453,
+      "step": 63715
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.782796209088282e-07,
+      "loss": 4.0172,
+      "step": 63720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.746382817108432e-07,
+      "loss": 3.9547,
+      "step": 63725
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.710054588186687e-07,
+      "loss": 3.975,
+      "step": 63730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.673811524396056e-07,
+      "loss": 3.9844,
+      "step": 63735
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.637653627805051e-07,
+      "loss": 3.8937,
+      "step": 63740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.601580900477356e-07,
+      "loss": 3.9375,
+      "step": 63745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.565593344471488e-07,
+      "loss": 3.9125,
+      "step": 63750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.529690961841306e-07,
+      "loss": 3.9297,
+      "step": 63755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.493873754635504e-07,
+      "loss": 3.8359,
+      "step": 63760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.458141724898615e-07,
+      "loss": 4.0187,
+      "step": 63765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.422494874669672e-07,
+      "loss": 3.9781,
+      "step": 63770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.386933205982881e-07,
+      "loss": 3.8438,
+      "step": 63775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.351456720867954e-07,
+      "loss": 3.8859,
+      "step": 63780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.316065421349437e-07,
+      "loss": 3.9219,
+      "step": 63785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.280759309447215e-07,
+      "loss": 3.9406,
+      "step": 63790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.245538387176175e-07,
+      "loss": 3.9219,
+      "step": 63795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.210402656546377e-07,
+      "loss": 3.9531,
+      "step": 63800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.175352119562883e-07,
+      "loss": 4.0187,
+      "step": 63805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.140386778226259e-07,
+      "loss": 3.9813,
+      "step": 63810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.105506634531744e-07,
+      "loss": 3.9438,
+      "step": 63815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.070711690470243e-07,
+      "loss": 3.8687,
+      "step": 63820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.036001948027336e-07,
+      "loss": 3.9969,
+      "step": 63825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.001377409183772e-07,
+      "loss": 3.9937,
+      "step": 63830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.966838075915804e-07,
+      "loss": 3.9891,
+      "step": 63835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.93238395019452e-07,
+      "loss": 3.9484,
+      "step": 63840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.898015033986181e-07,
+      "loss": 3.8844,
+      "step": 63845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.86373132925222e-07,
+      "loss": 3.9922,
+      "step": 63850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.829532837949236e-07,
+      "loss": 3.875,
+      "step": 63855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.79541956202917e-07,
+      "loss": 3.9344,
+      "step": 63860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.761391503438296e-07,
+      "loss": 3.9109,
+      "step": 63865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.727448664119229e-07,
+      "loss": 3.975,
+      "step": 63870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.693591046008751e-07,
+      "loss": 3.9141,
+      "step": 63875
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.659818651039151e-07,
+      "loss": 3.9062,
+      "step": 63880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.626131481137887e-07,
+      "loss": 3.9344,
+      "step": 63885
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.59252953822742e-07,
+      "loss": 3.9359,
+      "step": 63890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.559012824225551e-07,
+      "loss": 3.8359,
+      "step": 63895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.525581341044916e-07,
+      "loss": 3.9359,
+      "step": 63900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.49223509059349e-07,
+      "loss": 3.9516,
+      "step": 63905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.458974074774415e-07,
+      "loss": 3.9266,
+      "step": 63910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.425798295485841e-07,
+      "loss": 3.9859,
+      "step": 63915
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.392707754621085e-07,
+      "loss": 3.9828,
+      "step": 63920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.359702454068638e-07,
+      "loss": 3.9234,
+      "step": 63925
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.326782395712326e-07,
+      "loss": 3.9703,
+      "step": 63930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.293947581430647e-07,
+      "loss": 3.8922,
+      "step": 63935
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.261198013097601e-07,
+      "loss": 3.9703,
+      "step": 63940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.228533692582194e-07,
+      "loss": 3.9531,
+      "step": 63945
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.195954621748434e-07,
+      "loss": 3.95,
+      "step": 63950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.163460802455833e-07,
+      "loss": 3.8875,
+      "step": 63955
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.131052236558742e-07,
+      "loss": 4.0062,
+      "step": 63960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.098728925906682e-07,
+      "loss": 3.8781,
+      "step": 63965
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.066490872344343e-07,
+      "loss": 3.9094,
+      "step": 63970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.034338077711586e-07,
+      "loss": 3.9141,
+      "step": 63975
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.002270543843446e-07,
+      "loss": 3.9781,
+      "step": 63980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.970288272569956e-07,
+      "loss": 3.9906,
+      "step": 63985
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.938391265716158e-07,
+      "loss": 3.9359,
+      "step": 63990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.906579525102761e-07,
+      "loss": 3.9312,
+      "step": 63995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.874853052545148e-07,
+      "loss": 3.9484,
+      "step": 64000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.843211849853702e-07,
+      "loss": 3.9156,
+      "step": 64005
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.811655918834479e-07,
+      "loss": 3.8953,
+      "step": 64010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.780185261288373e-07,
+      "loss": 3.9641,
+      "step": 64015
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.748799879011279e-07,
+      "loss": 3.9937,
+      "step": 64020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.717499773794431e-07,
+      "loss": 3.925,
+      "step": 64025
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.686284947424069e-07,
+      "loss": 3.9672,
+      "step": 64030
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.655155401681599e-07,
+      "loss": 3.9906,
+      "step": 64035
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.624111138343768e-07,
+      "loss": 3.9109,
+      "step": 64040
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.593152159182157e-07,
+      "loss": 3.9578,
+      "step": 64045
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.562278465963521e-07,
+      "loss": 3.8922,
+      "step": 64050
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.531490060450117e-07,
+      "loss": 3.9859,
+      "step": 64055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.500786944398705e-07,
+      "loss": 3.8641,
+      "step": 64060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.470169119561552e-07,
+      "loss": 3.9531,
+      "step": 64065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.439636587686259e-07,
+      "loss": 4.025,
+      "step": 64070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.409189350515263e-07,
+      "loss": 3.9828,
+      "step": 64075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.378827409786013e-07,
+      "loss": 3.9453,
+      "step": 64080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.348550767231286e-07,
+      "loss": 3.9953,
+      "step": 64085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.318359424579033e-07,
+      "loss": 3.8859,
+      "step": 64090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.288253383552377e-07,
+      "loss": 3.975,
+      "step": 64095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.258232645869442e-07,
+      "loss": 3.9813,
+      "step": 64100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.22829721324336e-07,
+      "loss": 3.9375,
+      "step": 64105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.198447087382596e-07,
+      "loss": 3.9203,
+      "step": 64110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.168682269990954e-07,
+      "loss": 3.9766,
+      "step": 64115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.139002762766742e-07,
+      "loss": 4.0062,
+      "step": 64120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.109408567403939e-07,
+      "loss": 3.9156,
+      "step": 64125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.079899685591693e-07,
+      "loss": 3.9703,
+      "step": 64130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.050476119013658e-07,
+      "loss": 3.9484,
+      "step": 64135
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.021137869349323e-07,
+      "loss": 4.0047,
+      "step": 64140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.991884938273016e-07,
+      "loss": 3.9281,
+      "step": 64145
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.962717327454069e-07,
+      "loss": 3.975,
+      "step": 64150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.933635038557149e-07,
+      "loss": 3.9078,
+      "step": 64155
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.904638073242095e-07,
+      "loss": 3.9594,
+      "step": 64160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.875726433163585e-07,
+      "loss": 3.95,
+      "step": 64165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.846900119971797e-07,
+      "loss": 3.8875,
+      "step": 64170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.81815913531175e-07,
+      "loss": 3.9641,
+      "step": 64175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.789503480823631e-07,
+      "loss": 3.9109,
+      "step": 64180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.760933158142799e-07,
+      "loss": 3.9078,
+      "step": 64185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7324481688999496e-07,
+      "loss": 3.9,
+      "step": 64190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.704048514720615e-07,
+      "loss": 3.9797,
+      "step": 64195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6757341972256666e-07,
+      "loss": 3.9047,
+      "step": 64200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6475052180308116e-07,
+      "loss": 3.9703,
+      "step": 64205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.619361578747094e-07,
+      "loss": 3.8953,
+      "step": 64210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5913032809808957e-07,
+      "loss": 4.0187,
+      "step": 64215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.56333032633327e-07,
+      "loss": 3.9703,
+      "step": 64220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5354427164007723e-07,
+      "loss": 3.9609,
+      "step": 64225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.507640452774963e-07,
+      "loss": 3.9844,
+      "step": 64230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4799235370425733e-07,
+      "loss": 3.9641,
+      "step": 64235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.452291970785171e-07,
+      "loss": 3.9375,
+      "step": 64240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4247457555798284e-07,
+      "loss": 4.0125,
+      "step": 64245
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3972848929986206e-07,
+      "loss": 3.9922,
+      "step": 64250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3699093846087943e-07,
+      "loss": 3.9719,
+      "step": 64255
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3426192319725996e-07,
+      "loss": 3.9234,
+      "step": 64260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3154144366474574e-07,
+      "loss": 3.8469,
+      "step": 64265
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.288295000186126e-07,
+      "loss": 3.9438,
+      "step": 64270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.261260924136034e-07,
+      "loss": 3.9781,
+      "step": 64275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.23431221004028e-07,
+      "loss": 3.9797,
+      "step": 64280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.207448859436635e-07,
+      "loss": 3.9859,
+      "step": 64285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.180670873858205e-07,
+      "loss": 3.9672,
+      "step": 64290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.1539782548334344e-07,
+      "loss": 3.9719,
+      "step": 64295
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.127371003885438e-07,
+      "loss": 3.95,
+      "step": 64300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.1008491225328345e-07,
+      "loss": 4.0109,
+      "step": 64305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.07441261228908e-07,
+      "loss": 3.9156,
+      "step": 64310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.048061474663133e-07,
+      "loss": 4.0062,
+      "step": 64315
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.021795711158626e-07,
+      "loss": 3.9641,
+      "step": 64320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9956153232746925e-07,
+      "loss": 3.9516,
+      "step": 64325
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.969520312505303e-07,
+      "loss": 4.0016,
+      "step": 64330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9435106803399343e-07,
+      "loss": 3.9656,
+      "step": 64335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9175864282627313e-07,
+      "loss": 3.9781,
+      "step": 64340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.891747557753344e-07,
+      "loss": 3.8078,
+      "step": 64345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.865994070286427e-07,
+      "loss": 3.9578,
+      "step": 64350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.84032596733147e-07,
+      "loss": 3.9547,
+      "step": 64355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.814743250353636e-07,
+      "loss": 3.9625,
+      "step": 64360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7892459208129223e-07,
+      "loss": 4.0031,
+      "step": 64365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.763833980164166e-07,
+      "loss": 3.9172,
+      "step": 64370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7385074298580396e-07,
+      "loss": 3.9281,
+      "step": 64375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.713266271339554e-07,
+      "loss": 3.9266,
+      "step": 64380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.688110506049557e-07,
+      "loss": 3.9594,
+      "step": 64385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6630401354234005e-07,
+      "loss": 4.0656,
+      "step": 64390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.638055160892106e-07,
+      "loss": 3.9203,
+      "step": 64395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6131555838813664e-07,
+      "loss": 3.9562,
+      "step": 64400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5883414058123785e-07,
+      "loss": 3.8797,
+      "step": 64405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.563612628101176e-07,
+      "loss": 3.8266,
+      "step": 64410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.538969252158963e-07,
+      "loss": 3.9641,
+      "step": 64415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.514411279392282e-07,
+      "loss": 3.8891,
+      "step": 64420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.489938711202511e-07,
+      "loss": 3.9344,
+      "step": 64425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4655515489865336e-07,
+      "loss": 4.0219,
+      "step": 64430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.441249794136069e-07,
+      "loss": 3.8609,
+      "step": 64435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4170334480378424e-07,
+      "loss": 3.9234,
+      "step": 64440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.392902512073914e-07,
+      "loss": 4.0687,
+      "step": 64445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3688569876215157e-07,
+      "loss": 3.9328,
+      "step": 64450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3448968760530494e-07,
+      "loss": 3.9734,
+      "step": 64455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3210221787355885e-07,
+      "loss": 3.9562,
+      "step": 64460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2972328970320425e-07,
+      "loss": 3.9484,
+      "step": 64465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2735290322998265e-07,
+      "loss": 3.9016,
+      "step": 64470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.249910585891857e-07,
+      "loss": 3.9516,
+      "step": 64475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2263775591558903e-07,
+      "loss": 3.9625,
+      "step": 64480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.202929953435185e-07,
+      "loss": 3.8609,
+      "step": 64485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1795677700675035e-07,
+      "loss": 3.9156,
+      "step": 64490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1562910103866133e-07,
+      "loss": 3.9078,
+      "step": 64495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1330996757206184e-07,
+      "loss": 3.9391,
+      "step": 64500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.10999376739296e-07,
+      "loss": 3.9062,
+      "step": 64505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.086973286722749e-07,
+      "loss": 3.9094,
+      "step": 64510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0640382350232695e-07,
+      "loss": 3.9672,
+      "step": 64515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.04118861360364e-07,
+      "loss": 3.9234,
+      "step": 64520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0184244237679845e-07,
+      "loss": 3.9766,
+      "step": 64525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9957456668152633e-07,
+      "loss": 3.9484,
+      "step": 64530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9731523440397753e-07,
+      "loss": 3.9578,
+      "step": 64535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.950644456731155e-07,
+      "loss": 3.9562,
+      "step": 64540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9282220061737085e-07,
+      "loss": 3.9672,
+      "step": 64545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9058849936470784e-07,
+      "loss": 3.9047,
+      "step": 64550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8836334204262457e-07,
+      "loss": 3.9562,
+      "step": 64555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.861467287780861e-07,
+      "loss": 3.9953,
+      "step": 64560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8393865969760786e-07,
+      "loss": 3.9656,
+      "step": 64565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.817391349272058e-07,
+      "loss": 3.8813,
+      "step": 64570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.795481545923961e-07,
+      "loss": 3.9656,
+      "step": 64575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7736571881824545e-07,
+      "loss": 3.9859,
+      "step": 64580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7519182772927083e-07,
+      "loss": 3.9438,
+      "step": 64585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7302648144955643e-07,
+      "loss": 3.9219,
+      "step": 64590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7086968010266997e-07,
+      "loss": 3.8937,
+      "step": 64595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.687214238116964e-07,
+      "loss": 3.9219,
+      "step": 64600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.665817126992542e-07,
+      "loss": 3.8828,
+      "step": 64605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6445054688744583e-07,
+      "loss": 3.9297,
+      "step": 64610
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6232792649790724e-07,
+      "loss": 3.9719,
+      "step": 64615
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.602138516517749e-07,
+      "loss": 3.8219,
+      "step": 64620
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5810832246966894e-07,
+      "loss": 3.9906,
+      "step": 64625
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5601133907179326e-07,
+      "loss": 3.9516,
+      "step": 64630
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.539229015778022e-07,
+      "loss": 3.9656,
+      "step": 64635
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5184301010686694e-07,
+      "loss": 3.9469,
+      "step": 64640
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.497716647777259e-07,
+      "loss": 3.8859,
+      "step": 64645
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4770886570856794e-07,
+      "loss": 3.9547,
+      "step": 64650
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.456546130171322e-07,
+      "loss": 4.0359,
+      "step": 64655
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4360890682062484e-07,
+      "loss": 3.9609,
+      "step": 64660
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4157174723581916e-07,
+      "loss": 4.0328,
+      "step": 64665
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.395431343789722e-07,
+      "loss": 4.0,
+      "step": 64670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.37523068365858e-07,
+      "loss": 3.9391,
+      "step": 64675
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3551154931176763e-07,
+      "loss": 3.8062,
+      "step": 64680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.33508577331476e-07,
+      "loss": 3.9891,
+      "step": 64685
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3151415253930827e-07,
+      "loss": 3.9516,
+      "step": 64690
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.295282750490901e-07,
+      "loss": 3.9719,
+      "step": 64695
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.275509449741475e-07,
+      "loss": 3.8484,
+      "step": 64700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.255821624273235e-07,
+      "loss": 3.9578,
+      "step": 64705
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2362192752099495e-07,
+      "loss": 3.9219,
+      "step": 64710
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.216702403670223e-07,
+      "loss": 3.9172,
+      "step": 64715
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1972710107678315e-07,
+      "loss": 3.9672,
+      "step": 64720
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1779250976117212e-07,
+      "loss": 3.9609,
+      "step": 64725
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1586646653060093e-07,
+      "loss": 3.9078,
+      "step": 64730
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1394897149499823e-07,
+      "loss": 3.9328,
+      "step": 64735
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1204002476377656e-07,
+      "loss": 4.05,
+      "step": 64740
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1013962644589876e-07,
+      "loss": 3.9953,
+      "step": 64745
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.082477766498114e-07,
+      "loss": 3.925,
+      "step": 64750
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0636447548347813e-07,
+      "loss": 3.9281,
+      "step": 64755
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.044897230543796e-07,
+      "loss": 3.9203,
+      "step": 64760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.026235194695136e-07,
+      "loss": 3.9344,
+      "step": 64765
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0076586483539492e-07,
+      "loss": 3.8281,
+      "step": 64770
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.989167592580221e-07,
+      "loss": 3.8859,
+      "step": 64775
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.97076202842944e-07,
+      "loss": 3.8969,
+      "step": 64780
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9524419569516004e-07,
+      "loss": 3.8937,
+      "step": 64785
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9342073791926982e-07,
+      "loss": 3.9922,
+      "step": 64790
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9160582961932348e-07,
+      "loss": 3.9609,
+      "step": 64795
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.897994708988715e-07,
+      "loss": 3.9609,
+      "step": 64800
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8800166186104804e-07,
+      "loss": 3.9422,
+      "step": 64805
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8621240260842107e-07,
+      "loss": 3.8844,
+      "step": 64810
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.844316932431089e-07,
+      "loss": 3.9375,
+      "step": 64815
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8265953386674692e-07,
+      "loss": 3.9906,
+      "step": 64820
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8089592458047085e-07,
+      "loss": 3.8844,
+      "step": 64825
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7914086548493355e-07,
+      "loss": 3.9937,
+      "step": 64830
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7739435668027156e-07,
+      "loss": 3.9422,
+      "step": 64835
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.756563982661885e-07,
+      "loss": 3.9766,
+      "step": 64840
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.73926990341855e-07,
+      "loss": 3.9281,
+      "step": 64845
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.722061330059754e-07,
+      "loss": 4.025,
+      "step": 64850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7049382635673793e-07,
+      "loss": 4.0062,
+      "step": 64855
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6879007049189764e-07,
+      "loss": 3.8969,
+      "step": 64860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6709486550866013e-07,
+      "loss": 3.9984,
+      "step": 64865
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.65408211503798e-07,
+      "loss": 4.0062,
+      "step": 64870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.637301085735343e-07,
+      "loss": 3.9234,
+      "step": 64875
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.620605568136757e-07,
+      "loss": 4.0125,
+      "step": 64880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6039955631947932e-07,
+      "loss": 3.9469,
+      "step": 64885
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.58747107185736e-07,
+      "loss": 3.9562,
+      "step": 64890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.571032095067537e-07,
+      "loss": 4.0172,
+      "step": 64895
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5546786337637396e-07,
+      "loss": 3.8953,
+      "step": 64900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5384106888790548e-07,
+      "loss": 3.9609,
+      "step": 64905
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5222282613419067e-07,
+      "loss": 3.8891,
+      "step": 64910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5061313520757235e-07,
+      "loss": 3.9922,
+      "step": 64915
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4901199619994363e-07,
+      "loss": 4.0563,
+      "step": 64920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4741940920264817e-07,
+      "loss": 3.9844,
+      "step": 64925
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4583537430661318e-07,
+      "loss": 3.9672,
+      "step": 64930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4425989160219974e-07,
+      "loss": 4.0,
+      "step": 64935
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.426929611793526e-07,
+      "loss": 3.8531,
+      "step": 64940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4113458312748348e-07,
+      "loss": 3.9094,
+      "step": 64945
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3958475753552134e-07,
+      "loss": 3.9344,
+      "step": 64950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.380434844919287e-07,
+      "loss": 4.0578,
+      "step": 64955
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3651076408466855e-07,
+      "loss": 3.9625,
+      "step": 64960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3498659640118758e-07,
+      "loss": 3.9781,
+      "step": 64965
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3347098152849957e-07,
+      "loss": 3.9984,
+      "step": 64970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3196391955310193e-07,
+      "loss": 3.9172,
+      "step": 64975
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3046541056097592e-07,
+      "loss": 3.8891,
+      "step": 64980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2897545463765313e-07,
+      "loss": 4.0062,
+      "step": 64985
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2749405186816554e-07,
+      "loss": 3.9813,
+      "step": 64990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.260212023370788e-07,
+      "loss": 3.9688,
+      "step": 64995
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2455690612840908e-07,
+      "loss": 3.9719,
+      "step": 65000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2310116332575616e-07,
+      "loss": 3.9391,
+      "step": 65005
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.216539740121869e-07,
+      "loss": 3.9156,
+      "step": 65010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2021533827028528e-07,
+      "loss": 3.9375,
+      "step": 65015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1878525618216894e-07,
+      "loss": 3.9344,
+      "step": 65020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1736372782943926e-07,
+      "loss": 3.8594,
+      "step": 65025
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1595075329324799e-07,
+      "loss": 3.9422,
+      "step": 65030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1454633265419732e-07,
+      "loss": 4.0438,
+      "step": 65035
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1315046599247313e-07,
+      "loss": 3.9344,
+      "step": 65040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.117631533877117e-07,
+      "loss": 3.8984,
+      "step": 65045
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1038439491908302e-07,
+      "loss": 3.8844,
+      "step": 65050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0901419066530748e-07,
+      "loss": 3.9813,
+      "step": 65055
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0765254070453922e-07,
+      "loss": 3.8969,
+      "step": 65060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.062994451145327e-07,
+      "loss": 3.9203,
+      "step": 65065
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0495490397247619e-07,
+      "loss": 3.9047,
+      "step": 65070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.036189173551083e-07,
+      "loss": 3.9656,
+      "step": 65075
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0229148533868469e-07,
+      "loss": 3.9547,
+      "step": 65080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0097260799896145e-07,
+      "loss": 3.9312,
+      "step": 65085
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.966228541117838e-08,
+      "loss": 3.9766,
+      "step": 65090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.836051765015896e-08,
+      "loss": 3.9406,
+      "step": 65095
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.70673047901771e-08,
+      "loss": 3.9109,
+      "step": 65100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.578264690502379e-08,
+      "loss": 3.9562,
+      "step": 65105
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.450654406802372e-08,
+      "loss": 3.9406,
+      "step": 65110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.323899635201859e-08,
+      "loss": 4.0453,
+      "step": 65115
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.198000382933391e-08,
+      "loss": 3.9328,
+      "step": 65120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.072956657181219e-08,
+      "loss": 3.8828,
+      "step": 65125
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.948768465082967e-08,
+      "loss": 3.9516,
+      "step": 65130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.825435813724636e-08,
+      "loss": 3.8922,
+      "step": 65135
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.702958710147256e-08,
+      "loss": 3.9703,
+      "step": 65140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.581337161338575e-08,
+      "loss": 3.9922,
+      "step": 65145
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.460571174239706e-08,
+      "loss": 3.9031,
+      "step": 65150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.340660755743466e-08,
+      "loss": 3.9656,
+      "step": 65155
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.221605912692719e-08,
+      "loss": 4.0094,
+      "step": 65160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.103406651882028e-08,
+      "loss": 3.9828,
+      "step": 65165
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.986062980057661e-08,
+      "loss": 3.9266,
+      "step": 65170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.869574903915932e-08,
+      "loss": 3.9172,
+      "step": 65175
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.753942430104854e-08,
+      "loss": 3.8828,
+      "step": 65180
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.639165565222482e-08,
+      "loss": 3.9281,
+      "step": 65185
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.52524431582191e-08,
+      "loss": 3.9969,
+      "step": 65190
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.412178688401271e-08,
+      "loss": 3.9406,
+      "step": 65195
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.299968689415403e-08,
+      "loss": 3.9703,
+      "step": 65200
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.188614325267517e-08,
+      "loss": 3.8797,
+      "step": 65205
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.078115602312528e-08,
+      "loss": 3.9094,
+      "step": 65210
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.96847252685706e-08,
+      "loss": 4.0078,
+      "step": 65215
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.859685105157775e-08,
+      "loss": 3.8766,
+      "step": 65220
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.751753343424704e-08,
+      "loss": 3.9078,
+      "step": 65225
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.64467724781459e-08,
+      "loss": 3.8828,
+      "step": 65230
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.538456824442539e-08,
+      "loss": 3.9937,
+      "step": 65235
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.433092079365376e-08,
+      "loss": 3.9906,
+      "step": 65240
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.328583018601618e-08,
+      "loss": 3.9062,
+      "step": 65245
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.224929648111498e-08,
+      "loss": 3.9312,
+      "step": 65250
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.12213197381195e-08,
+      "loss": 3.9609,
+      "step": 65255
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.020190001569947e-08,
+      "loss": 3.9219,
+      "step": 65260
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9191037372025015e-08,
+      "loss": 3.9156,
+      "step": 65265
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.8188731864799995e-08,
+      "loss": 3.9094,
+      "step": 65270
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.719498355121199e-08,
+      "loss": 4.0047,
+      "step": 65275
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.6209792487982297e-08,
+      "loss": 3.9234,
+      "step": 65280
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.523315873134926e-08,
+      "loss": 3.9641,
+      "step": 65285
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.426508233701832e-08,
+      "loss": 3.8219,
+      "step": 65290
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3305563360245274e-08,
+      "loss": 4.0031,
+      "step": 65295
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2354601855819633e-08,
+      "loss": 3.8563,
+      "step": 65300
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.1412197877981345e-08,
+      "loss": 3.9766,
+      "step": 65305
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0478351480520706e-08,
+      "loss": 3.9453,
+      "step": 65310
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.9553062716745086e-08,
+      "loss": 3.9031,
+      "step": 65315
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.863633163944558e-08,
+      "loss": 3.9937,
+      "step": 65320
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.772815830093035e-08,
+      "loss": 3.9781,
+      "step": 65325
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.682854275305792e-08,
+      "loss": 3.9734,
+      "step": 65330
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.593748504717054e-08,
+      "loss": 3.9359,
+      "step": 65335
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.505498523409423e-08,
+      "loss": 3.9359,
+      "step": 65340
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4181043364188705e-08,
+      "loss": 3.7641,
+      "step": 65345
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.331565948736404e-08,
+      "loss": 3.925,
+      "step": 65350
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.2458833652997405e-08,
+      "loss": 3.8672,
+      "step": 65355
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.161056590996636e-08,
+      "loss": 3.9406,
+      "step": 65360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.077085630669885e-08,
+      "loss": 3.9453,
+      "step": 65365
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.993970489110654e-08,
+      "loss": 3.9094,
+      "step": 65370
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.911711171065146e-08,
+      "loss": 3.8953,
+      "step": 65375
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.830307681224609e-08,
+      "loss": 3.9406,
+      "step": 65380
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7497600242353266e-08,
+      "loss": 3.9203,
+      "step": 65385
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.670068204696952e-08,
+      "loss": 3.9922,
+      "step": 65390
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5912322271541836e-08,
+      "loss": 3.8563,
+      "step": 65395
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.513252096106756e-08,
+      "loss": 3.9469,
+      "step": 65400
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4361278160061065e-08,
+      "loss": 3.9594,
+      "step": 65405
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.359859391253716e-08,
+      "loss": 3.9281,
+      "step": 65410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.284446826201104e-08,
+      "loss": 3.8937,
+      "step": 65415
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.209890125153158e-08,
+      "loss": 3.9219,
+      "step": 65420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.13618929236481e-08,
+      "loss": 3.9875,
+      "step": 65425
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.063344332041029e-08,
+      "loss": 3.9953,
+      "step": 65430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9913552483401546e-08,
+      "loss": 3.9281,
+      "step": 65435
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9202220453705684e-08,
+      "loss": 3.8984,
+      "step": 65440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.849944727192355e-08,
+      "loss": 3.9188,
+      "step": 65445
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7805232978139745e-08,
+      "loss": 3.9406,
+      "step": 65450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7119577612005893e-08,
+      "loss": 3.9469,
+      "step": 65455
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6442481212624044e-08,
+      "loss": 3.9328,
+      "step": 65460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5773943818646615e-08,
+      "loss": 4.0203,
+      "step": 65465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5113965468226417e-08,
+      "loss": 3.9594,
+      "step": 65470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4462546199033317e-08,
+      "loss": 3.925,
+      "step": 65475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3819686048237587e-08,
+      "loss": 3.9703,
+      "step": 65480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.318538505252654e-08,
+      "loss": 3.9656,
+      "step": 65485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2559643248104553e-08,
+      "loss": 3.9656,
+      "step": 65490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1942460670693052e-08,
+      "loss": 3.9297,
+      "step": 65495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1333837355480553e-08,
+      "loss": 3.9828,
+      "step": 65500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0733773337239245e-08,
+      "loss": 3.9031,
+      "step": 65505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0142268650208403e-08,
+      "loss": 3.8656,
+      "step": 65510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9559323328111054e-08,
+      "loss": 3.9656,
+      "step": 65515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8984937404253886e-08,
+      "loss": 4.0578,
+      "step": 65520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8419110911410684e-08,
+      "loss": 3.8906,
+      "step": 65525
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.786184388185563e-08,
+      "loss": 3.8859,
+      "step": 65530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.731313634741327e-08,
+      "loss": 3.8766,
+      "step": 65535
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.677298833937524e-08,
+      "loss": 3.9891,
+      "step": 65540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.624139988860018e-08,
+      "loss": 3.8844,
+      "step": 65545
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5718371025397192e-08,
+      "loss": 3.9047,
+      "step": 65550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5203901779609062e-08,
+      "loss": 4.0109,
+      "step": 65555
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4697992180628948e-08,
+      "loss": 3.9141,
+      "step": 65560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4200642257317096e-08,
+      "loss": 4.0078,
+      "step": 65565
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3711852038050808e-08,
+      "loss": 4.0234,
+      "step": 65570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3231621550707782e-08,
+      "loss": 3.9578,
+      "step": 65575
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2759950822732734e-08,
+      "loss": 3.9438,
+      "step": 65580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2296839881020815e-08,
+      "loss": 3.8813,
+      "step": 65585
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1842288752017536e-08,
+      "loss": 3.975,
+      "step": 65590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1396297461635506e-08,
+      "loss": 3.9562,
+      "step": 65595
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.095886603535434e-08,
+      "loss": 3.9297,
+      "step": 65600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0529994498137407e-08,
+      "loss": 3.9297,
+      "step": 65605
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0109682874448466e-08,
+      "loss": 3.8656,
+      "step": 65610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.697931188268337e-09,
+      "loss": 3.9672,
+      "step": 65615
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.294739463111544e-09,
+      "loss": 4.0687,
+      "step": 65620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.90010772199301e-09,
+      "loss": 3.9688,
+      "step": 65625
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.514035987411406e-09,
+      "loss": 4.0,
+      "step": 65630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.136524281415757e-09,
+      "loss": 3.9594,
+      "step": 65635
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.767572625538842e-09,
+      "loss": 3.9813,
+      "step": 65640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.407181040863796e-09,
+      "loss": 3.9922,
+      "step": 65645
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.055349547924194e-09,
+      "loss": 3.9547,
+      "step": 65650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.712078166820623e-09,
+      "loss": 3.9422,
+      "step": 65655
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.377366917137416e-09,
+      "loss": 3.9016,
+      "step": 65660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.051215817992616e-09,
+      "loss": 3.9953,
+      "step": 65665
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.733624887954702e-09,
+      "loss": 3.9625,
+      "step": 65670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.4245941452091265e-09,
+      "loss": 3.8953,
+      "step": 65675
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.124123607341823e-09,
+      "loss": 3.8906,
+      "step": 65680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.832213291522391e-09,
+      "loss": 4.0062,
+      "step": 65685
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.548863214404175e-09,
+      "loss": 3.9172,
+      "step": 65690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.274073392174226e-09,
+      "loss": 3.8906,
+      "step": 65695
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.007843840486691e-09,
+      "loss": 3.9469,
+      "step": 65700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.750174574562725e-09,
+      "loss": 3.9125,
+      "step": 65705
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.501065609090581e-09,
+      "loss": 3.9562,
+      "step": 65710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.260516958292214e-09,
+      "loss": 4.1094,
+      "step": 65715
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0285286358899817e-09,
+      "loss": 3.9578,
+      "step": 65720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.805100655139947e-09,
+      "loss": 3.8578,
+      "step": 65725
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5902330287819183e-09,
+      "loss": 4.0125,
+      "step": 65730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3839257690727587e-09,
+      "loss": 3.9734,
+      "step": 65735
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.186178887786383e-09,
+      "loss": 3.8281,
+      "step": 65740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9969923962304123e-09,
+      "loss": 3.7969,
+      "step": 65745
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8163663051795619e-09,
+      "loss": 3.9891,
+      "step": 65750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6443006249422518e-09,
+      "loss": 4.0,
+      "step": 65755
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4807953653439564e-09,
+      "loss": 3.8641,
+      "step": 65760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.325850535727202e-09,
+      "loss": 4.0172,
+      "step": 65765
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1794661449182619e-09,
+      "loss": 3.9844,
+      "step": 65770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0416422012771154e-09,
+      "loss": 3.9578,
+      "step": 65775
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.123787126641413e-10,
+      "loss": 4.0141,
+      "step": 65780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.916756864567719e-10,
+      "loss": 3.8875,
+      "step": 65785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.795331295661454e-10,
+      "loss": 3.9234,
+      "step": 65790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.759510483538398e-10,
+      "loss": 3.9344,
+      "step": 65795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.809294487650994e-10,
+      "loss": 3.9609,
+      "step": 65800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9446833621226137e-10,
+      "loss": 3.925,
+      "step": 65805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.165677156080626e-10,
+      "loss": 3.9766,
+      "step": 65810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.47227591432253e-10,
+      "loss": 3.9734,
+      "step": 65815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8644796763167548e-10,
+      "loss": 4.0187,
+      "step": 65820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3422884765357243e-10,
+      "loss": 3.9406,
+      "step": 65825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.057023451219947e-11,
+      "loss": 4.0469,
+      "step": 65830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.547213067225165e-11,
+      "loss": 3.9516,
+      "step": 65835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.893453814878377e-11,
+      "loss": 4.0,
+      "step": 65840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0957458457250267e-11,
+      "loss": 3.9547,
+      "step": 65845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5408926135052068e-12,
+      "loss": 3.9344,
+      "step": 65850
+    }
+  ],
+  "max_steps": 65853,
+  "num_train_epochs": 1,
+  "total_flos": 3.202566251662541e+18,
+  "trial_name": null,
+  "trial_params": null
+}