diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,81352 @@
+{
+  "best_metric": 0.4053784906864166,
+  "best_model_checkpoint": "/work/twskvfb446/facebook/wav2vec2-large-lv60_voidful/phoneme_byt5_SpeechMixEEDT5_wav2vec2-large-lv60_t5lephone-small_textdecoderonly_bs64/checkpoint-54024",
+  "epoch": 29.993337774816787,
+  "global_step": 135060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.714285714285715e-07,
+      "loss": 7.8022,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.142857142857143e-06,
+      "loss": 7.0163,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7142857142857145e-06,
+      "loss": 5.7344,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.285714285714286e-06,
+      "loss": 4.5755,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8571428571428573e-06,
+      "loss": 3.8038,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.428571428571429e-06,
+      "loss": 3.3535,
+      "step": 60
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 3.0968,
+      "step": 70
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.571428571428572e-06,
+      "loss": 2.9693,
+      "step": 80
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.142857142857142e-06,
+      "loss": 2.8552,
+      "step": 90
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7142857142857145e-06,
+      "loss": 2.8047,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.285714285714286e-06,
+      "loss": 2.7601,
+      "step": 110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 6.857142857142858e-06,
+      "loss": 2.7412,
+      "step": 120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.428571428571429e-06,
+      "loss": 2.675,
+      "step": 130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 2.6704,
+      "step": 140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.571428571428571e-06,
+      "loss": 2.6137,
+      "step": 150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.142857142857144e-06,
+      "loss": 2.5653,
+      "step": 160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.714285714285715e-06,
+      "loss": 2.5024,
+      "step": 170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.0285714285714285e-05,
+      "loss": 2.3989,
+      "step": 180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.0857142857142858e-05,
+      "loss": 2.3257,
+      "step": 190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1428571428571429e-05,
+      "loss": 2.2989,
+      "step": 200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.2e-05,
+      "loss": 2.2369,
+      "step": 210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.2571428571428572e-05,
+      "loss": 2.1946,
+      "step": 220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3142857142857145e-05,
+      "loss": 2.1873,
+      "step": 230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.3714285714285716e-05,
+      "loss": 2.1545,
+      "step": 240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.4285714285714287e-05,
+      "loss": 2.119,
+      "step": 250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.4857142857142858e-05,
+      "loss": 2.1139,
+      "step": 260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.542857142857143e-05,
+      "loss": 2.1037,
+      "step": 270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 2.0736,
+      "step": 280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.6571428571428574e-05,
+      "loss": 2.0659,
+      "step": 290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.7142857142857142e-05,
+      "loss": 2.042,
+      "step": 300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.7714285714285717e-05,
+      "loss": 2.0267,
+      "step": 310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.8285714285714288e-05,
+      "loss": 2.0111,
+      "step": 320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.885714285714286e-05,
+      "loss": 2.0019,
+      "step": 330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.942857142857143e-05,
+      "loss": 1.9778,
+      "step": 340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2e-05,
+      "loss": 1.9627,
+      "step": 350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.057142857142857e-05,
+      "loss": 1.968,
+      "step": 360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.1142857142857144e-05,
+      "loss": 1.9698,
+      "step": 370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.1714285714285715e-05,
+      "loss": 1.9359,
+      "step": 380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.2285714285714287e-05,
+      "loss": 1.9309,
+      "step": 390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.2857142857142858e-05,
+      "loss": 1.9194,
+      "step": 400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.3428571428571433e-05,
+      "loss": 1.9101,
+      "step": 410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.4e-05,
+      "loss": 1.899,
+      "step": 420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.4571428571428575e-05,
+      "loss": 1.8803,
+      "step": 430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5142857142857143e-05,
+      "loss": 1.8826,
+      "step": 440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.5714285714285718e-05,
+      "loss": 1.8486,
+      "step": 450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.628571428571429e-05,
+      "loss": 1.867,
+      "step": 460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.6857142857142857e-05,
+      "loss": 1.8622,
+      "step": 470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.742857142857143e-05,
+      "loss": 1.8355,
+      "step": 480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.8e-05,
+      "loss": 1.8277,
+      "step": 490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.8571428571428574e-05,
+      "loss": 1.8371,
+      "step": 500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2.9142857142857146e-05,
+      "loss": 1.8251,
+      "step": 510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2.9714285714285717e-05,
+      "loss": 1.8091,
+      "step": 520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.0285714285714288e-05,
+      "loss": 1.7976,
+      "step": 530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.085714285714286e-05,
+      "loss": 1.802,
+      "step": 540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.142857142857143e-05,
+      "loss": 1.8056,
+      "step": 550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 1.7969,
+      "step": 560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.257142857142857e-05,
+      "loss": 1.7701,
+      "step": 570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.314285714285715e-05,
+      "loss": 1.7702,
+      "step": 580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.3714285714285716e-05,
+      "loss": 1.7569,
+      "step": 590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.4285714285714284e-05,
+      "loss": 1.7373,
+      "step": 600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.485714285714286e-05,
+      "loss": 1.7255,
+      "step": 610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.542857142857143e-05,
+      "loss": 1.7356,
+      "step": 620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.6e-05,
+      "loss": 1.7401,
+      "step": 630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.6571428571428576e-05,
+      "loss": 1.7477,
+      "step": 640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3.714285714285715e-05,
+      "loss": 1.7393,
+      "step": 650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.771428571428572e-05,
+      "loss": 1.7155,
+      "step": 660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.828571428571429e-05,
+      "loss": 1.7465,
+      "step": 670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.885714285714286e-05,
+      "loss": 1.72,
+      "step": 680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 3.9428571428571435e-05,
+      "loss": 1.7128,
+      "step": 690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4e-05,
+      "loss": 1.7142,
+      "step": 700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.999702358806459e-05,
+      "loss": 1.699,
+      "step": 710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.999404717612918e-05,
+      "loss": 1.6916,
+      "step": 720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.999107076419377e-05,
+      "loss": 1.6973,
+      "step": 730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 3.998809435225836e-05,
+      "loss": 1.692,
+      "step": 740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.9985117940322945e-05,
+      "loss": 1.6846,
+      "step": 750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.998214152838753e-05,
+      "loss": 1.6782,
+      "step": 760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.997916511645212e-05,
+      "loss": 1.6608,
+      "step": 770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 3.9976188704516705e-05,
+      "loss": 1.6506,
+      "step": 780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.99732122925813e-05,
+      "loss": 1.6559,
+      "step": 790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.9970235880645886e-05,
+      "loss": 1.6463,
+      "step": 800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.996725946871047e-05,
+      "loss": 1.6541,
+      "step": 810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.996428305677506e-05,
+      "loss": 1.663,
+      "step": 820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 3.996130664483965e-05,
+      "loss": 1.6349,
+      "step": 830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.995833023290424e-05,
+      "loss": 1.6434,
+      "step": 840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.995535382096883e-05,
+      "loss": 1.6352,
+      "step": 850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.9952377409033414e-05,
+      "loss": 1.6085,
+      "step": 860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 3.9949400997098e-05,
+      "loss": 1.6288,
+      "step": 870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.994642458516259e-05,
+      "loss": 1.6136,
+      "step": 880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.9943448173227175e-05,
+      "loss": 1.6128,
+      "step": 890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.994047176129177e-05,
+      "loss": 1.6157,
+      "step": 900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.9937495349356355e-05,
+      "loss": 1.6015,
+      "step": 910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 3.993451893742094e-05,
+      "loss": 1.613,
+      "step": 920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.993154252548553e-05,
+      "loss": 1.5895,
+      "step": 930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.992856611355012e-05,
+      "loss": 1.6169,
+      "step": 940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.992558970161471e-05,
+      "loss": 1.6038,
+      "step": 950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 3.992261328967929e-05,
+      "loss": 1.5896,
+      "step": 960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9919636877743883e-05,
+      "loss": 1.6021,
+      "step": 970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.991666046580847e-05,
+      "loss": 1.5939,
+      "step": 980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.991368405387306e-05,
+      "loss": 1.6007,
+      "step": 990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.9910707641937644e-05,
+      "loss": 1.5855,
+      "step": 1000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3.990773123000224e-05,
+      "loss": 1.5678,
+      "step": 1010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9904754818066825e-05,
+      "loss": 1.5765,
+      "step": 1020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.990177840613141e-05,
+      "loss": 1.5666,
+      "step": 1030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.9898801994196e-05,
+      "loss": 1.5884,
+      "step": 1040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 3.989582558226059e-05,
+      "loss": 1.5801,
+      "step": 1050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.989284917032517e-05,
+      "loss": 1.5526,
+      "step": 1060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.9889872758389766e-05,
+      "loss": 1.5741,
+      "step": 1070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.988689634645435e-05,
+      "loss": 1.5568,
+      "step": 1080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.988391993451894e-05,
+      "loss": 1.5466,
+      "step": 1090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 3.988094352258353e-05,
+      "loss": 1.5532,
+      "step": 1100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.987796711064812e-05,
+      "loss": 1.5633,
+      "step": 1110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.987499069871271e-05,
+      "loss": 1.5498,
+      "step": 1120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.9872014286777294e-05,
+      "loss": 1.5523,
+      "step": 1130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.986903787484188e-05,
+      "loss": 1.5505,
+      "step": 1140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.986606146290647e-05,
+      "loss": 1.5532,
+      "step": 1150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.986308505097106e-05,
+      "loss": 1.5584,
+      "step": 1160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.986010863903564e-05,
+      "loss": 1.5661,
+      "step": 1170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.9857132227100235e-05,
+      "loss": 1.5493,
+      "step": 1180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.985415581516482e-05,
+      "loss": 1.5475,
+      "step": 1190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.985117940322941e-05,
+      "loss": 1.5327,
+      "step": 1200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.9848202991293996e-05,
+      "loss": 1.539,
+      "step": 1210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.984522657935859e-05,
+      "loss": 1.5331,
+      "step": 1220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.984225016742318e-05,
+      "loss": 1.5121,
+      "step": 1230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.9839273755487764e-05,
+      "loss": 1.5282,
+      "step": 1240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.983629734355235e-05,
+      "loss": 1.5149,
+      "step": 1250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.9833320931616944e-05,
+      "loss": 1.5244,
+      "step": 1260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.9830344519681524e-05,
+      "loss": 1.5178,
+      "step": 1270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.982736810774611e-05,
+      "loss": 1.5257,
+      "step": 1280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.9824391695810705e-05,
+      "loss": 1.5224,
+      "step": 1290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.982141528387529e-05,
+      "loss": 1.5245,
+      "step": 1300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.981843887193988e-05,
+      "loss": 1.5284,
+      "step": 1310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.9815462460004466e-05,
+      "loss": 1.5121,
+      "step": 1320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.981248604806906e-05,
+      "loss": 1.494,
+      "step": 1330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.9809509636133646e-05,
+      "loss": 1.5179,
+      "step": 1340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.980653322419823e-05,
+      "loss": 1.5313,
+      "step": 1350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.980355681226282e-05,
+      "loss": 1.504,
+      "step": 1360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.980058040032741e-05,
+      "loss": 1.5139,
+      "step": 1370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9797603988391994e-05,
+      "loss": 1.5011,
+      "step": 1380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.979462757645658e-05,
+      "loss": 1.5035,
+      "step": 1390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9791651164521174e-05,
+      "loss": 1.4947,
+      "step": 1400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.978867475258576e-05,
+      "loss": 1.4967,
+      "step": 1410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.978569834065035e-05,
+      "loss": 1.4972,
+      "step": 1420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9782721928714935e-05,
+      "loss": 1.5109,
+      "step": 1430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.977974551677953e-05,
+      "loss": 1.4978,
+      "step": 1440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9776769104844116e-05,
+      "loss": 1.4854,
+      "step": 1450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.97737926929087e-05,
+      "loss": 1.5032,
+      "step": 1460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.977081628097329e-05,
+      "loss": 1.5016,
+      "step": 1470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9767839869037876e-05,
+      "loss": 1.4706,
+      "step": 1480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.976486345710246e-05,
+      "loss": 1.4977,
+      "step": 1490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.976188704516706e-05,
+      "loss": 1.4976,
+      "step": 1500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9758910633231644e-05,
+      "loss": 1.4964,
+      "step": 1510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.975593422129623e-05,
+      "loss": 1.5049,
+      "step": 1520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.975295780936082e-05,
+      "loss": 1.4811,
+      "step": 1530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.974998139742541e-05,
+      "loss": 1.487,
+      "step": 1540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.974700498549e-05,
+      "loss": 1.4786,
+      "step": 1550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.9744028573554585e-05,
+      "loss": 1.4905,
+      "step": 1560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.974105216161917e-05,
+      "loss": 1.475,
+      "step": 1570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.973807574968376e-05,
+      "loss": 1.4627,
+      "step": 1580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.9735099337748346e-05,
+      "loss": 1.4833,
+      "step": 1590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.973212292581293e-05,
+      "loss": 1.4901,
+      "step": 1600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.9729146513877526e-05,
+      "loss": 1.476,
+      "step": 1610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.972617010194211e-05,
+      "loss": 1.4712,
+      "step": 1620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.97231936900067e-05,
+      "loss": 1.4896,
+      "step": 1630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.972021727807129e-05,
+      "loss": 1.4721,
+      "step": 1640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.971724086613588e-05,
+      "loss": 1.4618,
+      "step": 1650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.971426445420047e-05,
+      "loss": 1.4772,
+      "step": 1660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.9711288042265054e-05,
+      "loss": 1.4783,
+      "step": 1670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.970831163032964e-05,
+      "loss": 1.4691,
+      "step": 1680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.970533521839423e-05,
+      "loss": 1.4909,
+      "step": 1690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.9702358806458815e-05,
+      "loss": 1.4828,
+      "step": 1700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.96993823945234e-05,
+      "loss": 1.4657,
+      "step": 1710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.9696405982587996e-05,
+      "loss": 1.4585,
+      "step": 1720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.969342957065258e-05,
+      "loss": 1.4757,
+      "step": 1730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.969045315871717e-05,
+      "loss": 1.4357,
+      "step": 1740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.9687476746781756e-05,
+      "loss": 1.4797,
+      "step": 1750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.968450033484635e-05,
+      "loss": 1.462,
+      "step": 1760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.968152392291094e-05,
+      "loss": 1.4541,
+      "step": 1770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.967854751097552e-05,
+      "loss": 1.4508,
+      "step": 1780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.967557109904011e-05,
+      "loss": 1.4468,
+      "step": 1790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.96725946871047e-05,
+      "loss": 1.4543,
+      "step": 1800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.9669618275169285e-05,
+      "loss": 1.4416,
+      "step": 1810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.966664186323387e-05,
+      "loss": 1.4553,
+      "step": 1820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.9663665451298465e-05,
+      "loss": 1.446,
+      "step": 1830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.966068903936305e-05,
+      "loss": 1.4542,
+      "step": 1840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.965771262742764e-05,
+      "loss": 1.4389,
+      "step": 1850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.9654736215492226e-05,
+      "loss": 1.4362,
+      "step": 1860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.965175980355682e-05,
+      "loss": 1.4453,
+      "step": 1870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.96487833916214e-05,
+      "loss": 1.4575,
+      "step": 1880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.964580697968599e-05,
+      "loss": 1.4512,
+      "step": 1890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.964283056775058e-05,
+      "loss": 1.4387,
+      "step": 1900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.963985415581517e-05,
+      "loss": 1.4478,
+      "step": 1910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.9636877743879754e-05,
+      "loss": 1.4276,
+      "step": 1920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.963390133194435e-05,
+      "loss": 1.4546,
+      "step": 1930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.9630924920008935e-05,
+      "loss": 1.4389,
+      "step": 1940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.962794850807352e-05,
+      "loss": 1.4419,
+      "step": 1950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.962497209613811e-05,
+      "loss": 1.4432,
+      "step": 1960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.96219956842027e-05,
+      "loss": 1.4387,
+      "step": 1970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.961901927226728e-05,
+      "loss": 1.4273,
+      "step": 1980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.961604286033187e-05,
+      "loss": 1.4348,
+      "step": 1990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.961306644839646e-05,
+      "loss": 1.4378,
+      "step": 2000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.961009003646105e-05,
+      "loss": 1.4348,
+      "step": 2010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.9607113624525637e-05,
+      "loss": 1.4382,
+      "step": 2020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.9604137212590223e-05,
+      "loss": 1.4335,
+      "step": 2030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.960116080065482e-05,
+      "loss": 1.4375,
+      "step": 2040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.9598184388719404e-05,
+      "loss": 1.4247,
+      "step": 2050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.959520797678399e-05,
+      "loss": 1.4303,
+      "step": 2060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.959223156484858e-05,
+      "loss": 1.434,
+      "step": 2070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.958925515291317e-05,
+      "loss": 1.4287,
+      "step": 2080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.958627874097775e-05,
+      "loss": 1.4254,
+      "step": 2090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.958330232904234e-05,
+      "loss": 1.4348,
+      "step": 2100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.958032591710693e-05,
+      "loss": 1.4258,
+      "step": 2110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.957734950517152e-05,
+      "loss": 1.4111,
+      "step": 2120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.9574373093236106e-05,
+      "loss": 1.4211,
+      "step": 2130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.957139668130069e-05,
+      "loss": 1.4326,
+      "step": 2140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.9568420269365287e-05,
+      "loss": 1.4162,
+      "step": 2150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.9565443857429873e-05,
+      "loss": 1.4215,
+      "step": 2160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.956246744549446e-05,
+      "loss": 1.4132,
+      "step": 2170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.955949103355905e-05,
+      "loss": 1.4224,
+      "step": 2180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.9556514621623634e-05,
+      "loss": 1.4126,
+      "step": 2190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.955353820968822e-05,
+      "loss": 1.4137,
+      "step": 2200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.955056179775281e-05,
+      "loss": 1.414,
+      "step": 2210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.95475853858174e-05,
+      "loss": 1.4152,
+      "step": 2220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.954460897388199e-05,
+      "loss": 1.4235,
+      "step": 2230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.9541632561946575e-05,
+      "loss": 1.4164,
+      "step": 2240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.953865615001116e-05,
+      "loss": 1.4077,
+      "step": 2250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.9535679738075756e-05,
+      "loss": 1.4187,
+      "step": 2260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.953270332614034e-05,
+      "loss": 1.4214,
+      "step": 2270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.952972691420493e-05,
+      "loss": 1.4158,
+      "step": 2280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.952675050226952e-05,
+      "loss": 1.406,
+      "step": 2290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.9523774090334104e-05,
+      "loss": 1.4066,
+      "step": 2300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.952079767839869e-05,
+      "loss": 1.4109,
+      "step": 2310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.9517821266463284e-05,
+      "loss": 1.3869,
+      "step": 2320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.951484485452787e-05,
+      "loss": 1.3991,
+      "step": 2330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.951186844259246e-05,
+      "loss": 1.4047,
+      "step": 2340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.9508892030657045e-05,
+      "loss": 1.392,
+      "step": 2350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.950591561872164e-05,
+      "loss": 1.4138,
+      "step": 2360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.9502939206786225e-05,
+      "loss": 1.387,
+      "step": 2370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.949996279485081e-05,
+      "loss": 1.3893,
+      "step": 2380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.94969863829154e-05,
+      "loss": 1.3858,
+      "step": 2390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.9494009970979986e-05,
+      "loss": 1.3999,
+      "step": 2400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.949103355904457e-05,
+      "loss": 1.3839,
+      "step": 2410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.948805714710916e-05,
+      "loss": 1.3866,
+      "step": 2420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.9485080735173754e-05,
+      "loss": 1.4105,
+      "step": 2430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.948210432323834e-05,
+      "loss": 1.412,
+      "step": 2440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.947912791130293e-05,
+      "loss": 1.3847,
+      "step": 2450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.9476151499367514e-05,
+      "loss": 1.3858,
+      "step": 2460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.947317508743211e-05,
+      "loss": 1.3936,
+      "step": 2470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.9470198675496695e-05,
+      "loss": 1.3922,
+      "step": 2480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.946722226356128e-05,
+      "loss": 1.3721,
+      "step": 2490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.946424585162587e-05,
+      "loss": 1.3869,
+      "step": 2500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.9461269439690456e-05,
+      "loss": 1.389,
+      "step": 2510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.945829302775504e-05,
+      "loss": 1.3825,
+      "step": 2520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.945531661581963e-05,
+      "loss": 1.3944,
+      "step": 2530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.945234020388422e-05,
+      "loss": 1.3923,
+      "step": 2540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.944936379194881e-05,
+      "loss": 1.3869,
+      "step": 2550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.94463873800134e-05,
+      "loss": 1.3856,
+      "step": 2560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9443410968077984e-05,
+      "loss": 1.3625,
+      "step": 2570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.944043455614258e-05,
+      "loss": 1.3861,
+      "step": 2580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9437458144207164e-05,
+      "loss": 1.3767,
+      "step": 2590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9434481732271744e-05,
+      "loss": 1.3757,
+      "step": 2600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.943150532033634e-05,
+      "loss": 1.3831,
+      "step": 2610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.9428528908400925e-05,
+      "loss": 1.3828,
+      "step": 2620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.942555249646551e-05,
+      "loss": 1.3783,
+      "step": 2630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.94225760845301e-05,
+      "loss": 1.3561,
+      "step": 2640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.941959967259469e-05,
+      "loss": 1.3824,
+      "step": 2650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.941662326065928e-05,
+      "loss": 1.3918,
+      "step": 2660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.9413646848723866e-05,
+      "loss": 1.375,
+      "step": 2670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.941067043678845e-05,
+      "loss": 1.3641,
+      "step": 2680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.940769402485305e-05,
+      "loss": 1.3672,
+      "step": 2690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.940471761291763e-05,
+      "loss": 1.3725,
+      "step": 2700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.940174120098222e-05,
+      "loss": 1.3606,
+      "step": 2710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.939876478904681e-05,
+      "loss": 1.3608,
+      "step": 2720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9395788377111394e-05,
+      "loss": 1.3657,
+      "step": 2730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.939281196517598e-05,
+      "loss": 1.3695,
+      "step": 2740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.9389835553240575e-05,
+      "loss": 1.3716,
+      "step": 2750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.938685914130516e-05,
+      "loss": 1.3776,
+      "step": 2760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.938388272936975e-05,
+      "loss": 1.3717,
+      "step": 2770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9380906317434336e-05,
+      "loss": 1.362,
+      "step": 2780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.937792990549893e-05,
+      "loss": 1.3613,
+      "step": 2790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.937495349356351e-05,
+      "loss": 1.3726,
+      "step": 2800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.9371977081628096e-05,
+      "loss": 1.3622,
+      "step": 2810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.936900066969269e-05,
+      "loss": 1.3581,
+      "step": 2820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.936602425775728e-05,
+      "loss": 1.3534,
+      "step": 2830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.9363047845821864e-05,
+      "loss": 1.3661,
+      "step": 2840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.936007143388645e-05,
+      "loss": 1.3656,
+      "step": 2850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.9357095021951044e-05,
+      "loss": 1.3615,
+      "step": 2860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.935411861001563e-05,
+      "loss": 1.3554,
+      "step": 2870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.935114219808022e-05,
+      "loss": 1.361,
+      "step": 2880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.9348165786144805e-05,
+      "loss": 1.3722,
+      "step": 2890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 3.93451893742094e-05,
+      "loss": 1.3491,
+      "step": 2900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.934221296227398e-05,
+      "loss": 1.3281,
+      "step": 2910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.9339236550338566e-05,
+      "loss": 1.3302,
+      "step": 2920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.933626013840316e-05,
+      "loss": 1.353,
+      "step": 2930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 3.9333283726467746e-05,
+      "loss": 1.354,
+      "step": 2940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.933030731453233e-05,
+      "loss": 1.3412,
+      "step": 2950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.932733090259692e-05,
+      "loss": 1.3432,
+      "step": 2960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.9324354490661514e-05,
+      "loss": 1.343,
+      "step": 2970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.93213780787261e-05,
+      "loss": 1.3466,
+      "step": 2980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 3.931840166679069e-05,
+      "loss": 1.3531,
+      "step": 2990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.9315425254855275e-05,
+      "loss": 1.3596,
+      "step": 3000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.931244884291986e-05,
+      "loss": 1.3413,
+      "step": 3010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.930947243098445e-05,
+      "loss": 1.342,
+      "step": 3020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 3.9306496019049035e-05,
+      "loss": 1.3503,
+      "step": 3030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.930351960711363e-05,
+      "loss": 1.354,
+      "step": 3040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.9300543195178216e-05,
+      "loss": 1.3366,
+      "step": 3050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.92975667832428e-05,
+      "loss": 1.342,
+      "step": 3060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.929459037130739e-05,
+      "loss": 1.325,
+      "step": 3070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 3.929161395937198e-05,
+      "loss": 1.3409,
+      "step": 3080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.928863754743657e-05,
+      "loss": 1.3492,
+      "step": 3090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.928566113550116e-05,
+      "loss": 1.3374,
+      "step": 3100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.9282684723565744e-05,
+      "loss": 1.3357,
+      "step": 3110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 3.927970831163033e-05,
+      "loss": 1.3375,
+      "step": 3120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.927673189969492e-05,
+      "loss": 1.3377,
+      "step": 3130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.927375548775951e-05,
+      "loss": 1.3387,
+      "step": 3140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.92707790758241e-05,
+      "loss": 1.3383,
+      "step": 3150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.9267802663888685e-05,
+      "loss": 1.3411,
+      "step": 3160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 3.926482625195327e-05,
+      "loss": 1.3345,
+      "step": 3170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9261849840017866e-05,
+      "loss": 1.3426,
+      "step": 3180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.925887342808245e-05,
+      "loss": 1.3348,
+      "step": 3190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.925589701614704e-05,
+      "loss": 1.3289,
+      "step": 3200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9252920604211627e-05,
+      "loss": 1.3116,
+      "step": 3210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9249944192276213e-05,
+      "loss": 1.3311,
+      "step": 3220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92469677803408e-05,
+      "loss": 1.3276,
+      "step": 3230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924399136840539e-05,
+      "loss": 1.3327,
+      "step": 3240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924101495646998e-05,
+      "loss": 1.3197,
+      "step": 3250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923803854453457e-05,
+      "loss": 1.3167,
+      "step": 3260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9235062132599155e-05,
+      "loss": 1.3248,
+      "step": 3270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.923208572066374e-05,
+      "loss": 1.3188,
+      "step": 3280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9229109308728335e-05,
+      "loss": 1.3134,
+      "step": 3290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.922613289679292e-05,
+      "loss": 1.3121,
+      "step": 3300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.922315648485751e-05,
+      "loss": 1.3161,
+      "step": 3310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.9220180072922096e-05,
+      "loss": 1.3254,
+      "step": 3320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.921720366098668e-05,
+      "loss": 1.3152,
+      "step": 3330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.921422724905127e-05,
+      "loss": 1.3116,
+      "step": 3340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.921125083711586e-05,
+      "loss": 1.3195,
+      "step": 3350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.920827442518045e-05,
+      "loss": 1.3123,
+      "step": 3360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.920529801324504e-05,
+      "loss": 1.3193,
+      "step": 3370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.9202321601309624e-05,
+      "loss": 1.3095,
+      "step": 3380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.919934518937421e-05,
+      "loss": 1.2992,
+      "step": 3390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9196368777438805e-05,
+      "loss": 1.3179,
+      "step": 3400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.919339236550339e-05,
+      "loss": 1.3108,
+      "step": 3410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.919041595356797e-05,
+      "loss": 1.3064,
+      "step": 3420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9187439541632565e-05,
+      "loss": 1.312,
+      "step": 3430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.918446312969715e-05,
+      "loss": 1.3049,
+      "step": 3440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.918148671776174e-05,
+      "loss": 1.3045,
+      "step": 3450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9178510305826326e-05,
+      "loss": 1.3079,
+      "step": 3460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.917553389389092e-05,
+      "loss": 1.3095,
+      "step": 3470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.917255748195551e-05,
+      "loss": 1.3004,
+      "step": 3480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.9169581070020094e-05,
+      "loss": 1.3061,
+      "step": 3490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.916660465808468e-05,
+      "loss": 1.3208,
+      "step": 3500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.9163628246149274e-05,
+      "loss": 1.3133,
+      "step": 3510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.9160651834213854e-05,
+      "loss": 1.3031,
+      "step": 3520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.915767542227845e-05,
+      "loss": 1.3035,
+      "step": 3530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.9154699010343035e-05,
+      "loss": 1.2928,
+      "step": 3540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.915172259840762e-05,
+      "loss": 1.3092,
+      "step": 3550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.914874618647221e-05,
+      "loss": 1.3017,
+      "step": 3560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.91457697745368e-05,
+      "loss": 1.3091,
+      "step": 3570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.914279336260139e-05,
+      "loss": 1.3191,
+      "step": 3580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.9139816950665976e-05,
+      "loss": 1.3088,
+      "step": 3590
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.913684053873056e-05,
+      "loss": 1.2881,
+      "step": 3600
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.913386412679516e-05,
+      "loss": 1.2863,
+      "step": 3610
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.913088771485974e-05,
+      "loss": 1.2982,
+      "step": 3620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.9127911302924324e-05,
+      "loss": 1.3128,
+      "step": 3630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.912493489098892e-05,
+      "loss": 1.3065,
+      "step": 3640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.9121958479053504e-05,
+      "loss": 1.3039,
+      "step": 3650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.911898206711809e-05,
+      "loss": 1.3036,
+      "step": 3660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.911600565518268e-05,
+      "loss": 1.3241,
+      "step": 3670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.911302924324727e-05,
+      "loss": 1.2943,
+      "step": 3680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.911005283131186e-05,
+      "loss": 1.295,
+      "step": 3690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.9107076419376446e-05,
+      "loss": 1.2992,
+      "step": 3700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.910410000744103e-05,
+      "loss": 1.2915,
+      "step": 3710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.910112359550562e-05,
+      "loss": 1.2924,
+      "step": 3720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.9098147183570206e-05,
+      "loss": 1.2883,
+      "step": 3730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.909517077163479e-05,
+      "loss": 1.3078,
+      "step": 3740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.909219435969939e-05,
+      "loss": 1.2953,
+      "step": 3750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.9089217947763974e-05,
+      "loss": 1.2922,
+      "step": 3760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.908624153582856e-05,
+      "loss": 1.2872,
+      "step": 3770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.908326512389315e-05,
+      "loss": 1.2885,
+      "step": 3780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.908028871195774e-05,
+      "loss": 1.2956,
+      "step": 3790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.907731230002233e-05,
+      "loss": 1.2911,
+      "step": 3800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.9074335888086915e-05,
+      "loss": 1.2861,
+      "step": 3810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.90713594761515e-05,
+      "loss": 1.2859,
+      "step": 3820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.906838306421609e-05,
+      "loss": 1.2991,
+      "step": 3830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.9065406652280676e-05,
+      "loss": 1.271,
+      "step": 3840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.906243024034526e-05,
+      "loss": 1.2789,
+      "step": 3850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.9059453828409856e-05,
+      "loss": 1.2753,
+      "step": 3860
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.905647741647444e-05,
+      "loss": 1.2797,
+      "step": 3870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.905350100453903e-05,
+      "loss": 1.285,
+      "step": 3880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.905052459260362e-05,
+      "loss": 1.278,
+      "step": 3890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.904754818066821e-05,
+      "loss": 1.276,
+      "step": 3900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.90445717687328e-05,
+      "loss": 1.2852,
+      "step": 3910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.9041595356797385e-05,
+      "loss": 1.2908,
+      "step": 3920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.903861894486197e-05,
+      "loss": 1.2723,
+      "step": 3930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.903564253292656e-05,
+      "loss": 1.2763,
+      "step": 3940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9032666120991145e-05,
+      "loss": 1.2795,
+      "step": 3950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.902968970905574e-05,
+      "loss": 1.2874,
+      "step": 3960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.9026713297120326e-05,
+      "loss": 1.2717,
+      "step": 3970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.902373688518491e-05,
+      "loss": 1.2759,
+      "step": 3980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.90207604732495e-05,
+      "loss": 1.2787,
+      "step": 3990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.901778406131409e-05,
+      "loss": 1.2616,
+      "step": 4000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.901480764937868e-05,
+      "loss": 1.2597,
+      "step": 4010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.901183123744327e-05,
+      "loss": 1.2799,
+      "step": 4020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.9008854825507854e-05,
+      "loss": 1.2656,
+      "step": 4030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.900587841357244e-05,
+      "loss": 1.2736,
+      "step": 4040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.900290200163703e-05,
+      "loss": 1.2597,
+      "step": 4050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.8999925589701615e-05,
+      "loss": 1.2715,
+      "step": 4060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.899694917776621e-05,
+      "loss": 1.268,
+      "step": 4070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8993972765830795e-05,
+      "loss": 1.2694,
+      "step": 4080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.899099635389538e-05,
+      "loss": 1.256,
+      "step": 4090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.898801994195997e-05,
+      "loss": 1.2605,
+      "step": 4100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.898504353002456e-05,
+      "loss": 1.2737,
+      "step": 4110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.898206711808915e-05,
+      "loss": 1.2591,
+      "step": 4120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.897909070615373e-05,
+      "loss": 1.2708,
+      "step": 4130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.897611429421832e-05,
+      "loss": 1.2542,
+      "step": 4140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.897313788228291e-05,
+      "loss": 1.2757,
+      "step": 4150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.89701614703475e-05,
+      "loss": 1.2636,
+      "step": 4160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8967185058412084e-05,
+      "loss": 1.2687,
+      "step": 4170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.896420864647668e-05,
+      "loss": 1.2432,
+      "step": 4180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8961232234541265e-05,
+      "loss": 1.2547,
+      "step": 4190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.895825582260585e-05,
+      "loss": 1.2744,
+      "step": 4200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.895527941067044e-05,
+      "loss": 1.269,
+      "step": 4210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.895230299873503e-05,
+      "loss": 1.2552,
+      "step": 4220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.894932658679962e-05,
+      "loss": 1.2669,
+      "step": 4230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.8946350174864206e-05,
+      "loss": 1.239,
+      "step": 4240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.894337376292879e-05,
+      "loss": 1.2557,
+      "step": 4250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.894039735099338e-05,
+      "loss": 1.2573,
+      "step": 4260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.893742093905797e-05,
+      "loss": 1.2478,
+      "step": 4270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8934444527122554e-05,
+      "loss": 1.2744,
+      "step": 4280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.893146811518715e-05,
+      "loss": 1.259,
+      "step": 4290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.8928491703251734e-05,
+      "loss": 1.2523,
+      "step": 4300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.892551529131632e-05,
+      "loss": 1.2579,
+      "step": 4310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.892253887938091e-05,
+      "loss": 1.2677,
+      "step": 4320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.89195624674455e-05,
+      "loss": 1.2308,
+      "step": 4330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.891658605551008e-05,
+      "loss": 1.2494,
+      "step": 4340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8913609643574675e-05,
+      "loss": 1.2365,
+      "step": 4350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.891063323163926e-05,
+      "loss": 1.2491,
+      "step": 4360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.890765681970385e-05,
+      "loss": 1.2561,
+      "step": 4370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8904680407768436e-05,
+      "loss": 1.2533,
+      "step": 4380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.890170399583303e-05,
+      "loss": 1.2362,
+      "step": 4390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.889872758389762e-05,
+      "loss": 1.2329,
+      "step": 4400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.8895751171962204e-05,
+      "loss": 1.2515,
+      "step": 4410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.889277476002679e-05,
+      "loss": 1.2515,
+      "step": 4420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.8889798348091384e-05,
+      "loss": 1.25,
+      "step": 4430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8886821936155964e-05,
+      "loss": 1.2617,
+      "step": 4440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.888384552422055e-05,
+      "loss": 1.2486,
+      "step": 4450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8880869112285145e-05,
+      "loss": 1.2305,
+      "step": 4460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.887789270034973e-05,
+      "loss": 1.2436,
+      "step": 4470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.887491628841432e-05,
+      "loss": 1.2498,
+      "step": 4480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8871939876478906e-05,
+      "loss": 1.2449,
+      "step": 4490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.88689634645435e-05,
+      "loss": 1.2448,
+      "step": 4500
+    },
+    {
+      "epoch": 1.0,
+      "eval_cer": 4.754273504273504,
+      "eval_loss": 1.085645079612732,
+      "eval_runtime": 1.5151,
+      "eval_samples_per_second": 6.6,
+      "eval_steps_per_second": 0.66,
+      "eval_wer": 0.948051948051948,
+      "step": 4502
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8865987052608086e-05,
+      "loss": 1.242,
+      "step": 4510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.886301064067267e-05,
+      "loss": 1.2529,
+      "step": 4520
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.886003422873726e-05,
+      "loss": 1.2312,
+      "step": 4530
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.885705781680185e-05,
+      "loss": 1.2562,
+      "step": 4540
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.8854081404866434e-05,
+      "loss": 1.2439,
+      "step": 4550
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.885110499293102e-05,
+      "loss": 1.2378,
+      "step": 4560
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.8848128580995614e-05,
+      "loss": 1.2408,
+      "step": 4570
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.88451521690602e-05,
+      "loss": 1.2279,
+      "step": 4580
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.884217575712479e-05,
+      "loss": 1.2308,
+      "step": 4590
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.8839199345189375e-05,
+      "loss": 1.2494,
+      "step": 4600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.883622293325397e-05,
+      "loss": 1.2294,
+      "step": 4610
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.8833246521318556e-05,
+      "loss": 1.2184,
+      "step": 4620
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.883027010938314e-05,
+      "loss": 1.2305,
+      "step": 4630
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.882729369744773e-05,
+      "loss": 1.2308,
+      "step": 4640
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.8824317285512316e-05,
+      "loss": 1.2425,
+      "step": 4650
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.88213408735769e-05,
+      "loss": 1.2233,
+      "step": 4660
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.88183644616415e-05,
+      "loss": 1.2468,
+      "step": 4670
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.8815388049706084e-05,
+      "loss": 1.2311,
+      "step": 4680
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.881241163777067e-05,
+      "loss": 1.2162,
+      "step": 4690
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.880943522583526e-05,
+      "loss": 1.2394,
+      "step": 4700
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.8806458813899844e-05,
+      "loss": 1.2118,
+      "step": 4710
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.880348240196444e-05,
+      "loss": 1.2264,
+      "step": 4720
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.8800505990029025e-05,
+      "loss": 1.2175,
+      "step": 4730
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.879752957809361e-05,
+      "loss": 1.2056,
+      "step": 4740
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.87945531661582e-05,
+      "loss": 1.2146,
+      "step": 4750
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.8791576754222786e-05,
+      "loss": 1.2292,
+      "step": 4760
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.878860034228737e-05,
+      "loss": 1.2178,
+      "step": 4770
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.8785623930351966e-05,
+      "loss": 1.2163,
+      "step": 4780
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.878264751841655e-05,
+      "loss": 1.213,
+      "step": 4790
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.877967110648114e-05,
+      "loss": 1.2196,
+      "step": 4800
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.877669469454573e-05,
+      "loss": 1.216,
+      "step": 4810
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.877371828261032e-05,
+      "loss": 1.2409,
+      "step": 4820
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.877074187067491e-05,
+      "loss": 1.2036,
+      "step": 4830
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.8767765458739494e-05,
+      "loss": 1.2071,
+      "step": 4840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.876478904680408e-05,
+      "loss": 1.233,
+      "step": 4850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.876181263486867e-05,
+      "loss": 1.2415,
+      "step": 4860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.8758836222933255e-05,
+      "loss": 1.2177,
+      "step": 4870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.875585981099784e-05,
+      "loss": 1.2187,
+      "step": 4880
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.8752883399062436e-05,
+      "loss": 1.2159,
+      "step": 4890
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.874990698712702e-05,
+      "loss": 1.2087,
+      "step": 4900
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.874693057519161e-05,
+      "loss": 1.22,
+      "step": 4910
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.8743954163256196e-05,
+      "loss": 1.2203,
+      "step": 4920
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.874097775132079e-05,
+      "loss": 1.2209,
+      "step": 4930
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.873800133938538e-05,
+      "loss": 1.2072,
+      "step": 4940
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.873502492744996e-05,
+      "loss": 1.214,
+      "step": 4950
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.873204851551455e-05,
+      "loss": 1.2158,
+      "step": 4960
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.872907210357914e-05,
+      "loss": 1.2072,
+      "step": 4970
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.8726095691643725e-05,
+      "loss": 1.2471,
+      "step": 4980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.872311927970831e-05,
+      "loss": 1.2117,
+      "step": 4990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.8720142867772905e-05,
+      "loss": 1.2051,
+      "step": 5000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.871716645583749e-05,
+      "loss": 1.2206,
+      "step": 5010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.871419004390208e-05,
+      "loss": 1.2174,
+      "step": 5020
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.8711213631966666e-05,
+      "loss": 1.1916,
+      "step": 5030
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.870823722003126e-05,
+      "loss": 1.211,
+      "step": 5040
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.8705260808095846e-05,
+      "loss": 1.2087,
+      "step": 5050
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.870228439616043e-05,
+      "loss": 1.2149,
+      "step": 5060
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.869930798422502e-05,
+      "loss": 1.2041,
+      "step": 5070
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.869633157228961e-05,
+      "loss": 1.1968,
+      "step": 5080
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.8693355160354194e-05,
+      "loss": 1.2136,
+      "step": 5090
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.869037874841879e-05,
+      "loss": 1.2065,
+      "step": 5100
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.8687402336483375e-05,
+      "loss": 1.2142,
+      "step": 5110
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.868442592454796e-05,
+      "loss": 1.2194,
+      "step": 5120
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.868144951261255e-05,
+      "loss": 1.1988,
+      "step": 5130
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.8678473100677135e-05,
+      "loss": 1.2101,
+      "step": 5140
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.867549668874173e-05,
+      "loss": 1.1966,
+      "step": 5150
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.867252027680631e-05,
+      "loss": 1.202,
+      "step": 5160
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.86695438648709e-05,
+      "loss": 1.1956,
+      "step": 5170
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.866656745293549e-05,
+      "loss": 1.2065,
+      "step": 5180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.8663591041000077e-05,
+      "loss": 1.1992,
+      "step": 5190
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.8660614629064663e-05,
+      "loss": 1.2015,
+      "step": 5200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.865763821712926e-05,
+      "loss": 1.1974,
+      "step": 5210
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.8654661805193844e-05,
+      "loss": 1.1899,
+      "step": 5220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.865168539325843e-05,
+      "loss": 1.2034,
+      "step": 5230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.864870898132302e-05,
+      "loss": 1.1922,
+      "step": 5240
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.864573256938761e-05,
+      "loss": 1.1894,
+      "step": 5250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.864275615745219e-05,
+      "loss": 1.1985,
+      "step": 5260
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.863977974551678e-05,
+      "loss": 1.1912,
+      "step": 5270
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.863680333358137e-05,
+      "loss": 1.1922,
+      "step": 5280
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.863382692164596e-05,
+      "loss": 1.1892,
+      "step": 5290
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.8630850509710546e-05,
+      "loss": 1.1919,
+      "step": 5300
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.862787409777513e-05,
+      "loss": 1.1961,
+      "step": 5310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.8624897685839727e-05,
+      "loss": 1.1764,
+      "step": 5320
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.8621921273904313e-05,
+      "loss": 1.1923,
+      "step": 5330
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.86189448619689e-05,
+      "loss": 1.1796,
+      "step": 5340
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.861596845003349e-05,
+      "loss": 1.1946,
+      "step": 5350
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.8612992038098074e-05,
+      "loss": 1.1791,
+      "step": 5360
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.861001562616266e-05,
+      "loss": 1.1898,
+      "step": 5370
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.860703921422725e-05,
+      "loss": 1.1776,
+      "step": 5380
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.860406280229184e-05,
+      "loss": 1.1841,
+      "step": 5390
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.860108639035643e-05,
+      "loss": 1.2027,
+      "step": 5400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.8598109978421015e-05,
+      "loss": 1.19,
+      "step": 5410
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.85951335664856e-05,
+      "loss": 1.1731,
+      "step": 5420
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.8592157154550196e-05,
+      "loss": 1.1836,
+      "step": 5430
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.858918074261478e-05,
+      "loss": 1.1793,
+      "step": 5440
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.858620433067937e-05,
+      "loss": 1.1975,
+      "step": 5450
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.858322791874396e-05,
+      "loss": 1.1888,
+      "step": 5460
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.8580251506808544e-05,
+      "loss": 1.1817,
+      "step": 5470
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.857727509487313e-05,
+      "loss": 1.1788,
+      "step": 5480
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.8574298682937724e-05,
+      "loss": 1.1872,
+      "step": 5490
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.857132227100231e-05,
+      "loss": 1.1736,
+      "step": 5500
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.85683458590669e-05,
+      "loss": 1.162,
+      "step": 5510
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8565369447131485e-05,
+      "loss": 1.176,
+      "step": 5520
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.856239303519608e-05,
+      "loss": 1.2021,
+      "step": 5530
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.8559416623260665e-05,
+      "loss": 1.1727,
+      "step": 5540
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.855644021132525e-05,
+      "loss": 1.1786,
+      "step": 5550
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.855346379938984e-05,
+      "loss": 1.1582,
+      "step": 5560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8550487387454426e-05,
+      "loss": 1.1623,
+      "step": 5570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.854751097551901e-05,
+      "loss": 1.1588,
+      "step": 5580
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.85445345635836e-05,
+      "loss": 1.1709,
+      "step": 5590
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.8541558151648194e-05,
+      "loss": 1.1573,
+      "step": 5600
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.853858173971278e-05,
+      "loss": 1.1646,
+      "step": 5610
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.853560532777737e-05,
+      "loss": 1.1699,
+      "step": 5620
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.8532628915841954e-05,
+      "loss": 1.1514,
+      "step": 5630
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.852965250390655e-05,
+      "loss": 1.1449,
+      "step": 5640
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.8526676091971135e-05,
+      "loss": 1.1587,
+      "step": 5650
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.852369968003572e-05,
+      "loss": 1.1662,
+      "step": 5660
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.852072326810031e-05,
+      "loss": 1.1885,
+      "step": 5670
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.8517746856164896e-05,
+      "loss": 1.1582,
+      "step": 5680
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.851477044422948e-05,
+      "loss": 1.173,
+      "step": 5690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.851179403229407e-05,
+      "loss": 1.1546,
+      "step": 5700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.850881762035866e-05,
+      "loss": 1.1769,
+      "step": 5710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.850584120842325e-05,
+      "loss": 1.1641,
+      "step": 5720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.850286479648784e-05,
+      "loss": 1.1584,
+      "step": 5730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.8499888384552424e-05,
+      "loss": 1.1524,
+      "step": 5740
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.849691197261702e-05,
+      "loss": 1.1628,
+      "step": 5750
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.8493935560681604e-05,
+      "loss": 1.1553,
+      "step": 5760
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.8490959148746184e-05,
+      "loss": 1.1592,
+      "step": 5770
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.848798273681078e-05,
+      "loss": 1.1648,
+      "step": 5780
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.8485006324875365e-05,
+      "loss": 1.1629,
+      "step": 5790
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.848202991293995e-05,
+      "loss": 1.1576,
+      "step": 5800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.847905350100454e-05,
+      "loss": 1.1585,
+      "step": 5810
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.847607708906913e-05,
+      "loss": 1.1571,
+      "step": 5820
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.847310067713372e-05,
+      "loss": 1.1392,
+      "step": 5830
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.8470124265198306e-05,
+      "loss": 1.1469,
+      "step": 5840
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.846714785326289e-05,
+      "loss": 1.1465,
+      "step": 5850
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.846417144132749e-05,
+      "loss": 1.1476,
+      "step": 5860
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.846119502939207e-05,
+      "loss": 1.1662,
+      "step": 5870
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.845821861745666e-05,
+      "loss": 1.1263,
+      "step": 5880
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.845524220552125e-05,
+      "loss": 1.154,
+      "step": 5890
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.8452265793585834e-05,
+      "loss": 1.1595,
+      "step": 5900
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.844928938165042e-05,
+      "loss": 1.1504,
+      "step": 5910
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 3.8446312969715015e-05,
+      "loss": 1.1441,
+      "step": 5920
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.84433365577796e-05,
+      "loss": 1.1481,
+      "step": 5930
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.844036014584419e-05,
+      "loss": 1.1386,
+      "step": 5940
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.8437383733908776e-05,
+      "loss": 1.1372,
+      "step": 5950
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 3.843440732197337e-05,
+      "loss": 1.1368,
+      "step": 5960
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.8431430910037956e-05,
+      "loss": 1.1494,
+      "step": 5970
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.8428454498102536e-05,
+      "loss": 1.148,
+      "step": 5980
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.842547808616713e-05,
+      "loss": 1.1337,
+      "step": 5990
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.842250167423172e-05,
+      "loss": 1.1465,
+      "step": 6000
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 3.8419525262296304e-05,
+      "loss": 1.1371,
+      "step": 6010
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.841654885036089e-05,
+      "loss": 1.1242,
+      "step": 6020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.8413572438425484e-05,
+      "loss": 1.1409,
+      "step": 6030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.841059602649007e-05,
+      "loss": 1.1452,
+      "step": 6040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 3.840761961455466e-05,
+      "loss": 1.1346,
+      "step": 6050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.8404643202619245e-05,
+      "loss": 1.1364,
+      "step": 6060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.840166679068384e-05,
+      "loss": 1.1436,
+      "step": 6070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.839869037874842e-05,
+      "loss": 1.1398,
+      "step": 6080
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.8395713966813006e-05,
+      "loss": 1.1574,
+      "step": 6090
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 3.83927375548776e-05,
+      "loss": 1.1414,
+      "step": 6100
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.8389761142942186e-05,
+      "loss": 1.1341,
+      "step": 6110
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.838678473100677e-05,
+      "loss": 1.141,
+      "step": 6120
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.838380831907136e-05,
+      "loss": 1.1101,
+      "step": 6130
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 3.8380831907135954e-05,
+      "loss": 1.1341,
+      "step": 6140
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.837785549520054e-05,
+      "loss": 1.1425,
+      "step": 6150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.837487908326513e-05,
+      "loss": 1.1246,
+      "step": 6160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.8371902671329715e-05,
+      "loss": 1.1381,
+      "step": 6170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.83689262593943e-05,
+      "loss": 1.1259,
+      "step": 6180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 3.836594984745889e-05,
+      "loss": 1.1512,
+      "step": 6190
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.8362973435523475e-05,
+      "loss": 1.1272,
+      "step": 6200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.835999702358807e-05,
+      "loss": 1.1403,
+      "step": 6210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.8357020611652656e-05,
+      "loss": 1.1433,
+      "step": 6220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.835404419971724e-05,
+      "loss": 1.1455,
+      "step": 6230
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.835106778778183e-05,
+      "loss": 1.1254,
+      "step": 6240
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.834809137584642e-05,
+      "loss": 1.1369,
+      "step": 6250
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.834511496391101e-05,
+      "loss": 1.1299,
+      "step": 6260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.83421385519756e-05,
+      "loss": 1.1196,
+      "step": 6270
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8339162140040184e-05,
+      "loss": 1.1497,
+      "step": 6280
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.833618572810477e-05,
+      "loss": 1.1128,
+      "step": 6290
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.833320931616936e-05,
+      "loss": 1.1267,
+      "step": 6300
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.833023290423395e-05,
+      "loss": 1.1305,
+      "step": 6310
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.832725649229854e-05,
+      "loss": 1.129,
+      "step": 6320
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.8324280080363125e-05,
+      "loss": 1.1043,
+      "step": 6330
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.832130366842771e-05,
+      "loss": 1.1118,
+      "step": 6340
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.8318327256492306e-05,
+      "loss": 1.1284,
+      "step": 6350
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.831535084455689e-05,
+      "loss": 1.1177,
+      "step": 6360
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.831237443262148e-05,
+      "loss": 1.1139,
+      "step": 6370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3.8309398020686067e-05,
+      "loss": 1.1071,
+      "step": 6380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3.8306421608750653e-05,
+      "loss": 1.1023,
+      "step": 6390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3.830344519681524e-05,
+      "loss": 1.1147,
+      "step": 6400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 3.830046878487983e-05,
+      "loss": 1.1229,
+      "step": 6410
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.829749237294442e-05,
+      "loss": 1.1109,
+      "step": 6420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.829451596100901e-05,
+      "loss": 1.1164,
+      "step": 6430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.8291539549073595e-05,
+      "loss": 1.1246,
+      "step": 6440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.828856313713818e-05,
+      "loss": 1.128,
+      "step": 6450
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 3.8285586725202775e-05,
+      "loss": 1.1092,
+      "step": 6460
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.828261031326736e-05,
+      "loss": 1.1121,
+      "step": 6470
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.827963390133195e-05,
+      "loss": 1.1167,
+      "step": 6480
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.8276657489396536e-05,
+      "loss": 1.1091,
+      "step": 6490
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 3.827368107746112e-05,
+      "loss": 1.1036,
+      "step": 6500
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.827070466552571e-05,
+      "loss": 1.1172,
+      "step": 6510
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.82677282535903e-05,
+      "loss": 1.1075,
+      "step": 6520
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.826475184165489e-05,
+      "loss": 1.1223,
+      "step": 6530
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.826177542971948e-05,
+      "loss": 1.1053,
+      "step": 6540
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 3.8258799017784064e-05,
+      "loss": 1.0968,
+      "step": 6550
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.825582260584865e-05,
+      "loss": 1.0947,
+      "step": 6560
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.8252846193913245e-05,
+      "loss": 1.1134,
+      "step": 6570
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.824986978197783e-05,
+      "loss": 1.1118,
+      "step": 6580
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 3.824689337004241e-05,
+      "loss": 1.1002,
+      "step": 6590
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.8243916958107005e-05,
+      "loss": 1.11,
+      "step": 6600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.824094054617159e-05,
+      "loss": 1.1034,
+      "step": 6610
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.823796413423618e-05,
+      "loss": 1.1032,
+      "step": 6620
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.8234987722300766e-05,
+      "loss": 1.1,
+      "step": 6630
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 3.823201131036536e-05,
+      "loss": 1.0981,
+      "step": 6640
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.822903489842995e-05,
+      "loss": 1.0943,
+      "step": 6650
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.8226058486494534e-05,
+      "loss": 1.1077,
+      "step": 6660
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.822308207455912e-05,
+      "loss": 1.0933,
+      "step": 6670
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.8220105662623714e-05,
+      "loss": 1.1062,
+      "step": 6680
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.8217129250688294e-05,
+      "loss": 1.0917,
+      "step": 6690
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.821415283875289e-05,
+      "loss": 1.1015,
+      "step": 6700
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.8211176426817475e-05,
+      "loss": 1.0839,
+      "step": 6710
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.820820001488206e-05,
+      "loss": 1.1201,
+      "step": 6720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.820522360294665e-05,
+      "loss": 1.0855,
+      "step": 6730
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.820224719101124e-05,
+      "loss": 1.1182,
+      "step": 6740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.819927077907583e-05,
+      "loss": 1.0862,
+      "step": 6750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.8196294367140416e-05,
+      "loss": 1.0984,
+      "step": 6760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.8193317955205e-05,
+      "loss": 1.0862,
+      "step": 6770
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.81903415432696e-05,
+      "loss": 1.0875,
+      "step": 6780
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.8187365131334184e-05,
+      "loss": 1.1108,
+      "step": 6790
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.8184388719398764e-05,
+      "loss": 1.0805,
+      "step": 6800
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.818141230746336e-05,
+      "loss": 1.0807,
+      "step": 6810
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 3.8178435895527944e-05,
+      "loss": 1.0862,
+      "step": 6820
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.817545948359253e-05,
+      "loss": 1.0746,
+      "step": 6830
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.817248307165712e-05,
+      "loss": 1.0796,
+      "step": 6840
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.816950665972171e-05,
+      "loss": 1.08,
+      "step": 6850
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.81665302477863e-05,
+      "loss": 1.052,
+      "step": 6860
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.8163553835850886e-05,
+      "loss": 1.0893,
+      "step": 6870
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.816057742391547e-05,
+      "loss": 1.0841,
+      "step": 6880
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.8157601011980066e-05,
+      "loss": 1.0858,
+      "step": 6890
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.8154624600044646e-05,
+      "loss": 1.0624,
+      "step": 6900
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 3.815164818810923e-05,
+      "loss": 1.0805,
+      "step": 6910
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.814867177617383e-05,
+      "loss": 1.0556,
+      "step": 6920
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.8145695364238414e-05,
+      "loss": 1.0876,
+      "step": 6930
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.8142718952303e-05,
+      "loss": 1.0858,
+      "step": 6940
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.813974254036759e-05,
+      "loss": 1.066,
+      "step": 6950
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.813676612843218e-05,
+      "loss": 1.0747,
+      "step": 6960
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.813378971649677e-05,
+      "loss": 1.0754,
+      "step": 6970
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.8130813304561355e-05,
+      "loss": 1.075,
+      "step": 6980
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.812783689262594e-05,
+      "loss": 1.0834,
+      "step": 6990
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.812486048069053e-05,
+      "loss": 1.0686,
+      "step": 7000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.8121884068755116e-05,
+      "loss": 1.077,
+      "step": 7010
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.81189076568197e-05,
+      "loss": 1.0611,
+      "step": 7020
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.8115931244884296e-05,
+      "loss": 1.0664,
+      "step": 7030
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 3.811295483294888e-05,
+      "loss": 1.0635,
+      "step": 7040
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.810997842101347e-05,
+      "loss": 1.0724,
+      "step": 7050
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.810700200907806e-05,
+      "loss": 1.0644,
+      "step": 7060
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.810402559714265e-05,
+      "loss": 1.0836,
+      "step": 7070
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.810104918520724e-05,
+      "loss": 1.0688,
+      "step": 7080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.8098072773271824e-05,
+      "loss": 1.0609,
+      "step": 7090
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.809509636133641e-05,
+      "loss": 1.0742,
+      "step": 7100
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.8092119949401e-05,
+      "loss": 1.0699,
+      "step": 7110
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.8089143537465585e-05,
+      "loss": 1.0624,
+      "step": 7120
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 3.808616712553018e-05,
+      "loss": 1.0493,
+      "step": 7130
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.8083190713594766e-05,
+      "loss": 1.0539,
+      "step": 7140
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.808021430165935e-05,
+      "loss": 1.0647,
+      "step": 7150
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.807723788972394e-05,
+      "loss": 1.0728,
+      "step": 7160
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.807426147778853e-05,
+      "loss": 1.069,
+      "step": 7170
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.807128506585312e-05,
+      "loss": 1.069,
+      "step": 7180
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.806830865391771e-05,
+      "loss": 1.0444,
+      "step": 7190
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.8065332241982294e-05,
+      "loss": 1.0522,
+      "step": 7200
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.806235583004688e-05,
+      "loss": 1.0402,
+      "step": 7210
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 3.805937941811147e-05,
+      "loss": 1.048,
+      "step": 7220
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.8056403006176055e-05,
+      "loss": 1.0653,
+      "step": 7230
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.805342659424065e-05,
+      "loss": 1.0674,
+      "step": 7240
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.8050450182305235e-05,
+      "loss": 1.0524,
+      "step": 7250
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.804747377036982e-05,
+      "loss": 1.0596,
+      "step": 7260
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.804449735843441e-05,
+      "loss": 1.0369,
+      "step": 7270
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8041520946499e-05,
+      "loss": 1.0516,
+      "step": 7280
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.803854453456359e-05,
+      "loss": 1.0487,
+      "step": 7290
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8035568122628176e-05,
+      "loss": 1.0408,
+      "step": 7300
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.803259171069276e-05,
+      "loss": 1.0429,
+      "step": 7310
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.802961529875735e-05,
+      "loss": 1.047,
+      "step": 7320
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.802663888682194e-05,
+      "loss": 1.0421,
+      "step": 7330
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8023662474886524e-05,
+      "loss": 1.0495,
+      "step": 7340
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.802068606295112e-05,
+      "loss": 1.0429,
+      "step": 7350
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8017709651015705e-05,
+      "loss": 1.026,
+      "step": 7360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.801473323908029e-05,
+      "loss": 1.0322,
+      "step": 7370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.801175682714488e-05,
+      "loss": 1.0428,
+      "step": 7380
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.800878041520947e-05,
+      "loss": 1.0481,
+      "step": 7390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.800580400327406e-05,
+      "loss": 1.0312,
+      "step": 7400
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.800282759133864e-05,
+      "loss": 1.0529,
+      "step": 7410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.799985117940323e-05,
+      "loss": 1.0374,
+      "step": 7420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.799687476746782e-05,
+      "loss": 1.035,
+      "step": 7430
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.7993898355532407e-05,
+      "loss": 1.0455,
+      "step": 7440
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.7990921943596993e-05,
+      "loss": 1.0227,
+      "step": 7450
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.798794553166159e-05,
+      "loss": 1.0469,
+      "step": 7460
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.7984969119726174e-05,
+      "loss": 1.0236,
+      "step": 7470
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.798199270779076e-05,
+      "loss": 1.0454,
+      "step": 7480
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.797901629585535e-05,
+      "loss": 1.0369,
+      "step": 7490
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.797603988391994e-05,
+      "loss": 1.0476,
+      "step": 7500
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.797306347198452e-05,
+      "loss": 1.04,
+      "step": 7510
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.7970087060049115e-05,
+      "loss": 1.0316,
+      "step": 7520
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.79671106481137e-05,
+      "loss": 1.0308,
+      "step": 7530
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.796413423617829e-05,
+      "loss": 1.0427,
+      "step": 7540
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.7961157824242876e-05,
+      "loss": 1.0347,
+      "step": 7550
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.795818141230747e-05,
+      "loss": 1.0281,
+      "step": 7560
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.7955205000372057e-05,
+      "loss": 1.0232,
+      "step": 7570
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.7952228588436644e-05,
+      "loss": 1.0331,
+      "step": 7580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.794925217650123e-05,
+      "loss": 1.0244,
+      "step": 7590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.7946275764565824e-05,
+      "loss": 1.0458,
+      "step": 7600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.7943299352630404e-05,
+      "loss": 1.0317,
+      "step": 7610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.794032294069499e-05,
+      "loss": 1.0197,
+      "step": 7620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.7937346528759585e-05,
+      "loss": 1.0193,
+      "step": 7630
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.793437011682417e-05,
+      "loss": 1.0155,
+      "step": 7640
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.793139370488876e-05,
+      "loss": 1.0388,
+      "step": 7650
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.7928417292953345e-05,
+      "loss": 1.0302,
+      "step": 7660
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.792544088101794e-05,
+      "loss": 1.0311,
+      "step": 7670
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.7922464469082526e-05,
+      "loss": 1.0158,
+      "step": 7680
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.791948805714711e-05,
+      "loss": 1.0224,
+      "step": 7690
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.79165116452117e-05,
+      "loss": 1.029,
+      "step": 7700
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.7913535233276294e-05,
+      "loss": 1.0184,
+      "step": 7710
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.7910558821340874e-05,
+      "loss": 1.0274,
+      "step": 7720
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.790758240940546e-05,
+      "loss": 1.0096,
+      "step": 7730
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.7904605997470054e-05,
+      "loss": 1.022,
+      "step": 7740
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.790162958553464e-05,
+      "loss": 1.0078,
+      "step": 7750
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 3.789865317359923e-05,
+      "loss": 1.0061,
+      "step": 7760
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.7895676761663815e-05,
+      "loss": 1.0035,
+      "step": 7770
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.789270034972841e-05,
+      "loss": 1.0234,
+      "step": 7780
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.7889723937792995e-05,
+      "loss": 0.9974,
+      "step": 7790
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.788674752585758e-05,
+      "loss": 1.011,
+      "step": 7800
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.788377111392217e-05,
+      "loss": 1.0217,
+      "step": 7810
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.7880794701986756e-05,
+      "loss": 0.9963,
+      "step": 7820
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.787781829005134e-05,
+      "loss": 1.012,
+      "step": 7830
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.787484187811593e-05,
+      "loss": 1.0093,
+      "step": 7840
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 3.7871865466180524e-05,
+      "loss": 1.0112,
+      "step": 7850
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.786888905424511e-05,
+      "loss": 1.0194,
+      "step": 7860
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.78659126423097e-05,
+      "loss": 1.0182,
+      "step": 7870
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7862936230374284e-05,
+      "loss": 1.0096,
+      "step": 7880
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.785995981843888e-05,
+      "loss": 1.0125,
+      "step": 7890
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.7856983406503465e-05,
+      "loss": 1.0079,
+      "step": 7900
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.785400699456805e-05,
+      "loss": 1.0032,
+      "step": 7910
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.785103058263264e-05,
+      "loss": 0.9947,
+      "step": 7920
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.7848054170697226e-05,
+      "loss": 0.992,
+      "step": 7930
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 3.784507775876181e-05,
+      "loss": 0.9898,
+      "step": 7940
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.7842101346826406e-05,
+      "loss": 0.98,
+      "step": 7950
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.783912493489099e-05,
+      "loss": 0.9815,
+      "step": 7960
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.783614852295558e-05,
+      "loss": 0.9942,
+      "step": 7970
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.783317211102017e-05,
+      "loss": 0.9851,
+      "step": 7980
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.783019569908476e-05,
+      "loss": 0.9984,
+      "step": 7990
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.782721928714935e-05,
+      "loss": 0.9957,
+      "step": 8000
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.7824242875213934e-05,
+      "loss": 1.0029,
+      "step": 8010
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.782126646327852e-05,
+      "loss": 0.9872,
+      "step": 8020
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 3.781829005134311e-05,
+      "loss": 0.978,
+      "step": 8030
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.7815313639407695e-05,
+      "loss": 0.9932,
+      "step": 8040
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.781233722747228e-05,
+      "loss": 0.9949,
+      "step": 8050
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.7809360815536876e-05,
+      "loss": 1.001,
+      "step": 8060
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.780638440360146e-05,
+      "loss": 0.9895,
+      "step": 8070
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.780340799166605e-05,
+      "loss": 0.9885,
+      "step": 8080
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.7800431579730636e-05,
+      "loss": 0.9989,
+      "step": 8090
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.779745516779523e-05,
+      "loss": 0.9817,
+      "step": 8100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.779447875585982e-05,
+      "loss": 0.9872,
+      "step": 8110
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.7791502343924404e-05,
+      "loss": 0.9949,
+      "step": 8120
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.778852593198899e-05,
+      "loss": 0.9782,
+      "step": 8130
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.778554952005358e-05,
+      "loss": 0.9708,
+      "step": 8140
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.7782573108118165e-05,
+      "loss": 0.9698,
+      "step": 8150
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.777959669618275e-05,
+      "loss": 0.9862,
+      "step": 8160
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 3.7776620284247345e-05,
+      "loss": 0.9901,
+      "step": 8170
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.777364387231193e-05,
+      "loss": 0.9854,
+      "step": 8180
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.777066746037652e-05,
+      "loss": 0.9839,
+      "step": 8190
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.7767691048441106e-05,
+      "loss": 0.9972,
+      "step": 8200
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.77647146365057e-05,
+      "loss": 0.9834,
+      "step": 8210
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7761738224570286e-05,
+      "loss": 0.9786,
+      "step": 8220
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7758761812634866e-05,
+      "loss": 0.9751,
+      "step": 8230
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.775578540069946e-05,
+      "loss": 0.9697,
+      "step": 8240
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.775280898876405e-05,
+      "loss": 0.9819,
+      "step": 8250
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 3.7749832576828634e-05,
+      "loss": 0.974,
+      "step": 8260
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.774685616489322e-05,
+      "loss": 0.9574,
+      "step": 8270
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.7743879752957815e-05,
+      "loss": 0.9729,
+      "step": 8280
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.77409033410224e-05,
+      "loss": 0.956,
+      "step": 8290
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.773792692908699e-05,
+      "loss": 0.9868,
+      "step": 8300
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.7734950517151575e-05,
+      "loss": 0.9841,
+      "step": 8310
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.773197410521617e-05,
+      "loss": 0.9652,
+      "step": 8320
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.772899769328075e-05,
+      "loss": 0.9819,
+      "step": 8330
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.772602128134534e-05,
+      "loss": 0.9667,
+      "step": 8340
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 3.772304486940993e-05,
+      "loss": 0.9775,
+      "step": 8350
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.7720068457474516e-05,
+      "loss": 0.975,
+      "step": 8360
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.77170920455391e-05,
+      "loss": 0.9593,
+      "step": 8370
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.77141156336037e-05,
+      "loss": 0.9752,
+      "step": 8380
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.7711139221668284e-05,
+      "loss": 0.967,
+      "step": 8390
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.770816280973287e-05,
+      "loss": 0.9687,
+      "step": 8400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.770518639779746e-05,
+      "loss": 0.9701,
+      "step": 8410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.770220998586205e-05,
+      "loss": 0.9732,
+      "step": 8420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.769923357392663e-05,
+      "loss": 0.9781,
+      "step": 8430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 3.769625716199122e-05,
+      "loss": 0.9708,
+      "step": 8440
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.769328075005581e-05,
+      "loss": 0.9513,
+      "step": 8450
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.76903043381204e-05,
+      "loss": 0.9511,
+      "step": 8460
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.7687327926184986e-05,
+      "loss": 0.9704,
+      "step": 8470
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.768435151424957e-05,
+      "loss": 0.9584,
+      "step": 8480
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.7681375102314166e-05,
+      "loss": 0.949,
+      "step": 8490
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.7678398690378753e-05,
+      "loss": 0.9671,
+      "step": 8500
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.767542227844334e-05,
+      "loss": 0.984,
+      "step": 8510
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.767244586650793e-05,
+      "loss": 0.9748,
+      "step": 8520
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.766946945457252e-05,
+      "loss": 0.9538,
+      "step": 8530
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.76664930426371e-05,
+      "loss": 0.9718,
+      "step": 8540
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.766351663070169e-05,
+      "loss": 0.9596,
+      "step": 8550
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.766054021876628e-05,
+      "loss": 0.9681,
+      "step": 8560
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 3.765756380683087e-05,
+      "loss": 0.9484,
+      "step": 8570
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.7654587394895455e-05,
+      "loss": 0.962,
+      "step": 8580
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.765161098296004e-05,
+      "loss": 0.9662,
+      "step": 8590
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.7648634571024636e-05,
+      "loss": 0.9754,
+      "step": 8600
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.764565815908922e-05,
+      "loss": 0.937,
+      "step": 8610
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.764268174715381e-05,
+      "loss": 0.9479,
+      "step": 8620
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.76397053352184e-05,
+      "loss": 0.9549,
+      "step": 8630
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.7636728923282984e-05,
+      "loss": 0.9426,
+      "step": 8640
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.763375251134757e-05,
+      "loss": 0.9424,
+      "step": 8650
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 3.763077609941216e-05,
+      "loss": 0.9594,
+      "step": 8660
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.762779968747675e-05,
+      "loss": 0.9533,
+      "step": 8670
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.762482327554134e-05,
+      "loss": 0.9449,
+      "step": 8680
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.7621846863605925e-05,
+      "loss": 0.9201,
+      "step": 8690
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.761887045167051e-05,
+      "loss": 0.9542,
+      "step": 8700
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.7615894039735105e-05,
+      "loss": 0.9345,
+      "step": 8710
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.761291762779969e-05,
+      "loss": 0.9438,
+      "step": 8720
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.760994121586428e-05,
+      "loss": 0.9487,
+      "step": 8730
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.7606964803928866e-05,
+      "loss": 0.9477,
+      "step": 8740
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 3.760398839199345e-05,
+      "loss": 0.9472,
+      "step": 8750
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.760101198005804e-05,
+      "loss": 0.9431,
+      "step": 8760
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7598035568122634e-05,
+      "loss": 0.9513,
+      "step": 8770
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.759505915618722e-05,
+      "loss": 0.9295,
+      "step": 8780
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.759208274425181e-05,
+      "loss": 0.9347,
+      "step": 8790
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.7589106332316394e-05,
+      "loss": 0.9342,
+      "step": 8800
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.758612992038099e-05,
+      "loss": 0.9261,
+      "step": 8810
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.7583153508445575e-05,
+      "loss": 0.9413,
+      "step": 8820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.758017709651016e-05,
+      "loss": 0.9242,
+      "step": 8830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.757720068457475e-05,
+      "loss": 0.9313,
+      "step": 8840
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.7574224272639336e-05,
+      "loss": 0.9425,
+      "step": 8850
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.757124786070392e-05,
+      "loss": 0.9263,
+      "step": 8860
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.756827144876851e-05,
+      "loss": 0.9401,
+      "step": 8870
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.75652950368331e-05,
+      "loss": 0.9473,
+      "step": 8880
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 3.756231862489769e-05,
+      "loss": 0.9045,
+      "step": 8890
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.755934221296228e-05,
+      "loss": 0.9203,
+      "step": 8900
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.7556365801026864e-05,
+      "loss": 0.9091,
+      "step": 8910
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.755338938909146e-05,
+      "loss": 0.9553,
+      "step": 8920
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.7550412977156044e-05,
+      "loss": 0.9083,
+      "step": 8930
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.7547436565220624e-05,
+      "loss": 0.9219,
+      "step": 8940
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.754446015328522e-05,
+      "loss": 0.9214,
+      "step": 8950
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.7541483741349805e-05,
+      "loss": 0.9329,
+      "step": 8960
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.753850732941439e-05,
+      "loss": 0.9268,
+      "step": 8970
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 3.753553091747898e-05,
+      "loss": 0.9159,
+      "step": 8980
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.753255450554357e-05,
+      "loss": 0.9437,
+      "step": 8990
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.752957809360816e-05,
+      "loss": 0.9188,
+      "step": 9000
+    },
+    {
+      "epoch": 2.0,
+      "eval_cer": 4.685897435897436,
+      "eval_loss": 0.8316300511360168,
+      "eval_runtime": 1.2974,
+      "eval_samples_per_second": 7.708,
+      "eval_steps_per_second": 0.771,
+      "eval_wer": 0.8311688311688312,
+      "step": 9004
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.7526601681672746e-05,
+      "loss": 0.909,
+      "step": 9010
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.752362526973733e-05,
+      "loss": 0.9058,
+      "step": 9020
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.752064885780193e-05,
+      "loss": 0.908,
+      "step": 9030
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.7517672445866514e-05,
+      "loss": 0.9152,
+      "step": 9040
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.75146960339311e-05,
+      "loss": 0.919,
+      "step": 9050
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.751171962199569e-05,
+      "loss": 0.9246,
+      "step": 9060
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 3.7508743210060274e-05,
+      "loss": 0.9139,
+      "step": 9070
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.750576679812486e-05,
+      "loss": 0.9105,
+      "step": 9080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.750279038618945e-05,
+      "loss": 0.9062,
+      "step": 9090
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.749981397425404e-05,
+      "loss": 0.9064,
+      "step": 9100
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.749683756231863e-05,
+      "loss": 0.8911,
+      "step": 9110
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.7493861150383216e-05,
+      "loss": 0.9365,
+      "step": 9120
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.74908847384478e-05,
+      "loss": 0.9221,
+      "step": 9130
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.7487908326512396e-05,
+      "loss": 0.918,
+      "step": 9140
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.7484931914576976e-05,
+      "loss": 0.9104,
+      "step": 9150
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 3.748195550264157e-05,
+      "loss": 0.9247,
+      "step": 9160
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.747897909070616e-05,
+      "loss": 0.9126,
+      "step": 9170
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.7476002678770744e-05,
+      "loss": 0.9248,
+      "step": 9180
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.747302626683533e-05,
+      "loss": 0.9224,
+      "step": 9190
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.7470049854899924e-05,
+      "loss": 0.9154,
+      "step": 9200
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.746707344296451e-05,
+      "loss": 0.9153,
+      "step": 9210
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.74640970310291e-05,
+      "loss": 0.9031,
+      "step": 9220
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.7461120619093685e-05,
+      "loss": 0.9152,
+      "step": 9230
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.745814420715828e-05,
+      "loss": 0.9025,
+      "step": 9240
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.745516779522286e-05,
+      "loss": 0.9115,
+      "step": 9250
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.7452191383287446e-05,
+      "loss": 0.9015,
+      "step": 9260
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.744921497135204e-05,
+      "loss": 0.9019,
+      "step": 9270
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.7446238559416626e-05,
+      "loss": 0.9118,
+      "step": 9280
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 3.744326214748121e-05,
+      "loss": 0.9137,
+      "step": 9290
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.74402857355458e-05,
+      "loss": 0.9006,
+      "step": 9300
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.7437309323610394e-05,
+      "loss": 0.894,
+      "step": 9310
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.743433291167498e-05,
+      "loss": 0.8886,
+      "step": 9320
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.743135649973957e-05,
+      "loss": 0.9091,
+      "step": 9330
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.7428380087804155e-05,
+      "loss": 0.8969,
+      "step": 9340
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 3.742540367586874e-05,
+      "loss": 0.8952,
+      "step": 9350
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 3.742242726393333e-05,
+      "loss": 0.8986,
+      "step": 9360
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 3.7419450851997915e-05,
+      "loss": 0.9088,
+      "step": 9370
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 3.741647444006251e-05,
+      "loss": 0.9057,
+      "step": 9380
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.7413498028127096e-05,
+      "loss": 0.8899,
+      "step": 9390
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.741052161619168e-05,
+      "loss": 0.8941,
+      "step": 9400
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.740754520425627e-05,
+      "loss": 0.9046,
+      "step": 9410
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.740456879232086e-05,
+      "loss": 0.8857,
+      "step": 9420
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.740159238038545e-05,
+      "loss": 0.8975,
+      "step": 9430
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.739861596845004e-05,
+      "loss": 0.8908,
+      "step": 9440
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.7395639556514624e-05,
+      "loss": 0.8687,
+      "step": 9450
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.739266314457921e-05,
+      "loss": 0.8881,
+      "step": 9460
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 3.73896867326438e-05,
+      "loss": 0.8776,
+      "step": 9470
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.738671032070839e-05,
+      "loss": 0.8909,
+      "step": 9480
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.738373390877298e-05,
+      "loss": 0.9086,
+      "step": 9490
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.7380757496837565e-05,
+      "loss": 0.8855,
+      "step": 9500
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.737778108490215e-05,
+      "loss": 0.8886,
+      "step": 9510
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3.737480467296674e-05,
+      "loss": 0.8911,
+      "step": 9520
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.737182826103133e-05,
+      "loss": 0.8964,
+      "step": 9530
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.736885184909592e-05,
+      "loss": 0.8926,
+      "step": 9540
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.7365875437160507e-05,
+      "loss": 0.8699,
+      "step": 9550
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.7362899025225093e-05,
+      "loss": 0.8837,
+      "step": 9560
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.735992261328968e-05,
+      "loss": 0.8827,
+      "step": 9570
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.735694620135427e-05,
+      "loss": 0.8796,
+      "step": 9580
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.735396978941886e-05,
+      "loss": 0.8829,
+      "step": 9590
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.735099337748345e-05,
+      "loss": 0.8814,
+      "step": 9600
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 3.7348016965548035e-05,
+      "loss": 0.8699,
+      "step": 9610
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.734504055361262e-05,
+      "loss": 0.8668,
+      "step": 9620
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.7342064141677215e-05,
+      "loss": 0.8735,
+      "step": 9630
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.73390877297418e-05,
+      "loss": 0.8924,
+      "step": 9640
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 3.733611131780639e-05,
+      "loss": 0.8847,
+      "step": 9650
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.7333134905870976e-05,
+      "loss": 0.8747,
+      "step": 9660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.733015849393556e-05,
+      "loss": 0.877,
+      "step": 9670
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.732718208200015e-05,
+      "loss": 0.8715,
+      "step": 9680
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.732420567006474e-05,
+      "loss": 0.8858,
+      "step": 9690
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 3.732122925812933e-05,
+      "loss": 0.8876,
+      "step": 9700
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.731825284619392e-05,
+      "loss": 0.8747,
+      "step": 9710
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.7315276434258504e-05,
+      "loss": 0.8906,
+      "step": 9720
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.731230002232309e-05,
+      "loss": 0.8806,
+      "step": 9730
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 3.7309323610387685e-05,
+      "loss": 0.879,
+      "step": 9740
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.730634719845227e-05,
+      "loss": 0.8781,
+      "step": 9750
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.730337078651685e-05,
+      "loss": 0.8691,
+      "step": 9760
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.7300394374581445e-05,
+      "loss": 0.8814,
+      "step": 9770
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.729741796264603e-05,
+      "loss": 0.898,
+      "step": 9780
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 3.729444155071062e-05,
+      "loss": 0.8613,
+      "step": 9790
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.7291465138775206e-05,
+      "loss": 0.8619,
+      "step": 9800
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.72884887268398e-05,
+      "loss": 0.8846,
+      "step": 9810
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.728551231490439e-05,
+      "loss": 0.863,
+      "step": 9820
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 3.7282535902968974e-05,
+      "loss": 0.8625,
+      "step": 9830
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.727955949103356e-05,
+      "loss": 0.8868,
+      "step": 9840
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7276583079098154e-05,
+      "loss": 0.8794,
+      "step": 9850
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.727360666716274e-05,
+      "loss": 0.8734,
+      "step": 9860
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.727063025522733e-05,
+      "loss": 0.8767,
+      "step": 9870
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 3.7267653843291915e-05,
+      "loss": 0.8649,
+      "step": 9880
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.72646774313565e-05,
+      "loss": 0.8798,
+      "step": 9890
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.726170101942109e-05,
+      "loss": 0.8717,
+      "step": 9900
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.725872460748568e-05,
+      "loss": 0.8685,
+      "step": 9910
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 3.725574819555027e-05,
+      "loss": 0.8622,
+      "step": 9920
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7252771783614856e-05,
+      "loss": 0.8753,
+      "step": 9930
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.724979537167944e-05,
+      "loss": 0.846,
+      "step": 9940
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.724681895974403e-05,
+      "loss": 0.8718,
+      "step": 9950
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7243842547808624e-05,
+      "loss": 0.8664,
+      "step": 9960
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 3.7240866135873204e-05,
+      "loss": 0.8674,
+      "step": 9970
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.72378897239378e-05,
+      "loss": 0.8659,
+      "step": 9980
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.7234913312002384e-05,
+      "loss": 0.8646,
+      "step": 9990
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.723193690006697e-05,
+      "loss": 0.8692,
+      "step": 10000
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 3.722896048813156e-05,
+      "loss": 0.8717,
+      "step": 10010
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.722598407619615e-05,
+      "loss": 0.8537,
+      "step": 10020
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.722300766426074e-05,
+      "loss": 0.8468,
+      "step": 10030
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.7220031252325326e-05,
+      "loss": 0.8571,
+      "step": 10040
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.721705484038991e-05,
+      "loss": 0.8618,
+      "step": 10050
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 3.7214078428454506e-05,
+      "loss": 0.8602,
+      "step": 10060
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.7211102016519086e-05,
+      "loss": 0.8524,
+      "step": 10070
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.720812560458367e-05,
+      "loss": 0.8541,
+      "step": 10080
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.720514919264827e-05,
+      "loss": 0.8578,
+      "step": 10090
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 3.7202172780712854e-05,
+      "loss": 0.8582,
+      "step": 10100
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.719919636877744e-05,
+      "loss": 0.8369,
+      "step": 10110
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.719621995684203e-05,
+      "loss": 0.8386,
+      "step": 10120
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.719324354490662e-05,
+      "loss": 0.8565,
+      "step": 10130
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.719026713297121e-05,
+      "loss": 0.8454,
+      "step": 10140
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 3.7187290721035795e-05,
+      "loss": 0.8456,
+      "step": 10150
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.718431430910038e-05,
+      "loss": 0.8701,
+      "step": 10160
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.718133789716497e-05,
+      "loss": 0.8573,
+      "step": 10170
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.7178361485229556e-05,
+      "loss": 0.8413,
+      "step": 10180
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 3.717538507329414e-05,
+      "loss": 0.8683,
+      "step": 10190
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.7172408661358736e-05,
+      "loss": 0.8373,
+      "step": 10200
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.716943224942332e-05,
+      "loss": 0.8287,
+      "step": 10210
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.716645583748791e-05,
+      "loss": 0.8613,
+      "step": 10220
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.71634794255525e-05,
+      "loss": 0.8366,
+      "step": 10230
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 3.716050301361709e-05,
+      "loss": 0.8401,
+      "step": 10240
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.715752660168168e-05,
+      "loss": 0.8672,
+      "step": 10250
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.7154550189746264e-05,
+      "loss": 0.8361,
+      "step": 10260
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.715157377781085e-05,
+      "loss": 0.8516,
+      "step": 10270
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 3.714859736587544e-05,
+      "loss": 0.8361,
+      "step": 10280
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.7145620953940025e-05,
+      "loss": 0.8408,
+      "step": 10290
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.714264454200462e-05,
+      "loss": 0.8284,
+      "step": 10300
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.7139668130069206e-05,
+      "loss": 0.8442,
+      "step": 10310
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.713669171813379e-05,
+      "loss": 0.8487,
+      "step": 10320
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 3.713371530619838e-05,
+      "loss": 0.8469,
+      "step": 10330
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.713073889426297e-05,
+      "loss": 0.8553,
+      "step": 10340
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.712776248232756e-05,
+      "loss": 0.8392,
+      "step": 10350
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.712478607039215e-05,
+      "loss": 0.8377,
+      "step": 10360
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 3.7121809658456734e-05,
+      "loss": 0.839,
+      "step": 10370
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.711883324652132e-05,
+      "loss": 0.8557,
+      "step": 10380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.711585683458591e-05,
+      "loss": 0.8299,
+      "step": 10390
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.7112880422650495e-05,
+      "loss": 0.8427,
+      "step": 10400
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.710990401071509e-05,
+      "loss": 0.8536,
+      "step": 10410
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 3.7106927598779675e-05,
+      "loss": 0.857,
+      "step": 10420
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.710395118684426e-05,
+      "loss": 0.8291,
+      "step": 10430
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.710097477490885e-05,
+      "loss": 0.8443,
+      "step": 10440
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.709799836297344e-05,
+      "loss": 0.8472,
+      "step": 10450
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 3.709502195103803e-05,
+      "loss": 0.8314,
+      "step": 10460
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.7092045539102616e-05,
+      "loss": 0.8473,
+      "step": 10470
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.70890691271672e-05,
+      "loss": 0.842,
+      "step": 10480
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.708609271523179e-05,
+      "loss": 0.8405,
+      "step": 10490
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.708311630329638e-05,
+      "loss": 0.8296,
+      "step": 10500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 3.7080139891360964e-05,
+      "loss": 0.831,
+      "step": 10510
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.707716347942556e-05,
+      "loss": 0.8335,
+      "step": 10520
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.7074187067490145e-05,
+      "loss": 0.8287,
+      "step": 10530
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.707121065555473e-05,
+      "loss": 0.8356,
+      "step": 10540
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 3.706823424361932e-05,
+      "loss": 0.8302,
+      "step": 10550
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.706525783168391e-05,
+      "loss": 0.8084,
+      "step": 10560
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.70622814197485e-05,
+      "loss": 0.8171,
+      "step": 10570
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.705930500781308e-05,
+      "loss": 0.8418,
+      "step": 10580
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.705632859587767e-05,
+      "loss": 0.8116,
+      "step": 10590
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 3.705335218394226e-05,
+      "loss": 0.8125,
+      "step": 10600
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.7050375772006847e-05,
+      "loss": 0.8129,
+      "step": 10610
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.7047399360071433e-05,
+      "loss": 0.8246,
+      "step": 10620
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.704442294813603e-05,
+      "loss": 0.8263,
+      "step": 10630
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 3.7041446536200614e-05,
+      "loss": 0.8357,
+      "step": 10640
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.70384701242652e-05,
+      "loss": 0.8328,
+      "step": 10650
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.703549371232979e-05,
+      "loss": 0.8297,
+      "step": 10660
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.703251730039438e-05,
+      "loss": 0.839,
+      "step": 10670
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.702954088845896e-05,
+      "loss": 0.821,
+      "step": 10680
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.7026564476523555e-05,
+      "loss": 0.8227,
+      "step": 10690
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.702358806458814e-05,
+      "loss": 0.8311,
+      "step": 10700
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.702061165265273e-05,
+      "loss": 0.814,
+      "step": 10710
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.7017635240717316e-05,
+      "loss": 0.8367,
+      "step": 10720
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.701465882878191e-05,
+      "loss": 0.8181,
+      "step": 10730
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.7011682416846497e-05,
+      "loss": 0.8313,
+      "step": 10740
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.7008706004911083e-05,
+      "loss": 0.8,
+      "step": 10750
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.700572959297567e-05,
+      "loss": 0.8231,
+      "step": 10760
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.7002753181040264e-05,
+      "loss": 0.8129,
+      "step": 10770
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.699977676910485e-05,
+      "loss": 0.8214,
+      "step": 10780
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.699680035716943e-05,
+      "loss": 0.8266,
+      "step": 10790
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.6993823945234025e-05,
+      "loss": 0.8263,
+      "step": 10800
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.699084753329861e-05,
+      "loss": 0.8121,
+      "step": 10810
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.69878711213632e-05,
+      "loss": 0.8123,
+      "step": 10820
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.6984894709427785e-05,
+      "loss": 0.8424,
+      "step": 10830
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.698191829749238e-05,
+      "loss": 0.8144,
+      "step": 10840
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.6978941885556966e-05,
+      "loss": 0.8157,
+      "step": 10850
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.697596547362155e-05,
+      "loss": 0.8318,
+      "step": 10860
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.697298906168614e-05,
+      "loss": 0.8234,
+      "step": 10870
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.6970012649750733e-05,
+      "loss": 0.825,
+      "step": 10880
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.6967036237815314e-05,
+      "loss": 0.8197,
+      "step": 10890
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.69640598258799e-05,
+      "loss": 0.8118,
+      "step": 10900
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.6961083413944494e-05,
+      "loss": 0.8093,
+      "step": 10910
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.695810700200908e-05,
+      "loss": 0.8172,
+      "step": 10920
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.695513059007367e-05,
+      "loss": 0.8101,
+      "step": 10930
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.6952154178138255e-05,
+      "loss": 0.8243,
+      "step": 10940
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.694917776620285e-05,
+      "loss": 0.8235,
+      "step": 10950
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.6946201354267435e-05,
+      "loss": 0.8218,
+      "step": 10960
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.694322494233202e-05,
+      "loss": 0.8224,
+      "step": 10970
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.694024853039661e-05,
+      "loss": 0.835,
+      "step": 10980
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.6937272118461196e-05,
+      "loss": 0.8168,
+      "step": 10990
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.693429570652578e-05,
+      "loss": 0.8166,
+      "step": 11000
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.693131929459037e-05,
+      "loss": 0.8283,
+      "step": 11010
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6928342882654964e-05,
+      "loss": 0.8296,
+      "step": 11020
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.692536647071955e-05,
+      "loss": 0.8144,
+      "step": 11030
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.692239005878414e-05,
+      "loss": 0.7934,
+      "step": 11040
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6919413646848724e-05,
+      "loss": 0.8059,
+      "step": 11050
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.691643723491332e-05,
+      "loss": 0.8028,
+      "step": 11060
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6913460822977905e-05,
+      "loss": 0.8141,
+      "step": 11070
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.691048441104249e-05,
+      "loss": 0.8136,
+      "step": 11080
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.690750799910708e-05,
+      "loss": 0.8101,
+      "step": 11090
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.6904531587171666e-05,
+      "loss": 0.8141,
+      "step": 11100
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.690155517523625e-05,
+      "loss": 0.8001,
+      "step": 11110
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.6898578763300846e-05,
+      "loss": 0.8283,
+      "step": 11120
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.689560235136543e-05,
+      "loss": 0.7853,
+      "step": 11130
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.689262593943002e-05,
+      "loss": 0.8152,
+      "step": 11140
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.688964952749461e-05,
+      "loss": 0.8077,
+      "step": 11150
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.68866731155592e-05,
+      "loss": 0.7942,
+      "step": 11160
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.688369670362379e-05,
+      "loss": 0.7932,
+      "step": 11170
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.6880720291688374e-05,
+      "loss": 0.7932,
+      "step": 11180
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.687774387975296e-05,
+      "loss": 0.7895,
+      "step": 11190
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.687476746781755e-05,
+      "loss": 0.7933,
+      "step": 11200
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.6871791055882135e-05,
+      "loss": 0.7913,
+      "step": 11210
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.686881464394672e-05,
+      "loss": 0.8118,
+      "step": 11220
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.6865838232011316e-05,
+      "loss": 0.7873,
+      "step": 11230
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.68628618200759e-05,
+      "loss": 0.8194,
+      "step": 11240
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.685988540814049e-05,
+      "loss": 0.7974,
+      "step": 11250
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.6856908996205076e-05,
+      "loss": 0.8208,
+      "step": 11260
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.685393258426967e-05,
+      "loss": 0.7982,
+      "step": 11270
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.685095617233426e-05,
+      "loss": 0.8169,
+      "step": 11280
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.6847979760398844e-05,
+      "loss": 0.805,
+      "step": 11290
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.684500334846343e-05,
+      "loss": 0.8036,
+      "step": 11300
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.684202693652802e-05,
+      "loss": 0.8048,
+      "step": 11310
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.6839050524592604e-05,
+      "loss": 0.8022,
+      "step": 11320
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.683607411265719e-05,
+      "loss": 0.7992,
+      "step": 11330
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.6833097700721785e-05,
+      "loss": 0.7795,
+      "step": 11340
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.683012128878637e-05,
+      "loss": 0.7977,
+      "step": 11350
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.682714487685096e-05,
+      "loss": 0.7903,
+      "step": 11360
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.6824168464915546e-05,
+      "loss": 0.7944,
+      "step": 11370
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.682119205298014e-05,
+      "loss": 0.8031,
+      "step": 11380
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.6818215641044726e-05,
+      "loss": 0.7959,
+      "step": 11390
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.6815239229109306e-05,
+      "loss": 0.7921,
+      "step": 11400
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.68122628171739e-05,
+      "loss": 0.8035,
+      "step": 11410
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.680928640523849e-05,
+      "loss": 0.801,
+      "step": 11420
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.6806309993303074e-05,
+      "loss": 0.7948,
+      "step": 11430
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.680333358136766e-05,
+      "loss": 0.7868,
+      "step": 11440
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.6800357169432254e-05,
+      "loss": 0.7876,
+      "step": 11450
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.679738075749684e-05,
+      "loss": 0.7819,
+      "step": 11460
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.679440434556143e-05,
+      "loss": 0.8033,
+      "step": 11470
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.6791427933626015e-05,
+      "loss": 0.7938,
+      "step": 11480
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.678845152169061e-05,
+      "loss": 0.7997,
+      "step": 11490
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.678547510975519e-05,
+      "loss": 0.7924,
+      "step": 11500
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.678249869781978e-05,
+      "loss": 0.7972,
+      "step": 11510
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.677952228588437e-05,
+      "loss": 0.798,
+      "step": 11520
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.6776545873948956e-05,
+      "loss": 0.7714,
+      "step": 11530
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.677356946201354e-05,
+      "loss": 0.7895,
+      "step": 11540
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.677059305007814e-05,
+      "loss": 0.7986,
+      "step": 11550
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.6767616638142724e-05,
+      "loss": 0.7899,
+      "step": 11560
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.676464022620731e-05,
+      "loss": 0.7925,
+      "step": 11570
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.67616638142719e-05,
+      "loss": 0.7984,
+      "step": 11580
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.675868740233649e-05,
+      "loss": 0.7946,
+      "step": 11590
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.675571099040108e-05,
+      "loss": 0.801,
+      "step": 11600
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.675273457846566e-05,
+      "loss": 0.7926,
+      "step": 11610
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.674975816653025e-05,
+      "loss": 0.7875,
+      "step": 11620
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.674678175459484e-05,
+      "loss": 0.7815,
+      "step": 11630
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.6743805342659426e-05,
+      "loss": 0.7612,
+      "step": 11640
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.674082893072401e-05,
+      "loss": 0.7873,
+      "step": 11650
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.6737852518788606e-05,
+      "loss": 0.7843,
+      "step": 11660
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.673487610685319e-05,
+      "loss": 0.7812,
+      "step": 11670
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.673189969491778e-05,
+      "loss": 0.8103,
+      "step": 11680
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.672892328298237e-05,
+      "loss": 0.7969,
+      "step": 11690
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.672594687104696e-05,
+      "loss": 0.797,
+      "step": 11700
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.672297045911154e-05,
+      "loss": 0.7901,
+      "step": 11710
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.671999404717613e-05,
+      "loss": 0.7776,
+      "step": 11720
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.671701763524072e-05,
+      "loss": 0.782,
+      "step": 11730
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.671404122330531e-05,
+      "loss": 0.7839,
+      "step": 11740
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.6711064811369895e-05,
+      "loss": 0.778,
+      "step": 11750
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.670808839943448e-05,
+      "loss": 0.7909,
+      "step": 11760
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.6705111987499076e-05,
+      "loss": 0.7674,
+      "step": 11770
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.670213557556366e-05,
+      "loss": 0.8012,
+      "step": 11780
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.669915916362825e-05,
+      "loss": 0.7703,
+      "step": 11790
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.6696182751692837e-05,
+      "loss": 0.7827,
+      "step": 11800
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.6693206339757423e-05,
+      "loss": 0.7734,
+      "step": 11810
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.669022992782201e-05,
+      "loss": 0.7817,
+      "step": 11820
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.66872535158866e-05,
+      "loss": 0.7929,
+      "step": 11830
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.668427710395119e-05,
+      "loss": 0.7771,
+      "step": 11840
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.668130069201578e-05,
+      "loss": 0.7691,
+      "step": 11850
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.6678324280080365e-05,
+      "loss": 0.7667,
+      "step": 11860
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.667534786814495e-05,
+      "loss": 0.7777,
+      "step": 11870
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.6672371456209545e-05,
+      "loss": 0.7634,
+      "step": 11880
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.666939504427413e-05,
+      "loss": 0.7784,
+      "step": 11890
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.666641863233872e-05,
+      "loss": 0.7793,
+      "step": 11900
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.6663442220403306e-05,
+      "loss": 0.768,
+      "step": 11910
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.666046580846789e-05,
+      "loss": 0.7888,
+      "step": 11920
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.665748939653248e-05,
+      "loss": 0.789,
+      "step": 11930
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.6654512984597074e-05,
+      "loss": 0.7616,
+      "step": 11940
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.665153657266166e-05,
+      "loss": 0.7855,
+      "step": 11950
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.664856016072625e-05,
+      "loss": 0.7694,
+      "step": 11960
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.6645583748790834e-05,
+      "loss": 0.7769,
+      "step": 11970
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.664260733685543e-05,
+      "loss": 0.7877,
+      "step": 11980
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.6639630924920015e-05,
+      "loss": 0.7973,
+      "step": 11990
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.66366545129846e-05,
+      "loss": 0.7838,
+      "step": 12000
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.663367810104919e-05,
+      "loss": 0.7836,
+      "step": 12010
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.6630701689113775e-05,
+      "loss": 0.7739,
+      "step": 12020
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.662772527717836e-05,
+      "loss": 0.7933,
+      "step": 12030
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.662474886524295e-05,
+      "loss": 0.776,
+      "step": 12040
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.662177245330754e-05,
+      "loss": 0.7757,
+      "step": 12050
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.661879604137213e-05,
+      "loss": 0.7933,
+      "step": 12060
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.661581962943672e-05,
+      "loss": 0.7951,
+      "step": 12070
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.6612843217501304e-05,
+      "loss": 0.7771,
+      "step": 12080
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.66098668055659e-05,
+      "loss": 0.7852,
+      "step": 12090
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.6606890393630484e-05,
+      "loss": 0.7592,
+      "step": 12100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.660391398169507e-05,
+      "loss": 0.7868,
+      "step": 12110
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.660093756975966e-05,
+      "loss": 0.7769,
+      "step": 12120
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.6597961157824245e-05,
+      "loss": 0.7741,
+      "step": 12130
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.659498474588883e-05,
+      "loss": 0.7722,
+      "step": 12140
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.659200833395342e-05,
+      "loss": 0.7512,
+      "step": 12150
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.658903192201801e-05,
+      "loss": 0.7533,
+      "step": 12160
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.65860555100826e-05,
+      "loss": 0.7752,
+      "step": 12170
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.6583079098147186e-05,
+      "loss": 0.7623,
+      "step": 12180
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.658010268621177e-05,
+      "loss": 0.7648,
+      "step": 12190
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.657712627427637e-05,
+      "loss": 0.7886,
+      "step": 12200
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.6574149862340954e-05,
+      "loss": 0.7608,
+      "step": 12210
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.6571173450405534e-05,
+      "loss": 0.7551,
+      "step": 12220
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.656819703847013e-05,
+      "loss": 0.7625,
+      "step": 12230
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.6565220626534714e-05,
+      "loss": 0.7608,
+      "step": 12240
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.65622442145993e-05,
+      "loss": 0.7803,
+      "step": 12250
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.655926780266389e-05,
+      "loss": 0.7703,
+      "step": 12260
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.655629139072848e-05,
+      "loss": 0.7397,
+      "step": 12270
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.655331497879307e-05,
+      "loss": 0.7476,
+      "step": 12280
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.6550338566857656e-05,
+      "loss": 0.7734,
+      "step": 12290
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.654736215492224e-05,
+      "loss": 0.7791,
+      "step": 12300
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.6544385742986836e-05,
+      "loss": 0.7575,
+      "step": 12310
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6541409331051416e-05,
+      "loss": 0.7671,
+      "step": 12320
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.653843291911601e-05,
+      "loss": 0.7554,
+      "step": 12330
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.65354565071806e-05,
+      "loss": 0.7431,
+      "step": 12340
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.6532480095245184e-05,
+      "loss": 0.7779,
+      "step": 12350
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.652950368330977e-05,
+      "loss": 0.7576,
+      "step": 12360
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.6526527271374364e-05,
+      "loss": 0.7411,
+      "step": 12370
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.652355085943895e-05,
+      "loss": 0.7497,
+      "step": 12380
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.652057444750354e-05,
+      "loss": 0.7743,
+      "step": 12390
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.6517598035568125e-05,
+      "loss": 0.7598,
+      "step": 12400
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.651462162363272e-05,
+      "loss": 0.7647,
+      "step": 12410
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.65116452116973e-05,
+      "loss": 0.7623,
+      "step": 12420
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.6508668799761886e-05,
+      "loss": 0.7593,
+      "step": 12430
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.650569238782648e-05,
+      "loss": 0.7559,
+      "step": 12440
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.6502715975891066e-05,
+      "loss": 0.7627,
+      "step": 12450
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.649973956395565e-05,
+      "loss": 0.7549,
+      "step": 12460
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.649676315202024e-05,
+      "loss": 0.7548,
+      "step": 12470
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.6493786740084834e-05,
+      "loss": 0.7616,
+      "step": 12480
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.649081032814942e-05,
+      "loss": 0.7709,
+      "step": 12490
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.648783391621401e-05,
+      "loss": 0.7636,
+      "step": 12500
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.6484857504278595e-05,
+      "loss": 0.7649,
+      "step": 12510
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.648188109234319e-05,
+      "loss": 0.757,
+      "step": 12520
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.647890468040777e-05,
+      "loss": 0.7568,
+      "step": 12530
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.6475928268472355e-05,
+      "loss": 0.7446,
+      "step": 12540
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.647295185653695e-05,
+      "loss": 0.7633,
+      "step": 12550
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.6469975444601536e-05,
+      "loss": 0.7467,
+      "step": 12560
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.646699903266612e-05,
+      "loss": 0.7482,
+      "step": 12570
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.646402262073071e-05,
+      "loss": 0.7633,
+      "step": 12580
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.64610462087953e-05,
+      "loss": 0.7606,
+      "step": 12590
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.645806979685989e-05,
+      "loss": 0.7539,
+      "step": 12600
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.645509338492448e-05,
+      "loss": 0.7573,
+      "step": 12610
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.6452116972989064e-05,
+      "loss": 0.7447,
+      "step": 12620
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.644914056105365e-05,
+      "loss": 0.7432,
+      "step": 12630
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.644616414911824e-05,
+      "loss": 0.7511,
+      "step": 12640
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.6443187737182825e-05,
+      "loss": 0.7379,
+      "step": 12650
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.644021132524742e-05,
+      "loss": 0.7633,
+      "step": 12660
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.6437234913312005e-05,
+      "loss": 0.764,
+      "step": 12670
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.643425850137659e-05,
+      "loss": 0.7319,
+      "step": 12680
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.643128208944118e-05,
+      "loss": 0.7458,
+      "step": 12690
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.642830567750577e-05,
+      "loss": 0.7541,
+      "step": 12700
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.642532926557036e-05,
+      "loss": 0.7553,
+      "step": 12710
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.6422352853634946e-05,
+      "loss": 0.757,
+      "step": 12720
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.6419376441699533e-05,
+      "loss": 0.7498,
+      "step": 12730
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.641640002976412e-05,
+      "loss": 0.7544,
+      "step": 12740
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.641342361782871e-05,
+      "loss": 0.7535,
+      "step": 12750
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.64104472058933e-05,
+      "loss": 0.7599,
+      "step": 12760
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.640747079395789e-05,
+      "loss": 0.7624,
+      "step": 12770
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.6404494382022475e-05,
+      "loss": 0.7572,
+      "step": 12780
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.640151797008706e-05,
+      "loss": 0.7483,
+      "step": 12790
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.6398541558151655e-05,
+      "loss": 0.7502,
+      "step": 12800
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.639556514621624e-05,
+      "loss": 0.7249,
+      "step": 12810
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.639258873428083e-05,
+      "loss": 0.7338,
+      "step": 12820
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.6389612322345416e-05,
+      "loss": 0.7416,
+      "step": 12830
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.638663591041e-05,
+      "loss": 0.7636,
+      "step": 12840
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.638365949847459e-05,
+      "loss": 0.7545,
+      "step": 12850
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.638068308653918e-05,
+      "loss": 0.757,
+      "step": 12860
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.637770667460377e-05,
+      "loss": 0.7371,
+      "step": 12870
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.637473026266836e-05,
+      "loss": 0.7404,
+      "step": 12880
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.6371753850732944e-05,
+      "loss": 0.7495,
+      "step": 12890
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.636877743879753e-05,
+      "loss": 0.7382,
+      "step": 12900
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.6365801026862125e-05,
+      "loss": 0.7563,
+      "step": 12910
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.636282461492671e-05,
+      "loss": 0.7549,
+      "step": 12920
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.63598482029913e-05,
+      "loss": 0.7536,
+      "step": 12930
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.6356871791055885e-05,
+      "loss": 0.7303,
+      "step": 12940
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.635389537912047e-05,
+      "loss": 0.7526,
+      "step": 12950
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.635091896718506e-05,
+      "loss": 0.7433,
+      "step": 12960
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.6347942555249646e-05,
+      "loss": 0.729,
+      "step": 12970
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.634496614331424e-05,
+      "loss": 0.7503,
+      "step": 12980
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.634198973137883e-05,
+      "loss": 0.7457,
+      "step": 12990
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.6339013319443414e-05,
+      "loss": 0.7439,
+      "step": 13000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.6336036907508e-05,
+      "loss": 0.7546,
+      "step": 13010
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.6333060495572594e-05,
+      "loss": 0.7454,
+      "step": 13020
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.633008408363718e-05,
+      "loss": 0.7406,
+      "step": 13030
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.632710767170177e-05,
+      "loss": 0.7298,
+      "step": 13040
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.6324131259766355e-05,
+      "loss": 0.7303,
+      "step": 13050
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.632115484783094e-05,
+      "loss": 0.7542,
+      "step": 13060
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.631817843589553e-05,
+      "loss": 0.741,
+      "step": 13070
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.6315202023960116e-05,
+      "loss": 0.754,
+      "step": 13080
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.631222561202471e-05,
+      "loss": 0.7584,
+      "step": 13090
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.6309249200089296e-05,
+      "loss": 0.7389,
+      "step": 13100
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.630627278815388e-05,
+      "loss": 0.728,
+      "step": 13110
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.630329637621847e-05,
+      "loss": 0.7258,
+      "step": 13120
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6300319964283064e-05,
+      "loss": 0.7317,
+      "step": 13130
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6297343552347644e-05,
+      "loss": 0.7206,
+      "step": 13140
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.629436714041224e-05,
+      "loss": 0.7458,
+      "step": 13150
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6291390728476824e-05,
+      "loss": 0.7483,
+      "step": 13160
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.628841431654141e-05,
+      "loss": 0.7426,
+      "step": 13170
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.6285437904606e-05,
+      "loss": 0.7468,
+      "step": 13180
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.628246149267059e-05,
+      "loss": 0.7361,
+      "step": 13190
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.627948508073518e-05,
+      "loss": 0.7249,
+      "step": 13200
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.6276508668799766e-05,
+      "loss": 0.743,
+      "step": 13210
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.627353225686435e-05,
+      "loss": 0.7321,
+      "step": 13220
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.6270555844928946e-05,
+      "loss": 0.7295,
+      "step": 13230
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.6267579432993526e-05,
+      "loss": 0.7385,
+      "step": 13240
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.626460302105811e-05,
+      "loss": 0.7192,
+      "step": 13250
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.626162660912271e-05,
+      "loss": 0.7407,
+      "step": 13260
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.6258650197187294e-05,
+      "loss": 0.7251,
+      "step": 13270
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.625567378525188e-05,
+      "loss": 0.7378,
+      "step": 13280
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.625269737331647e-05,
+      "loss": 0.7407,
+      "step": 13290
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.624972096138106e-05,
+      "loss": 0.7175,
+      "step": 13300
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.624674454944565e-05,
+      "loss": 0.732,
+      "step": 13310
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.6243768137510235e-05,
+      "loss": 0.7404,
+      "step": 13320
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.624079172557482e-05,
+      "loss": 0.7427,
+      "step": 13330
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.6237815313639416e-05,
+      "loss": 0.715,
+      "step": 13340
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.6234838901703996e-05,
+      "loss": 0.7203,
+      "step": 13350
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.623186248976858e-05,
+      "loss": 0.7511,
+      "step": 13360
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.6228886077833176e-05,
+      "loss": 0.7298,
+      "step": 13370
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.622590966589776e-05,
+      "loss": 0.7224,
+      "step": 13380
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.622293325396235e-05,
+      "loss": 0.7222,
+      "step": 13390
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.621995684202694e-05,
+      "loss": 0.742,
+      "step": 13400
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.621698043009153e-05,
+      "loss": 0.7206,
+      "step": 13410
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.621400401815612e-05,
+      "loss": 0.7144,
+      "step": 13420
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.6211027606220704e-05,
+      "loss": 0.7168,
+      "step": 13430
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.620805119428529e-05,
+      "loss": 0.7442,
+      "step": 13440
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.620507478234988e-05,
+      "loss": 0.7263,
+      "step": 13450
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.6202098370414465e-05,
+      "loss": 0.7136,
+      "step": 13460
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.619912195847906e-05,
+      "loss": 0.7317,
+      "step": 13470
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.6196145546543646e-05,
+      "loss": 0.7159,
+      "step": 13480
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.619316913460823e-05,
+      "loss": 0.7217,
+      "step": 13490
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.619019272267282e-05,
+      "loss": 0.747,
+      "step": 13500
+    },
+    {
+      "epoch": 3.0,
+      "eval_cer": 4.632478632478633,
+      "eval_loss": 0.6618424654006958,
+      "eval_runtime": 1.2987,
+      "eval_samples_per_second": 7.7,
+      "eval_steps_per_second": 0.77,
+      "eval_wer": 0.6883116883116883,
+      "step": 13506
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.6187216310737406e-05,
+      "loss": 0.7374,
+      "step": 13510
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.6184239898802e-05,
+      "loss": 0.7292,
+      "step": 13520
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.618126348686659e-05,
+      "loss": 0.7153,
+      "step": 13530
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.6178287074931174e-05,
+      "loss": 0.7207,
+      "step": 13540
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.617531066299576e-05,
+      "loss": 0.7038,
+      "step": 13550
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.617233425106035e-05,
+      "loss": 0.7265,
+      "step": 13560
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.6169357839124935e-05,
+      "loss": 0.7151,
+      "step": 13570
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.616638142718953e-05,
+      "loss": 0.7313,
+      "step": 13580
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.6163405015254115e-05,
+      "loss": 0.7177,
+      "step": 13590
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.61604286033187e-05,
+      "loss": 0.7052,
+      "step": 13600
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.615745219138329e-05,
+      "loss": 0.715,
+      "step": 13610
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.615447577944788e-05,
+      "loss": 0.7208,
+      "step": 13620
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.615149936751247e-05,
+      "loss": 0.7075,
+      "step": 13630
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.6148522955577056e-05,
+      "loss": 0.724,
+      "step": 13640
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.614554654364164e-05,
+      "loss": 0.7243,
+      "step": 13650
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.614257013170623e-05,
+      "loss": 0.7231,
+      "step": 13660
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.613959371977082e-05,
+      "loss": 0.7079,
+      "step": 13670
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.6136617307835404e-05,
+      "loss": 0.7279,
+      "step": 13680
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.61336408959e-05,
+      "loss": 0.7097,
+      "step": 13690
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.6130664483964585e-05,
+      "loss": 0.7165,
+      "step": 13700
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.612768807202917e-05,
+      "loss": 0.7167,
+      "step": 13710
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.612471166009376e-05,
+      "loss": 0.7148,
+      "step": 13720
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.612173524815835e-05,
+      "loss": 0.7177,
+      "step": 13730
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.611875883622294e-05,
+      "loss": 0.6976,
+      "step": 13740
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.6115782424287526e-05,
+      "loss": 0.704,
+      "step": 13750
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.611280601235211e-05,
+      "loss": 0.728,
+      "step": 13760
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.61098296004167e-05,
+      "loss": 0.7042,
+      "step": 13770
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.6106853188481287e-05,
+      "loss": 0.7248,
+      "step": 13780
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.6103876776545873e-05,
+      "loss": 0.6996,
+      "step": 13790
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.610090036461047e-05,
+      "loss": 0.6952,
+      "step": 13800
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.6097923952675054e-05,
+      "loss": 0.7226,
+      "step": 13810
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.609494754073964e-05,
+      "loss": 0.7269,
+      "step": 13820
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.609197112880423e-05,
+      "loss": 0.6953,
+      "step": 13830
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.608899471686882e-05,
+      "loss": 0.6997,
+      "step": 13840
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.608601830493341e-05,
+      "loss": 0.7123,
+      "step": 13850
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.6083041892997995e-05,
+      "loss": 0.7104,
+      "step": 13860
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.608006548106258e-05,
+      "loss": 0.7204,
+      "step": 13870
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.607708906912717e-05,
+      "loss": 0.7154,
+      "step": 13880
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.6074112657191756e-05,
+      "loss": 0.6915,
+      "step": 13890
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.607113624525635e-05,
+      "loss": 0.7091,
+      "step": 13900
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.6068159833320937e-05,
+      "loss": 0.7074,
+      "step": 13910
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.6065183421385523e-05,
+      "loss": 0.7121,
+      "step": 13920
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.606220700945011e-05,
+      "loss": 0.696,
+      "step": 13930
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.60592305975147e-05,
+      "loss": 0.7069,
+      "step": 13940
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.605625418557929e-05,
+      "loss": 0.7037,
+      "step": 13950
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.605327777364387e-05,
+      "loss": 0.706,
+      "step": 13960
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.6050301361708465e-05,
+      "loss": 0.6966,
+      "step": 13970
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.604732494977305e-05,
+      "loss": 0.7001,
+      "step": 13980
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.604434853783764e-05,
+      "loss": 0.7084,
+      "step": 13990
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.6041372125902225e-05,
+      "loss": 0.7108,
+      "step": 14000
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.603839571396682e-05,
+      "loss": 0.6941,
+      "step": 14010
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.6035419302031406e-05,
+      "loss": 0.7097,
+      "step": 14020
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.603244289009599e-05,
+      "loss": 0.7191,
+      "step": 14030
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.602946647816058e-05,
+      "loss": 0.7069,
+      "step": 14040
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.6026490066225173e-05,
+      "loss": 0.7049,
+      "step": 14050
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.6023513654289754e-05,
+      "loss": 0.7223,
+      "step": 14060
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.602053724235434e-05,
+      "loss": 0.7037,
+      "step": 14070
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.6017560830418934e-05,
+      "loss": 0.703,
+      "step": 14080
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.601458441848352e-05,
+      "loss": 0.7073,
+      "step": 14090
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.601160800654811e-05,
+      "loss": 0.701,
+      "step": 14100
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.6008631594612695e-05,
+      "loss": 0.7064,
+      "step": 14110
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.600565518267729e-05,
+      "loss": 0.7026,
+      "step": 14120
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.6002678770741875e-05,
+      "loss": 0.7093,
+      "step": 14130
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.599970235880646e-05,
+      "loss": 0.7025,
+      "step": 14140
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.599672594687105e-05,
+      "loss": 0.6948,
+      "step": 14150
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.5993749534935636e-05,
+      "loss": 0.6936,
+      "step": 14160
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.599077312300022e-05,
+      "loss": 0.6834,
+      "step": 14170
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.598779671106481e-05,
+      "loss": 0.696,
+      "step": 14180
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.5984820299129404e-05,
+      "loss": 0.7043,
+      "step": 14190
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.598184388719399e-05,
+      "loss": 0.7079,
+      "step": 14200
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.597886747525858e-05,
+      "loss": 0.7077,
+      "step": 14210
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.5975891063323164e-05,
+      "loss": 0.6969,
+      "step": 14220
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.597291465138776e-05,
+      "loss": 0.7032,
+      "step": 14230
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.5969938239452345e-05,
+      "loss": 0.7116,
+      "step": 14240
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.596696182751693e-05,
+      "loss": 0.69,
+      "step": 14250
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 3.596398541558152e-05,
+      "loss": 0.7224,
+      "step": 14260
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 3.5961009003646106e-05,
+      "loss": 0.7065,
+      "step": 14270
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 3.595803259171069e-05,
+      "loss": 0.6929,
+      "step": 14280
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 3.5955056179775286e-05,
+      "loss": 0.6974,
+      "step": 14290
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.595207976783987e-05,
+      "loss": 0.6925,
+      "step": 14300
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.594910335590446e-05,
+      "loss": 0.7095,
+      "step": 14310
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.594612694396905e-05,
+      "loss": 0.6868,
+      "step": 14320
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.594315053203364e-05,
+      "loss": 0.7132,
+      "step": 14330
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.594017412009823e-05,
+      "loss": 0.7057,
+      "step": 14340
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 3.5937197708162814e-05,
+      "loss": 0.6972,
+      "step": 14350
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 3.59342212962274e-05,
+      "loss": 0.6996,
+      "step": 14360
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 3.593124488429199e-05,
+      "loss": 0.6888,
+      "step": 14370
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 3.5928268472356575e-05,
+      "loss": 0.7176,
+      "step": 14380
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.592529206042116e-05,
+      "loss": 0.6993,
+      "step": 14390
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.5922315648485756e-05,
+      "loss": 0.7111,
+      "step": 14400
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.591933923655034e-05,
+      "loss": 0.6974,
+      "step": 14410
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.591636282461493e-05,
+      "loss": 0.6944,
+      "step": 14420
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.5913386412679516e-05,
+      "loss": 0.6847,
+      "step": 14430
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.591041000074411e-05,
+      "loss": 0.6971,
+      "step": 14440
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.59074335888087e-05,
+      "loss": 0.6975,
+      "step": 14450
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.5904457176873284e-05,
+      "loss": 0.6997,
+      "step": 14460
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 3.590148076493787e-05,
+      "loss": 0.7066,
+      "step": 14470
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.589850435300246e-05,
+      "loss": 0.6849,
+      "step": 14480
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.5895527941067044e-05,
+      "loss": 0.6948,
+      "step": 14490
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.589255152913163e-05,
+      "loss": 0.6943,
+      "step": 14500
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.5889575117196225e-05,
+      "loss": 0.6845,
+      "step": 14510
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 3.588659870526081e-05,
+      "loss": 0.6862,
+      "step": 14520
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.58836222933254e-05,
+      "loss": 0.6948,
+      "step": 14530
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.5880645881389986e-05,
+      "loss": 0.7014,
+      "step": 14540
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.587766946945458e-05,
+      "loss": 0.6916,
+      "step": 14550
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 3.5874693057519166e-05,
+      "loss": 0.6838,
+      "step": 14560
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.5871716645583746e-05,
+      "loss": 0.6837,
+      "step": 14570
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.586874023364834e-05,
+      "loss": 0.6954,
+      "step": 14580
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.586576382171293e-05,
+      "loss": 0.6918,
+      "step": 14590
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.5862787409777514e-05,
+      "loss": 0.7183,
+      "step": 14600
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 3.58598109978421e-05,
+      "loss": 0.7236,
+      "step": 14610
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.5856834585906694e-05,
+      "loss": 0.7035,
+      "step": 14620
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.585385817397128e-05,
+      "loss": 0.6796,
+      "step": 14630
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.585088176203587e-05,
+      "loss": 0.6846,
+      "step": 14640
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 3.5847905350100455e-05,
+      "loss": 0.7049,
+      "step": 14650
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.584492893816505e-05,
+      "loss": 0.6983,
+      "step": 14660
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.5841952526229636e-05,
+      "loss": 0.6895,
+      "step": 14670
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.583897611429422e-05,
+      "loss": 0.7045,
+      "step": 14680
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.583599970235881e-05,
+      "loss": 0.6874,
+      "step": 14690
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 3.5833023290423396e-05,
+      "loss": 0.6793,
+      "step": 14700
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.583004687848798e-05,
+      "loss": 0.686,
+      "step": 14710
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.582707046655258e-05,
+      "loss": 0.7117,
+      "step": 14720
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.5824094054617164e-05,
+      "loss": 0.7033,
+      "step": 14730
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 3.582111764268175e-05,
+      "loss": 0.6759,
+      "step": 14740
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.581814123074634e-05,
+      "loss": 0.687,
+      "step": 14750
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.581516481881093e-05,
+      "loss": 0.6893,
+      "step": 14760
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.581218840687552e-05,
+      "loss": 0.6827,
+      "step": 14770
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.58092119949401e-05,
+      "loss": 0.7097,
+      "step": 14780
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 3.580623558300469e-05,
+      "loss": 0.6668,
+      "step": 14790
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.580325917106928e-05,
+      "loss": 0.7137,
+      "step": 14800
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5800282759133866e-05,
+      "loss": 0.7047,
+      "step": 14810
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.579730634719845e-05,
+      "loss": 0.6977,
+      "step": 14820
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 3.5794329935263046e-05,
+      "loss": 0.7102,
+      "step": 14830
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.579135352332763e-05,
+      "loss": 0.6922,
+      "step": 14840
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.578837711139222e-05,
+      "loss": 0.6757,
+      "step": 14850
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.578540069945681e-05,
+      "loss": 0.6793,
+      "step": 14860
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.57824242875214e-05,
+      "loss": 0.6819,
+      "step": 14870
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 3.577944787558598e-05,
+      "loss": 0.6991,
+      "step": 14880
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.577647146365057e-05,
+      "loss": 0.677,
+      "step": 14890
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.577349505171516e-05,
+      "loss": 0.674,
+      "step": 14900
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.577051863977975e-05,
+      "loss": 0.6865,
+      "step": 14910
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 3.5767542227844335e-05,
+      "loss": 0.6908,
+      "step": 14920
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.576456581590892e-05,
+      "loss": 0.6824,
+      "step": 14930
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.5761589403973516e-05,
+      "loss": 0.6722,
+      "step": 14940
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.57586129920381e-05,
+      "loss": 0.6913,
+      "step": 14950
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.575563658010269e-05,
+      "loss": 0.6802,
+      "step": 14960
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 3.5752660168167277e-05,
+      "loss": 0.6831,
+      "step": 14970
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.5749683756231863e-05,
+      "loss": 0.6784,
+      "step": 14980
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.574670734429645e-05,
+      "loss": 0.6737,
+      "step": 14990
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.574373093236104e-05,
+      "loss": 0.687,
+      "step": 15000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 3.574075452042563e-05,
+      "loss": 0.6856,
+      "step": 15010
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.573777810849022e-05,
+      "loss": 0.6709,
+      "step": 15020
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.5734801696554805e-05,
+      "loss": 0.6926,
+      "step": 15030
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.573182528461939e-05,
+      "loss": 0.6857,
+      "step": 15040
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.5728848872683985e-05,
+      "loss": 0.6821,
+      "step": 15050
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 3.572587246074857e-05,
+      "loss": 0.6828,
+      "step": 15060
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.572289604881316e-05,
+      "loss": 0.6948,
+      "step": 15070
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.5719919636877746e-05,
+      "loss": 0.6799,
+      "step": 15080
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.571694322494233e-05,
+      "loss": 0.6786,
+      "step": 15090
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 3.571396681300692e-05,
+      "loss": 0.6824,
+      "step": 15100
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.5710990401071513e-05,
+      "loss": 0.702,
+      "step": 15110
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.57080139891361e-05,
+      "loss": 0.6858,
+      "step": 15120
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.570503757720069e-05,
+      "loss": 0.6585,
+      "step": 15130
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.5702061165265274e-05,
+      "loss": 0.6743,
+      "step": 15140
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 3.569908475332987e-05,
+      "loss": 0.7009,
+      "step": 15150
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.5696108341394455e-05,
+      "loss": 0.6763,
+      "step": 15160
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.569313192945904e-05,
+      "loss": 0.6849,
+      "step": 15170
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.569015551752363e-05,
+      "loss": 0.6701,
+      "step": 15180
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 3.5687179105588215e-05,
+      "loss": 0.6876,
+      "step": 15190
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.56842026936528e-05,
+      "loss": 0.6893,
+      "step": 15200
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.568122628171739e-05,
+      "loss": 0.6815,
+      "step": 15210
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.567824986978198e-05,
+      "loss": 0.6734,
+      "step": 15220
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.567527345784657e-05,
+      "loss": 0.655,
+      "step": 15230
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 3.567229704591116e-05,
+      "loss": 0.6843,
+      "step": 15240
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.5669320633975744e-05,
+      "loss": 0.6959,
+      "step": 15250
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.566634422204034e-05,
+      "loss": 0.6811,
+      "step": 15260
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.5663367810104924e-05,
+      "loss": 0.6965,
+      "step": 15270
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 3.566039139816951e-05,
+      "loss": 0.7011,
+      "step": 15280
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.56574149862341e-05,
+      "loss": 0.671,
+      "step": 15290
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.5654438574298685e-05,
+      "loss": 0.6865,
+      "step": 15300
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.565146216236327e-05,
+      "loss": 0.7035,
+      "step": 15310
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.564848575042786e-05,
+      "loss": 0.6859,
+      "step": 15320
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 3.564550933849245e-05,
+      "loss": 0.6829,
+      "step": 15330
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 3.564253292655704e-05,
+      "loss": 0.6709,
+      "step": 15340
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 3.5639556514621626e-05,
+      "loss": 0.6828,
+      "step": 15350
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 3.563658010268621e-05,
+      "loss": 0.6686,
+      "step": 15360
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 3.563360369075081e-05,
+      "loss": 0.6644,
+      "step": 15370
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 3.5630627278815394e-05,
+      "loss": 0.6681,
+      "step": 15380
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 3.5627650866879974e-05,
+      "loss": 0.6768,
+      "step": 15390
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 3.562467445494457e-05,
+      "loss": 0.6668,
+      "step": 15400
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 3.5621698043009154e-05,
+      "loss": 0.6824,
+      "step": 15410
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 3.561872163107374e-05,
+      "loss": 0.6491,
+      "step": 15420
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 3.561574521913833e-05,
+      "loss": 0.6817,
+      "step": 15430
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 3.561276880720292e-05,
+      "loss": 0.6841,
+      "step": 15440
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 3.560979239526751e-05,
+      "loss": 0.6842,
+      "step": 15450
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 3.5606815983332096e-05,
+      "loss": 0.6876,
+      "step": 15460
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 3.560383957139668e-05,
+      "loss": 0.6825,
+      "step": 15470
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 3.5600863159461276e-05,
+      "loss": 0.6692,
+      "step": 15480
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 3.559788674752586e-05,
+      "loss": 0.6807,
+      "step": 15490
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 3.559491033559045e-05,
+      "loss": 0.6547,
+      "step": 15500
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 3.559193392365504e-05,
+      "loss": 0.6886,
+      "step": 15510
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 3.5588957511719624e-05,
+      "loss": 0.6624,
+      "step": 15520
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 3.558598109978421e-05,
+      "loss": 0.6786,
+      "step": 15530
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 3.5583004687848804e-05,
+      "loss": 0.6897,
+      "step": 15540
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 3.558002827591339e-05,
+      "loss": 0.6661,
+      "step": 15550
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 3.557705186397798e-05,
+      "loss": 0.674,
+      "step": 15560
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 3.5574075452042565e-05,
+      "loss": 0.6705,
+      "step": 15570
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 3.557109904010716e-05,
+      "loss": 0.6608,
+      "step": 15580
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 3.5568122628171746e-05,
+      "loss": 0.6651,
+      "step": 15590
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 3.5565146216236326e-05,
+      "loss": 0.6718,
+      "step": 15600
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 3.556216980430092e-05,
+      "loss": 0.6684,
+      "step": 15610
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 3.5559193392365506e-05,
+      "loss": 0.6933,
+      "step": 15620
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 3.555621698043009e-05,
+      "loss": 0.6685,
+      "step": 15630
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 3.555324056849468e-05,
+      "loss": 0.6854,
+      "step": 15640
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 3.5550264156559274e-05,
+      "loss": 0.6791,
+      "step": 15650
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 3.554728774462386e-05,
+      "loss": 0.6822,
+      "step": 15660
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 3.554431133268845e-05,
+      "loss": 0.6656,
+      "step": 15670
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 3.5541334920753034e-05,
+      "loss": 0.6807,
+      "step": 15680
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 3.553835850881763e-05,
+      "loss": 0.6618,
+      "step": 15690
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 3.553538209688221e-05,
+      "loss": 0.671,
+      "step": 15700
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 3.5532405684946795e-05,
+      "loss": 0.6857,
+      "step": 15710
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 3.552942927301139e-05,
+      "loss": 0.6949,
+      "step": 15720
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 3.5526452861075976e-05,
+      "loss": 0.6723,
+      "step": 15730
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 3.552347644914056e-05,
+      "loss": 0.666,
+      "step": 15740
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 3.552050003720515e-05,
+      "loss": 0.6689,
+      "step": 15750
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 3.551752362526974e-05,
+      "loss": 0.67,
+      "step": 15760
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 3.551454721333433e-05,
+      "loss": 0.6828,
+      "step": 15770
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 3.551157080139892e-05,
+      "loss": 0.6725,
+      "step": 15780
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 3.5508594389463504e-05,
+      "loss": 0.6613,
+      "step": 15790
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 3.550561797752809e-05,
+      "loss": 0.6648,
+      "step": 15800
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 3.550264156559268e-05,
+      "loss": 0.6517,
+      "step": 15810
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 3.5499665153657265e-05,
+      "loss": 0.6839,
+      "step": 15820
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 3.549668874172186e-05,
+      "loss": 0.6656,
+      "step": 15830
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 3.5493712329786445e-05,
+      "loss": 0.6791,
+      "step": 15840
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 3.549073591785103e-05,
+      "loss": 0.6622,
+      "step": 15850
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 3.548775950591562e-05,
+      "loss": 0.6541,
+      "step": 15860
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 3.548478309398021e-05,
+      "loss": 0.6713,
+      "step": 15870
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 3.54818066820448e-05,
+      "loss": 0.6705,
+      "step": 15880
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 3.5478830270109386e-05,
+      "loss": 0.684,
+      "step": 15890
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 3.547585385817397e-05,
+      "loss": 0.668,
+      "step": 15900
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 3.547287744623856e-05,
+      "loss": 0.6643,
+      "step": 15910
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 3.546990103430315e-05,
+      "loss": 0.6756,
+      "step": 15920
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 3.546692462236774e-05,
+      "loss": 0.6662,
+      "step": 15930
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 3.546394821043233e-05,
+      "loss": 0.6502,
+      "step": 15940
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 3.5460971798496915e-05,
+      "loss": 0.6677,
+      "step": 15950
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 3.54579953865615e-05,
+      "loss": 0.6725,
+      "step": 15960
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 3.5455018974626095e-05,
+      "loss": 0.6689,
+      "step": 15970
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 3.545204256269068e-05,
+      "loss": 0.6772,
+      "step": 15980
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 3.544906615075527e-05,
+      "loss": 0.6777,
+      "step": 15990
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 3.5446089738819856e-05,
+      "loss": 0.6918,
+      "step": 16000
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 3.544311332688444e-05,
+      "loss": 0.6658,
+      "step": 16010
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 3.544013691494903e-05,
+      "loss": 0.6734,
+      "step": 16020
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 3.5437160503013617e-05,
+      "loss": 0.6607,
+      "step": 16030
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 3.543418409107821e-05,
+      "loss": 0.6649,
+      "step": 16040
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 3.54312076791428e-05,
+      "loss": 0.6605,
+      "step": 16050
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 3.5428231267207384e-05,
+      "loss": 0.6437,
+      "step": 16060
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 3.542525485527197e-05,
+      "loss": 0.6778,
+      "step": 16070
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 3.5422278443336565e-05,
+      "loss": 0.6724,
+      "step": 16080
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 3.541930203140115e-05,
+      "loss": 0.667,
+      "step": 16090
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 3.541632561946574e-05,
+      "loss": 0.6674,
+      "step": 16100
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 3.5413349207530325e-05,
+      "loss": 0.6585,
+      "step": 16110
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 3.541037279559491e-05,
+      "loss": 0.6721,
+      "step": 16120
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 3.54073963836595e-05,
+      "loss": 0.666,
+      "step": 16130
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 3.5404419971724086e-05,
+      "loss": 0.6571,
+      "step": 16140
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 3.540144355978868e-05,
+      "loss": 0.668,
+      "step": 16150
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 3.5398467147853267e-05,
+      "loss": 0.6523,
+      "step": 16160
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 3.5395490735917854e-05,
+      "loss": 0.6696,
+      "step": 16170
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 3.539251432398244e-05,
+      "loss": 0.6592,
+      "step": 16180
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 3.5389537912047034e-05,
+      "loss": 0.6774,
+      "step": 16190
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 3.538656150011162e-05,
+      "loss": 0.6636,
+      "step": 16200
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 3.53835850881762e-05,
+      "loss": 0.6389,
+      "step": 16210
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 3.5380608676240795e-05,
+      "loss": 0.6529,
+      "step": 16220
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 3.537763226430538e-05,
+      "loss": 0.6666,
+      "step": 16230
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 3.537465585236997e-05,
+      "loss": 0.6593,
+      "step": 16240
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 3.5371679440434555e-05,
+      "loss": 0.6566,
+      "step": 16250
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 3.536870302849915e-05,
+      "loss": 0.6597,
+      "step": 16260
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 3.5365726616563736e-05,
+      "loss": 0.6631,
+      "step": 16270
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 3.536275020462832e-05,
+      "loss": 0.6685,
+      "step": 16280
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 3.535977379269291e-05,
+      "loss": 0.6746,
+      "step": 16290
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 3.5356797380757504e-05,
+      "loss": 0.657,
+      "step": 16300
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 3.5353820968822084e-05,
+      "loss": 0.6735,
+      "step": 16310
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 3.535084455688668e-05,
+      "loss": 0.6674,
+      "step": 16320
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 3.5347868144951264e-05,
+      "loss": 0.652,
+      "step": 16330
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 3.534489173301585e-05,
+      "loss": 0.6571,
+      "step": 16340
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 3.534191532108044e-05,
+      "loss": 0.6553,
+      "step": 16350
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 3.533893890914503e-05,
+      "loss": 0.6527,
+      "step": 16360
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 3.533596249720962e-05,
+      "loss": 0.637,
+      "step": 16370
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 3.5332986085274205e-05,
+      "loss": 0.6588,
+      "step": 16380
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 3.533000967333879e-05,
+      "loss": 0.6688,
+      "step": 16390
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 3.5327033261403386e-05,
+      "loss": 0.6489,
+      "step": 16400
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 3.532405684946797e-05,
+      "loss": 0.6641,
+      "step": 16410
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 3.532108043753255e-05,
+      "loss": 0.6663,
+      "step": 16420
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 3.531810402559715e-05,
+      "loss": 0.6579,
+      "step": 16430
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 3.5315127613661734e-05,
+      "loss": 0.6526,
+      "step": 16440
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 3.531215120172632e-05,
+      "loss": 0.658,
+      "step": 16450
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 3.530917478979091e-05,
+      "loss": 0.6457,
+      "step": 16460
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 3.53061983778555e-05,
+      "loss": 0.6486,
+      "step": 16470
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 3.530322196592009e-05,
+      "loss": 0.6649,
+      "step": 16480
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 3.5300245553984675e-05,
+      "loss": 0.6427,
+      "step": 16490
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 3.529726914204926e-05,
+      "loss": 0.6523,
+      "step": 16500
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 3.5294292730113855e-05,
+      "loss": 0.6558,
+      "step": 16510
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 3.5291316318178436e-05,
+      "loss": 0.6475,
+      "step": 16520
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 3.528833990624302e-05,
+      "loss": 0.6647,
+      "step": 16530
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 3.5285363494307616e-05,
+      "loss": 0.661,
+      "step": 16540
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 3.52823870823722e-05,
+      "loss": 0.6417,
+      "step": 16550
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 3.527941067043679e-05,
+      "loss": 0.6668,
+      "step": 16560
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 3.527643425850138e-05,
+      "loss": 0.6628,
+      "step": 16570
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 3.527345784656597e-05,
+      "loss": 0.6599,
+      "step": 16580
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 3.527048143463056e-05,
+      "loss": 0.6557,
+      "step": 16590
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 3.5267505022695144e-05,
+      "loss": 0.658,
+      "step": 16600
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 3.526452861075973e-05,
+      "loss": 0.6593,
+      "step": 16610
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 3.526155219882432e-05,
+      "loss": 0.6507,
+      "step": 16620
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 3.5258575786888905e-05,
+      "loss": 0.6501,
+      "step": 16630
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 3.525559937495349e-05,
+      "loss": 0.6675,
+      "step": 16640
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 3.5252622963018086e-05,
+      "loss": 0.6603,
+      "step": 16650
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 3.524964655108267e-05,
+      "loss": 0.64,
+      "step": 16660
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 3.524667013914726e-05,
+      "loss": 0.6559,
+      "step": 16670
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 3.5243693727211846e-05,
+      "loss": 0.6404,
+      "step": 16680
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 3.524071731527644e-05,
+      "loss": 0.6562,
+      "step": 16690
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 3.523774090334103e-05,
+      "loss": 0.656,
+      "step": 16700
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 3.5234764491405614e-05,
+      "loss": 0.6461,
+      "step": 16710
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 3.52317880794702e-05,
+      "loss": 0.6541,
+      "step": 16720
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 3.522881166753479e-05,
+      "loss": 0.6542,
+      "step": 16730
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 3.5225835255599375e-05,
+      "loss": 0.6431,
+      "step": 16740
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 3.522285884366397e-05,
+      "loss": 0.6554,
+      "step": 16750
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 3.5219882431728555e-05,
+      "loss": 0.6499,
+      "step": 16760
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 3.521690601979314e-05,
+      "loss": 0.6585,
+      "step": 16770
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 3.521392960785773e-05,
+      "loss": 0.6504,
+      "step": 16780
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 3.521095319592232e-05,
+      "loss": 0.6552,
+      "step": 16790
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 3.520797678398691e-05,
+      "loss": 0.6562,
+      "step": 16800
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 3.5205000372051496e-05,
+      "loss": 0.6698,
+      "step": 16810
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 3.520202396011608e-05,
+      "loss": 0.655,
+      "step": 16820
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 3.519904754818067e-05,
+      "loss": 0.6716,
+      "step": 16830
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 3.519607113624526e-05,
+      "loss": 0.6506,
+      "step": 16840
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 3.5193094724309844e-05,
+      "loss": 0.6519,
+      "step": 16850
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 3.519011831237444e-05,
+      "loss": 0.6487,
+      "step": 16860
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 3.5187141900439025e-05,
+      "loss": 0.6443,
+      "step": 16870
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 3.518416548850361e-05,
+      "loss": 0.6544,
+      "step": 16880
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 3.51811890765682e-05,
+      "loss": 0.6542,
+      "step": 16890
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 3.517821266463279e-05,
+      "loss": 0.649,
+      "step": 16900
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 3.517523625269738e-05,
+      "loss": 0.6604,
+      "step": 16910
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 3.5172259840761966e-05,
+      "loss": 0.6341,
+      "step": 16920
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 3.516928342882655e-05,
+      "loss": 0.6606,
+      "step": 16930
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 3.516630701689114e-05,
+      "loss": 0.6571,
+      "step": 16940
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 3.5163330604955726e-05,
+      "loss": 0.6524,
+      "step": 16950
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 3.516035419302031e-05,
+      "loss": 0.6477,
+      "step": 16960
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 3.515737778108491e-05,
+      "loss": 0.6452,
+      "step": 16970
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 3.5154401369149494e-05,
+      "loss": 0.6382,
+      "step": 16980
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 3.515142495721408e-05,
+      "loss": 0.6644,
+      "step": 16990
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 3.514844854527867e-05,
+      "loss": 0.6597,
+      "step": 17000
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 3.514547213334326e-05,
+      "loss": 0.637,
+      "step": 17010
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 3.514249572140785e-05,
+      "loss": 0.6529,
+      "step": 17020
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 3.5139519309472435e-05,
+      "loss": 0.6269,
+      "step": 17030
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 3.513654289753702e-05,
+      "loss": 0.647,
+      "step": 17040
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 3.513356648560161e-05,
+      "loss": 0.6409,
+      "step": 17050
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 3.5130590073666196e-05,
+      "loss": 0.6362,
+      "step": 17060
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 3.512761366173078e-05,
+      "loss": 0.6597,
+      "step": 17070
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 3.5124637249795376e-05,
+      "loss": 0.648,
+      "step": 17080
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.5121660837859963e-05,
+      "loss": 0.6436,
+      "step": 17090
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.511868442592455e-05,
+      "loss": 0.6577,
+      "step": 17100
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.511570801398914e-05,
+      "loss": 0.633,
+      "step": 17110
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.511273160205373e-05,
+      "loss": 0.6379,
+      "step": 17120
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.510975519011831e-05,
+      "loss": 0.6573,
+      "step": 17130
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 3.5106778778182905e-05,
+      "loss": 0.6388,
+      "step": 17140
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 3.510380236624749e-05,
+      "loss": 0.6497,
+      "step": 17150
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 3.510082595431208e-05,
+      "loss": 0.645,
+      "step": 17160
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 3.5097849542376665e-05,
+      "loss": 0.635,
+      "step": 17170
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.509487313044126e-05,
+      "loss": 0.6461,
+      "step": 17180
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.5091896718505846e-05,
+      "loss": 0.6511,
+      "step": 17190
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.508892030657043e-05,
+      "loss": 0.659,
+      "step": 17200
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.508594389463502e-05,
+      "loss": 0.6342,
+      "step": 17210
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.5082967482699613e-05,
+      "loss": 0.6484,
+      "step": 17220
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 3.50799910707642e-05,
+      "loss": 0.6494,
+      "step": 17230
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 3.507701465882878e-05,
+      "loss": 0.6495,
+      "step": 17240
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 3.5074038246893374e-05,
+      "loss": 0.6558,
+      "step": 17250
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 3.507106183495796e-05,
+      "loss": 0.643,
+      "step": 17260
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 3.506808542302255e-05,
+      "loss": 0.6475,
+      "step": 17270
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 3.5065109011087135e-05,
+      "loss": 0.6355,
+      "step": 17280
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 3.506213259915173e-05,
+      "loss": 0.6438,
+      "step": 17290
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 3.5059156187216315e-05,
+      "loss": 0.6511,
+      "step": 17300
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 3.50561797752809e-05,
+      "loss": 0.6395,
+      "step": 17310
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 3.505320336334549e-05,
+      "loss": 0.6521,
+      "step": 17320
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 3.505022695141008e-05,
+      "loss": 0.6555,
+      "step": 17330
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 3.504725053947466e-05,
+      "loss": 0.656,
+      "step": 17340
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 3.504427412753925e-05,
+      "loss": 0.659,
+      "step": 17350
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 3.5041297715603844e-05,
+      "loss": 0.6448,
+      "step": 17360
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 3.503832130366843e-05,
+      "loss": 0.6678,
+      "step": 17370
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 3.503534489173302e-05,
+      "loss": 0.6442,
+      "step": 17380
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 3.5032368479797604e-05,
+      "loss": 0.6358,
+      "step": 17390
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 3.50293920678622e-05,
+      "loss": 0.6373,
+      "step": 17400
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 3.5026415655926785e-05,
+      "loss": 0.6251,
+      "step": 17410
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 3.502343924399137e-05,
+      "loss": 0.6416,
+      "step": 17420
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 3.502046283205596e-05,
+      "loss": 0.6277,
+      "step": 17430
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 3.5017486420120546e-05,
+      "loss": 0.6471,
+      "step": 17440
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 3.501451000818513e-05,
+      "loss": 0.641,
+      "step": 17450
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 3.501153359624972e-05,
+      "loss": 0.6537,
+      "step": 17460
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 3.500855718431431e-05,
+      "loss": 0.6494,
+      "step": 17470
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 3.50055807723789e-05,
+      "loss": 0.6344,
+      "step": 17480
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 3.500260436044349e-05,
+      "loss": 0.6401,
+      "step": 17490
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 3.4999627948508074e-05,
+      "loss": 0.6531,
+      "step": 17500
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 3.499665153657267e-05,
+      "loss": 0.6373,
+      "step": 17510
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 3.4993675124637254e-05,
+      "loss": 0.6408,
+      "step": 17520
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 3.499069871270184e-05,
+      "loss": 0.6235,
+      "step": 17530
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 3.498772230076643e-05,
+      "loss": 0.6425,
+      "step": 17540
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 3.4984745888831015e-05,
+      "loss": 0.6324,
+      "step": 17550
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 3.49817694768956e-05,
+      "loss": 0.6293,
+      "step": 17560
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 3.4978793064960196e-05,
+      "loss": 0.6446,
+      "step": 17570
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 3.497581665302478e-05,
+      "loss": 0.6505,
+      "step": 17580
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 3.497284024108937e-05,
+      "loss": 0.6311,
+      "step": 17590
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 3.4969863829153956e-05,
+      "loss": 0.6448,
+      "step": 17600
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 3.496688741721855e-05,
+      "loss": 0.6441,
+      "step": 17610
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 3.496391100528314e-05,
+      "loss": 0.652,
+      "step": 17620
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 3.4960934593347724e-05,
+      "loss": 0.627,
+      "step": 17630
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 3.495795818141231e-05,
+      "loss": 0.631,
+      "step": 17640
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 3.49549817694769e-05,
+      "loss": 0.6348,
+      "step": 17650
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 3.4952005357541484e-05,
+      "loss": 0.6417,
+      "step": 17660
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 3.494902894560607e-05,
+      "loss": 0.6262,
+      "step": 17670
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 3.4946052533670665e-05,
+      "loss": 0.6365,
+      "step": 17680
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 3.494307612173525e-05,
+      "loss": 0.6105,
+      "step": 17690
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 3.494009970979984e-05,
+      "loss": 0.6467,
+      "step": 17700
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 3.4937123297864426e-05,
+      "loss": 0.6342,
+      "step": 17710
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 3.493414688592902e-05,
+      "loss": 0.6359,
+      "step": 17720
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 3.4931170473993606e-05,
+      "loss": 0.6289,
+      "step": 17730
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 3.492819406205819e-05,
+      "loss": 0.6446,
+      "step": 17740
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 3.492521765012278e-05,
+      "loss": 0.6279,
+      "step": 17750
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 3.492224123818737e-05,
+      "loss": 0.6483,
+      "step": 17760
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 3.4919264826251954e-05,
+      "loss": 0.6093,
+      "step": 17770
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 3.491628841431654e-05,
+      "loss": 0.6389,
+      "step": 17780
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 3.4913312002381134e-05,
+      "loss": 0.6413,
+      "step": 17790
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 3.491033559044572e-05,
+      "loss": 0.6579,
+      "step": 17800
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 3.490735917851031e-05,
+      "loss": 0.62,
+      "step": 17810
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 3.4904382766574895e-05,
+      "loss": 0.636,
+      "step": 17820
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 3.490140635463949e-05,
+      "loss": 0.6474,
+      "step": 17830
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 3.4898429942704076e-05,
+      "loss": 0.6331,
+      "step": 17840
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 3.489545353076866e-05,
+      "loss": 0.6427,
+      "step": 17850
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 3.489247711883325e-05,
+      "loss": 0.6322,
+      "step": 17860
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 3.4889500706897836e-05,
+      "loss": 0.6339,
+      "step": 17870
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 3.488652429496242e-05,
+      "loss": 0.6154,
+      "step": 17880
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 3.488354788302701e-05,
+      "loss": 0.6371,
+      "step": 17890
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 3.4880571471091604e-05,
+      "loss": 0.6106,
+      "step": 17900
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 3.487759505915619e-05,
+      "loss": 0.6276,
+      "step": 17910
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 3.487461864722078e-05,
+      "loss": 0.634,
+      "step": 17920
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 3.4871642235285365e-05,
+      "loss": 0.6195,
+      "step": 17930
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 3.486866582334996e-05,
+      "loss": 0.6404,
+      "step": 17940
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 3.486568941141454e-05,
+      "loss": 0.642,
+      "step": 17950
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 3.486271299947913e-05,
+      "loss": 0.6403,
+      "step": 17960
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 3.485973658754372e-05,
+      "loss": 0.6357,
+      "step": 17970
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 3.4856760175608306e-05,
+      "loss": 0.6318,
+      "step": 17980
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.485378376367289e-05,
+      "loss": 0.6502,
+      "step": 17990
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.4850807351737486e-05,
+      "loss": 0.6197,
+      "step": 18000
+    },
+    {
+      "epoch": 4.0,
+      "eval_cer": 4.608974358974359,
+      "eval_loss": 0.5299192667007446,
+      "eval_runtime": 1.3141,
+      "eval_samples_per_second": 7.61,
+      "eval_steps_per_second": 0.761,
+      "eval_wer": 0.6883116883116883,
+      "step": 18008
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.484783093980207e-05,
+      "loss": 0.6455,
+      "step": 18010
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.484485452786666e-05,
+      "loss": 0.6097,
+      "step": 18020
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 3.484187811593125e-05,
+      "loss": 0.6018,
+      "step": 18030
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 3.483890170399584e-05,
+      "loss": 0.6084,
+      "step": 18040
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 3.483592529206042e-05,
+      "loss": 0.6076,
+      "step": 18050
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 3.483294888012501e-05,
+      "loss": 0.6289,
+      "step": 18060
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 3.48299724681896e-05,
+      "loss": 0.624,
+      "step": 18070
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 3.482699605625419e-05,
+      "loss": 0.6306,
+      "step": 18080
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 3.4824019644318775e-05,
+      "loss": 0.6199,
+      "step": 18090
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 3.482104323238336e-05,
+      "loss": 0.6248,
+      "step": 18100
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 3.4818066820447956e-05,
+      "loss": 0.62,
+      "step": 18110
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 3.481509040851254e-05,
+      "loss": 0.621,
+      "step": 18120
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 3.481211399657713e-05,
+      "loss": 0.6236,
+      "step": 18130
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 3.4809137584641717e-05,
+      "loss": 0.6122,
+      "step": 18140
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 3.480616117270631e-05,
+      "loss": 0.6059,
+      "step": 18150
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 3.480318476077089e-05,
+      "loss": 0.6129,
+      "step": 18160
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.480020834883548e-05,
+      "loss": 0.6135,
+      "step": 18170
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.479723193690007e-05,
+      "loss": 0.6002,
+      "step": 18180
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.479425552496466e-05,
+      "loss": 0.6053,
+      "step": 18190
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.4791279113029245e-05,
+      "loss": 0.6081,
+      "step": 18200
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.478830270109383e-05,
+      "loss": 0.6304,
+      "step": 18210
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.4785326289158425e-05,
+      "loss": 0.6016,
+      "step": 18220
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.478234987722301e-05,
+      "loss": 0.6238,
+      "step": 18230
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.47793734652876e-05,
+      "loss": 0.6133,
+      "step": 18240
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.4776397053352186e-05,
+      "loss": 0.6112,
+      "step": 18250
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.477342064141677e-05,
+      "loss": 0.6008,
+      "step": 18260
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.477044422948136e-05,
+      "loss": 0.6304,
+      "step": 18270
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.4767467817545953e-05,
+      "loss": 0.6029,
+      "step": 18280
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.476449140561054e-05,
+      "loss": 0.6159,
+      "step": 18290
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.476151499367513e-05,
+      "loss": 0.6116,
+      "step": 18300
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.4758538581739714e-05,
+      "loss": 0.6432,
+      "step": 18310
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.47555621698043e-05,
+      "loss": 0.622,
+      "step": 18320
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.4752585757868895e-05,
+      "loss": 0.6254,
+      "step": 18330
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.474960934593348e-05,
+      "loss": 0.599,
+      "step": 18340
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.474663293399807e-05,
+      "loss": 0.6154,
+      "step": 18350
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.4743656522062655e-05,
+      "loss": 0.6216,
+      "step": 18360
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.474068011012724e-05,
+      "loss": 0.6373,
+      "step": 18370
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.473770369819183e-05,
+      "loss": 0.6059,
+      "step": 18380
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.473472728625642e-05,
+      "loss": 0.6138,
+      "step": 18390
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.473175087432101e-05,
+      "loss": 0.6062,
+      "step": 18400
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.47287744623856e-05,
+      "loss": 0.6117,
+      "step": 18410
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.4725798050450184e-05,
+      "loss": 0.6196,
+      "step": 18420
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.472282163851478e-05,
+      "loss": 0.6065,
+      "step": 18430
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.4719845226579364e-05,
+      "loss": 0.6044,
+      "step": 18440
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.471686881464395e-05,
+      "loss": 0.6296,
+      "step": 18450
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.471389240270854e-05,
+      "loss": 0.6245,
+      "step": 18460
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.4710915990773125e-05,
+      "loss": 0.6191,
+      "step": 18470
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.470793957883771e-05,
+      "loss": 0.6156,
+      "step": 18480
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.47049631669023e-05,
+      "loss": 0.6191,
+      "step": 18490
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.470198675496689e-05,
+      "loss": 0.6026,
+      "step": 18500
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.469901034303148e-05,
+      "loss": 0.6326,
+      "step": 18510
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.4696033931096066e-05,
+      "loss": 0.5977,
+      "step": 18520
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.469305751916065e-05,
+      "loss": 0.6433,
+      "step": 18530
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.469008110722525e-05,
+      "loss": 0.595,
+      "step": 18540
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.4687104695289834e-05,
+      "loss": 0.6028,
+      "step": 18550
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.468412828335442e-05,
+      "loss": 0.6148,
+      "step": 18560
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.468115187141901e-05,
+      "loss": 0.6375,
+      "step": 18570
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.4678175459483594e-05,
+      "loss": 0.6285,
+      "step": 18580
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.467519904754818e-05,
+      "loss": 0.6118,
+      "step": 18590
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.467222263561277e-05,
+      "loss": 0.6245,
+      "step": 18600
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.466924622367736e-05,
+      "loss": 0.6173,
+      "step": 18610
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.466626981174195e-05,
+      "loss": 0.6064,
+      "step": 18620
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.4663293399806536e-05,
+      "loss": 0.6023,
+      "step": 18630
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.466031698787112e-05,
+      "loss": 0.618,
+      "step": 18640
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.4657340575935716e-05,
+      "loss": 0.6209,
+      "step": 18650
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.46543641640003e-05,
+      "loss": 0.6016,
+      "step": 18660
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.465138775206489e-05,
+      "loss": 0.6144,
+      "step": 18670
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.464841134012948e-05,
+      "loss": 0.6222,
+      "step": 18680
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.4645434928194064e-05,
+      "loss": 0.617,
+      "step": 18690
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.464245851625865e-05,
+      "loss": 0.6256,
+      "step": 18700
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.4639482104323244e-05,
+      "loss": 0.6356,
+      "step": 18710
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.463650569238783e-05,
+      "loss": 0.6142,
+      "step": 18720
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.463352928045242e-05,
+      "loss": 0.604,
+      "step": 18730
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.4630552868517005e-05,
+      "loss": 0.6198,
+      "step": 18740
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.462757645658159e-05,
+      "loss": 0.6231,
+      "step": 18750
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.4624600044646186e-05,
+      "loss": 0.6047,
+      "step": 18760
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.4621623632710766e-05,
+      "loss": 0.6294,
+      "step": 18770
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.461864722077536e-05,
+      "loss": 0.6256,
+      "step": 18780
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.4615670808839946e-05,
+      "loss": 0.5932,
+      "step": 18790
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.461269439690453e-05,
+      "loss": 0.6095,
+      "step": 18800
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.460971798496912e-05,
+      "loss": 0.6274,
+      "step": 18810
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.4606741573033714e-05,
+      "loss": 0.6327,
+      "step": 18820
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.46037651610983e-05,
+      "loss": 0.6134,
+      "step": 18830
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.460078874916289e-05,
+      "loss": 0.6081,
+      "step": 18840
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.4597812337227474e-05,
+      "loss": 0.6097,
+      "step": 18850
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.459483592529207e-05,
+      "loss": 0.6157,
+      "step": 18860
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.459185951335665e-05,
+      "loss": 0.6081,
+      "step": 18870
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.4588883101421235e-05,
+      "loss": 0.6054,
+      "step": 18880
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.458590668948583e-05,
+      "loss": 0.616,
+      "step": 18890
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.4582930277550416e-05,
+      "loss": 0.6113,
+      "step": 18900
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.4579953865615e-05,
+      "loss": 0.6226,
+      "step": 18910
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.457697745367959e-05,
+      "loss": 0.637,
+      "step": 18920
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.457400104174418e-05,
+      "loss": 0.5885,
+      "step": 18930
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.457102462980877e-05,
+      "loss": 0.5999,
+      "step": 18940
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.456804821787336e-05,
+      "loss": 0.6163,
+      "step": 18950
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.4565071805937944e-05,
+      "loss": 0.597,
+      "step": 18960
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.456209539400254e-05,
+      "loss": 0.6118,
+      "step": 18970
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.455911898206712e-05,
+      "loss": 0.5954,
+      "step": 18980
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.4556142570131705e-05,
+      "loss": 0.6168,
+      "step": 18990
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.45531661581963e-05,
+      "loss": 0.6181,
+      "step": 19000
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.4550189746260885e-05,
+      "loss": 0.6169,
+      "step": 19010
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.454721333432547e-05,
+      "loss": 0.6227,
+      "step": 19020
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.454423692239006e-05,
+      "loss": 0.6246,
+      "step": 19030
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.454126051045465e-05,
+      "loss": 0.6227,
+      "step": 19040
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.453828409851924e-05,
+      "loss": 0.6094,
+      "step": 19050
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.4535307686583826e-05,
+      "loss": 0.5929,
+      "step": 19060
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.453233127464841e-05,
+      "loss": 0.606,
+      "step": 19070
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.4529354862713e-05,
+      "loss": 0.6056,
+      "step": 19080
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.452637845077759e-05,
+      "loss": 0.6011,
+      "step": 19090
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.452340203884218e-05,
+      "loss": 0.6016,
+      "step": 19100
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.452042562690677e-05,
+      "loss": 0.6077,
+      "step": 19110
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.4517449214971355e-05,
+      "loss": 0.5837,
+      "step": 19120
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.451447280303594e-05,
+      "loss": 0.6027,
+      "step": 19130
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.4511496391100535e-05,
+      "loss": 0.6108,
+      "step": 19140
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.450851997916512e-05,
+      "loss": 0.6179,
+      "step": 19150
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.450554356722971e-05,
+      "loss": 0.6083,
+      "step": 19160
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.4502567155294296e-05,
+      "loss": 0.6123,
+      "step": 19170
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.449959074335888e-05,
+      "loss": 0.6137,
+      "step": 19180
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.449661433142347e-05,
+      "loss": 0.6112,
+      "step": 19190
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.4493637919488057e-05,
+      "loss": 0.6281,
+      "step": 19200
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.449066150755265e-05,
+      "loss": 0.5944,
+      "step": 19210
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.448768509561724e-05,
+      "loss": 0.6028,
+      "step": 19220
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.4484708683681824e-05,
+      "loss": 0.6085,
+      "step": 19230
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.448173227174641e-05,
+      "loss": 0.6002,
+      "step": 19240
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.4478755859811005e-05,
+      "loss": 0.6008,
+      "step": 19250
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.447577944787559e-05,
+      "loss": 0.5906,
+      "step": 19260
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.447280303594018e-05,
+      "loss": 0.6179,
+      "step": 19270
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.4469826624004765e-05,
+      "loss": 0.6053,
+      "step": 19280
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.446685021206935e-05,
+      "loss": 0.6223,
+      "step": 19290
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.446387380013394e-05,
+      "loss": 0.6022,
+      "step": 19300
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.4460897388198526e-05,
+      "loss": 0.6169,
+      "step": 19310
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.445792097626312e-05,
+      "loss": 0.5868,
+      "step": 19320
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.4454944564327707e-05,
+      "loss": 0.6035,
+      "step": 19330
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.4451968152392293e-05,
+      "loss": 0.6156,
+      "step": 19340
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.444899174045688e-05,
+      "loss": 0.5954,
+      "step": 19350
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.4446015328521474e-05,
+      "loss": 0.5971,
+      "step": 19360
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.444303891658606e-05,
+      "loss": 0.5905,
+      "step": 19370
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.444006250465064e-05,
+      "loss": 0.6154,
+      "step": 19380
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.4437086092715235e-05,
+      "loss": 0.6032,
+      "step": 19390
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.443410968077982e-05,
+      "loss": 0.5907,
+      "step": 19400
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.443113326884441e-05,
+      "loss": 0.6009,
+      "step": 19410
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.4428156856908995e-05,
+      "loss": 0.5938,
+      "step": 19420
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.442518044497359e-05,
+      "loss": 0.6152,
+      "step": 19430
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.4422204033038176e-05,
+      "loss": 0.6133,
+      "step": 19440
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.441922762110276e-05,
+      "loss": 0.6096,
+      "step": 19450
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.441625120916735e-05,
+      "loss": 0.6125,
+      "step": 19460
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.4413274797231943e-05,
+      "loss": 0.6054,
+      "step": 19470
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.441029838529653e-05,
+      "loss": 0.5995,
+      "step": 19480
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.440732197336112e-05,
+      "loss": 0.6142,
+      "step": 19490
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.4404345561425704e-05,
+      "loss": 0.5897,
+      "step": 19500
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.440136914949029e-05,
+      "loss": 0.6043,
+      "step": 19510
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.439839273755488e-05,
+      "loss": 0.5865,
+      "step": 19520
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.439541632561947e-05,
+      "loss": 0.5938,
+      "step": 19530
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.439243991368406e-05,
+      "loss": 0.6136,
+      "step": 19540
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.4389463501748645e-05,
+      "loss": 0.5923,
+      "step": 19550
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.438648708981323e-05,
+      "loss": 0.5925,
+      "step": 19560
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.4383510677877826e-05,
+      "loss": 0.5948,
+      "step": 19570
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.438053426594241e-05,
+      "loss": 0.6129,
+      "step": 19580
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.437755785400699e-05,
+      "loss": 0.5916,
+      "step": 19590
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.437458144207159e-05,
+      "loss": 0.6051,
+      "step": 19600
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.4371605030136174e-05,
+      "loss": 0.6112,
+      "step": 19610
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.436862861820076e-05,
+      "loss": 0.5972,
+      "step": 19620
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.436565220626535e-05,
+      "loss": 0.6045,
+      "step": 19630
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.436267579432994e-05,
+      "loss": 0.5961,
+      "step": 19640
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.435969938239453e-05,
+      "loss": 0.6007,
+      "step": 19650
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.4356722970459115e-05,
+      "loss": 0.6085,
+      "step": 19660
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.43537465585237e-05,
+      "loss": 0.6087,
+      "step": 19670
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.4350770146588295e-05,
+      "loss": 0.6093,
+      "step": 19680
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.4347793734652876e-05,
+      "loss": 0.613,
+      "step": 19690
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.434481732271746e-05,
+      "loss": 0.5887,
+      "step": 19700
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.4341840910782056e-05,
+      "loss": 0.5945,
+      "step": 19710
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.433886449884664e-05,
+      "loss": 0.5972,
+      "step": 19720
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.433588808691123e-05,
+      "loss": 0.5971,
+      "step": 19730
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.433291167497582e-05,
+      "loss": 0.581,
+      "step": 19740
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.432993526304041e-05,
+      "loss": 0.5893,
+      "step": 19750
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.4326958851105e-05,
+      "loss": 0.5932,
+      "step": 19760
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.4323982439169584e-05,
+      "loss": 0.5964,
+      "step": 19770
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.432100602723417e-05,
+      "loss": 0.5937,
+      "step": 19780
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.431802961529876e-05,
+      "loss": 0.5961,
+      "step": 19790
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.4315053203363345e-05,
+      "loss": 0.5953,
+      "step": 19800
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.431207679142793e-05,
+      "loss": 0.613,
+      "step": 19810
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.4309100379492526e-05,
+      "loss": 0.5958,
+      "step": 19820
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.430612396755711e-05,
+      "loss": 0.5737,
+      "step": 19830
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.43031475556217e-05,
+      "loss": 0.6035,
+      "step": 19840
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.4300171143686286e-05,
+      "loss": 0.5883,
+      "step": 19850
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.429719473175088e-05,
+      "loss": 0.5979,
+      "step": 19860
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.429421831981547e-05,
+      "loss": 0.5923,
+      "step": 19870
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.4291241907880054e-05,
+      "loss": 0.6,
+      "step": 19880
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.428826549594464e-05,
+      "loss": 0.6034,
+      "step": 19890
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.428528908400923e-05,
+      "loss": 0.592,
+      "step": 19900
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.4282312672073814e-05,
+      "loss": 0.597,
+      "step": 19910
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.427933626013841e-05,
+      "loss": 0.583,
+      "step": 19920
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.4276359848202995e-05,
+      "loss": 0.5995,
+      "step": 19930
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.427338343626758e-05,
+      "loss": 0.5874,
+      "step": 19940
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.427040702433217e-05,
+      "loss": 0.5948,
+      "step": 19950
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.426743061239676e-05,
+      "loss": 0.6082,
+      "step": 19960
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.426445420046135e-05,
+      "loss": 0.5983,
+      "step": 19970
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.4261477788525936e-05,
+      "loss": 0.5908,
+      "step": 19980
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.425850137659052e-05,
+      "loss": 0.5867,
+      "step": 19990
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.425552496465511e-05,
+      "loss": 0.5789,
+      "step": 20000
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.42525485527197e-05,
+      "loss": 0.5989,
+      "step": 20010
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.4249572140784284e-05,
+      "loss": 0.5961,
+      "step": 20020
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.424659572884888e-05,
+      "loss": 0.618,
+      "step": 20030
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.4243619316913464e-05,
+      "loss": 0.5896,
+      "step": 20040
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.424064290497805e-05,
+      "loss": 0.5926,
+      "step": 20050
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.423766649304264e-05,
+      "loss": 0.5755,
+      "step": 20060
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.423469008110723e-05,
+      "loss": 0.6,
+      "step": 20070
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.423171366917182e-05,
+      "loss": 0.5851,
+      "step": 20080
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.4228737257236406e-05,
+      "loss": 0.6003,
+      "step": 20090
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.422576084530099e-05,
+      "loss": 0.5955,
+      "step": 20100
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.422278443336558e-05,
+      "loss": 0.586,
+      "step": 20110
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.4219808021430166e-05,
+      "loss": 0.6076,
+      "step": 20120
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.421683160949475e-05,
+      "loss": 0.6003,
+      "step": 20130
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.421385519755935e-05,
+      "loss": 0.5811,
+      "step": 20140
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.4210878785623934e-05,
+      "loss": 0.6098,
+      "step": 20150
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.420790237368852e-05,
+      "loss": 0.6064,
+      "step": 20160
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.420492596175311e-05,
+      "loss": 0.605,
+      "step": 20170
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.42019495498177e-05,
+      "loss": 0.5737,
+      "step": 20180
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.419897313788229e-05,
+      "loss": 0.5892,
+      "step": 20190
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.419599672594687e-05,
+      "loss": 0.5664,
+      "step": 20200
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.419302031401146e-05,
+      "loss": 0.5797,
+      "step": 20210
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.419004390207605e-05,
+      "loss": 0.6065,
+      "step": 20220
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.4187067490140636e-05,
+      "loss": 0.6111,
+      "step": 20230
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.418409107820522e-05,
+      "loss": 0.5812,
+      "step": 20240
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.4181114666269816e-05,
+      "loss": 0.5791,
+      "step": 20250
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.41781382543344e-05,
+      "loss": 0.5918,
+      "step": 20260
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.417516184239899e-05,
+      "loss": 0.6077,
+      "step": 20270
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.417218543046358e-05,
+      "loss": 0.5785,
+      "step": 20280
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.416920901852817e-05,
+      "loss": 0.5999,
+      "step": 20290
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.416623260659276e-05,
+      "loss": 0.5907,
+      "step": 20300
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.4163256194657345e-05,
+      "loss": 0.586,
+      "step": 20310
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.416027978272193e-05,
+      "loss": 0.595,
+      "step": 20320
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.415730337078652e-05,
+      "loss": 0.5912,
+      "step": 20330
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.4154326958851105e-05,
+      "loss": 0.5973,
+      "step": 20340
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.41513505469157e-05,
+      "loss": 0.6032,
+      "step": 20350
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.4148374134980286e-05,
+      "loss": 0.605,
+      "step": 20360
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.414539772304487e-05,
+      "loss": 0.5945,
+      "step": 20370
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.414242131110946e-05,
+      "loss": 0.6001,
+      "step": 20380
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.413944489917405e-05,
+      "loss": 0.6011,
+      "step": 20390
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.413646848723864e-05,
+      "loss": 0.5899,
+      "step": 20400
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.413349207530322e-05,
+      "loss": 0.588,
+      "step": 20410
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.4130515663367814e-05,
+      "loss": 0.5769,
+      "step": 20420
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.41275392514324e-05,
+      "loss": 0.5998,
+      "step": 20430
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.412456283949699e-05,
+      "loss": 0.5988,
+      "step": 20440
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.4121586427561575e-05,
+      "loss": 0.5857,
+      "step": 20450
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.411861001562617e-05,
+      "loss": 0.5863,
+      "step": 20460
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.4115633603690755e-05,
+      "loss": 0.5906,
+      "step": 20470
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.411265719175534e-05,
+      "loss": 0.6076,
+      "step": 20480
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.410968077981993e-05,
+      "loss": 0.5842,
+      "step": 20490
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.410670436788452e-05,
+      "loss": 0.5779,
+      "step": 20500
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.41037279559491e-05,
+      "loss": 0.5742,
+      "step": 20510
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.410075154401369e-05,
+      "loss": 0.5827,
+      "step": 20520
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.4097775132078284e-05,
+      "loss": 0.5951,
+      "step": 20530
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.409479872014287e-05,
+      "loss": 0.5954,
+      "step": 20540
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.409182230820746e-05,
+      "loss": 0.5832,
+      "step": 20550
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.4088845896272044e-05,
+      "loss": 0.5884,
+      "step": 20560
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.408586948433664e-05,
+      "loss": 0.6,
+      "step": 20570
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.4082893072401225e-05,
+      "loss": 0.6037,
+      "step": 20580
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.407991666046581e-05,
+      "loss": 0.5842,
+      "step": 20590
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.40769402485304e-05,
+      "loss": 0.5886,
+      "step": 20600
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.4073963836594985e-05,
+      "loss": 0.5975,
+      "step": 20610
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.407098742465957e-05,
+      "loss": 0.5764,
+      "step": 20620
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.406801101272416e-05,
+      "loss": 0.5823,
+      "step": 20630
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.406503460078875e-05,
+      "loss": 0.5921,
+      "step": 20640
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.406205818885334e-05,
+      "loss": 0.5864,
+      "step": 20650
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.405908177691793e-05,
+      "loss": 0.6049,
+      "step": 20660
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.4056105364982514e-05,
+      "loss": 0.593,
+      "step": 20670
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.405312895304711e-05,
+      "loss": 0.6005,
+      "step": 20680
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.4050152541111694e-05,
+      "loss": 0.5995,
+      "step": 20690
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.404717612917628e-05,
+      "loss": 0.5764,
+      "step": 20700
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.404419971724087e-05,
+      "loss": 0.5846,
+      "step": 20710
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.4041223305305455e-05,
+      "loss": 0.5853,
+      "step": 20720
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.403824689337004e-05,
+      "loss": 0.6057,
+      "step": 20730
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.4035270481434635e-05,
+      "loss": 0.5728,
+      "step": 20740
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.403229406949922e-05,
+      "loss": 0.5904,
+      "step": 20750
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.402931765756381e-05,
+      "loss": 0.587,
+      "step": 20760
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.4026341245628396e-05,
+      "loss": 0.5767,
+      "step": 20770
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.402336483369299e-05,
+      "loss": 0.584,
+      "step": 20780
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.402038842175758e-05,
+      "loss": 0.5846,
+      "step": 20790
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.4017412009822164e-05,
+      "loss": 0.5782,
+      "step": 20800
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.401443559788675e-05,
+      "loss": 0.5892,
+      "step": 20810
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.401145918595134e-05,
+      "loss": 0.5975,
+      "step": 20820
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.4008482774015924e-05,
+      "loss": 0.5996,
+      "step": 20830
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.400550636208051e-05,
+      "loss": 0.5886,
+      "step": 20840
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.4002529950145105e-05,
+      "loss": 0.5933,
+      "step": 20850
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.399955353820969e-05,
+      "loss": 0.5902,
+      "step": 20860
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.399657712627428e-05,
+      "loss": 0.5923,
+      "step": 20870
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.3993600714338866e-05,
+      "loss": 0.596,
+      "step": 20880
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.399062430240346e-05,
+      "loss": 0.5877,
+      "step": 20890
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.3987647890468046e-05,
+      "loss": 0.604,
+      "step": 20900
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.398467147853263e-05,
+      "loss": 0.5852,
+      "step": 20910
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.398169506659722e-05,
+      "loss": 0.5936,
+      "step": 20920
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.397871865466181e-05,
+      "loss": 0.5678,
+      "step": 20930
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.3975742242726394e-05,
+      "loss": 0.5863,
+      "step": 20940
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.397276583079098e-05,
+      "loss": 0.5921,
+      "step": 20950
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.3969789418855574e-05,
+      "loss": 0.5982,
+      "step": 20960
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.396681300692016e-05,
+      "loss": 0.597,
+      "step": 20970
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.396383659498475e-05,
+      "loss": 0.5768,
+      "step": 20980
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.3960860183049335e-05,
+      "loss": 0.5819,
+      "step": 20990
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.395788377111393e-05,
+      "loss": 0.5815,
+      "step": 21000
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.3954907359178516e-05,
+      "loss": 0.5803,
+      "step": 21010
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.3951930947243096e-05,
+      "loss": 0.5779,
+      "step": 21020
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.394895453530769e-05,
+      "loss": 0.5923,
+      "step": 21030
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.3945978123372276e-05,
+      "loss": 0.5911,
+      "step": 21040
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.394300171143686e-05,
+      "loss": 0.5816,
+      "step": 21050
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.394002529950145e-05,
+      "loss": 0.591,
+      "step": 21060
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.3937048887566044e-05,
+      "loss": 0.6105,
+      "step": 21070
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.393407247563063e-05,
+      "loss": 0.5677,
+      "step": 21080
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.393109606369522e-05,
+      "loss": 0.579,
+      "step": 21090
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.3928119651759805e-05,
+      "loss": 0.5835,
+      "step": 21100
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.39251432398244e-05,
+      "loss": 0.5808,
+      "step": 21110
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.392216682788898e-05,
+      "loss": 0.577,
+      "step": 21120
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.391919041595357e-05,
+      "loss": 0.5796,
+      "step": 21130
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.391621400401816e-05,
+      "loss": 0.5736,
+      "step": 21140
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.3913237592082746e-05,
+      "loss": 0.5715,
+      "step": 21150
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.391026118014733e-05,
+      "loss": 0.5867,
+      "step": 21160
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.3907284768211926e-05,
+      "loss": 0.5628,
+      "step": 21170
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.390430835627651e-05,
+      "loss": 0.6013,
+      "step": 21180
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.39013319443411e-05,
+      "loss": 0.5999,
+      "step": 21190
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.389835553240569e-05,
+      "loss": 0.5918,
+      "step": 21200
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.389537912047028e-05,
+      "loss": 0.5859,
+      "step": 21210
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.389240270853487e-05,
+      "loss": 0.5666,
+      "step": 21220
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.388942629659945e-05,
+      "loss": 0.5629,
+      "step": 21230
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.388644988466404e-05,
+      "loss": 0.5846,
+      "step": 21240
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.388347347272863e-05,
+      "loss": 0.567,
+      "step": 21250
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.3880497060793215e-05,
+      "loss": 0.5893,
+      "step": 21260
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.38775206488578e-05,
+      "loss": 0.5729,
+      "step": 21270
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.3874544236922396e-05,
+      "loss": 0.5842,
+      "step": 21280
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.387156782498698e-05,
+      "loss": 0.5798,
+      "step": 21290
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.386859141305157e-05,
+      "loss": 0.5966,
+      "step": 21300
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.3865615001116156e-05,
+      "loss": 0.5844,
+      "step": 21310
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.386263858918075e-05,
+      "loss": 0.5725,
+      "step": 21320
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.385966217724533e-05,
+      "loss": 0.6005,
+      "step": 21330
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.385668576530992e-05,
+      "loss": 0.5817,
+      "step": 21340
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.385370935337451e-05,
+      "loss": 0.58,
+      "step": 21350
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.38507329414391e-05,
+      "loss": 0.5885,
+      "step": 21360
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.3847756529503685e-05,
+      "loss": 0.5738,
+      "step": 21370
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.384478011756827e-05,
+      "loss": 0.5792,
+      "step": 21380
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.3841803705632865e-05,
+      "loss": 0.5962,
+      "step": 21390
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.383882729369745e-05,
+      "loss": 0.5818,
+      "step": 21400
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.383585088176204e-05,
+      "loss": 0.5995,
+      "step": 21410
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.3832874469826626e-05,
+      "loss": 0.5889,
+      "step": 21420
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.382989805789121e-05,
+      "loss": 0.5853,
+      "step": 21430
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.38269216459558e-05,
+      "loss": 0.5744,
+      "step": 21440
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.382394523402039e-05,
+      "loss": 0.5829,
+      "step": 21450
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.382096882208498e-05,
+      "loss": 0.6011,
+      "step": 21460
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.381799241014957e-05,
+      "loss": 0.5877,
+      "step": 21470
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.3815015998214154e-05,
+      "loss": 0.5729,
+      "step": 21480
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.381203958627874e-05,
+      "loss": 0.585,
+      "step": 21490
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.3809063174343335e-05,
+      "loss": 0.5829,
+      "step": 21500
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.380608676240792e-05,
+      "loss": 0.5797,
+      "step": 21510
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.380311035047251e-05,
+      "loss": 0.5887,
+      "step": 21520
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.3800133938537095e-05,
+      "loss": 0.5771,
+      "step": 21530
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.379715752660168e-05,
+      "loss": 0.578,
+      "step": 21540
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.379418111466627e-05,
+      "loss": 0.5799,
+      "step": 21550
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.379120470273086e-05,
+      "loss": 0.584,
+      "step": 21560
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.378822829079545e-05,
+      "loss": 0.5707,
+      "step": 21570
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.378525187886004e-05,
+      "loss": 0.5757,
+      "step": 21580
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.3782275466924624e-05,
+      "loss": 0.5797,
+      "step": 21590
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.377929905498922e-05,
+      "loss": 0.5851,
+      "step": 21600
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.3776322643053804e-05,
+      "loss": 0.5729,
+      "step": 21610
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.377334623111839e-05,
+      "loss": 0.5821,
+      "step": 21620
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.377036981918298e-05,
+      "loss": 0.5793,
+      "step": 21630
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.3767393407247565e-05,
+      "loss": 0.5921,
+      "step": 21640
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.376441699531215e-05,
+      "loss": 0.5771,
+      "step": 21650
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.376144058337674e-05,
+      "loss": 0.5867,
+      "step": 21660
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.375846417144133e-05,
+      "loss": 0.5843,
+      "step": 21670
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.375548775950592e-05,
+      "loss": 0.5563,
+      "step": 21680
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.3752511347570506e-05,
+      "loss": 0.5541,
+      "step": 21690
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.374953493563509e-05,
+      "loss": 0.5869,
+      "step": 21700
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.374655852369969e-05,
+      "loss": 0.5642,
+      "step": 21710
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.3743582111764274e-05,
+      "loss": 0.5673,
+      "step": 21720
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.374060569982886e-05,
+      "loss": 0.5746,
+      "step": 21730
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.373762928789345e-05,
+      "loss": 0.5799,
+      "step": 21740
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.3734652875958034e-05,
+      "loss": 0.5856,
+      "step": 21750
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.373167646402262e-05,
+      "loss": 0.573,
+      "step": 21760
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.372870005208721e-05,
+      "loss": 0.5615,
+      "step": 21770
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.37257236401518e-05,
+      "loss": 0.5776,
+      "step": 21780
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.372274722821639e-05,
+      "loss": 0.5856,
+      "step": 21790
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.3719770816280976e-05,
+      "loss": 0.5591,
+      "step": 21800
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.371679440434556e-05,
+      "loss": 0.5645,
+      "step": 21810
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.3713817992410156e-05,
+      "loss": 0.5874,
+      "step": 21820
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.371084158047474e-05,
+      "loss": 0.5564,
+      "step": 21830
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.370786516853933e-05,
+      "loss": 0.5894,
+      "step": 21840
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.370488875660392e-05,
+      "loss": 0.5821,
+      "step": 21850
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.3701912344668504e-05,
+      "loss": 0.5813,
+      "step": 21860
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.369893593273309e-05,
+      "loss": 0.5676,
+      "step": 21870
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.369595952079768e-05,
+      "loss": 0.5776,
+      "step": 21880
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.369298310886227e-05,
+      "loss": 0.5857,
+      "step": 21890
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.369000669692686e-05,
+      "loss": 0.5785,
+      "step": 21900
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.3687030284991445e-05,
+      "loss": 0.586,
+      "step": 21910
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.368405387305603e-05,
+      "loss": 0.5781,
+      "step": 21920
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.3681077461120626e-05,
+      "loss": 0.5927,
+      "step": 21930
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.3678101049185206e-05,
+      "loss": 0.58,
+      "step": 21940
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.36751246372498e-05,
+      "loss": 0.5737,
+      "step": 21950
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.3672148225314386e-05,
+      "loss": 0.5894,
+      "step": 21960
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.366917181337897e-05,
+      "loss": 0.5804,
+      "step": 21970
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.366619540144356e-05,
+      "loss": 0.5644,
+      "step": 21980
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.3663218989508154e-05,
+      "loss": 0.5826,
+      "step": 21990
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.366024257757274e-05,
+      "loss": 0.5825,
+      "step": 22000
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.365726616563733e-05,
+      "loss": 0.579,
+      "step": 22010
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.3654289753701914e-05,
+      "loss": 0.5848,
+      "step": 22020
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.365131334176651e-05,
+      "loss": 0.5628,
+      "step": 22030
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.3648336929831095e-05,
+      "loss": 0.5857,
+      "step": 22040
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.3645360517895675e-05,
+      "loss": 0.5611,
+      "step": 22050
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.364238410596027e-05,
+      "loss": 0.5715,
+      "step": 22060
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.3639407694024856e-05,
+      "loss": 0.5701,
+      "step": 22070
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.363643128208944e-05,
+      "loss": 0.5716,
+      "step": 22080
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.363345487015403e-05,
+      "loss": 0.601,
+      "step": 22090
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.363047845821862e-05,
+      "loss": 0.5699,
+      "step": 22100
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.362750204628321e-05,
+      "loss": 0.5862,
+      "step": 22110
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.36245256343478e-05,
+      "loss": 0.5866,
+      "step": 22120
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.3621549222412384e-05,
+      "loss": 0.5664,
+      "step": 22130
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.361857281047698e-05,
+      "loss": 0.5806,
+      "step": 22140
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.361559639854156e-05,
+      "loss": 0.564,
+      "step": 22150
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.3612619986606145e-05,
+      "loss": 0.5631,
+      "step": 22160
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.360964357467074e-05,
+      "loss": 0.5833,
+      "step": 22170
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.3606667162735325e-05,
+      "loss": 0.5669,
+      "step": 22180
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.360369075079991e-05,
+      "loss": 0.5642,
+      "step": 22190
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.36007143388645e-05,
+      "loss": 0.5742,
+      "step": 22200
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.359773792692909e-05,
+      "loss": 0.5668,
+      "step": 22210
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.359476151499368e-05,
+      "loss": 0.5683,
+      "step": 22220
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.3591785103058266e-05,
+      "loss": 0.5726,
+      "step": 22230
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.358880869112285e-05,
+      "loss": 0.5909,
+      "step": 22240
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.358583227918744e-05,
+      "loss": 0.5606,
+      "step": 22250
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.358285586725203e-05,
+      "loss": 0.5549,
+      "step": 22260
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.357987945531662e-05,
+      "loss": 0.5615,
+      "step": 22270
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.357690304338121e-05,
+      "loss": 0.553,
+      "step": 22280
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.3573926631445795e-05,
+      "loss": 0.5814,
+      "step": 22290
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.357095021951038e-05,
+      "loss": 0.5717,
+      "step": 22300
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.356797380757497e-05,
+      "loss": 0.5657,
+      "step": 22310
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.356499739563956e-05,
+      "loss": 0.5566,
+      "step": 22320
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.356202098370415e-05,
+      "loss": 0.5797,
+      "step": 22330
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.3559044571768736e-05,
+      "loss": 0.5918,
+      "step": 22340
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.355606815983332e-05,
+      "loss": 0.5791,
+      "step": 22350
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.355309174789791e-05,
+      "loss": 0.5588,
+      "step": 22360
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.3550115335962497e-05,
+      "loss": 0.5819,
+      "step": 22370
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.354713892402709e-05,
+      "loss": 0.5686,
+      "step": 22380
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.354416251209168e-05,
+      "loss": 0.5748,
+      "step": 22390
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.3541186100156264e-05,
+      "loss": 0.5655,
+      "step": 22400
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.353820968822085e-05,
+      "loss": 0.5929,
+      "step": 22410
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.3535233276285445e-05,
+      "loss": 0.5365,
+      "step": 22420
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.353225686435003e-05,
+      "loss": 0.5714,
+      "step": 22430
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.352928045241462e-05,
+      "loss": 0.5686,
+      "step": 22440
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.3526304040479205e-05,
+      "loss": 0.5615,
+      "step": 22450
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.352332762854379e-05,
+      "loss": 0.5737,
+      "step": 22460
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.352035121660838e-05,
+      "loss": 0.5675,
+      "step": 22470
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.3517374804672966e-05,
+      "loss": 0.5572,
+      "step": 22480
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.351439839273756e-05,
+      "loss": 0.5767,
+      "step": 22490
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.3511421980802147e-05,
+      "loss": 0.5921,
+      "step": 22500
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.3508445568866733e-05,
+      "loss": 0.5501,
+      "step": 22510
+    },
+    {
+      "epoch": 5.0,
+      "eval_cer": 4.594017094017094,
+      "eval_loss": 0.5387598276138306,
+      "eval_runtime": 1.2701,
+      "eval_samples_per_second": 7.873,
+      "eval_steps_per_second": 0.787,
+      "eval_wer": 0.6363636363636364,
+      "step": 22510
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.350546915693132e-05,
+      "loss": 0.5822,
+      "step": 22520
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.3502492744995914e-05,
+      "loss": 0.5735,
+      "step": 22530
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.34995163330605e-05,
+      "loss": 0.5491,
+      "step": 22540
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.349653992112509e-05,
+      "loss": 0.5569,
+      "step": 22550
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.3493563509189675e-05,
+      "loss": 0.5428,
+      "step": 22560
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.349058709725426e-05,
+      "loss": 0.5538,
+      "step": 22570
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.348761068531885e-05,
+      "loss": 0.5603,
+      "step": 22580
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.3484634273383435e-05,
+      "loss": 0.5313,
+      "step": 22590
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.348165786144803e-05,
+      "loss": 0.5634,
+      "step": 22600
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.3478681449512616e-05,
+      "loss": 0.5656,
+      "step": 22610
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.34757050375772e-05,
+      "loss": 0.5514,
+      "step": 22620
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.347272862564179e-05,
+      "loss": 0.5628,
+      "step": 22630
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.3469752213706383e-05,
+      "loss": 0.5504,
+      "step": 22640
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.346677580177097e-05,
+      "loss": 0.5831,
+      "step": 22650
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.346379938983556e-05,
+      "loss": 0.5497,
+      "step": 22660
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.3460822977900144e-05,
+      "loss": 0.5626,
+      "step": 22670
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.345784656596473e-05,
+      "loss": 0.5634,
+      "step": 22680
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.345487015402932e-05,
+      "loss": 0.5552,
+      "step": 22690
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.345189374209391e-05,
+      "loss": 0.5604,
+      "step": 22700
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.34489173301585e-05,
+      "loss": 0.5466,
+      "step": 22710
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.3445940918223085e-05,
+      "loss": 0.5592,
+      "step": 22720
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.344296450628767e-05,
+      "loss": 0.5698,
+      "step": 22730
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.343998809435226e-05,
+      "loss": 0.5556,
+      "step": 22740
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.343701168241685e-05,
+      "loss": 0.5645,
+      "step": 22750
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.343403527048143e-05,
+      "loss": 0.5623,
+      "step": 22760
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.343105885854603e-05,
+      "loss": 0.5568,
+      "step": 22770
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.3428082446610614e-05,
+      "loss": 0.5578,
+      "step": 22780
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.34251060346752e-05,
+      "loss": 0.5504,
+      "step": 22790
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.342212962273979e-05,
+      "loss": 0.5519,
+      "step": 22800
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.341915321080438e-05,
+      "loss": 0.5543,
+      "step": 22810
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.341617679886897e-05,
+      "loss": 0.5383,
+      "step": 22820
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.3413200386933555e-05,
+      "loss": 0.5495,
+      "step": 22830
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.341022397499814e-05,
+      "loss": 0.5549,
+      "step": 22840
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.3407247563062735e-05,
+      "loss": 0.5531,
+      "step": 22850
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.3404271151127316e-05,
+      "loss": 0.5398,
+      "step": 22860
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.34012947391919e-05,
+      "loss": 0.5572,
+      "step": 22870
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.3398318327256496e-05,
+      "loss": 0.5762,
+      "step": 22880
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.339534191532108e-05,
+      "loss": 0.5476,
+      "step": 22890
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.339236550338567e-05,
+      "loss": 0.5609,
+      "step": 22900
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.338938909145026e-05,
+      "loss": 0.5483,
+      "step": 22910
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.338641267951485e-05,
+      "loss": 0.5609,
+      "step": 22920
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.338343626757944e-05,
+      "loss": 0.5798,
+      "step": 22930
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.3380459855644024e-05,
+      "loss": 0.5322,
+      "step": 22940
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.337748344370861e-05,
+      "loss": 0.5599,
+      "step": 22950
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.3374507031773205e-05,
+      "loss": 0.5586,
+      "step": 22960
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.3371530619837785e-05,
+      "loss": 0.5542,
+      "step": 22970
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.336855420790237e-05,
+      "loss": 0.5525,
+      "step": 22980
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.3365577795966966e-05,
+      "loss": 0.5418,
+      "step": 22990
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.336260138403155e-05,
+      "loss": 0.553,
+      "step": 23000
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.335962497209614e-05,
+      "loss": 0.5563,
+      "step": 23010
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.3356648560160726e-05,
+      "loss": 0.549,
+      "step": 23020
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.335367214822532e-05,
+      "loss": 0.5469,
+      "step": 23030
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.335069573628991e-05,
+      "loss": 0.5566,
+      "step": 23040
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.3347719324354494e-05,
+      "loss": 0.5828,
+      "step": 23050
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.334474291241908e-05,
+      "loss": 0.5474,
+      "step": 23060
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.334176650048367e-05,
+      "loss": 0.5496,
+      "step": 23070
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.3338790088548254e-05,
+      "loss": 0.555,
+      "step": 23080
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.333581367661285e-05,
+      "loss": 0.5472,
+      "step": 23090
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.3332837264677435e-05,
+      "loss": 0.5379,
+      "step": 23100
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.332986085274202e-05,
+      "loss": 0.5476,
+      "step": 23110
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.332688444080661e-05,
+      "loss": 0.5535,
+      "step": 23120
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.33239080288712e-05,
+      "loss": 0.5412,
+      "step": 23130
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.332093161693579e-05,
+      "loss": 0.55,
+      "step": 23140
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.3317955205000376e-05,
+      "loss": 0.5389,
+      "step": 23150
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.331497879306496e-05,
+      "loss": 0.548,
+      "step": 23160
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.331200238112955e-05,
+      "loss": 0.5484,
+      "step": 23170
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.330902596919414e-05,
+      "loss": 0.5515,
+      "step": 23180
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.3306049557258724e-05,
+      "loss": 0.5349,
+      "step": 23190
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.330307314532332e-05,
+      "loss": 0.5601,
+      "step": 23200
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.3300096733387904e-05,
+      "loss": 0.5629,
+      "step": 23210
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.329712032145249e-05,
+      "loss": 0.5609,
+      "step": 23220
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.329414390951708e-05,
+      "loss": 0.5636,
+      "step": 23230
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.329116749758167e-05,
+      "loss": 0.5433,
+      "step": 23240
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.328819108564626e-05,
+      "loss": 0.557,
+      "step": 23250
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.3285214673710846e-05,
+      "loss": 0.5498,
+      "step": 23260
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.328223826177543e-05,
+      "loss": 0.5652,
+      "step": 23270
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.327926184984002e-05,
+      "loss": 0.5598,
+      "step": 23280
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.3276285437904606e-05,
+      "loss": 0.5445,
+      "step": 23290
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.327330902596919e-05,
+      "loss": 0.5501,
+      "step": 23300
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.327033261403379e-05,
+      "loss": 0.5389,
+      "step": 23310
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.3267356202098374e-05,
+      "loss": 0.5583,
+      "step": 23320
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.326437979016296e-05,
+      "loss": 0.5284,
+      "step": 23330
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.326140337822755e-05,
+      "loss": 0.5611,
+      "step": 23340
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.325842696629214e-05,
+      "loss": 0.5563,
+      "step": 23350
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.325545055435673e-05,
+      "loss": 0.5488,
+      "step": 23360
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.3252474142421315e-05,
+      "loss": 0.549,
+      "step": 23370
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.32494977304859e-05,
+      "loss": 0.5608,
+      "step": 23380
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.324652131855049e-05,
+      "loss": 0.5538,
+      "step": 23390
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.3243544906615076e-05,
+      "loss": 0.5658,
+      "step": 23400
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.324056849467966e-05,
+      "loss": 0.558,
+      "step": 23410
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.3237592082744256e-05,
+      "loss": 0.5446,
+      "step": 23420
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.323461567080884e-05,
+      "loss": 0.5594,
+      "step": 23430
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.323163925887343e-05,
+      "loss": 0.564,
+      "step": 23440
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.322866284693802e-05,
+      "loss": 0.5645,
+      "step": 23450
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.322568643500261e-05,
+      "loss": 0.5578,
+      "step": 23460
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.32227100230672e-05,
+      "loss": 0.5499,
+      "step": 23470
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.3219733611131785e-05,
+      "loss": 0.5573,
+      "step": 23480
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.321675719919637e-05,
+      "loss": 0.5454,
+      "step": 23490
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.321378078726096e-05,
+      "loss": 0.5517,
+      "step": 23500
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.3210804375325545e-05,
+      "loss": 0.5573,
+      "step": 23510
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.320782796339014e-05,
+      "loss": 0.5419,
+      "step": 23520
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.3204851551454726e-05,
+      "loss": 0.5634,
+      "step": 23530
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.320187513951931e-05,
+      "loss": 0.5444,
+      "step": 23540
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.31988987275839e-05,
+      "loss": 0.5573,
+      "step": 23550
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.319592231564849e-05,
+      "loss": 0.562,
+      "step": 23560
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.319294590371308e-05,
+      "loss": 0.5629,
+      "step": 23570
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.318996949177766e-05,
+      "loss": 0.5324,
+      "step": 23580
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.3186993079842254e-05,
+      "loss": 0.5491,
+      "step": 23590
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.318401666790684e-05,
+      "loss": 0.5498,
+      "step": 23600
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.318104025597143e-05,
+      "loss": 0.5424,
+      "step": 23610
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.3178063844036015e-05,
+      "loss": 0.5693,
+      "step": 23620
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.317508743210061e-05,
+      "loss": 0.545,
+      "step": 23630
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.3172111020165195e-05,
+      "loss": 0.5546,
+      "step": 23640
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.316913460822978e-05,
+      "loss": 0.5662,
+      "step": 23650
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.316615819629437e-05,
+      "loss": 0.5744,
+      "step": 23660
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.316318178435896e-05,
+      "loss": 0.5443,
+      "step": 23670
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.316020537242354e-05,
+      "loss": 0.5402,
+      "step": 23680
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.315722896048813e-05,
+      "loss": 0.5474,
+      "step": 23690
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.3154252548552723e-05,
+      "loss": 0.5657,
+      "step": 23700
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.315127613661731e-05,
+      "loss": 0.5512,
+      "step": 23710
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.31482997246819e-05,
+      "loss": 0.5343,
+      "step": 23720
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.3145323312746484e-05,
+      "loss": 0.5373,
+      "step": 23730
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.314234690081108e-05,
+      "loss": 0.5381,
+      "step": 23740
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.3139370488875665e-05,
+      "loss": 0.5444,
+      "step": 23750
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.313639407694025e-05,
+      "loss": 0.5401,
+      "step": 23760
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.313341766500484e-05,
+      "loss": 0.5611,
+      "step": 23770
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.313044125306943e-05,
+      "loss": 0.5483,
+      "step": 23780
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.312746484113401e-05,
+      "loss": 0.5444,
+      "step": 23790
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.31244884291986e-05,
+      "loss": 0.5529,
+      "step": 23800
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.312151201726319e-05,
+      "loss": 0.5546,
+      "step": 23810
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.311853560532778e-05,
+      "loss": 0.5503,
+      "step": 23820
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.311555919339237e-05,
+      "loss": 0.5491,
+      "step": 23830
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.3112582781456954e-05,
+      "loss": 0.5434,
+      "step": 23840
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.310960636952155e-05,
+      "loss": 0.5553,
+      "step": 23850
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.3106629957586134e-05,
+      "loss": 0.5523,
+      "step": 23860
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.310365354565072e-05,
+      "loss": 0.5462,
+      "step": 23870
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.310067713371531e-05,
+      "loss": 0.5533,
+      "step": 23880
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.3097700721779895e-05,
+      "loss": 0.5417,
+      "step": 23890
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.309472430984448e-05,
+      "loss": 0.5462,
+      "step": 23900
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.3091747897909075e-05,
+      "loss": 0.5546,
+      "step": 23910
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.308877148597366e-05,
+      "loss": 0.538,
+      "step": 23920
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.308579507403825e-05,
+      "loss": 0.5524,
+      "step": 23930
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.3082818662102836e-05,
+      "loss": 0.5478,
+      "step": 23940
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.307984225016743e-05,
+      "loss": 0.5668,
+      "step": 23950
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.307686583823202e-05,
+      "loss": 0.5464,
+      "step": 23960
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.3073889426296604e-05,
+      "loss": 0.5311,
+      "step": 23970
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.307091301436119e-05,
+      "loss": 0.5462,
+      "step": 23980
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.306793660242578e-05,
+      "loss": 0.5479,
+      "step": 23990
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.3064960190490364e-05,
+      "loss": 0.5394,
+      "step": 24000
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.306198377855495e-05,
+      "loss": 0.5522,
+      "step": 24010
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.3059007366619545e-05,
+      "loss": 0.5454,
+      "step": 24020
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.305603095468413e-05,
+      "loss": 0.5499,
+      "step": 24030
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.305305454274872e-05,
+      "loss": 0.5331,
+      "step": 24040
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.3050078130813306e-05,
+      "loss": 0.545,
+      "step": 24050
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.30471017188779e-05,
+      "loss": 0.5428,
+      "step": 24060
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.3044125306942486e-05,
+      "loss": 0.5372,
+      "step": 24070
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.304114889500707e-05,
+      "loss": 0.5483,
+      "step": 24080
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.303817248307166e-05,
+      "loss": 0.5434,
+      "step": 24090
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.303519607113625e-05,
+      "loss": 0.5364,
+      "step": 24100
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.3032219659200834e-05,
+      "loss": 0.551,
+      "step": 24110
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.302924324726542e-05,
+      "loss": 0.5365,
+      "step": 24120
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.3026266835330014e-05,
+      "loss": 0.5522,
+      "step": 24130
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.30232904233946e-05,
+      "loss": 0.537,
+      "step": 24140
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.302031401145919e-05,
+      "loss": 0.5396,
+      "step": 24150
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.3017337599523775e-05,
+      "loss": 0.5379,
+      "step": 24160
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.301436118758837e-05,
+      "loss": 0.5387,
+      "step": 24170
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.3011384775652956e-05,
+      "loss": 0.5517,
+      "step": 24180
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.300840836371754e-05,
+      "loss": 0.5552,
+      "step": 24190
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.300543195178213e-05,
+      "loss": 0.5444,
+      "step": 24200
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.3002455539846716e-05,
+      "loss": 0.5399,
+      "step": 24210
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.29994791279113e-05,
+      "loss": 0.5548,
+      "step": 24220
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.299650271597589e-05,
+      "loss": 0.5254,
+      "step": 24230
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.2993526304040484e-05,
+      "loss": 0.5555,
+      "step": 24240
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.299054989210507e-05,
+      "loss": 0.5476,
+      "step": 24250
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.298757348016966e-05,
+      "loss": 0.5345,
+      "step": 24260
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.2984597068234244e-05,
+      "loss": 0.5326,
+      "step": 24270
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.298162065629884e-05,
+      "loss": 0.5477,
+      "step": 24280
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.2978644244363425e-05,
+      "loss": 0.5511,
+      "step": 24290
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.297566783242801e-05,
+      "loss": 0.5582,
+      "step": 24300
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.29726914204926e-05,
+      "loss": 0.5453,
+      "step": 24310
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.2969715008557186e-05,
+      "loss": 0.5541,
+      "step": 24320
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.296673859662177e-05,
+      "loss": 0.541,
+      "step": 24330
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.2963762184686366e-05,
+      "loss": 0.5549,
+      "step": 24340
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.296078577275095e-05,
+      "loss": 0.5339,
+      "step": 24350
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.295780936081554e-05,
+      "loss": 0.5569,
+      "step": 24360
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.295483294888013e-05,
+      "loss": 0.5488,
+      "step": 24370
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.295185653694472e-05,
+      "loss": 0.5588,
+      "step": 24380
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.294888012500931e-05,
+      "loss": 0.5629,
+      "step": 24390
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.294590371307389e-05,
+      "loss": 0.5409,
+      "step": 24400
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.294292730113848e-05,
+      "loss": 0.5406,
+      "step": 24410
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.293995088920307e-05,
+      "loss": 0.5323,
+      "step": 24420
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.2936974477267655e-05,
+      "loss": 0.5422,
+      "step": 24430
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.293399806533224e-05,
+      "loss": 0.5467,
+      "step": 24440
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.2931021653396836e-05,
+      "loss": 0.5332,
+      "step": 24450
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.292804524146142e-05,
+      "loss": 0.552,
+      "step": 24460
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.292506882952601e-05,
+      "loss": 0.554,
+      "step": 24470
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.2922092417590596e-05,
+      "loss": 0.5354,
+      "step": 24480
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.291911600565519e-05,
+      "loss": 0.5525,
+      "step": 24490
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.291613959371977e-05,
+      "loss": 0.5658,
+      "step": 24500
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.291316318178436e-05,
+      "loss": 0.5493,
+      "step": 24510
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.291018676984895e-05,
+      "loss": 0.5467,
+      "step": 24520
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.290721035791354e-05,
+      "loss": 0.5434,
+      "step": 24530
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.2904233945978125e-05,
+      "loss": 0.5379,
+      "step": 24540
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.290125753404271e-05,
+      "loss": 0.5283,
+      "step": 24550
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.2898281122107305e-05,
+      "loss": 0.5161,
+      "step": 24560
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.289530471017189e-05,
+      "loss": 0.544,
+      "step": 24570
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.289232829823648e-05,
+      "loss": 0.5356,
+      "step": 24580
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.2889351886301066e-05,
+      "loss": 0.5373,
+      "step": 24590
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.288637547436565e-05,
+      "loss": 0.551,
+      "step": 24600
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.288339906243024e-05,
+      "loss": 0.5432,
+      "step": 24610
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.2880422650494827e-05,
+      "loss": 0.5414,
+      "step": 24620
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.287744623855942e-05,
+      "loss": 0.5375,
+      "step": 24630
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.287446982662401e-05,
+      "loss": 0.5182,
+      "step": 24640
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.2871493414688594e-05,
+      "loss": 0.54,
+      "step": 24650
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.286851700275318e-05,
+      "loss": 0.5317,
+      "step": 24660
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.2865540590817775e-05,
+      "loss": 0.5343,
+      "step": 24670
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.286256417888236e-05,
+      "loss": 0.5302,
+      "step": 24680
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.285958776694695e-05,
+      "loss": 0.5326,
+      "step": 24690
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.2856611355011535e-05,
+      "loss": 0.5682,
+      "step": 24700
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.285363494307612e-05,
+      "loss": 0.5418,
+      "step": 24710
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.285065853114071e-05,
+      "loss": 0.5411,
+      "step": 24720
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.28476821192053e-05,
+      "loss": 0.5544,
+      "step": 24730
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.284470570726989e-05,
+      "loss": 0.5379,
+      "step": 24740
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.2841729295334477e-05,
+      "loss": 0.536,
+      "step": 24750
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.2838752883399064e-05,
+      "loss": 0.5455,
+      "step": 24760
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.283577647146366e-05,
+      "loss": 0.546,
+      "step": 24770
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.2832800059528244e-05,
+      "loss": 0.5499,
+      "step": 24780
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.282982364759283e-05,
+      "loss": 0.5461,
+      "step": 24790
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.282684723565742e-05,
+      "loss": 0.5295,
+      "step": 24800
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.2823870823722005e-05,
+      "loss": 0.5368,
+      "step": 24810
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.282089441178659e-05,
+      "loss": 0.5547,
+      "step": 24820
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.281791799985118e-05,
+      "loss": 0.5273,
+      "step": 24830
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.281494158791577e-05,
+      "loss": 0.5242,
+      "step": 24840
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.281196517598036e-05,
+      "loss": 0.5334,
+      "step": 24850
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.2808988764044946e-05,
+      "loss": 0.5306,
+      "step": 24860
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.280601235210953e-05,
+      "loss": 0.5438,
+      "step": 24870
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.280303594017413e-05,
+      "loss": 0.5507,
+      "step": 24880
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.2800059528238714e-05,
+      "loss": 0.5587,
+      "step": 24890
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.27970831163033e-05,
+      "loss": 0.528,
+      "step": 24900
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.279410670436789e-05,
+      "loss": 0.5288,
+      "step": 24910
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.2791130292432474e-05,
+      "loss": 0.5414,
+      "step": 24920
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.278815388049706e-05,
+      "loss": 0.5514,
+      "step": 24930
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.278517746856165e-05,
+      "loss": 0.5344,
+      "step": 24940
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.278220105662624e-05,
+      "loss": 0.5169,
+      "step": 24950
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.277922464469083e-05,
+      "loss": 0.5425,
+      "step": 24960
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.2776248232755415e-05,
+      "loss": 0.5329,
+      "step": 24970
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.277327182082e-05,
+      "loss": 0.5342,
+      "step": 24980
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.2770295408884596e-05,
+      "loss": 0.5357,
+      "step": 24990
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.276731899694918e-05,
+      "loss": 0.5385,
+      "step": 25000
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.276434258501376e-05,
+      "loss": 0.5071,
+      "step": 25010
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.276136617307836e-05,
+      "loss": 0.5486,
+      "step": 25020
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.2758389761142944e-05,
+      "loss": 0.5402,
+      "step": 25030
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.275541334920753e-05,
+      "loss": 0.536,
+      "step": 25040
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.275243693727212e-05,
+      "loss": 0.5419,
+      "step": 25050
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.274946052533671e-05,
+      "loss": 0.5349,
+      "step": 25060
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.27464841134013e-05,
+      "loss": 0.5285,
+      "step": 25070
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.2743507701465885e-05,
+      "loss": 0.5495,
+      "step": 25080
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.274053128953047e-05,
+      "loss": 0.5533,
+      "step": 25090
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.2737554877595066e-05,
+      "loss": 0.538,
+      "step": 25100
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.273457846565965e-05,
+      "loss": 0.5417,
+      "step": 25110
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.273160205372424e-05,
+      "loss": 0.536,
+      "step": 25120
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.2728625641788826e-05,
+      "loss": 0.5249,
+      "step": 25130
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.272564922985341e-05,
+      "loss": 0.5529,
+      "step": 25140
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.2722672817918e-05,
+      "loss": 0.5239,
+      "step": 25150
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.2719696405982594e-05,
+      "loss": 0.5421,
+      "step": 25160
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.271671999404718e-05,
+      "loss": 0.5589,
+      "step": 25170
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.271374358211177e-05,
+      "loss": 0.5378,
+      "step": 25180
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.2710767170176354e-05,
+      "loss": 0.5336,
+      "step": 25190
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.270779075824095e-05,
+      "loss": 0.5188,
+      "step": 25200
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.2704814346305535e-05,
+      "loss": 0.5252,
+      "step": 25210
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.2701837934370115e-05,
+      "loss": 0.5384,
+      "step": 25220
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.269886152243471e-05,
+      "loss": 0.5443,
+      "step": 25230
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.2695885110499296e-05,
+      "loss": 0.5385,
+      "step": 25240
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.269290869856388e-05,
+      "loss": 0.5329,
+      "step": 25250
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.268993228662847e-05,
+      "loss": 0.5385,
+      "step": 25260
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.268695587469306e-05,
+      "loss": 0.5415,
+      "step": 25270
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.268397946275765e-05,
+      "loss": 0.5221,
+      "step": 25280
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.268100305082224e-05,
+      "loss": 0.5408,
+      "step": 25290
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.2678026638886824e-05,
+      "loss": 0.5555,
+      "step": 25300
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.267505022695142e-05,
+      "loss": 0.5344,
+      "step": 25310
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.2672073815016e-05,
+      "loss": 0.5397,
+      "step": 25320
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.2669097403080585e-05,
+      "loss": 0.541,
+      "step": 25330
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.266612099114518e-05,
+      "loss": 0.5456,
+      "step": 25340
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.2663144579209765e-05,
+      "loss": 0.5326,
+      "step": 25350
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.266016816727435e-05,
+      "loss": 0.5242,
+      "step": 25360
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.265719175533894e-05,
+      "loss": 0.5311,
+      "step": 25370
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.265421534340353e-05,
+      "loss": 0.5331,
+      "step": 25380
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.265123893146812e-05,
+      "loss": 0.5316,
+      "step": 25390
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.2648262519532706e-05,
+      "loss": 0.5485,
+      "step": 25400
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.264528610759729e-05,
+      "loss": 0.5473,
+      "step": 25410
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.264230969566188e-05,
+      "loss": 0.5282,
+      "step": 25420
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.263933328372647e-05,
+      "loss": 0.5388,
+      "step": 25430
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.2636356871791054e-05,
+      "loss": 0.5343,
+      "step": 25440
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.263338045985565e-05,
+      "loss": 0.5265,
+      "step": 25450
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.2630404047920235e-05,
+      "loss": 0.5287,
+      "step": 25460
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.262742763598482e-05,
+      "loss": 0.5242,
+      "step": 25470
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.262445122404941e-05,
+      "loss": 0.5241,
+      "step": 25480
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.2621474812114e-05,
+      "loss": 0.5275,
+      "step": 25490
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.261849840017859e-05,
+      "loss": 0.5417,
+      "step": 25500
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.2615521988243176e-05,
+      "loss": 0.5435,
+      "step": 25510
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.261254557630776e-05,
+      "loss": 0.5438,
+      "step": 25520
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.260956916437235e-05,
+      "loss": 0.5312,
+      "step": 25530
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.2606592752436936e-05,
+      "loss": 0.5419,
+      "step": 25540
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.260361634050153e-05,
+      "loss": 0.5376,
+      "step": 25550
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.260063992856612e-05,
+      "loss": 0.5346,
+      "step": 25560
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.2597663516630704e-05,
+      "loss": 0.5376,
+      "step": 25570
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.259468710469529e-05,
+      "loss": 0.5462,
+      "step": 25580
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.2591710692759885e-05,
+      "loss": 0.5455,
+      "step": 25590
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.258873428082447e-05,
+      "loss": 0.547,
+      "step": 25600
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.258575786888906e-05,
+      "loss": 0.5314,
+      "step": 25610
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.2582781456953645e-05,
+      "loss": 0.5334,
+      "step": 25620
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.257980504501823e-05,
+      "loss": 0.5325,
+      "step": 25630
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.257682863308282e-05,
+      "loss": 0.5479,
+      "step": 25640
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.2573852221147406e-05,
+      "loss": 0.5411,
+      "step": 25650
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.2570875809212e-05,
+      "loss": 0.5147,
+      "step": 25660
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.2567899397276586e-05,
+      "loss": 0.5322,
+      "step": 25670
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.2564922985341173e-05,
+      "loss": 0.5373,
+      "step": 25680
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.256194657340576e-05,
+      "loss": 0.553,
+      "step": 25690
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.2558970161470354e-05,
+      "loss": 0.5375,
+      "step": 25700
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.255599374953494e-05,
+      "loss": 0.5413,
+      "step": 25710
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.255301733759953e-05,
+      "loss": 0.5485,
+      "step": 25720
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.2550040925664115e-05,
+      "loss": 0.5194,
+      "step": 25730
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.25470645137287e-05,
+      "loss": 0.5213,
+      "step": 25740
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.254408810179329e-05,
+      "loss": 0.5318,
+      "step": 25750
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.2541111689857875e-05,
+      "loss": 0.5564,
+      "step": 25760
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.253813527792247e-05,
+      "loss": 0.5323,
+      "step": 25770
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.2535158865987056e-05,
+      "loss": 0.5224,
+      "step": 25780
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.253218245405164e-05,
+      "loss": 0.5252,
+      "step": 25790
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.252920604211623e-05,
+      "loss": 0.5441,
+      "step": 25800
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.2526229630180823e-05,
+      "loss": 0.5398,
+      "step": 25810
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.252325321824541e-05,
+      "loss": 0.5326,
+      "step": 25820
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.252027680631e-05,
+      "loss": 0.5436,
+      "step": 25830
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.2517300394374584e-05,
+      "loss": 0.555,
+      "step": 25840
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.251432398243917e-05,
+      "loss": 0.5454,
+      "step": 25850
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.251134757050376e-05,
+      "loss": 0.5132,
+      "step": 25860
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.2508371158568345e-05,
+      "loss": 0.5318,
+      "step": 25870
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.250539474663294e-05,
+      "loss": 0.5458,
+      "step": 25880
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.2502418334697525e-05,
+      "loss": 0.5251,
+      "step": 25890
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.249944192276211e-05,
+      "loss": 0.5341,
+      "step": 25900
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.24964655108267e-05,
+      "loss": 0.5267,
+      "step": 25910
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.249348909889129e-05,
+      "loss": 0.5303,
+      "step": 25920
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.249051268695588e-05,
+      "loss": 0.532,
+      "step": 25930
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.248753627502047e-05,
+      "loss": 0.547,
+      "step": 25940
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.2484559863085054e-05,
+      "loss": 0.5101,
+      "step": 25950
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.248158345114964e-05,
+      "loss": 0.5415,
+      "step": 25960
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.247860703921423e-05,
+      "loss": 0.5368,
+      "step": 25970
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.247563062727882e-05,
+      "loss": 0.543,
+      "step": 25980
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.247265421534341e-05,
+      "loss": 0.5355,
+      "step": 25990
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.2469677803407995e-05,
+      "loss": 0.5299,
+      "step": 26000
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.246670139147258e-05,
+      "loss": 0.5289,
+      "step": 26010
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.2463724979537175e-05,
+      "loss": 0.5518,
+      "step": 26020
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.246074856760176e-05,
+      "loss": 0.5289,
+      "step": 26030
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.245777215566634e-05,
+      "loss": 0.5367,
+      "step": 26040
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.2454795743730936e-05,
+      "loss": 0.5147,
+      "step": 26050
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.245181933179552e-05,
+      "loss": 0.5354,
+      "step": 26060
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.244884291986011e-05,
+      "loss": 0.5193,
+      "step": 26070
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.24458665079247e-05,
+      "loss": 0.5358,
+      "step": 26080
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.244289009598929e-05,
+      "loss": 0.5358,
+      "step": 26090
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.243991368405388e-05,
+      "loss": 0.5424,
+      "step": 26100
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.2436937272118464e-05,
+      "loss": 0.5317,
+      "step": 26110
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.243396086018305e-05,
+      "loss": 0.5324,
+      "step": 26120
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.2430984448247645e-05,
+      "loss": 0.5328,
+      "step": 26130
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.2428008036312225e-05,
+      "loss": 0.5373,
+      "step": 26140
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.242503162437681e-05,
+      "loss": 0.5208,
+      "step": 26150
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.2422055212441406e-05,
+      "loss": 0.5373,
+      "step": 26160
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.241907880050599e-05,
+      "loss": 0.5358,
+      "step": 26170
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.241610238857058e-05,
+      "loss": 0.5324,
+      "step": 26180
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.2413125976635166e-05,
+      "loss": 0.5307,
+      "step": 26190
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.241014956469976e-05,
+      "loss": 0.5344,
+      "step": 26200
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.240717315276435e-05,
+      "loss": 0.5373,
+      "step": 26210
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.2404196740828934e-05,
+      "loss": 0.5286,
+      "step": 26220
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.240122032889352e-05,
+      "loss": 0.5248,
+      "step": 26230
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.239824391695811e-05,
+      "loss": 0.5203,
+      "step": 26240
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.2395267505022694e-05,
+      "loss": 0.5208,
+      "step": 26250
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.239229109308729e-05,
+      "loss": 0.5263,
+      "step": 26260
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.2389314681151875e-05,
+      "loss": 0.5507,
+      "step": 26270
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.238633826921646e-05,
+      "loss": 0.5349,
+      "step": 26280
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.238336185728105e-05,
+      "loss": 0.5333,
+      "step": 26290
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.2380385445345636e-05,
+      "loss": 0.5238,
+      "step": 26300
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.237740903341023e-05,
+      "loss": 0.5345,
+      "step": 26310
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.2374432621474816e-05,
+      "loss": 0.5348,
+      "step": 26320
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.23714562095394e-05,
+      "loss": 0.5227,
+      "step": 26330
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.236847979760399e-05,
+      "loss": 0.5273,
+      "step": 26340
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.236550338566858e-05,
+      "loss": 0.5384,
+      "step": 26350
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.2362526973733164e-05,
+      "loss": 0.5048,
+      "step": 26360
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.235955056179776e-05,
+      "loss": 0.5397,
+      "step": 26370
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.2356574149862344e-05,
+      "loss": 0.5236,
+      "step": 26380
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.235359773792693e-05,
+      "loss": 0.5318,
+      "step": 26390
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.235062132599152e-05,
+      "loss": 0.5401,
+      "step": 26400
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.234764491405611e-05,
+      "loss": 0.5208,
+      "step": 26410
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.23446685021207e-05,
+      "loss": 0.527,
+      "step": 26420
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.2341692090185286e-05,
+      "loss": 0.5404,
+      "step": 26430
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.233871567824987e-05,
+      "loss": 0.5371,
+      "step": 26440
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.233573926631446e-05,
+      "loss": 0.5392,
+      "step": 26450
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.2332762854379046e-05,
+      "loss": 0.5303,
+      "step": 26460
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.232978644244363e-05,
+      "loss": 0.5273,
+      "step": 26470
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.232681003050823e-05,
+      "loss": 0.5279,
+      "step": 26480
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.2323833618572814e-05,
+      "loss": 0.5325,
+      "step": 26490
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.23208572066374e-05,
+      "loss": 0.513,
+      "step": 26500
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.231788079470199e-05,
+      "loss": 0.5211,
+      "step": 26510
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.231490438276658e-05,
+      "loss": 0.5056,
+      "step": 26520
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.231192797083117e-05,
+      "loss": 0.5308,
+      "step": 26530
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.2308951558895755e-05,
+      "loss": 0.5426,
+      "step": 26540
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.230597514696034e-05,
+      "loss": 0.5072,
+      "step": 26550
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.230299873502493e-05,
+      "loss": 0.5299,
+      "step": 26560
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.2300022323089516e-05,
+      "loss": 0.5481,
+      "step": 26570
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.22970459111541e-05,
+      "loss": 0.5367,
+      "step": 26580
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.2294069499218696e-05,
+      "loss": 0.5356,
+      "step": 26590
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.229109308728328e-05,
+      "loss": 0.5369,
+      "step": 26600
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.228811667534787e-05,
+      "loss": 0.5276,
+      "step": 26610
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.228514026341246e-05,
+      "loss": 0.5257,
+      "step": 26620
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.228216385147705e-05,
+      "loss": 0.5233,
+      "step": 26630
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.227918743954164e-05,
+      "loss": 0.543,
+      "step": 26640
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.2276211027606225e-05,
+      "loss": 0.5282,
+      "step": 26650
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.227323461567081e-05,
+      "loss": 0.5393,
+      "step": 26660
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.22702582037354e-05,
+      "loss": 0.5217,
+      "step": 26670
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.2267281791799985e-05,
+      "loss": 0.5438,
+      "step": 26680
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.226430537986458e-05,
+      "loss": 0.5245,
+      "step": 26690
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.2261328967929166e-05,
+      "loss": 0.5187,
+      "step": 26700
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.225835255599375e-05,
+      "loss": 0.529,
+      "step": 26710
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.225537614405834e-05,
+      "loss": 0.5251,
+      "step": 26720
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.2252399732122927e-05,
+      "loss": 0.5334,
+      "step": 26730
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.224942332018752e-05,
+      "loss": 0.5129,
+      "step": 26740
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.22464469082521e-05,
+      "loss": 0.5186,
+      "step": 26750
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.2243470496316694e-05,
+      "loss": 0.5187,
+      "step": 26760
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.224049408438128e-05,
+      "loss": 0.5292,
+      "step": 26770
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.223751767244587e-05,
+      "loss": 0.5277,
+      "step": 26780
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.2234541260510455e-05,
+      "loss": 0.5236,
+      "step": 26790
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.223156484857505e-05,
+      "loss": 0.5432,
+      "step": 26800
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.2228588436639635e-05,
+      "loss": 0.5339,
+      "step": 26810
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.222561202470422e-05,
+      "loss": 0.5212,
+      "step": 26820
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.222263561276881e-05,
+      "loss": 0.5319,
+      "step": 26830
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.22196592008334e-05,
+      "loss": 0.5335,
+      "step": 26840
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.221668278889799e-05,
+      "loss": 0.5257,
+      "step": 26850
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.221370637696257e-05,
+      "loss": 0.5373,
+      "step": 26860
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.2210729965027163e-05,
+      "loss": 0.526,
+      "step": 26870
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.220775355309175e-05,
+      "loss": 0.5328,
+      "step": 26880
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.220477714115634e-05,
+      "loss": 0.5183,
+      "step": 26890
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.2201800729220924e-05,
+      "loss": 0.5391,
+      "step": 26900
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.219882431728552e-05,
+      "loss": 0.5202,
+      "step": 26910
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.2195847905350105e-05,
+      "loss": 0.5278,
+      "step": 26920
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.219287149341469e-05,
+      "loss": 0.5336,
+      "step": 26930
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.218989508147928e-05,
+      "loss": 0.5384,
+      "step": 26940
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.218691866954387e-05,
+      "loss": 0.5137,
+      "step": 26950
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.218394225760845e-05,
+      "loss": 0.5183,
+      "step": 26960
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.218096584567304e-05,
+      "loss": 0.5247,
+      "step": 26970
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.217798943373763e-05,
+      "loss": 0.5284,
+      "step": 26980
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.217501302180222e-05,
+      "loss": 0.5244,
+      "step": 26990
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.217203660986681e-05,
+      "loss": 0.5286,
+      "step": 27000
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.2169060197931394e-05,
+      "loss": 0.5335,
+      "step": 27010
+    },
+    {
+      "epoch": 6.0,
+      "eval_cer": 4.583333333333333,
+      "eval_loss": 0.4708133339881897,
+      "eval_runtime": 1.2839,
+      "eval_samples_per_second": 7.789,
+      "eval_steps_per_second": 0.779,
+      "eval_wer": 0.6103896103896104,
+      "step": 27012
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.216608378599599e-05,
+      "loss": 0.5109,
+      "step": 27020
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.2163107374060574e-05,
+      "loss": 0.5161,
+      "step": 27030
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.216013096212516e-05,
+      "loss": 0.5243,
+      "step": 27040
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.215715455018975e-05,
+      "loss": 0.5074,
+      "step": 27050
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.2154178138254335e-05,
+      "loss": 0.5312,
+      "step": 27060
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.215120172631892e-05,
+      "loss": 0.5159,
+      "step": 27070
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.2148225314383515e-05,
+      "loss": 0.5044,
+      "step": 27080
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.21452489024481e-05,
+      "loss": 0.5083,
+      "step": 27090
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.214227249051269e-05,
+      "loss": 0.5077,
+      "step": 27100
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.2139296078577276e-05,
+      "loss": 0.5231,
+      "step": 27110
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.213631966664187e-05,
+      "loss": 0.5148,
+      "step": 27120
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.213334325470646e-05,
+      "loss": 0.5286,
+      "step": 27130
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.2130366842771044e-05,
+      "loss": 0.5059,
+      "step": 27140
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.212739043083563e-05,
+      "loss": 0.5166,
+      "step": 27150
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.212441401890022e-05,
+      "loss": 0.5181,
+      "step": 27160
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.2121437606964804e-05,
+      "loss": 0.5112,
+      "step": 27170
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.211846119502939e-05,
+      "loss": 0.4966,
+      "step": 27180
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.2115484783093985e-05,
+      "loss": 0.5166,
+      "step": 27190
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.211250837115857e-05,
+      "loss": 0.5026,
+      "step": 27200
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.210953195922316e-05,
+      "loss": 0.5013,
+      "step": 27210
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.2106555547287746e-05,
+      "loss": 0.5104,
+      "step": 27220
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.210357913535234e-05,
+      "loss": 0.5052,
+      "step": 27230
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.2100602723416926e-05,
+      "loss": 0.5177,
+      "step": 27240
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.209762631148151e-05,
+      "loss": 0.5177,
+      "step": 27250
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.20946498995461e-05,
+      "loss": 0.5219,
+      "step": 27260
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.209167348761069e-05,
+      "loss": 0.5062,
+      "step": 27270
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.2088697075675274e-05,
+      "loss": 0.5136,
+      "step": 27280
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.208572066373986e-05,
+      "loss": 0.5131,
+      "step": 27290
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.2082744251804454e-05,
+      "loss": 0.5197,
+      "step": 27300
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.207976783986904e-05,
+      "loss": 0.4922,
+      "step": 27310
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.207679142793363e-05,
+      "loss": 0.5243,
+      "step": 27320
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.2073815015998215e-05,
+      "loss": 0.492,
+      "step": 27330
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.207083860406281e-05,
+      "loss": 0.5169,
+      "step": 27340
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.2067862192127396e-05,
+      "loss": 0.4876,
+      "step": 27350
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.206488578019198e-05,
+      "loss": 0.5017,
+      "step": 27360
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.206190936825657e-05,
+      "loss": 0.511,
+      "step": 27370
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.2058932956321156e-05,
+      "loss": 0.4955,
+      "step": 27380
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.205595654438574e-05,
+      "loss": 0.5152,
+      "step": 27390
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.205298013245033e-05,
+      "loss": 0.4971,
+      "step": 27400
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.2050003720514924e-05,
+      "loss": 0.5004,
+      "step": 27410
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.204702730857951e-05,
+      "loss": 0.5173,
+      "step": 27420
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.20440508966441e-05,
+      "loss": 0.518,
+      "step": 27430
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.2041074484708684e-05,
+      "loss": 0.5274,
+      "step": 27440
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.203809807277328e-05,
+      "loss": 0.5096,
+      "step": 27450
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.2035121660837865e-05,
+      "loss": 0.5048,
+      "step": 27460
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.203214524890245e-05,
+      "loss": 0.506,
+      "step": 27470
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.202916883696704e-05,
+      "loss": 0.5187,
+      "step": 27480
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.2026192425031626e-05,
+      "loss": 0.5101,
+      "step": 27490
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.202321601309621e-05,
+      "loss": 0.5202,
+      "step": 27500
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.2020239601160806e-05,
+      "loss": 0.4862,
+      "step": 27510
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.201726318922539e-05,
+      "loss": 0.5066,
+      "step": 27520
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.201428677728998e-05,
+      "loss": 0.5161,
+      "step": 27530
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.201131036535457e-05,
+      "loss": 0.496,
+      "step": 27540
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.200833395341916e-05,
+      "loss": 0.5159,
+      "step": 27550
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.200535754148375e-05,
+      "loss": 0.5179,
+      "step": 27560
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.200238112954833e-05,
+      "loss": 0.5166,
+      "step": 27570
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.199940471761292e-05,
+      "loss": 0.503,
+      "step": 27580
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.199642830567751e-05,
+      "loss": 0.5115,
+      "step": 27590
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.1993451893742095e-05,
+      "loss": 0.5188,
+      "step": 27600
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.199047548180668e-05,
+      "loss": 0.5101,
+      "step": 27610
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.1987499069871276e-05,
+      "loss": 0.5194,
+      "step": 27620
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.198452265793586e-05,
+      "loss": 0.5228,
+      "step": 27630
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.198154624600045e-05,
+      "loss": 0.52,
+      "step": 27640
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.1978569834065036e-05,
+      "loss": 0.5072,
+      "step": 27650
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.197559342212963e-05,
+      "loss": 0.5012,
+      "step": 27660
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.197261701019422e-05,
+      "loss": 0.5247,
+      "step": 27670
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.19696405982588e-05,
+      "loss": 0.505,
+      "step": 27680
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.196666418632339e-05,
+      "loss": 0.5111,
+      "step": 27690
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.196368777438798e-05,
+      "loss": 0.5085,
+      "step": 27700
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.1960711362452565e-05,
+      "loss": 0.5145,
+      "step": 27710
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.195773495051715e-05,
+      "loss": 0.5081,
+      "step": 27720
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.1954758538581745e-05,
+      "loss": 0.5122,
+      "step": 27730
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.195178212664633e-05,
+      "loss": 0.5216,
+      "step": 27740
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.194880571471092e-05,
+      "loss": 0.515,
+      "step": 27750
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.1945829302775506e-05,
+      "loss": 0.5092,
+      "step": 27760
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.19428528908401e-05,
+      "loss": 0.5135,
+      "step": 27770
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.193987647890468e-05,
+      "loss": 0.4965,
+      "step": 27780
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.1936900066969267e-05,
+      "loss": 0.5119,
+      "step": 27790
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.193392365503386e-05,
+      "loss": 0.4941,
+      "step": 27800
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.193094724309845e-05,
+      "loss": 0.5123,
+      "step": 27810
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.1927970831163034e-05,
+      "loss": 0.5156,
+      "step": 27820
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.192499441922762e-05,
+      "loss": 0.5205,
+      "step": 27830
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.1922018007292215e-05,
+      "loss": 0.5161,
+      "step": 27840
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.19190415953568e-05,
+      "loss": 0.5135,
+      "step": 27850
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.191606518342139e-05,
+      "loss": 0.5158,
+      "step": 27860
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.1913088771485975e-05,
+      "loss": 0.5211,
+      "step": 27870
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.191011235955056e-05,
+      "loss": 0.4928,
+      "step": 27880
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.190713594761515e-05,
+      "loss": 0.51,
+      "step": 27890
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.190415953567974e-05,
+      "loss": 0.5162,
+      "step": 27900
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.190118312374433e-05,
+      "loss": 0.507,
+      "step": 27910
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.1898206711808917e-05,
+      "loss": 0.5084,
+      "step": 27920
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.1895230299873503e-05,
+      "loss": 0.5201,
+      "step": 27930
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.18922538879381e-05,
+      "loss": 0.5368,
+      "step": 27940
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.1889277476002684e-05,
+      "loss": 0.5123,
+      "step": 27950
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.188630106406727e-05,
+      "loss": 0.5035,
+      "step": 27960
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.188332465213186e-05,
+      "loss": 0.5124,
+      "step": 27970
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.1880348240196445e-05,
+      "loss": 0.5079,
+      "step": 27980
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.187737182826103e-05,
+      "loss": 0.5089,
+      "step": 27990
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.187439541632562e-05,
+      "loss": 0.5284,
+      "step": 28000
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.187141900439021e-05,
+      "loss": 0.5204,
+      "step": 28010
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.18684425924548e-05,
+      "loss": 0.5116,
+      "step": 28020
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.1865466180519386e-05,
+      "loss": 0.4997,
+      "step": 28030
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.186248976858397e-05,
+      "loss": 0.4938,
+      "step": 28040
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.1859513356648567e-05,
+      "loss": 0.5179,
+      "step": 28050
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.1856536944713153e-05,
+      "loss": 0.504,
+      "step": 28060
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.185356053277774e-05,
+      "loss": 0.5219,
+      "step": 28070
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.185058412084233e-05,
+      "loss": 0.5008,
+      "step": 28080
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.1847607708906914e-05,
+      "loss": 0.4977,
+      "step": 28090
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.18446312969715e-05,
+      "loss": 0.5088,
+      "step": 28100
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.184165488503609e-05,
+      "loss": 0.5292,
+      "step": 28110
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.183867847310068e-05,
+      "loss": 0.497,
+      "step": 28120
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.183570206116527e-05,
+      "loss": 0.5151,
+      "step": 28130
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.1832725649229855e-05,
+      "loss": 0.5046,
+      "step": 28140
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.182974923729444e-05,
+      "loss": 0.5013,
+      "step": 28150
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.1826772825359036e-05,
+      "loss": 0.5187,
+      "step": 28160
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.182379641342362e-05,
+      "loss": 0.5008,
+      "step": 28170
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.182082000148821e-05,
+      "loss": 0.5043,
+      "step": 28180
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.18178435895528e-05,
+      "loss": 0.5,
+      "step": 28190
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.1814867177617384e-05,
+      "loss": 0.4959,
+      "step": 28200
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.181189076568197e-05,
+      "loss": 0.5131,
+      "step": 28210
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.180891435374656e-05,
+      "loss": 0.5127,
+      "step": 28220
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.180593794181115e-05,
+      "loss": 0.512,
+      "step": 28230
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.180296152987574e-05,
+      "loss": 0.5129,
+      "step": 28240
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.1799985117940325e-05,
+      "loss": 0.4979,
+      "step": 28250
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.179700870600491e-05,
+      "loss": 0.505,
+      "step": 28260
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.1794032294069505e-05,
+      "loss": 0.5084,
+      "step": 28270
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.179105588213409e-05,
+      "loss": 0.5091,
+      "step": 28280
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.178807947019868e-05,
+      "loss": 0.5039,
+      "step": 28290
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.1785103058263266e-05,
+      "loss": 0.5116,
+      "step": 28300
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.178212664632785e-05,
+      "loss": 0.5112,
+      "step": 28310
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.177915023439244e-05,
+      "loss": 0.5069,
+      "step": 28320
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.1776173822457034e-05,
+      "loss": 0.4924,
+      "step": 28330
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.177319741052162e-05,
+      "loss": 0.5199,
+      "step": 28340
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.177022099858621e-05,
+      "loss": 0.5059,
+      "step": 28350
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.1767244586650794e-05,
+      "loss": 0.5014,
+      "step": 28360
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.176426817471539e-05,
+      "loss": 0.5232,
+      "step": 28370
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.1761291762779975e-05,
+      "loss": 0.4995,
+      "step": 28380
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.1758315350844555e-05,
+      "loss": 0.5267,
+      "step": 28390
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.175533893890915e-05,
+      "loss": 0.5196,
+      "step": 28400
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.1752362526973736e-05,
+      "loss": 0.5125,
+      "step": 28410
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.174938611503832e-05,
+      "loss": 0.5054,
+      "step": 28420
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.174640970310291e-05,
+      "loss": 0.5089,
+      "step": 28430
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.17434332911675e-05,
+      "loss": 0.4946,
+      "step": 28440
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.174045687923209e-05,
+      "loss": 0.5149,
+      "step": 28450
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.173748046729668e-05,
+      "loss": 0.5108,
+      "step": 28460
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.1734504055361264e-05,
+      "loss": 0.5039,
+      "step": 28470
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.173152764342586e-05,
+      "loss": 0.5133,
+      "step": 28480
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.172855123149044e-05,
+      "loss": 0.5107,
+      "step": 28490
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.1725574819555024e-05,
+      "loss": 0.5131,
+      "step": 28500
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.172259840761962e-05,
+      "loss": 0.5174,
+      "step": 28510
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.1719621995684205e-05,
+      "loss": 0.5055,
+      "step": 28520
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.171664558374879e-05,
+      "loss": 0.4868,
+      "step": 28530
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.171366917181338e-05,
+      "loss": 0.5033,
+      "step": 28540
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.171069275987797e-05,
+      "loss": 0.4938,
+      "step": 28550
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.170771634794256e-05,
+      "loss": 0.506,
+      "step": 28560
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.1704739936007146e-05,
+      "loss": 0.5157,
+      "step": 28570
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.170176352407173e-05,
+      "loss": 0.4938,
+      "step": 28580
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.169878711213633e-05,
+      "loss": 0.514,
+      "step": 28590
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.169581070020091e-05,
+      "loss": 0.5062,
+      "step": 28600
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.1692834288265494e-05,
+      "loss": 0.5164,
+      "step": 28610
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.168985787633009e-05,
+      "loss": 0.5143,
+      "step": 28620
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.1686881464394674e-05,
+      "loss": 0.5179,
+      "step": 28630
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.168390505245926e-05,
+      "loss": 0.5045,
+      "step": 28640
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.168092864052385e-05,
+      "loss": 0.5046,
+      "step": 28650
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.167795222858844e-05,
+      "loss": 0.505,
+      "step": 28660
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.167497581665303e-05,
+      "loss": 0.5088,
+      "step": 28670
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.1671999404717616e-05,
+      "loss": 0.5018,
+      "step": 28680
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.16690229927822e-05,
+      "loss": 0.5127,
+      "step": 28690
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.166604658084679e-05,
+      "loss": 0.5067,
+      "step": 28700
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.1663070168911376e-05,
+      "loss": 0.5063,
+      "step": 28710
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.166009375697597e-05,
+      "loss": 0.5057,
+      "step": 28720
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.165711734504056e-05,
+      "loss": 0.4904,
+      "step": 28730
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.1654140933105144e-05,
+      "loss": 0.5003,
+      "step": 28740
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.165116452116973e-05,
+      "loss": 0.5083,
+      "step": 28750
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.1648188109234324e-05,
+      "loss": 0.4993,
+      "step": 28760
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.164521169729891e-05,
+      "loss": 0.4964,
+      "step": 28770
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.16422352853635e-05,
+      "loss": 0.5058,
+      "step": 28780
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.1639258873428085e-05,
+      "loss": 0.4973,
+      "step": 28790
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.163628246149267e-05,
+      "loss": 0.4992,
+      "step": 28800
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.163330604955726e-05,
+      "loss": 0.5034,
+      "step": 28810
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.1630329637621846e-05,
+      "loss": 0.5008,
+      "step": 28820
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.162735322568644e-05,
+      "loss": 0.5075,
+      "step": 28830
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.1624376813751026e-05,
+      "loss": 0.5071,
+      "step": 28840
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.162140040181561e-05,
+      "loss": 0.4767,
+      "step": 28850
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.16184239898802e-05,
+      "loss": 0.5063,
+      "step": 28860
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.1615447577944794e-05,
+      "loss": 0.5134,
+      "step": 28870
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.161247116600938e-05,
+      "loss": 0.5014,
+      "step": 28880
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.160949475407397e-05,
+      "loss": 0.5101,
+      "step": 28890
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.1606518342138555e-05,
+      "loss": 0.4876,
+      "step": 28900
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.160354193020314e-05,
+      "loss": 0.483,
+      "step": 28910
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.160056551826773e-05,
+      "loss": 0.4899,
+      "step": 28920
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.1597589106332315e-05,
+      "loss": 0.4963,
+      "step": 28930
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.159461269439691e-05,
+      "loss": 0.503,
+      "step": 28940
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.1591636282461496e-05,
+      "loss": 0.5085,
+      "step": 28950
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.158865987052608e-05,
+      "loss": 0.4985,
+      "step": 28960
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.158568345859067e-05,
+      "loss": 0.4954,
+      "step": 28970
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.158270704665526e-05,
+      "loss": 0.513,
+      "step": 28980
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.157973063471985e-05,
+      "loss": 0.5066,
+      "step": 28990
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.157675422278444e-05,
+      "loss": 0.518,
+      "step": 29000
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.1573777810849024e-05,
+      "loss": 0.5077,
+      "step": 29010
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.157080139891361e-05,
+      "loss": 0.4944,
+      "step": 29020
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.15678249869782e-05,
+      "loss": 0.513,
+      "step": 29030
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.1564848575042785e-05,
+      "loss": 0.5061,
+      "step": 29040
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.156187216310738e-05,
+      "loss": 0.507,
+      "step": 29050
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.1558895751171965e-05,
+      "loss": 0.4933,
+      "step": 29060
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.155591933923655e-05,
+      "loss": 0.497,
+      "step": 29070
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.155294292730114e-05,
+      "loss": 0.5045,
+      "step": 29080
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.154996651536573e-05,
+      "loss": 0.485,
+      "step": 29090
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.154699010343032e-05,
+      "loss": 0.491,
+      "step": 29100
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.154401369149491e-05,
+      "loss": 0.5075,
+      "step": 29110
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.1541037279559494e-05,
+      "loss": 0.5012,
+      "step": 29120
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.153806086762408e-05,
+      "loss": 0.5239,
+      "step": 29130
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.153508445568867e-05,
+      "loss": 0.504,
+      "step": 29140
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.153210804375326e-05,
+      "loss": 0.5087,
+      "step": 29150
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.152913163181785e-05,
+      "loss": 0.5132,
+      "step": 29160
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.1526155219882435e-05,
+      "loss": 0.4957,
+      "step": 29170
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.152317880794702e-05,
+      "loss": 0.5089,
+      "step": 29180
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.1520202396011615e-05,
+      "loss": 0.4981,
+      "step": 29190
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.15172259840762e-05,
+      "loss": 0.4999,
+      "step": 29200
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.151424957214078e-05,
+      "loss": 0.5148,
+      "step": 29210
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.1511273160205376e-05,
+      "loss": 0.4915,
+      "step": 29220
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.150829674826996e-05,
+      "loss": 0.5012,
+      "step": 29230
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.150532033633455e-05,
+      "loss": 0.4929,
+      "step": 29240
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.150234392439914e-05,
+      "loss": 0.5084,
+      "step": 29250
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.149936751246373e-05,
+      "loss": 0.5056,
+      "step": 29260
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.149639110052832e-05,
+      "loss": 0.5035,
+      "step": 29270
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.1493414688592904e-05,
+      "loss": 0.4937,
+      "step": 29280
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.149043827665749e-05,
+      "loss": 0.4893,
+      "step": 29290
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.1487461864722085e-05,
+      "loss": 0.5109,
+      "step": 29300
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.1484485452786665e-05,
+      "loss": 0.4956,
+      "step": 29310
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.148150904085125e-05,
+      "loss": 0.5089,
+      "step": 29320
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.1478532628915845e-05,
+      "loss": 0.5055,
+      "step": 29330
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.147555621698043e-05,
+      "loss": 0.5016,
+      "step": 29340
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.147257980504502e-05,
+      "loss": 0.4926,
+      "step": 29350
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.1469603393109606e-05,
+      "loss": 0.4904,
+      "step": 29360
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.14666269811742e-05,
+      "loss": 0.5324,
+      "step": 29370
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.146365056923879e-05,
+      "loss": 0.484,
+      "step": 29380
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.1460674157303374e-05,
+      "loss": 0.4983,
+      "step": 29390
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.145769774536796e-05,
+      "loss": 0.5325,
+      "step": 29400
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.1454721333432554e-05,
+      "loss": 0.4983,
+      "step": 29410
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.1451744921497134e-05,
+      "loss": 0.4889,
+      "step": 29420
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.144876850956172e-05,
+      "loss": 0.4954,
+      "step": 29430
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.1445792097626315e-05,
+      "loss": 0.51,
+      "step": 29440
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.14428156856909e-05,
+      "loss": 0.5122,
+      "step": 29450
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.143983927375549e-05,
+      "loss": 0.5115,
+      "step": 29460
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.1436862861820076e-05,
+      "loss": 0.5092,
+      "step": 29470
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.143388644988467e-05,
+      "loss": 0.4918,
+      "step": 29480
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.1430910037949256e-05,
+      "loss": 0.5092,
+      "step": 29490
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.142793362601384e-05,
+      "loss": 0.4772,
+      "step": 29500
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.142495721407843e-05,
+      "loss": 0.4946,
+      "step": 29510
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.142198080214302e-05,
+      "loss": 0.5048,
+      "step": 29520
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.1419004390207604e-05,
+      "loss": 0.4952,
+      "step": 29530
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.14160279782722e-05,
+      "loss": 0.5106,
+      "step": 29540
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.1413051566336784e-05,
+      "loss": 0.4989,
+      "step": 29550
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.141007515440137e-05,
+      "loss": 0.4893,
+      "step": 29560
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.140709874246596e-05,
+      "loss": 0.4981,
+      "step": 29570
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.140412233053055e-05,
+      "loss": 0.5117,
+      "step": 29580
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.140114591859514e-05,
+      "loss": 0.5166,
+      "step": 29590
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.1398169506659726e-05,
+      "loss": 0.5222,
+      "step": 29600
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.139519309472431e-05,
+      "loss": 0.5169,
+      "step": 29610
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.13922166827889e-05,
+      "loss": 0.483,
+      "step": 29620
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.1389240270853486e-05,
+      "loss": 0.5095,
+      "step": 29630
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.138626385891807e-05,
+      "loss": 0.5038,
+      "step": 29640
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.138328744698267e-05,
+      "loss": 0.5011,
+      "step": 29650
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.1380311035047254e-05,
+      "loss": 0.4803,
+      "step": 29660
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.137733462311184e-05,
+      "loss": 0.5053,
+      "step": 29670
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.137435821117643e-05,
+      "loss": 0.4917,
+      "step": 29680
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.137138179924102e-05,
+      "loss": 0.5126,
+      "step": 29690
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.136840538730561e-05,
+      "loss": 0.4807,
+      "step": 29700
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.1365428975370195e-05,
+      "loss": 0.5014,
+      "step": 29710
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.136245256343478e-05,
+      "loss": 0.4955,
+      "step": 29720
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.135947615149937e-05,
+      "loss": 0.5144,
+      "step": 29730
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.1356499739563956e-05,
+      "loss": 0.5027,
+      "step": 29740
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.135352332762854e-05,
+      "loss": 0.4984,
+      "step": 29750
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.1350546915693136e-05,
+      "loss": 0.5023,
+      "step": 29760
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.134757050375772e-05,
+      "loss": 0.4827,
+      "step": 29770
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.134459409182231e-05,
+      "loss": 0.4973,
+      "step": 29780
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.13416176798869e-05,
+      "loss": 0.4989,
+      "step": 29790
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.133864126795149e-05,
+      "loss": 0.5182,
+      "step": 29800
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.133566485601608e-05,
+      "loss": 0.4951,
+      "step": 29810
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.133268844408066e-05,
+      "loss": 0.5151,
+      "step": 29820
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.132971203214525e-05,
+      "loss": 0.5088,
+      "step": 29830
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.132673562020984e-05,
+      "loss": 0.4924,
+      "step": 29840
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.1323759208274425e-05,
+      "loss": 0.4987,
+      "step": 29850
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.132078279633901e-05,
+      "loss": 0.4792,
+      "step": 29860
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.1317806384403606e-05,
+      "loss": 0.4968,
+      "step": 29870
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.131482997246819e-05,
+      "loss": 0.5085,
+      "step": 29880
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.131185356053278e-05,
+      "loss": 0.5062,
+      "step": 29890
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.1308877148597366e-05,
+      "loss": 0.4969,
+      "step": 29900
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.130590073666196e-05,
+      "loss": 0.5079,
+      "step": 29910
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.130292432472655e-05,
+      "loss": 0.4951,
+      "step": 29920
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.1299947912791134e-05,
+      "loss": 0.4904,
+      "step": 29930
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.129697150085572e-05,
+      "loss": 0.4918,
+      "step": 29940
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.129399508892031e-05,
+      "loss": 0.4732,
+      "step": 29950
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.1291018676984895e-05,
+      "loss": 0.5056,
+      "step": 29960
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.128804226504949e-05,
+      "loss": 0.4893,
+      "step": 29970
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.1285065853114075e-05,
+      "loss": 0.4883,
+      "step": 29980
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.128208944117866e-05,
+      "loss": 0.5195,
+      "step": 29990
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.127911302924325e-05,
+      "loss": 0.4967,
+      "step": 30000
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.127613661730784e-05,
+      "loss": 0.5104,
+      "step": 30010
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.127316020537243e-05,
+      "loss": 0.4955,
+      "step": 30020
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.127018379343701e-05,
+      "loss": 0.4881,
+      "step": 30030
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.1267207381501603e-05,
+      "loss": 0.4945,
+      "step": 30040
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.126423096956619e-05,
+      "loss": 0.4938,
+      "step": 30050
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.126125455763078e-05,
+      "loss": 0.4991,
+      "step": 30060
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.1258278145695364e-05,
+      "loss": 0.4996,
+      "step": 30070
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.125530173375996e-05,
+      "loss": 0.5221,
+      "step": 30080
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.1252325321824545e-05,
+      "loss": 0.5003,
+      "step": 30090
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.124934890988913e-05,
+      "loss": 0.4975,
+      "step": 30100
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.124637249795372e-05,
+      "loss": 0.4938,
+      "step": 30110
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.124339608601831e-05,
+      "loss": 0.4998,
+      "step": 30120
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.124041967408289e-05,
+      "loss": 0.4942,
+      "step": 30130
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.123744326214748e-05,
+      "loss": 0.5009,
+      "step": 30140
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.123446685021207e-05,
+      "loss": 0.4888,
+      "step": 30150
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.123149043827666e-05,
+      "loss": 0.4919,
+      "step": 30160
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.122851402634125e-05,
+      "loss": 0.4971,
+      "step": 30170
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.1225537614405834e-05,
+      "loss": 0.5042,
+      "step": 30180
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.122256120247043e-05,
+      "loss": 0.4985,
+      "step": 30190
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.1219584790535014e-05,
+      "loss": 0.4965,
+      "step": 30200
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.12166083785996e-05,
+      "loss": 0.4834,
+      "step": 30210
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.121363196666419e-05,
+      "loss": 0.4873,
+      "step": 30220
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.1210655554728775e-05,
+      "loss": 0.4873,
+      "step": 30230
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.120767914279336e-05,
+      "loss": 0.5001,
+      "step": 30240
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.120470273085795e-05,
+      "loss": 0.4978,
+      "step": 30250
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.120172631892254e-05,
+      "loss": 0.4861,
+      "step": 30260
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.119874990698713e-05,
+      "loss": 0.5259,
+      "step": 30270
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.1195773495051716e-05,
+      "loss": 0.5015,
+      "step": 30280
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.11927970831163e-05,
+      "loss": 0.4876,
+      "step": 30290
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.11898206711809e-05,
+      "loss": 0.5045,
+      "step": 30300
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.1186844259245484e-05,
+      "loss": 0.5066,
+      "step": 30310
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.118386784731007e-05,
+      "loss": 0.4993,
+      "step": 30320
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.118089143537466e-05,
+      "loss": 0.4996,
+      "step": 30330
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.1177915023439244e-05,
+      "loss": 0.4992,
+      "step": 30340
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.117493861150383e-05,
+      "loss": 0.4784,
+      "step": 30350
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.1171962199568425e-05,
+      "loss": 0.4963,
+      "step": 30360
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.116898578763301e-05,
+      "loss": 0.4999,
+      "step": 30370
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.11660093756976e-05,
+      "loss": 0.5002,
+      "step": 30380
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.1163032963762186e-05,
+      "loss": 0.4911,
+      "step": 30390
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.116005655182678e-05,
+      "loss": 0.5028,
+      "step": 30400
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.1157080139891366e-05,
+      "loss": 0.4917,
+      "step": 30410
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.115410372795595e-05,
+      "loss": 0.5071,
+      "step": 30420
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.115112731602054e-05,
+      "loss": 0.4907,
+      "step": 30430
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.114815090408513e-05,
+      "loss": 0.5022,
+      "step": 30440
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.1145174492149714e-05,
+      "loss": 0.4998,
+      "step": 30450
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.11421980802143e-05,
+      "loss": 0.4999,
+      "step": 30460
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.1139221668278894e-05,
+      "loss": 0.4932,
+      "step": 30470
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.113624525634348e-05,
+      "loss": 0.5038,
+      "step": 30480
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.113326884440807e-05,
+      "loss": 0.4913,
+      "step": 30490
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.1130292432472655e-05,
+      "loss": 0.5178,
+      "step": 30500
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.112731602053725e-05,
+      "loss": 0.49,
+      "step": 30510
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.1124339608601836e-05,
+      "loss": 0.4998,
+      "step": 30520
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.112136319666642e-05,
+      "loss": 0.4915,
+      "step": 30530
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.111838678473101e-05,
+      "loss": 0.5085,
+      "step": 30540
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.1115410372795596e-05,
+      "loss": 0.5028,
+      "step": 30550
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.111243396086018e-05,
+      "loss": 0.4902,
+      "step": 30560
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.110945754892477e-05,
+      "loss": 0.4715,
+      "step": 30570
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.1106481136989364e-05,
+      "loss": 0.4898,
+      "step": 30580
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.110350472505395e-05,
+      "loss": 0.5051,
+      "step": 30590
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.110052831311854e-05,
+      "loss": 0.4859,
+      "step": 30600
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.1097551901183124e-05,
+      "loss": 0.4984,
+      "step": 30610
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.109457548924772e-05,
+      "loss": 0.4986,
+      "step": 30620
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.1091599077312305e-05,
+      "loss": 0.4835,
+      "step": 30630
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.108862266537689e-05,
+      "loss": 0.4938,
+      "step": 30640
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.108564625344148e-05,
+      "loss": 0.4848,
+      "step": 30650
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1082669841506066e-05,
+      "loss": 0.4861,
+      "step": 30660
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.107969342957065e-05,
+      "loss": 0.5,
+      "step": 30670
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.107671701763524e-05,
+      "loss": 0.5038,
+      "step": 30680
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.107374060569983e-05,
+      "loss": 0.4758,
+      "step": 30690
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.107076419376442e-05,
+      "loss": 0.4974,
+      "step": 30700
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.106778778182901e-05,
+      "loss": 0.4917,
+      "step": 30710
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1064811369893594e-05,
+      "loss": 0.4811,
+      "step": 30720
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.106183495795819e-05,
+      "loss": 0.482,
+      "step": 30730
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1058858546022774e-05,
+      "loss": 0.4936,
+      "step": 30740
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.105588213408736e-05,
+      "loss": 0.5012,
+      "step": 30750
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.105290572215195e-05,
+      "loss": 0.5119,
+      "step": 30760
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1049929310216535e-05,
+      "loss": 0.4969,
+      "step": 30770
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.104695289828112e-05,
+      "loss": 0.5007,
+      "step": 30780
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1043976486345716e-05,
+      "loss": 0.4844,
+      "step": 30790
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.10410000744103e-05,
+      "loss": 0.4833,
+      "step": 30800
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.103802366247489e-05,
+      "loss": 0.4793,
+      "step": 30810
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1035047250539476e-05,
+      "loss": 0.4911,
+      "step": 30820
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.103207083860407e-05,
+      "loss": 0.4802,
+      "step": 30830
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.102909442666866e-05,
+      "loss": 0.4898,
+      "step": 30840
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.102611801473324e-05,
+      "loss": 0.4896,
+      "step": 30850
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.102314160279783e-05,
+      "loss": 0.5032,
+      "step": 30860
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.102016519086242e-05,
+      "loss": 0.4947,
+      "step": 30870
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1017188778927005e-05,
+      "loss": 0.5066,
+      "step": 30880
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.101421236699159e-05,
+      "loss": 0.4962,
+      "step": 30890
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1011235955056185e-05,
+      "loss": 0.4856,
+      "step": 30900
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.100825954312077e-05,
+      "loss": 0.5151,
+      "step": 30910
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.100528313118536e-05,
+      "loss": 0.4809,
+      "step": 30920
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1002306719249946e-05,
+      "loss": 0.4966,
+      "step": 30930
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.099933030731454e-05,
+      "loss": 0.5074,
+      "step": 30940
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.099635389537912e-05,
+      "loss": 0.4918,
+      "step": 30950
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.0993377483443707e-05,
+      "loss": 0.4994,
+      "step": 30960
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.09904010715083e-05,
+      "loss": 0.514,
+      "step": 30970
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.098742465957289e-05,
+      "loss": 0.4765,
+      "step": 30980
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.0984448247637474e-05,
+      "loss": 0.4852,
+      "step": 30990
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.098147183570206e-05,
+      "loss": 0.4985,
+      "step": 31000
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.0978495423766655e-05,
+      "loss": 0.5006,
+      "step": 31010
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.097551901183124e-05,
+      "loss": 0.5023,
+      "step": 31020
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.097254259989583e-05,
+      "loss": 0.5051,
+      "step": 31030
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.0969566187960415e-05,
+      "loss": 0.4941,
+      "step": 31040
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.0966589776025e-05,
+      "loss": 0.4932,
+      "step": 31050
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.096361336408959e-05,
+      "loss": 0.4836,
+      "step": 31060
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.096063695215418e-05,
+      "loss": 0.4836,
+      "step": 31070
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.095766054021877e-05,
+      "loss": 0.4875,
+      "step": 31080
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.0954684128283357e-05,
+      "loss": 0.4964,
+      "step": 31090
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0951707716347943e-05,
+      "loss": 0.5026,
+      "step": 31100
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.094873130441253e-05,
+      "loss": 0.511,
+      "step": 31110
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0945754892477124e-05,
+      "loss": 0.5002,
+      "step": 31120
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.094277848054171e-05,
+      "loss": 0.4813,
+      "step": 31130
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.09398020686063e-05,
+      "loss": 0.4931,
+      "step": 31140
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0936825656670885e-05,
+      "loss": 0.4988,
+      "step": 31150
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.093384924473547e-05,
+      "loss": 0.4832,
+      "step": 31160
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.093087283280006e-05,
+      "loss": 0.4851,
+      "step": 31170
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.092789642086465e-05,
+      "loss": 0.5112,
+      "step": 31180
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.092492000892924e-05,
+      "loss": 0.4997,
+      "step": 31190
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0921943596993826e-05,
+      "loss": 0.4684,
+      "step": 31200
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.091896718505841e-05,
+      "loss": 0.4945,
+      "step": 31210
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0915990773123007e-05,
+      "loss": 0.4951,
+      "step": 31220
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0913014361187593e-05,
+      "loss": 0.4794,
+      "step": 31230
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.091003794925218e-05,
+      "loss": 0.4725,
+      "step": 31240
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.090706153731677e-05,
+      "loss": 0.4896,
+      "step": 31250
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0904085125381354e-05,
+      "loss": 0.4949,
+      "step": 31260
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.090110871344594e-05,
+      "loss": 0.5032,
+      "step": 31270
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.089813230151053e-05,
+      "loss": 0.5003,
+      "step": 31280
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.089515588957512e-05,
+      "loss": 0.4998,
+      "step": 31290
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.089217947763971e-05,
+      "loss": 0.4896,
+      "step": 31300
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0889203065704295e-05,
+      "loss": 0.4934,
+      "step": 31310
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.088622665376888e-05,
+      "loss": 0.482,
+      "step": 31320
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0883250241833476e-05,
+      "loss": 0.4864,
+      "step": 31330
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.088027382989806e-05,
+      "loss": 0.4913,
+      "step": 31340
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.087729741796265e-05,
+      "loss": 0.4705,
+      "step": 31350
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.087432100602724e-05,
+      "loss": 0.4954,
+      "step": 31360
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0871344594091824e-05,
+      "loss": 0.4911,
+      "step": 31370
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.086836818215641e-05,
+      "loss": 0.4983,
+      "step": 31380
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0865391770221e-05,
+      "loss": 0.5001,
+      "step": 31390
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.086241535828559e-05,
+      "loss": 0.4947,
+      "step": 31400
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.085943894635018e-05,
+      "loss": 0.5049,
+      "step": 31410
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0856462534414765e-05,
+      "loss": 0.4898,
+      "step": 31420
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.085348612247935e-05,
+      "loss": 0.4807,
+      "step": 31430
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0850509710543945e-05,
+      "loss": 0.4874,
+      "step": 31440
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.084753329860853e-05,
+      "loss": 0.4953,
+      "step": 31450
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.084455688667312e-05,
+      "loss": 0.4981,
+      "step": 31460
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0841580474737706e-05,
+      "loss": 0.4919,
+      "step": 31470
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.083860406280229e-05,
+      "loss": 0.5035,
+      "step": 31480
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.083562765086688e-05,
+      "loss": 0.4773,
+      "step": 31490
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0832651238931474e-05,
+      "loss": 0.4661,
+      "step": 31500
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.082967482699606e-05,
+      "loss": 0.4993,
+      "step": 31510
+    },
+    {
+      "epoch": 7.0,
+      "eval_cer": 4.5747863247863245,
+      "eval_loss": 0.43723177909851074,
+      "eval_runtime": 1.3118,
+      "eval_samples_per_second": 7.623,
+      "eval_steps_per_second": 0.762,
+      "eval_wer": 0.6103896103896104,
+      "step": 31514
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.082669841506065e-05,
+      "loss": 0.4742,
+      "step": 31520
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0823722003125234e-05,
+      "loss": 0.4862,
+      "step": 31530
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.082074559118982e-05,
+      "loss": 0.4803,
+      "step": 31540
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0817769179254415e-05,
+      "loss": 0.4716,
+      "step": 31550
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0814792767318995e-05,
+      "loss": 0.4963,
+      "step": 31560
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.081181635538359e-05,
+      "loss": 0.473,
+      "step": 31570
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.0808839943448176e-05,
+      "loss": 0.4775,
+      "step": 31580
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.080586353151276e-05,
+      "loss": 0.471,
+      "step": 31590
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.080288711957735e-05,
+      "loss": 0.4772,
+      "step": 31600
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.079991070764194e-05,
+      "loss": 0.4756,
+      "step": 31610
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.079693429570653e-05,
+      "loss": 0.4837,
+      "step": 31620
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.079395788377112e-05,
+      "loss": 0.487,
+      "step": 31630
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.0790981471835704e-05,
+      "loss": 0.4868,
+      "step": 31640
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.07880050599003e-05,
+      "loss": 0.4776,
+      "step": 31650
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.0785028647964884e-05,
+      "loss": 0.4884,
+      "step": 31660
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.0782052236029464e-05,
+      "loss": 0.4923,
+      "step": 31670
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.077907582409406e-05,
+      "loss": 0.4789,
+      "step": 31680
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.0776099412158645e-05,
+      "loss": 0.4881,
+      "step": 31690
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.077312300022323e-05,
+      "loss": 0.4675,
+      "step": 31700
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.077014658828782e-05,
+      "loss": 0.4861,
+      "step": 31710
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.076717017635241e-05,
+      "loss": 0.4801,
+      "step": 31720
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.0764193764417e-05,
+      "loss": 0.4917,
+      "step": 31730
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.0761217352481586e-05,
+      "loss": 0.4886,
+      "step": 31740
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.075824094054617e-05,
+      "loss": 0.4638,
+      "step": 31750
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.075526452861077e-05,
+      "loss": 0.4677,
+      "step": 31760
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.075228811667535e-05,
+      "loss": 0.5037,
+      "step": 31770
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.0749311704739934e-05,
+      "loss": 0.4789,
+      "step": 31780
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.074633529280453e-05,
+      "loss": 0.4873,
+      "step": 31790
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.0743358880869114e-05,
+      "loss": 0.4768,
+      "step": 31800
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.07403824689337e-05,
+      "loss": 0.4561,
+      "step": 31810
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.073740605699829e-05,
+      "loss": 0.4852,
+      "step": 31820
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.073442964506288e-05,
+      "loss": 0.4632,
+      "step": 31830
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.073145323312747e-05,
+      "loss": 0.4803,
+      "step": 31840
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.0728476821192056e-05,
+      "loss": 0.4896,
+      "step": 31850
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.072550040925664e-05,
+      "loss": 0.4833,
+      "step": 31860
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.072252399732123e-05,
+      "loss": 0.4727,
+      "step": 31870
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.0719547585385816e-05,
+      "loss": 0.4709,
+      "step": 31880
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.071657117345041e-05,
+      "loss": 0.4732,
+      "step": 31890
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.0713594761515e-05,
+      "loss": 0.4839,
+      "step": 31900
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.0710618349579584e-05,
+      "loss": 0.4938,
+      "step": 31910
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.070764193764417e-05,
+      "loss": 0.4811,
+      "step": 31920
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.0704665525708764e-05,
+      "loss": 0.4878,
+      "step": 31930
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.070168911377335e-05,
+      "loss": 0.4791,
+      "step": 31940
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.069871270183794e-05,
+      "loss": 0.485,
+      "step": 31950
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.0695736289902525e-05,
+      "loss": 0.474,
+      "step": 31960
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.069275987796711e-05,
+      "loss": 0.4745,
+      "step": 31970
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.06897834660317e-05,
+      "loss": 0.4703,
+      "step": 31980
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.0686807054096286e-05,
+      "loss": 0.4842,
+      "step": 31990
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.068383064216088e-05,
+      "loss": 0.5012,
+      "step": 32000
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.0680854230225466e-05,
+      "loss": 0.4825,
+      "step": 32010
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.067787781829005e-05,
+      "loss": 0.4946,
+      "step": 32020
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.067490140635464e-05,
+      "loss": 0.4757,
+      "step": 32030
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.0671924994419234e-05,
+      "loss": 0.4673,
+      "step": 32040
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.066894858248382e-05,
+      "loss": 0.4634,
+      "step": 32050
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.066597217054841e-05,
+      "loss": 0.484,
+      "step": 32060
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.0662995758612995e-05,
+      "loss": 0.4911,
+      "step": 32070
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.066001934667758e-05,
+      "loss": 0.473,
+      "step": 32080
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.065704293474217e-05,
+      "loss": 0.4672,
+      "step": 32090
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.0654066522806755e-05,
+      "loss": 0.4678,
+      "step": 32100
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.065109011087135e-05,
+      "loss": 0.4761,
+      "step": 32110
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.0648113698935936e-05,
+      "loss": 0.4774,
+      "step": 32120
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.064513728700052e-05,
+      "loss": 0.4931,
+      "step": 32130
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.064216087506511e-05,
+      "loss": 0.4709,
+      "step": 32140
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.06391844631297e-05,
+      "loss": 0.4705,
+      "step": 32150
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.063620805119429e-05,
+      "loss": 0.4665,
+      "step": 32160
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.063323163925888e-05,
+      "loss": 0.4844,
+      "step": 32170
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.0630255227323464e-05,
+      "loss": 0.4721,
+      "step": 32180
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.062727881538805e-05,
+      "loss": 0.4757,
+      "step": 32190
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.062430240345264e-05,
+      "loss": 0.4728,
+      "step": 32200
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.0621325991517225e-05,
+      "loss": 0.4681,
+      "step": 32210
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.061834957958182e-05,
+      "loss": 0.4745,
+      "step": 32220
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.0615373167646405e-05,
+      "loss": 0.4779,
+      "step": 32230
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.061239675571099e-05,
+      "loss": 0.473,
+      "step": 32240
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.060942034377558e-05,
+      "loss": 0.4839,
+      "step": 32250
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.060644393184017e-05,
+      "loss": 0.4797,
+      "step": 32260
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.060346751990476e-05,
+      "loss": 0.4771,
+      "step": 32270
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.0600491107969347e-05,
+      "loss": 0.4883,
+      "step": 32280
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.0597514696033933e-05,
+      "loss": 0.4804,
+      "step": 32290
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.059453828409852e-05,
+      "loss": 0.4688,
+      "step": 32300
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.059156187216311e-05,
+      "loss": 0.4756,
+      "step": 32310
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.05885854602277e-05,
+      "loss": 0.4763,
+      "step": 32320
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.058560904829229e-05,
+      "loss": 0.488,
+      "step": 32330
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.0582632636356875e-05,
+      "loss": 0.4656,
+      "step": 32340
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.057965622442146e-05,
+      "loss": 0.4914,
+      "step": 32350
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.0576679812486055e-05,
+      "loss": 0.4652,
+      "step": 32360
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.057370340055064e-05,
+      "loss": 0.4816,
+      "step": 32370
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.057072698861522e-05,
+      "loss": 0.492,
+      "step": 32380
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.0567750576679816e-05,
+      "loss": 0.4731,
+      "step": 32390
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.05647741647444e-05,
+      "loss": 0.4618,
+      "step": 32400
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.056179775280899e-05,
+      "loss": 0.4796,
+      "step": 32410
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.055882134087358e-05,
+      "loss": 0.4934,
+      "step": 32420
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.055584492893817e-05,
+      "loss": 0.4678,
+      "step": 32430
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.055286851700276e-05,
+      "loss": 0.4741,
+      "step": 32440
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.0549892105067344e-05,
+      "loss": 0.4786,
+      "step": 32450
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.054691569313193e-05,
+      "loss": 0.484,
+      "step": 32460
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.0543939281196525e-05,
+      "loss": 0.4752,
+      "step": 32470
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.054096286926111e-05,
+      "loss": 0.4775,
+      "step": 32480
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.053798645732569e-05,
+      "loss": 0.4829,
+      "step": 32490
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.0535010045390285e-05,
+      "loss": 0.4679,
+      "step": 32500
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.053203363345487e-05,
+      "loss": 0.4943,
+      "step": 32510
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.052905722151946e-05,
+      "loss": 0.4686,
+      "step": 32520
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.0526080809584046e-05,
+      "loss": 0.4852,
+      "step": 32530
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.052310439764864e-05,
+      "loss": 0.4721,
+      "step": 32540
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.052012798571323e-05,
+      "loss": 0.495,
+      "step": 32550
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.0517151573777814e-05,
+      "loss": 0.4621,
+      "step": 32560
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.0514175161842404e-05,
+      "loss": 0.4824,
+      "step": 32570
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.051119874990699e-05,
+      "loss": 0.4792,
+      "step": 32580
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.0508222337971574e-05,
+      "loss": 0.4693,
+      "step": 32590
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.0505245926036165e-05,
+      "loss": 0.4876,
+      "step": 32600
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.050226951410075e-05,
+      "loss": 0.4888,
+      "step": 32610
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.0499293102165342e-05,
+      "loss": 0.5031,
+      "step": 32620
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.049631669022993e-05,
+      "loss": 0.4583,
+      "step": 32630
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.049334027829452e-05,
+      "loss": 0.4697,
+      "step": 32640
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.0490363866359106e-05,
+      "loss": 0.4717,
+      "step": 32650
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.0487387454423696e-05,
+      "loss": 0.4791,
+      "step": 32660
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.0484411042488283e-05,
+      "loss": 0.4701,
+      "step": 32670
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.0481434630552873e-05,
+      "loss": 0.4741,
+      "step": 32680
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.0478458218617457e-05,
+      "loss": 0.4628,
+      "step": 32690
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.0475481806682047e-05,
+      "loss": 0.4943,
+      "step": 32700
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.0472505394746634e-05,
+      "loss": 0.481,
+      "step": 32710
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.0469528982811224e-05,
+      "loss": 0.4584,
+      "step": 32720
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.046655257087581e-05,
+      "loss": 0.4817,
+      "step": 32730
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.04635761589404e-05,
+      "loss": 0.468,
+      "step": 32740
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.046059974700499e-05,
+      "loss": 0.4648,
+      "step": 32750
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.045762333506958e-05,
+      "loss": 0.4691,
+      "step": 32760
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.0454646923134166e-05,
+      "loss": 0.4663,
+      "step": 32770
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.0451670511198756e-05,
+      "loss": 0.5023,
+      "step": 32780
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.044869409926334e-05,
+      "loss": 0.475,
+      "step": 32790
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.0445717687327926e-05,
+      "loss": 0.4671,
+      "step": 32800
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.0442741275392517e-05,
+      "loss": 0.473,
+      "step": 32810
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.0439764863457103e-05,
+      "loss": 0.4994,
+      "step": 32820
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.0436788451521694e-05,
+      "loss": 0.4687,
+      "step": 32830
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.043381203958628e-05,
+      "loss": 0.4711,
+      "step": 32840
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.043083562765087e-05,
+      "loss": 0.4761,
+      "step": 32850
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.0427859215715458e-05,
+      "loss": 0.4755,
+      "step": 32860
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.0424882803780048e-05,
+      "loss": 0.4866,
+      "step": 32870
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.0421906391844635e-05,
+      "loss": 0.4839,
+      "step": 32880
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.0418929979909225e-05,
+      "loss": 0.4833,
+      "step": 32890
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.041595356797381e-05,
+      "loss": 0.481,
+      "step": 32900
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.0412977156038396e-05,
+      "loss": 0.478,
+      "step": 32910
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.0410000744102986e-05,
+      "loss": 0.4709,
+      "step": 32920
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.0407024332167573e-05,
+      "loss": 0.4791,
+      "step": 32930
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.0404047920232163e-05,
+      "loss": 0.4592,
+      "step": 32940
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.040107150829675e-05,
+      "loss": 0.4727,
+      "step": 32950
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.039809509636134e-05,
+      "loss": 0.4819,
+      "step": 32960
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.0395118684425927e-05,
+      "loss": 0.4629,
+      "step": 32970
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.0392142272490518e-05,
+      "loss": 0.4704,
+      "step": 32980
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.0389165860555104e-05,
+      "loss": 0.471,
+      "step": 32990
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.0386189448619688e-05,
+      "loss": 0.4778,
+      "step": 33000
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.0383213036684278e-05,
+      "loss": 0.466,
+      "step": 33010
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.0380236624748865e-05,
+      "loss": 0.4821,
+      "step": 33020
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.0377260212813455e-05,
+      "loss": 0.466,
+      "step": 33030
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.0374283800878042e-05,
+      "loss": 0.4673,
+      "step": 33040
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.0371307388942633e-05,
+      "loss": 0.4819,
+      "step": 33050
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.036833097700722e-05,
+      "loss": 0.4767,
+      "step": 33060
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.036535456507181e-05,
+      "loss": 0.4881,
+      "step": 33070
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.0362378153136397e-05,
+      "loss": 0.4854,
+      "step": 33080
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.0359401741200987e-05,
+      "loss": 0.4812,
+      "step": 33090
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.035642532926557e-05,
+      "loss": 0.4707,
+      "step": 33100
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.035344891733016e-05,
+      "loss": 0.465,
+      "step": 33110
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.0350472505394748e-05,
+      "loss": 0.4804,
+      "step": 33120
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.0347496093459338e-05,
+      "loss": 0.4686,
+      "step": 33130
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.0344519681523925e-05,
+      "loss": 0.468,
+      "step": 33140
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.0341543269588515e-05,
+      "loss": 0.4681,
+      "step": 33150
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.0338566857653102e-05,
+      "loss": 0.4674,
+      "step": 33160
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.0335590445717692e-05,
+      "loss": 0.4848,
+      "step": 33170
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.033261403378228e-05,
+      "loss": 0.4743,
+      "step": 33180
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.032963762184687e-05,
+      "loss": 0.4753,
+      "step": 33190
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.0326661209911453e-05,
+      "loss": 0.4632,
+      "step": 33200
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.032368479797604e-05,
+      "loss": 0.4777,
+      "step": 33210
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.032070838604063e-05,
+      "loss": 0.4693,
+      "step": 33220
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.0317731974105217e-05,
+      "loss": 0.4747,
+      "step": 33230
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.0314755562169807e-05,
+      "loss": 0.4673,
+      "step": 33240
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.0311779150234394e-05,
+      "loss": 0.4906,
+      "step": 33250
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.0308802738298985e-05,
+      "loss": 0.4715,
+      "step": 33260
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.030582632636357e-05,
+      "loss": 0.4748,
+      "step": 33270
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.0302849914428162e-05,
+      "loss": 0.4759,
+      "step": 33280
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.029987350249275e-05,
+      "loss": 0.481,
+      "step": 33290
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.0296897090557332e-05,
+      "loss": 0.4887,
+      "step": 33300
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.0293920678621923e-05,
+      "loss": 0.4807,
+      "step": 33310
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.029094426668651e-05,
+      "loss": 0.4767,
+      "step": 33320
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.02879678547511e-05,
+      "loss": 0.4806,
+      "step": 33330
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.0284991442815687e-05,
+      "loss": 0.4827,
+      "step": 33340
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.0282015030880277e-05,
+      "loss": 0.4724,
+      "step": 33350
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.0279038618944864e-05,
+      "loss": 0.4695,
+      "step": 33360
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.0276062207009454e-05,
+      "loss": 0.488,
+      "step": 33370
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.027308579507404e-05,
+      "loss": 0.4777,
+      "step": 33380
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.027010938313863e-05,
+      "loss": 0.4684,
+      "step": 33390
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.0267132971203218e-05,
+      "loss": 0.4655,
+      "step": 33400
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.02641565592678e-05,
+      "loss": 0.4716,
+      "step": 33410
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.0261180147332392e-05,
+      "loss": 0.4921,
+      "step": 33420
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.025820373539698e-05,
+      "loss": 0.451,
+      "step": 33430
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.025522732346157e-05,
+      "loss": 0.4674,
+      "step": 33440
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.0252250911526156e-05,
+      "loss": 0.4728,
+      "step": 33450
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.0249274499590746e-05,
+      "loss": 0.4675,
+      "step": 33460
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.0246298087655333e-05,
+      "loss": 0.4878,
+      "step": 33470
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.0243321675719924e-05,
+      "loss": 0.4643,
+      "step": 33480
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.024034526378451e-05,
+      "loss": 0.4606,
+      "step": 33490
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.02373688518491e-05,
+      "loss": 0.4682,
+      "step": 33500
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.0234392439913684e-05,
+      "loss": 0.4726,
+      "step": 33510
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.0231416027978275e-05,
+      "loss": 0.4861,
+      "step": 33520
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.022843961604286e-05,
+      "loss": 0.4878,
+      "step": 33530
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.022546320410745e-05,
+      "loss": 0.4715,
+      "step": 33540
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.022248679217204e-05,
+      "loss": 0.471,
+      "step": 33550
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.021951038023663e-05,
+      "loss": 0.47,
+      "step": 33560
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.0216533968301216e-05,
+      "loss": 0.484,
+      "step": 33570
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.0213557556365806e-05,
+      "loss": 0.458,
+      "step": 33580
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.0210581144430393e-05,
+      "loss": 0.4591,
+      "step": 33590
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.0207604732494983e-05,
+      "loss": 0.4636,
+      "step": 33600
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.0204628320559567e-05,
+      "loss": 0.4581,
+      "step": 33610
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.0201651908624154e-05,
+      "loss": 0.4818,
+      "step": 33620
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.0198675496688744e-05,
+      "loss": 0.4693,
+      "step": 33630
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.019569908475333e-05,
+      "loss": 0.4575,
+      "step": 33640
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.019272267281792e-05,
+      "loss": 0.4807,
+      "step": 33650
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.0189746260882508e-05,
+      "loss": 0.4647,
+      "step": 33660
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.01867698489471e-05,
+      "loss": 0.4641,
+      "step": 33670
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.0183793437011685e-05,
+      "loss": 0.4629,
+      "step": 33680
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.0180817025076276e-05,
+      "loss": 0.4663,
+      "step": 33690
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.0177840613140862e-05,
+      "loss": 0.4807,
+      "step": 33700
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.0174864201205446e-05,
+      "loss": 0.4732,
+      "step": 33710
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.0171887789270036e-05,
+      "loss": 0.4818,
+      "step": 33720
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.0168911377334623e-05,
+      "loss": 0.4713,
+      "step": 33730
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.0165934965399213e-05,
+      "loss": 0.4773,
+      "step": 33740
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.01629585534638e-05,
+      "loss": 0.4676,
+      "step": 33750
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.015998214152839e-05,
+      "loss": 0.4693,
+      "step": 33760
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.0157005729592977e-05,
+      "loss": 0.4559,
+      "step": 33770
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.0154029317657568e-05,
+      "loss": 0.4809,
+      "step": 33780
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.0151052905722155e-05,
+      "loss": 0.4632,
+      "step": 33790
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.0148076493786745e-05,
+      "loss": 0.4781,
+      "step": 33800
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.0145100081851332e-05,
+      "loss": 0.4712,
+      "step": 33810
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.0142123669915915e-05,
+      "loss": 0.4793,
+      "step": 33820
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.0139147257980506e-05,
+      "loss": 0.4701,
+      "step": 33830
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.0136170846045093e-05,
+      "loss": 0.4739,
+      "step": 33840
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.0133194434109683e-05,
+      "loss": 0.4676,
+      "step": 33850
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.013021802217427e-05,
+      "loss": 0.4828,
+      "step": 33860
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.012724161023886e-05,
+      "loss": 0.462,
+      "step": 33870
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.0124265198303447e-05,
+      "loss": 0.4706,
+      "step": 33880
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.0121288786368037e-05,
+      "loss": 0.4822,
+      "step": 33890
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.0118312374432624e-05,
+      "loss": 0.4811,
+      "step": 33900
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.0115335962497214e-05,
+      "loss": 0.4691,
+      "step": 33910
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.0112359550561798e-05,
+      "loss": 0.4613,
+      "step": 33920
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.0109383138626388e-05,
+      "loss": 0.4699,
+      "step": 33930
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.0106406726690975e-05,
+      "loss": 0.4686,
+      "step": 33940
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.0103430314755565e-05,
+      "loss": 0.4907,
+      "step": 33950
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.0100453902820152e-05,
+      "loss": 0.4677,
+      "step": 33960
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.0097477490884743e-05,
+      "loss": 0.4775,
+      "step": 33970
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.009450107894933e-05,
+      "loss": 0.4671,
+      "step": 33980
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.009152466701392e-05,
+      "loss": 0.4659,
+      "step": 33990
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.0088548255078507e-05,
+      "loss": 0.4614,
+      "step": 34000
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.0085571843143097e-05,
+      "loss": 0.4732,
+      "step": 34010
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.008259543120768e-05,
+      "loss": 0.4677,
+      "step": 34020
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.0079619019272267e-05,
+      "loss": 0.4627,
+      "step": 34030
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.0076642607336858e-05,
+      "loss": 0.4577,
+      "step": 34040
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.0073666195401445e-05,
+      "loss": 0.4653,
+      "step": 34050
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.0070689783466035e-05,
+      "loss": 0.4762,
+      "step": 34060
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.0067713371530622e-05,
+      "loss": 0.4734,
+      "step": 34070
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.0064736959595212e-05,
+      "loss": 0.4774,
+      "step": 34080
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.00617605476598e-05,
+      "loss": 0.4765,
+      "step": 34090
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.005878413572439e-05,
+      "loss": 0.4588,
+      "step": 34100
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.0055807723788976e-05,
+      "loss": 0.4577,
+      "step": 34110
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.005283131185356e-05,
+      "loss": 0.4644,
+      "step": 34120
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.004985489991815e-05,
+      "loss": 0.4666,
+      "step": 34130
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.0046878487982737e-05,
+      "loss": 0.482,
+      "step": 34140
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.0043902076047327e-05,
+      "loss": 0.4823,
+      "step": 34150
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.0040925664111914e-05,
+      "loss": 0.474,
+      "step": 34160
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.0037949252176504e-05,
+      "loss": 0.4802,
+      "step": 34170
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.003497284024109e-05,
+      "loss": 0.4665,
+      "step": 34180
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.003199642830568e-05,
+      "loss": 0.4681,
+      "step": 34190
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.002902001637027e-05,
+      "loss": 0.4662,
+      "step": 34200
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.002604360443486e-05,
+      "loss": 0.4677,
+      "step": 34210
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.0023067192499446e-05,
+      "loss": 0.4729,
+      "step": 34220
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.002009078056403e-05,
+      "loss": 0.4595,
+      "step": 34230
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.001711436862862e-05,
+      "loss": 0.4768,
+      "step": 34240
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.0014137956693206e-05,
+      "loss": 0.4621,
+      "step": 34250
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.0011161544757797e-05,
+      "loss": 0.4687,
+      "step": 34260
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.0008185132822383e-05,
+      "loss": 0.4712,
+      "step": 34270
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.0005208720886974e-05,
+      "loss": 0.4807,
+      "step": 34280
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.000223230895156e-05,
+      "loss": 0.4768,
+      "step": 34290
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.999925589701615e-05,
+      "loss": 0.4587,
+      "step": 34300
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.9996279485080738e-05,
+      "loss": 0.4671,
+      "step": 34310
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.9993303073145328e-05,
+      "loss": 0.4688,
+      "step": 34320
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.999032666120991e-05,
+      "loss": 0.4828,
+      "step": 34330
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.9987350249274502e-05,
+      "loss": 0.4629,
+      "step": 34340
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.998437383733909e-05,
+      "loss": 0.4608,
+      "step": 34350
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.998139742540368e-05,
+      "loss": 0.4725,
+      "step": 34360
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.9978421013468266e-05,
+      "loss": 0.4892,
+      "step": 34370
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.9975444601532856e-05,
+      "loss": 0.4807,
+      "step": 34380
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.9972468189597443e-05,
+      "loss": 0.4675,
+      "step": 34390
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.9969491777662033e-05,
+      "loss": 0.4606,
+      "step": 34400
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.996651536572662e-05,
+      "loss": 0.476,
+      "step": 34410
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.996353895379121e-05,
+      "loss": 0.4668,
+      "step": 34420
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.9960562541855794e-05,
+      "loss": 0.4558,
+      "step": 34430
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.995758612992038e-05,
+      "loss": 0.4758,
+      "step": 34440
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.995460971798497e-05,
+      "loss": 0.4604,
+      "step": 34450
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.9951633306049558e-05,
+      "loss": 0.4817,
+      "step": 34460
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.994865689411415e-05,
+      "loss": 0.471,
+      "step": 34470
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.9945680482178735e-05,
+      "loss": 0.4615,
+      "step": 34480
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.9942704070243326e-05,
+      "loss": 0.4838,
+      "step": 34490
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.9939727658307913e-05,
+      "loss": 0.4488,
+      "step": 34500
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.9936751246372503e-05,
+      "loss": 0.4642,
+      "step": 34510
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.993377483443709e-05,
+      "loss": 0.4702,
+      "step": 34520
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.9930798422501673e-05,
+      "loss": 0.4777,
+      "step": 34530
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.9927822010566264e-05,
+      "loss": 0.482,
+      "step": 34540
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.992484559863085e-05,
+      "loss": 0.475,
+      "step": 34550
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.992186918669544e-05,
+      "loss": 0.4703,
+      "step": 34560
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.9918892774760028e-05,
+      "loss": 0.4597,
+      "step": 34570
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.9915916362824618e-05,
+      "loss": 0.4589,
+      "step": 34580
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.9912939950889205e-05,
+      "loss": 0.4691,
+      "step": 34590
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.9909963538953795e-05,
+      "loss": 0.4598,
+      "step": 34600
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.9906987127018382e-05,
+      "loss": 0.4727,
+      "step": 34610
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.9904010715082972e-05,
+      "loss": 0.4779,
+      "step": 34620
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.990103430314756e-05,
+      "loss": 0.4703,
+      "step": 34630
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.9898057891212146e-05,
+      "loss": 0.4603,
+      "step": 34640
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.9895081479276733e-05,
+      "loss": 0.4611,
+      "step": 34650
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.989210506734132e-05,
+      "loss": 0.461,
+      "step": 34660
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.988912865540591e-05,
+      "loss": 0.4723,
+      "step": 34670
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.9886152243470497e-05,
+      "loss": 0.4703,
+      "step": 34680
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.9883175831535087e-05,
+      "loss": 0.469,
+      "step": 34690
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.9880199419599674e-05,
+      "loss": 0.4838,
+      "step": 34700
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.9877223007664265e-05,
+      "loss": 0.4589,
+      "step": 34710
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.987424659572885e-05,
+      "loss": 0.4762,
+      "step": 34720
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.9871270183793442e-05,
+      "loss": 0.4629,
+      "step": 34730
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.9868293771858025e-05,
+      "loss": 0.484,
+      "step": 34740
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.9865317359922616e-05,
+      "loss": 0.4574,
+      "step": 34750
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.9862340947987202e-05,
+      "loss": 0.4612,
+      "step": 34760
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.9859364536051793e-05,
+      "loss": 0.4733,
+      "step": 34770
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.985638812411638e-05,
+      "loss": 0.4735,
+      "step": 34780
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.985341171218097e-05,
+      "loss": 0.4649,
+      "step": 34790
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.9850435300245557e-05,
+      "loss": 0.4623,
+      "step": 34800
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.9847458888310147e-05,
+      "loss": 0.478,
+      "step": 34810
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.9844482476374734e-05,
+      "loss": 0.466,
+      "step": 34820
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.9841506064439324e-05,
+      "loss": 0.4662,
+      "step": 34830
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.9838529652503908e-05,
+      "loss": 0.4691,
+      "step": 34840
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.9835553240568495e-05,
+      "loss": 0.473,
+      "step": 34850
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.9832576828633085e-05,
+      "loss": 0.4576,
+      "step": 34860
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.9829600416697672e-05,
+      "loss": 0.4576,
+      "step": 34870
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.9826624004762262e-05,
+      "loss": 0.4723,
+      "step": 34880
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.982364759282685e-05,
+      "loss": 0.4663,
+      "step": 34890
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.982067118089144e-05,
+      "loss": 0.4471,
+      "step": 34900
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.9817694768956026e-05,
+      "loss": 0.4638,
+      "step": 34910
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.9814718357020617e-05,
+      "loss": 0.4698,
+      "step": 34920
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.9811741945085203e-05,
+      "loss": 0.4626,
+      "step": 34930
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.9808765533149787e-05,
+      "loss": 0.4748,
+      "step": 34940
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.9805789121214377e-05,
+      "loss": 0.4578,
+      "step": 34950
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.9802812709278964e-05,
+      "loss": 0.4614,
+      "step": 34960
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.9799836297343554e-05,
+      "loss": 0.4766,
+      "step": 34970
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.979685988540814e-05,
+      "loss": 0.4817,
+      "step": 34980
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.979388347347273e-05,
+      "loss": 0.4527,
+      "step": 34990
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.979090706153732e-05,
+      "loss": 0.469,
+      "step": 35000
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.978793064960191e-05,
+      "loss": 0.4847,
+      "step": 35010
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.9784954237666496e-05,
+      "loss": 0.4706,
+      "step": 35020
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.9781977825731086e-05,
+      "loss": 0.4682,
+      "step": 35030
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.977900141379567e-05,
+      "loss": 0.4647,
+      "step": 35040
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.977602500186026e-05,
+      "loss": 0.4708,
+      "step": 35050
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.9773048589924847e-05,
+      "loss": 0.4812,
+      "step": 35060
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.9770072177989437e-05,
+      "loss": 0.4525,
+      "step": 35070
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.9767095766054024e-05,
+      "loss": 0.4721,
+      "step": 35080
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.976411935411861e-05,
+      "loss": 0.4599,
+      "step": 35090
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.97611429421832e-05,
+      "loss": 0.4826,
+      "step": 35100
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.9758166530247788e-05,
+      "loss": 0.4611,
+      "step": 35110
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.9755190118312378e-05,
+      "loss": 0.4575,
+      "step": 35120
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.9752213706376965e-05,
+      "loss": 0.4789,
+      "step": 35130
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.9749237294441555e-05,
+      "loss": 0.4682,
+      "step": 35140
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.974626088250614e-05,
+      "loss": 0.4729,
+      "step": 35150
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.974328447057073e-05,
+      "loss": 0.464,
+      "step": 35160
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.9740308058635316e-05,
+      "loss": 0.4804,
+      "step": 35170
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.9737331646699906e-05,
+      "loss": 0.469,
+      "step": 35180
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.9734355234764493e-05,
+      "loss": 0.4783,
+      "step": 35190
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.9731378822829084e-05,
+      "loss": 0.4797,
+      "step": 35200
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.972840241089367e-05,
+      "loss": 0.4629,
+      "step": 35210
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.972542599895826e-05,
+      "loss": 0.4737,
+      "step": 35220
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.9722449587022848e-05,
+      "loss": 0.4762,
+      "step": 35230
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.9719473175087438e-05,
+      "loss": 0.4692,
+      "step": 35240
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.971649676315202e-05,
+      "loss": 0.474,
+      "step": 35250
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.971352035121661e-05,
+      "loss": 0.4473,
+      "step": 35260
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.97105439392812e-05,
+      "loss": 0.4655,
+      "step": 35270
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.9707567527345786e-05,
+      "loss": 0.46,
+      "step": 35280
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.9704591115410376e-05,
+      "loss": 0.4722,
+      "step": 35290
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.9701614703474963e-05,
+      "loss": 0.4869,
+      "step": 35300
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.9698638291539553e-05,
+      "loss": 0.472,
+      "step": 35310
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.969566187960414e-05,
+      "loss": 0.4684,
+      "step": 35320
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.969268546766873e-05,
+      "loss": 0.4601,
+      "step": 35330
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.9689709055733317e-05,
+      "loss": 0.4764,
+      "step": 35340
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.96867326437979e-05,
+      "loss": 0.4571,
+      "step": 35350
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.968375623186249e-05,
+      "loss": 0.4537,
+      "step": 35360
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.9680779819927078e-05,
+      "loss": 0.463,
+      "step": 35370
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.9677803407991668e-05,
+      "loss": 0.456,
+      "step": 35380
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.9674826996056255e-05,
+      "loss": 0.4781,
+      "step": 35390
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.9671850584120845e-05,
+      "loss": 0.4602,
+      "step": 35400
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.9668874172185432e-05,
+      "loss": 0.4595,
+      "step": 35410
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.9665897760250022e-05,
+      "loss": 0.4463,
+      "step": 35420
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.966292134831461e-05,
+      "loss": 0.4858,
+      "step": 35430
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.96599449363792e-05,
+      "loss": 0.4807,
+      "step": 35440
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.9656968524443783e-05,
+      "loss": 0.48,
+      "step": 35450
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.9653992112508373e-05,
+      "loss": 0.4499,
+      "step": 35460
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.965101570057296e-05,
+      "loss": 0.455,
+      "step": 35470
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.964803928863755e-05,
+      "loss": 0.472,
+      "step": 35480
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.9645062876702138e-05,
+      "loss": 0.467,
+      "step": 35490
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.9642086464766728e-05,
+      "loss": 0.4612,
+      "step": 35500
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.9639110052831315e-05,
+      "loss": 0.468,
+      "step": 35510
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.96361336408959e-05,
+      "loss": 0.4813,
+      "step": 35520
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.9633157228960492e-05,
+      "loss": 0.4664,
+      "step": 35530
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.963018081702508e-05,
+      "loss": 0.4562,
+      "step": 35540
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.962720440508967e-05,
+      "loss": 0.4773,
+      "step": 35550
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.9624227993154253e-05,
+      "loss": 0.4724,
+      "step": 35560
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.9621251581218843e-05,
+      "loss": 0.4447,
+      "step": 35570
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.961827516928343e-05,
+      "loss": 0.4574,
+      "step": 35580
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.961529875734802e-05,
+      "loss": 0.4771,
+      "step": 35590
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.9612322345412607e-05,
+      "loss": 0.4681,
+      "step": 35600
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.9609345933477197e-05,
+      "loss": 0.4609,
+      "step": 35610
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.9606369521541784e-05,
+      "loss": 0.4751,
+      "step": 35620
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.9603393109606374e-05,
+      "loss": 0.4745,
+      "step": 35630
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.960041669767096e-05,
+      "loss": 0.4583,
+      "step": 35640
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.959744028573555e-05,
+      "loss": 0.4549,
+      "step": 35650
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.9594463873800135e-05,
+      "loss": 0.4609,
+      "step": 35660
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.9591487461864722e-05,
+      "loss": 0.4383,
+      "step": 35670
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.9588511049929312e-05,
+      "loss": 0.4444,
+      "step": 35680
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.95855346379939e-05,
+      "loss": 0.4709,
+      "step": 35690
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.958255822605849e-05,
+      "loss": 0.476,
+      "step": 35700
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.9579581814123076e-05,
+      "loss": 0.4635,
+      "step": 35710
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.9576605402187667e-05,
+      "loss": 0.4656,
+      "step": 35720
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.9573628990252254e-05,
+      "loss": 0.4746,
+      "step": 35730
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.9570652578316844e-05,
+      "loss": 0.4589,
+      "step": 35740
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.956767616638143e-05,
+      "loss": 0.4445,
+      "step": 35750
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.9564699754446014e-05,
+      "loss": 0.4672,
+      "step": 35760
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.9561723342510605e-05,
+      "loss": 0.4711,
+      "step": 35770
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.955874693057519e-05,
+      "loss": 0.4623,
+      "step": 35780
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.9555770518639782e-05,
+      "loss": 0.4549,
+      "step": 35790
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.955279410670437e-05,
+      "loss": 0.4606,
+      "step": 35800
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.954981769476896e-05,
+      "loss": 0.4781,
+      "step": 35810
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.9546841282833546e-05,
+      "loss": 0.4632,
+      "step": 35820
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.9543864870898136e-05,
+      "loss": 0.4666,
+      "step": 35830
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.9540888458962723e-05,
+      "loss": 0.4761,
+      "step": 35840
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.9537912047027313e-05,
+      "loss": 0.4527,
+      "step": 35850
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.9534935635091897e-05,
+      "loss": 0.4457,
+      "step": 35860
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.9531959223156487e-05,
+      "loss": 0.4598,
+      "step": 35870
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.9528982811221074e-05,
+      "loss": 0.4524,
+      "step": 35880
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.9526006399285664e-05,
+      "loss": 0.4626,
+      "step": 35890
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.952302998735025e-05,
+      "loss": 0.4728,
+      "step": 35900
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.952005357541484e-05,
+      "loss": 0.4888,
+      "step": 35910
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.951707716347943e-05,
+      "loss": 0.4715,
+      "step": 35920
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.951410075154402e-05,
+      "loss": 0.457,
+      "step": 35930
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.9511124339608606e-05,
+      "loss": 0.4643,
+      "step": 35940
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.9508147927673192e-05,
+      "loss": 0.4732,
+      "step": 35950
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.9505171515737783e-05,
+      "loss": 0.4572,
+      "step": 35960
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.9502195103802366e-05,
+      "loss": 0.4711,
+      "step": 35970
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.9499218691866957e-05,
+      "loss": 0.4641,
+      "step": 35980
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.9496242279931543e-05,
+      "loss": 0.4738,
+      "step": 35990
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.9493265867996134e-05,
+      "loss": 0.4382,
+      "step": 36000
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.949028945606072e-05,
+      "loss": 0.4734,
+      "step": 36010
+    },
+    {
+      "epoch": 8.0,
+      "eval_cer": 4.564102564102564,
+      "eval_loss": 0.4123542904853821,
+      "eval_runtime": 1.3042,
+      "eval_samples_per_second": 7.668,
+      "eval_steps_per_second": 0.767,
+      "eval_wer": 0.6103896103896104,
+      "step": 36016
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.948731304412531e-05,
+      "loss": 0.4509,
+      "step": 36020
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.9484336632189898e-05,
+      "loss": 0.4696,
+      "step": 36030
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.9481360220254488e-05,
+      "loss": 0.4508,
+      "step": 36040
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.9478383808319075e-05,
+      "loss": 0.4654,
+      "step": 36050
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.9475407396383665e-05,
+      "loss": 0.453,
+      "step": 36060
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.947243098444825e-05,
+      "loss": 0.4369,
+      "step": 36070
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.9469454572512836e-05,
+      "loss": 0.4439,
+      "step": 36080
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 2.9466478160577426e-05,
+      "loss": 0.4549,
+      "step": 36090
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.9463501748642013e-05,
+      "loss": 0.4425,
+      "step": 36100
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.9460525336706603e-05,
+      "loss": 0.4638,
+      "step": 36110
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.945754892477119e-05,
+      "loss": 0.4426,
+      "step": 36120
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 2.945457251283578e-05,
+      "loss": 0.4647,
+      "step": 36130
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9451596100900367e-05,
+      "loss": 0.4472,
+      "step": 36140
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9448619688964958e-05,
+      "loss": 0.4495,
+      "step": 36150
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9445643277029544e-05,
+      "loss": 0.4401,
+      "step": 36160
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9442666865094128e-05,
+      "loss": 0.4595,
+      "step": 36170
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9439690453158718e-05,
+      "loss": 0.4447,
+      "step": 36180
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9436714041223305e-05,
+      "loss": 0.4453,
+      "step": 36190
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9433737629287895e-05,
+      "loss": 0.454,
+      "step": 36200
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9430761217352482e-05,
+      "loss": 0.4514,
+      "step": 36210
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9427784805417073e-05,
+      "loss": 0.4409,
+      "step": 36220
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.942480839348166e-05,
+      "loss": 0.4504,
+      "step": 36230
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.942183198154625e-05,
+      "loss": 0.4705,
+      "step": 36240
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9418855569610837e-05,
+      "loss": 0.4522,
+      "step": 36250
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9415879157675427e-05,
+      "loss": 0.4408,
+      "step": 36260
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.941290274574001e-05,
+      "loss": 0.4262,
+      "step": 36270
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.94099263338046e-05,
+      "loss": 0.4756,
+      "step": 36280
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9406949921869188e-05,
+      "loss": 0.4317,
+      "step": 36290
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9403973509933778e-05,
+      "loss": 0.4529,
+      "step": 36300
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9400997097998365e-05,
+      "loss": 0.4526,
+      "step": 36310
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9398020686062955e-05,
+      "loss": 0.4521,
+      "step": 36320
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9395044274127542e-05,
+      "loss": 0.4478,
+      "step": 36330
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9392067862192132e-05,
+      "loss": 0.4575,
+      "step": 36340
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.938909145025672e-05,
+      "loss": 0.4476,
+      "step": 36350
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.938611503832131e-05,
+      "loss": 0.4575,
+      "step": 36360
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9383138626385896e-05,
+      "loss": 0.4511,
+      "step": 36370
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.938016221445048e-05,
+      "loss": 0.4364,
+      "step": 36380
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.937718580251507e-05,
+      "loss": 0.4526,
+      "step": 36390
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9374209390579657e-05,
+      "loss": 0.4598,
+      "step": 36400
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9371232978644247e-05,
+      "loss": 0.4737,
+      "step": 36410
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9368256566708834e-05,
+      "loss": 0.4603,
+      "step": 36420
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9365280154773425e-05,
+      "loss": 0.4625,
+      "step": 36430
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.936230374283801e-05,
+      "loss": 0.4526,
+      "step": 36440
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9359327330902602e-05,
+      "loss": 0.4374,
+      "step": 36450
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.935635091896719e-05,
+      "loss": 0.4497,
+      "step": 36460
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.935337450703178e-05,
+      "loss": 0.4731,
+      "step": 36470
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9350398095096362e-05,
+      "loss": 0.4492,
+      "step": 36480
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.934742168316095e-05,
+      "loss": 0.4517,
+      "step": 36490
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.934444527122554e-05,
+      "loss": 0.4515,
+      "step": 36500
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9341468859290127e-05,
+      "loss": 0.4678,
+      "step": 36510
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9338492447354717e-05,
+      "loss": 0.4534,
+      "step": 36520
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9335516035419304e-05,
+      "loss": 0.4576,
+      "step": 36530
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9332539623483894e-05,
+      "loss": 0.4448,
+      "step": 36540
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.932956321154848e-05,
+      "loss": 0.4384,
+      "step": 36550
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.932658679961307e-05,
+      "loss": 0.4752,
+      "step": 36560
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9323610387677658e-05,
+      "loss": 0.4529,
+      "step": 36570
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.932063397574224e-05,
+      "loss": 0.4536,
+      "step": 36580
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9317657563806832e-05,
+      "loss": 0.448,
+      "step": 36590
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.931468115187142e-05,
+      "loss": 0.4681,
+      "step": 36600
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.931170473993601e-05,
+      "loss": 0.4588,
+      "step": 36610
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9308728328000596e-05,
+      "loss": 0.4434,
+      "step": 36620
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9305751916065186e-05,
+      "loss": 0.4483,
+      "step": 36630
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9302775504129773e-05,
+      "loss": 0.4462,
+      "step": 36640
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9299799092194363e-05,
+      "loss": 0.4511,
+      "step": 36650
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.929682268025895e-05,
+      "loss": 0.4608,
+      "step": 36660
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.929384626832354e-05,
+      "loss": 0.4715,
+      "step": 36670
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9290869856388124e-05,
+      "loss": 0.4489,
+      "step": 36680
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9287893444452714e-05,
+      "loss": 0.4478,
+      "step": 36690
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.92849170325173e-05,
+      "loss": 0.4369,
+      "step": 36700
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.928194062058189e-05,
+      "loss": 0.4548,
+      "step": 36710
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.927896420864648e-05,
+      "loss": 0.4668,
+      "step": 36720
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.927598779671107e-05,
+      "loss": 0.4475,
+      "step": 36730
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9273011384775656e-05,
+      "loss": 0.4423,
+      "step": 36740
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9270034972840246e-05,
+      "loss": 0.4512,
+      "step": 36750
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9267058560904833e-05,
+      "loss": 0.4579,
+      "step": 36760
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9264082148969423e-05,
+      "loss": 0.458,
+      "step": 36770
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9261105737034007e-05,
+      "loss": 0.4621,
+      "step": 36780
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9258129325098594e-05,
+      "loss": 0.4583,
+      "step": 36790
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9255152913163184e-05,
+      "loss": 0.4556,
+      "step": 36800
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.925217650122777e-05,
+      "loss": 0.451,
+      "step": 36810
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.924920008929236e-05,
+      "loss": 0.4721,
+      "step": 36820
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9246223677356948e-05,
+      "loss": 0.4521,
+      "step": 36830
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9243247265421538e-05,
+      "loss": 0.4433,
+      "step": 36840
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9240270853486125e-05,
+      "loss": 0.447,
+      "step": 36850
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9237294441550715e-05,
+      "loss": 0.4517,
+      "step": 36860
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9234318029615302e-05,
+      "loss": 0.4603,
+      "step": 36870
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9231341617679893e-05,
+      "loss": 0.4499,
+      "step": 36880
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9228365205744476e-05,
+      "loss": 0.4616,
+      "step": 36890
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9225388793809063e-05,
+      "loss": 0.46,
+      "step": 36900
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9222412381873653e-05,
+      "loss": 0.4528,
+      "step": 36910
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.921943596993824e-05,
+      "loss": 0.4416,
+      "step": 36920
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.921645955800283e-05,
+      "loss": 0.4398,
+      "step": 36930
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9213483146067417e-05,
+      "loss": 0.4553,
+      "step": 36940
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9210506734132008e-05,
+      "loss": 0.4322,
+      "step": 36950
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9207530322196595e-05,
+      "loss": 0.4604,
+      "step": 36960
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9204553910261185e-05,
+      "loss": 0.4505,
+      "step": 36970
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9201577498325772e-05,
+      "loss": 0.4455,
+      "step": 36980
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9198601086390355e-05,
+      "loss": 0.4498,
+      "step": 36990
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9195624674454946e-05,
+      "loss": 0.463,
+      "step": 37000
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9192648262519533e-05,
+      "loss": 0.464,
+      "step": 37010
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9189671850584123e-05,
+      "loss": 0.4764,
+      "step": 37020
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.918669543864871e-05,
+      "loss": 0.4482,
+      "step": 37030
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.91837190267133e-05,
+      "loss": 0.446,
+      "step": 37040
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9180742614777887e-05,
+      "loss": 0.4485,
+      "step": 37050
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9177766202842477e-05,
+      "loss": 0.4491,
+      "step": 37060
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9174789790907064e-05,
+      "loss": 0.4545,
+      "step": 37070
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9171813378971654e-05,
+      "loss": 0.4481,
+      "step": 37080
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9168836967036238e-05,
+      "loss": 0.4561,
+      "step": 37090
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9165860555100828e-05,
+      "loss": 0.4566,
+      "step": 37100
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9162884143165415e-05,
+      "loss": 0.4706,
+      "step": 37110
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9159907731230005e-05,
+      "loss": 0.4573,
+      "step": 37120
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.9156931319294592e-05,
+      "loss": 0.4489,
+      "step": 37130
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.9153954907359183e-05,
+      "loss": 0.463,
+      "step": 37140
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.915097849542377e-05,
+      "loss": 0.449,
+      "step": 37150
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.914800208348836e-05,
+      "loss": 0.4443,
+      "step": 37160
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.9145025671552947e-05,
+      "loss": 0.4495,
+      "step": 37170
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9142049259617537e-05,
+      "loss": 0.4545,
+      "step": 37180
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.913907284768212e-05,
+      "loss": 0.4672,
+      "step": 37190
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9136096435746707e-05,
+      "loss": 0.4419,
+      "step": 37200
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9133120023811298e-05,
+      "loss": 0.4288,
+      "step": 37210
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9130143611875884e-05,
+      "loss": 0.4634,
+      "step": 37220
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9127167199940475e-05,
+      "loss": 0.4426,
+      "step": 37230
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.912419078800506e-05,
+      "loss": 0.4655,
+      "step": 37240
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9121214376069652e-05,
+      "loss": 0.452,
+      "step": 37250
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.911823796413424e-05,
+      "loss": 0.4503,
+      "step": 37260
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.911526155219883e-05,
+      "loss": 0.4563,
+      "step": 37270
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9112285140263416e-05,
+      "loss": 0.4605,
+      "step": 37280
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9109308728328006e-05,
+      "loss": 0.4518,
+      "step": 37290
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.910633231639259e-05,
+      "loss": 0.4458,
+      "step": 37300
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9103355904457177e-05,
+      "loss": 0.4554,
+      "step": 37310
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9100379492521767e-05,
+      "loss": 0.4387,
+      "step": 37320
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9097403080586354e-05,
+      "loss": 0.4565,
+      "step": 37330
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9094426668650944e-05,
+      "loss": 0.4487,
+      "step": 37340
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.909145025671553e-05,
+      "loss": 0.4619,
+      "step": 37350
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.908847384478012e-05,
+      "loss": 0.4385,
+      "step": 37360
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9085497432844708e-05,
+      "loss": 0.4563,
+      "step": 37370
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.90825210209093e-05,
+      "loss": 0.4487,
+      "step": 37380
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9079544608973885e-05,
+      "loss": 0.4456,
+      "step": 37390
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.907656819703847e-05,
+      "loss": 0.4486,
+      "step": 37400
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.907359178510306e-05,
+      "loss": 0.4478,
+      "step": 37410
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9070615373167646e-05,
+      "loss": 0.452,
+      "step": 37420
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9067638961232236e-05,
+      "loss": 0.4463,
+      "step": 37430
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9064662549296823e-05,
+      "loss": 0.4388,
+      "step": 37440
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9061686137361414e-05,
+      "loss": 0.4401,
+      "step": 37450
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9058709725426e-05,
+      "loss": 0.447,
+      "step": 37460
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.905573331349059e-05,
+      "loss": 0.4519,
+      "step": 37470
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9052756901555178e-05,
+      "loss": 0.4407,
+      "step": 37480
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9049780489619768e-05,
+      "loss": 0.4457,
+      "step": 37490
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.904680407768435e-05,
+      "loss": 0.4439,
+      "step": 37500
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9043827665748942e-05,
+      "loss": 0.4354,
+      "step": 37510
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.904085125381353e-05,
+      "loss": 0.4495,
+      "step": 37520
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.903787484187812e-05,
+      "loss": 0.4372,
+      "step": 37530
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9034898429942706e-05,
+      "loss": 0.4361,
+      "step": 37540
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9031922018007296e-05,
+      "loss": 0.4657,
+      "step": 37550
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9028945606071883e-05,
+      "loss": 0.4479,
+      "step": 37560
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9025969194136473e-05,
+      "loss": 0.4453,
+      "step": 37570
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.902299278220106e-05,
+      "loss": 0.4598,
+      "step": 37580
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.902001637026565e-05,
+      "loss": 0.4526,
+      "step": 37590
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9017039958330234e-05,
+      "loss": 0.4439,
+      "step": 37600
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.901406354639482e-05,
+      "loss": 0.4564,
+      "step": 37610
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.901108713445941e-05,
+      "loss": 0.4643,
+      "step": 37620
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9008110722523998e-05,
+      "loss": 0.4415,
+      "step": 37630
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.900513431058859e-05,
+      "loss": 0.4513,
+      "step": 37640
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9002157898653175e-05,
+      "loss": 0.4474,
+      "step": 37650
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.8999181486717766e-05,
+      "loss": 0.4384,
+      "step": 37660
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.8996205074782353e-05,
+      "loss": 0.4509,
+      "step": 37670
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.8993228662846943e-05,
+      "loss": 0.4506,
+      "step": 37680
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.899025225091153e-05,
+      "loss": 0.4389,
+      "step": 37690
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.898727583897612e-05,
+      "loss": 0.448,
+      "step": 37700
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.8984299427040704e-05,
+      "loss": 0.4509,
+      "step": 37710
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.898132301510529e-05,
+      "loss": 0.4619,
+      "step": 37720
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.897834660316988e-05,
+      "loss": 0.4495,
+      "step": 37730
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.8975370191234468e-05,
+      "loss": 0.4539,
+      "step": 37740
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.8972393779299058e-05,
+      "loss": 0.4516,
+      "step": 37750
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.8969417367363645e-05,
+      "loss": 0.452,
+      "step": 37760
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.8966440955428235e-05,
+      "loss": 0.4392,
+      "step": 37770
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.8963464543492822e-05,
+      "loss": 0.4569,
+      "step": 37780
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.8960488131557412e-05,
+      "loss": 0.444,
+      "step": 37790
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.8957511719622e-05,
+      "loss": 0.4428,
+      "step": 37800
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.8954535307686583e-05,
+      "loss": 0.4541,
+      "step": 37810
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.8951558895751173e-05,
+      "loss": 0.4502,
+      "step": 37820
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.894858248381576e-05,
+      "loss": 0.4463,
+      "step": 37830
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.894560607188035e-05,
+      "loss": 0.448,
+      "step": 37840
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.8942629659944937e-05,
+      "loss": 0.4397,
+      "step": 37850
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.8939653248009527e-05,
+      "loss": 0.4363,
+      "step": 37860
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.8936676836074114e-05,
+      "loss": 0.4648,
+      "step": 37870
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.8933700424138705e-05,
+      "loss": 0.4404,
+      "step": 37880
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.893072401220329e-05,
+      "loss": 0.4534,
+      "step": 37890
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.8927747600267882e-05,
+      "loss": 0.4593,
+      "step": 37900
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.8924771188332465e-05,
+      "loss": 0.4451,
+      "step": 37910
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.8921794776397055e-05,
+      "loss": 0.4548,
+      "step": 37920
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.8918818364461642e-05,
+      "loss": 0.4495,
+      "step": 37930
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8915841952526233e-05,
+      "loss": 0.4379,
+      "step": 37940
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.891286554059082e-05,
+      "loss": 0.4552,
+      "step": 37950
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.890988912865541e-05,
+      "loss": 0.4459,
+      "step": 37960
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8906912716719997e-05,
+      "loss": 0.4393,
+      "step": 37970
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8903936304784587e-05,
+      "loss": 0.4355,
+      "step": 37980
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8900959892849174e-05,
+      "loss": 0.4448,
+      "step": 37990
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8897983480913764e-05,
+      "loss": 0.449,
+      "step": 38000
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8895007068978348e-05,
+      "loss": 0.4353,
+      "step": 38010
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8892030657042935e-05,
+      "loss": 0.4519,
+      "step": 38020
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8889054245107525e-05,
+      "loss": 0.4317,
+      "step": 38030
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8886077833172112e-05,
+      "loss": 0.4294,
+      "step": 38040
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8883101421236702e-05,
+      "loss": 0.462,
+      "step": 38050
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.888012500930129e-05,
+      "loss": 0.4416,
+      "step": 38060
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.887714859736588e-05,
+      "loss": 0.4507,
+      "step": 38070
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8874172185430466e-05,
+      "loss": 0.4411,
+      "step": 38080
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8871195773495056e-05,
+      "loss": 0.4504,
+      "step": 38090
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8868219361559643e-05,
+      "loss": 0.4442,
+      "step": 38100
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8865242949624234e-05,
+      "loss": 0.4529,
+      "step": 38110
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8862266537688817e-05,
+      "loss": 0.4555,
+      "step": 38120
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8859290125753404e-05,
+      "loss": 0.4356,
+      "step": 38130
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8856313713817994e-05,
+      "loss": 0.4429,
+      "step": 38140
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.885333730188258e-05,
+      "loss": 0.471,
+      "step": 38150
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.885036088994717e-05,
+      "loss": 0.4327,
+      "step": 38160
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.884738447801176e-05,
+      "loss": 0.4439,
+      "step": 38170
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.884440806607635e-05,
+      "loss": 0.4413,
+      "step": 38180
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8841431654140936e-05,
+      "loss": 0.446,
+      "step": 38190
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8838455242205526e-05,
+      "loss": 0.442,
+      "step": 38200
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8835478830270113e-05,
+      "loss": 0.4375,
+      "step": 38210
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8832502418334696e-05,
+      "loss": 0.4428,
+      "step": 38220
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8829526006399287e-05,
+      "loss": 0.4445,
+      "step": 38230
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8826549594463874e-05,
+      "loss": 0.4495,
+      "step": 38240
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8823573182528464e-05,
+      "loss": 0.4458,
+      "step": 38250
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.882059677059305e-05,
+      "loss": 0.4602,
+      "step": 38260
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.881762035865764e-05,
+      "loss": 0.4487,
+      "step": 38270
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8814643946722228e-05,
+      "loss": 0.4465,
+      "step": 38280
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8811667534786818e-05,
+      "loss": 0.4481,
+      "step": 38290
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8808691122851405e-05,
+      "loss": 0.4637,
+      "step": 38300
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8805714710915995e-05,
+      "loss": 0.4587,
+      "step": 38310
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.880273829898058e-05,
+      "loss": 0.4481,
+      "step": 38320
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.879976188704517e-05,
+      "loss": 0.4317,
+      "step": 38330
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8796785475109756e-05,
+      "loss": 0.4491,
+      "step": 38340
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8793809063174346e-05,
+      "loss": 0.4456,
+      "step": 38350
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8790832651238933e-05,
+      "loss": 0.454,
+      "step": 38360
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8787856239303524e-05,
+      "loss": 0.4414,
+      "step": 38370
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.878487982736811e-05,
+      "loss": 0.4404,
+      "step": 38380
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.87819034154327e-05,
+      "loss": 0.4526,
+      "step": 38390
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8778927003497288e-05,
+      "loss": 0.4423,
+      "step": 38400
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8775950591561878e-05,
+      "loss": 0.4344,
+      "step": 38410
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.877297417962646e-05,
+      "loss": 0.4531,
+      "step": 38420
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.876999776769105e-05,
+      "loss": 0.4418,
+      "step": 38430
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.876702135575564e-05,
+      "loss": 0.454,
+      "step": 38440
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8764044943820226e-05,
+      "loss": 0.4422,
+      "step": 38450
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8761068531884816e-05,
+      "loss": 0.4442,
+      "step": 38460
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8758092119949403e-05,
+      "loss": 0.4396,
+      "step": 38470
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8755115708013993e-05,
+      "loss": 0.4571,
+      "step": 38480
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.875213929607858e-05,
+      "loss": 0.4561,
+      "step": 38490
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.874916288414317e-05,
+      "loss": 0.4458,
+      "step": 38500
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8746186472207757e-05,
+      "loss": 0.4484,
+      "step": 38510
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.874321006027234e-05,
+      "loss": 0.4473,
+      "step": 38520
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.874023364833693e-05,
+      "loss": 0.4391,
+      "step": 38530
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8737257236401518e-05,
+      "loss": 0.4492,
+      "step": 38540
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8734280824466108e-05,
+      "loss": 0.4496,
+      "step": 38550
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8731304412530695e-05,
+      "loss": 0.438,
+      "step": 38560
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8728328000595285e-05,
+      "loss": 0.4412,
+      "step": 38570
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8725351588659872e-05,
+      "loss": 0.4508,
+      "step": 38580
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8722375176724462e-05,
+      "loss": 0.4421,
+      "step": 38590
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.871939876478905e-05,
+      "loss": 0.4421,
+      "step": 38600
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.871642235285364e-05,
+      "loss": 0.4587,
+      "step": 38610
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8713445940918227e-05,
+      "loss": 0.4422,
+      "step": 38620
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.871046952898281e-05,
+      "loss": 0.4332,
+      "step": 38630
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.87074931170474e-05,
+      "loss": 0.4593,
+      "step": 38640
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8704516705111987e-05,
+      "loss": 0.4284,
+      "step": 38650
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8701540293176577e-05,
+      "loss": 0.4342,
+      "step": 38660
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8698563881241164e-05,
+      "loss": 0.4369,
+      "step": 38670
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8695587469305755e-05,
+      "loss": 0.4466,
+      "step": 38680
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.869261105737034e-05,
+      "loss": 0.4724,
+      "step": 38690
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8689634645434932e-05,
+      "loss": 0.4452,
+      "step": 38700
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.868665823349952e-05,
+      "loss": 0.4438,
+      "step": 38710
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.868368182156411e-05,
+      "loss": 0.4369,
+      "step": 38720
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8680705409628693e-05,
+      "loss": 0.4397,
+      "step": 38730
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8677728997693283e-05,
+      "loss": 0.4473,
+      "step": 38740
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.867475258575787e-05,
+      "loss": 0.4571,
+      "step": 38750
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.867177617382246e-05,
+      "loss": 0.4482,
+      "step": 38760
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8668799761887047e-05,
+      "loss": 0.4351,
+      "step": 38770
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8665823349951637e-05,
+      "loss": 0.4484,
+      "step": 38780
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8662846938016224e-05,
+      "loss": 0.4373,
+      "step": 38790
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.8659870526080814e-05,
+      "loss": 0.448,
+      "step": 38800
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.86568941141454e-05,
+      "loss": 0.4385,
+      "step": 38810
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.865391770220999e-05,
+      "loss": 0.4511,
+      "step": 38820
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.8650941290274575e-05,
+      "loss": 0.4457,
+      "step": 38830
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8647964878339162e-05,
+      "loss": 0.449,
+      "step": 38840
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8644988466403752e-05,
+      "loss": 0.4448,
+      "step": 38850
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.864201205446834e-05,
+      "loss": 0.4478,
+      "step": 38860
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.863903564253293e-05,
+      "loss": 0.4442,
+      "step": 38870
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8636059230597516e-05,
+      "loss": 0.4438,
+      "step": 38880
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8633082818662107e-05,
+      "loss": 0.4628,
+      "step": 38890
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8630106406726694e-05,
+      "loss": 0.4533,
+      "step": 38900
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8627129994791284e-05,
+      "loss": 0.444,
+      "step": 38910
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.862415358285587e-05,
+      "loss": 0.4465,
+      "step": 38920
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8621177170920454e-05,
+      "loss": 0.4527,
+      "step": 38930
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8618200758985045e-05,
+      "loss": 0.4436,
+      "step": 38940
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.861522434704963e-05,
+      "loss": 0.4401,
+      "step": 38950
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8612247935114222e-05,
+      "loss": 0.4382,
+      "step": 38960
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.860927152317881e-05,
+      "loss": 0.4628,
+      "step": 38970
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.86062951112434e-05,
+      "loss": 0.4501,
+      "step": 38980
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8603318699307986e-05,
+      "loss": 0.4431,
+      "step": 38990
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8600342287372576e-05,
+      "loss": 0.4246,
+      "step": 39000
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8597365875437163e-05,
+      "loss": 0.4319,
+      "step": 39010
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8594389463501753e-05,
+      "loss": 0.4569,
+      "step": 39020
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.859141305156634e-05,
+      "loss": 0.4631,
+      "step": 39030
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8588436639630927e-05,
+      "loss": 0.43,
+      "step": 39040
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8585460227695514e-05,
+      "loss": 0.4541,
+      "step": 39050
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.85824838157601e-05,
+      "loss": 0.4481,
+      "step": 39060
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.857950740382469e-05,
+      "loss": 0.466,
+      "step": 39070
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8576530991889278e-05,
+      "loss": 0.4609,
+      "step": 39080
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.857355457995387e-05,
+      "loss": 0.4299,
+      "step": 39090
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8570578168018455e-05,
+      "loss": 0.4422,
+      "step": 39100
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8567601756083046e-05,
+      "loss": 0.439,
+      "step": 39110
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8564625344147632e-05,
+      "loss": 0.4425,
+      "step": 39120
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8561648932212223e-05,
+      "loss": 0.4494,
+      "step": 39130
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8558672520276806e-05,
+      "loss": 0.4456,
+      "step": 39140
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8555696108341397e-05,
+      "loss": 0.446,
+      "step": 39150
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8552719696405983e-05,
+      "loss": 0.4397,
+      "step": 39160
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8549743284470574e-05,
+      "loss": 0.4492,
+      "step": 39170
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.854676687253516e-05,
+      "loss": 0.4367,
+      "step": 39180
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.854379046059975e-05,
+      "loss": 0.4575,
+      "step": 39190
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8540814048664338e-05,
+      "loss": 0.4517,
+      "step": 39200
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8537837636728928e-05,
+      "loss": 0.4595,
+      "step": 39210
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8534861224793515e-05,
+      "loss": 0.4329,
+      "step": 39220
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8531884812858105e-05,
+      "loss": 0.4503,
+      "step": 39230
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.852890840092269e-05,
+      "loss": 0.442,
+      "step": 39240
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8525931988987276e-05,
+      "loss": 0.4503,
+      "step": 39250
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8522955577051866e-05,
+      "loss": 0.4484,
+      "step": 39260
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8519979165116453e-05,
+      "loss": 0.4533,
+      "step": 39270
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8517002753181043e-05,
+      "loss": 0.436,
+      "step": 39280
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.851402634124563e-05,
+      "loss": 0.4296,
+      "step": 39290
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.851104992931022e-05,
+      "loss": 0.4522,
+      "step": 39300
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8508073517374807e-05,
+      "loss": 0.4547,
+      "step": 39310
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8505097105439398e-05,
+      "loss": 0.4442,
+      "step": 39320
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8502120693503984e-05,
+      "loss": 0.442,
+      "step": 39330
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8499144281568568e-05,
+      "loss": 0.4384,
+      "step": 39340
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8496167869633158e-05,
+      "loss": 0.4463,
+      "step": 39350
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8493191457697745e-05,
+      "loss": 0.4323,
+      "step": 39360
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8490215045762335e-05,
+      "loss": 0.4287,
+      "step": 39370
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.8487238633826922e-05,
+      "loss": 0.4164,
+      "step": 39380
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.8484262221891513e-05,
+      "loss": 0.4473,
+      "step": 39390
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.84812858099561e-05,
+      "loss": 0.4384,
+      "step": 39400
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.847830939802069e-05,
+      "loss": 0.4432,
+      "step": 39410
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.8475332986085277e-05,
+      "loss": 0.4296,
+      "step": 39420
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8472356574149867e-05,
+      "loss": 0.4542,
+      "step": 39430
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8469380162214454e-05,
+      "loss": 0.4436,
+      "step": 39440
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.846640375027904e-05,
+      "loss": 0.4475,
+      "step": 39450
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8463427338343628e-05,
+      "loss": 0.4357,
+      "step": 39460
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8460450926408218e-05,
+      "loss": 0.4434,
+      "step": 39470
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8457474514472805e-05,
+      "loss": 0.4428,
+      "step": 39480
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8454498102537392e-05,
+      "loss": 0.4418,
+      "step": 39490
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8451521690601982e-05,
+      "loss": 0.4408,
+      "step": 39500
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.844854527866657e-05,
+      "loss": 0.4448,
+      "step": 39510
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.844556886673116e-05,
+      "loss": 0.4192,
+      "step": 39520
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8442592454795746e-05,
+      "loss": 0.4642,
+      "step": 39530
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8439616042860336e-05,
+      "loss": 0.4414,
+      "step": 39540
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.843663963092492e-05,
+      "loss": 0.4557,
+      "step": 39550
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.843366321898951e-05,
+      "loss": 0.4568,
+      "step": 39560
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8430686807054097e-05,
+      "loss": 0.4403,
+      "step": 39570
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8427710395118687e-05,
+      "loss": 0.4422,
+      "step": 39580
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8424733983183274e-05,
+      "loss": 0.4466,
+      "step": 39590
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8421757571247865e-05,
+      "loss": 0.4275,
+      "step": 39600
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.841878115931245e-05,
+      "loss": 0.4406,
+      "step": 39610
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8415804747377042e-05,
+      "loss": 0.4498,
+      "step": 39620
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.841282833544163e-05,
+      "loss": 0.4426,
+      "step": 39630
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.840985192350622e-05,
+      "loss": 0.4317,
+      "step": 39640
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8406875511570802e-05,
+      "loss": 0.4605,
+      "step": 39650
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.840389909963539e-05,
+      "loss": 0.4578,
+      "step": 39660
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.840092268769998e-05,
+      "loss": 0.4291,
+      "step": 39670
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8397946275764567e-05,
+      "loss": 0.4332,
+      "step": 39680
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8394969863829157e-05,
+      "loss": 0.4304,
+      "step": 39690
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8391993451893744e-05,
+      "loss": 0.4721,
+      "step": 39700
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8389017039958334e-05,
+      "loss": 0.4421,
+      "step": 39710
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.838604062802292e-05,
+      "loss": 0.4296,
+      "step": 39720
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.838306421608751e-05,
+      "loss": 0.4343,
+      "step": 39730
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.8380087804152098e-05,
+      "loss": 0.4646,
+      "step": 39740
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.837711139221668e-05,
+      "loss": 0.4433,
+      "step": 39750
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.8374134980281272e-05,
+      "loss": 0.4287,
+      "step": 39760
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.837115856834586e-05,
+      "loss": 0.4393,
+      "step": 39770
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.836818215641045e-05,
+      "loss": 0.4254,
+      "step": 39780
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.8365205744475036e-05,
+      "loss": 0.437,
+      "step": 39790
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.8362229332539626e-05,
+      "loss": 0.4543,
+      "step": 39800
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.8359252920604213e-05,
+      "loss": 0.4406,
+      "step": 39810
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.8356276508668803e-05,
+      "loss": 0.4444,
+      "step": 39820
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.835330009673339e-05,
+      "loss": 0.433,
+      "step": 39830
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.835032368479798e-05,
+      "loss": 0.4353,
+      "step": 39840
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.8347347272862568e-05,
+      "loss": 0.437,
+      "step": 39850
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.8344370860927154e-05,
+      "loss": 0.4235,
+      "step": 39860
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.834139444899174e-05,
+      "loss": 0.4243,
+      "step": 39870
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.833841803705633e-05,
+      "loss": 0.4384,
+      "step": 39880
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.833544162512092e-05,
+      "loss": 0.4386,
+      "step": 39890
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.833246521318551e-05,
+      "loss": 0.4368,
+      "step": 39900
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.8329488801250096e-05,
+      "loss": 0.4501,
+      "step": 39910
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.8326512389314683e-05,
+      "loss": 0.4567,
+      "step": 39920
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.8323535977379273e-05,
+      "loss": 0.4443,
+      "step": 39930
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.832055956544386e-05,
+      "loss": 0.4411,
+      "step": 39940
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.831758315350845e-05,
+      "loss": 0.4346,
+      "step": 39950
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.8314606741573034e-05,
+      "loss": 0.4349,
+      "step": 39960
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.8311630329637624e-05,
+      "loss": 0.437,
+      "step": 39970
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.830865391770221e-05,
+      "loss": 0.4499,
+      "step": 39980
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.83056775057668e-05,
+      "loss": 0.4322,
+      "step": 39990
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.8302701093831388e-05,
+      "loss": 0.4537,
+      "step": 40000
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.8299724681895978e-05,
+      "loss": 0.4447,
+      "step": 40010
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.8296748269960565e-05,
+      "loss": 0.45,
+      "step": 40020
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.8293771858025155e-05,
+      "loss": 0.4606,
+      "step": 40030
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.8290795446089742e-05,
+      "loss": 0.4357,
+      "step": 40040
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.8287819034154333e-05,
+      "loss": 0.4424,
+      "step": 40050
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.8284842622218916e-05,
+      "loss": 0.4323,
+      "step": 40060
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.8281866210283503e-05,
+      "loss": 0.4499,
+      "step": 40070
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.8278889798348093e-05,
+      "loss": 0.4526,
+      "step": 40080
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.827591338641268e-05,
+      "loss": 0.4337,
+      "step": 40090
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.827293697447727e-05,
+      "loss": 0.4596,
+      "step": 40100
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.8269960562541857e-05,
+      "loss": 0.4283,
+      "step": 40110
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.8266984150606448e-05,
+      "loss": 0.4419,
+      "step": 40120
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.8264007738671035e-05,
+      "loss": 0.4409,
+      "step": 40130
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.8261031326735625e-05,
+      "loss": 0.4396,
+      "step": 40140
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.8258054914800212e-05,
+      "loss": 0.4418,
+      "step": 40150
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.8255078502864795e-05,
+      "loss": 0.4487,
+      "step": 40160
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.8252102090929386e-05,
+      "loss": 0.4621,
+      "step": 40170
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.8249125678993972e-05,
+      "loss": 0.4489,
+      "step": 40180
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.8246149267058563e-05,
+      "loss": 0.4392,
+      "step": 40190
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.824317285512315e-05,
+      "loss": 0.4373,
+      "step": 40200
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.824019644318774e-05,
+      "loss": 0.4445,
+      "step": 40210
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.8237220031252327e-05,
+      "loss": 0.427,
+      "step": 40220
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.8234243619316917e-05,
+      "loss": 0.4401,
+      "step": 40230
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.8231267207381504e-05,
+      "loss": 0.4431,
+      "step": 40240
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.8228290795446094e-05,
+      "loss": 0.4434,
+      "step": 40250
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.8225314383510678e-05,
+      "loss": 0.4292,
+      "step": 40260
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.8222337971575268e-05,
+      "loss": 0.4385,
+      "step": 40270
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.8219361559639855e-05,
+      "loss": 0.451,
+      "step": 40280
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.8216385147704445e-05,
+      "loss": 0.436,
+      "step": 40290
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.8213408735769032e-05,
+      "loss": 0.4226,
+      "step": 40300
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.8210432323833622e-05,
+      "loss": 0.4381,
+      "step": 40310
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.820745591189821e-05,
+      "loss": 0.4305,
+      "step": 40320
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.82044794999628e-05,
+      "loss": 0.4336,
+      "step": 40330
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.8201503088027387e-05,
+      "loss": 0.4418,
+      "step": 40340
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.8198526676091973e-05,
+      "loss": 0.4425,
+      "step": 40350
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.8195550264156564e-05,
+      "loss": 0.4527,
+      "step": 40360
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.8192573852221147e-05,
+      "loss": 0.4451,
+      "step": 40370
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.8189597440285738e-05,
+      "loss": 0.4395,
+      "step": 40380
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.8186621028350324e-05,
+      "loss": 0.4314,
+      "step": 40390
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.8183644616414915e-05,
+      "loss": 0.4327,
+      "step": 40400
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.81806682044795e-05,
+      "loss": 0.4443,
+      "step": 40410
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.8177691792544092e-05,
+      "loss": 0.4261,
+      "step": 40420
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.817471538060868e-05,
+      "loss": 0.4448,
+      "step": 40430
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.817173896867327e-05,
+      "loss": 0.4378,
+      "step": 40440
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.8168762556737856e-05,
+      "loss": 0.4626,
+      "step": 40450
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.8165786144802446e-05,
+      "loss": 0.449,
+      "step": 40460
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.816280973286703e-05,
+      "loss": 0.4453,
+      "step": 40470
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.8159833320931617e-05,
+      "loss": 0.4282,
+      "step": 40480
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.8156856908996207e-05,
+      "loss": 0.4405,
+      "step": 40490
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.8153880497060794e-05,
+      "loss": 0.4444,
+      "step": 40500
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.8150904085125384e-05,
+      "loss": 0.4577,
+      "step": 40510
+    },
+    {
+      "epoch": 9.0,
+      "eval_cer": 4.561965811965812,
+      "eval_loss": 0.4203878343105316,
+      "eval_runtime": 1.2863,
+      "eval_samples_per_second": 7.774,
+      "eval_steps_per_second": 0.777,
+      "eval_wer": 0.6363636363636364,
+      "step": 40518
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.814792767318997e-05,
+      "loss": 0.4589,
+      "step": 40520
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.814495126125456e-05,
+      "loss": 0.4349,
+      "step": 40530
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.8141974849319148e-05,
+      "loss": 0.4414,
+      "step": 40540
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.813899843738374e-05,
+      "loss": 0.4315,
+      "step": 40550
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.8136022025448325e-05,
+      "loss": 0.4296,
+      "step": 40560
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.813304561351291e-05,
+      "loss": 0.4355,
+      "step": 40570
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.81300692015775e-05,
+      "loss": 0.4408,
+      "step": 40580
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.8127092789642086e-05,
+      "loss": 0.4346,
+      "step": 40590
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.8124116377706676e-05,
+      "loss": 0.4453,
+      "step": 40600
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.8121139965771263e-05,
+      "loss": 0.4315,
+      "step": 40610
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.8118163553835854e-05,
+      "loss": 0.4174,
+      "step": 40620
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.811518714190044e-05,
+      "loss": 0.4178,
+      "step": 40630
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.811221072996503e-05,
+      "loss": 0.4264,
+      "step": 40640
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8109234318029618e-05,
+      "loss": 0.4162,
+      "step": 40650
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8106257906094208e-05,
+      "loss": 0.44,
+      "step": 40660
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.810328149415879e-05,
+      "loss": 0.429,
+      "step": 40670
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.8100305082223382e-05,
+      "loss": 0.4239,
+      "step": 40680
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.809732867028797e-05,
+      "loss": 0.4254,
+      "step": 40690
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.809435225835256e-05,
+      "loss": 0.4278,
+      "step": 40700
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8091375846417146e-05,
+      "loss": 0.4429,
+      "step": 40710
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.8088399434481736e-05,
+      "loss": 0.4276,
+      "step": 40720
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8085423022546323e-05,
+      "loss": 0.4213,
+      "step": 40730
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8082446610610913e-05,
+      "loss": 0.43,
+      "step": 40740
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.80794701986755e-05,
+      "loss": 0.4285,
+      "step": 40750
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.807649378674009e-05,
+      "loss": 0.4279,
+      "step": 40760
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.8073517374804677e-05,
+      "loss": 0.428,
+      "step": 40770
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.807054096286926e-05,
+      "loss": 0.4195,
+      "step": 40780
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.806756455093385e-05,
+      "loss": 0.4307,
+      "step": 40790
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.8064588138998438e-05,
+      "loss": 0.447,
+      "step": 40800
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.806161172706303e-05,
+      "loss": 0.4173,
+      "step": 40810
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8058635315127615e-05,
+      "loss": 0.4269,
+      "step": 40820
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8055658903192206e-05,
+      "loss": 0.4183,
+      "step": 40830
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8052682491256792e-05,
+      "loss": 0.433,
+      "step": 40840
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.8049706079321383e-05,
+      "loss": 0.4342,
+      "step": 40850
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.804672966738597e-05,
+      "loss": 0.4226,
+      "step": 40860
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.804375325545056e-05,
+      "loss": 0.4267,
+      "step": 40870
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.8040776843515143e-05,
+      "loss": 0.443,
+      "step": 40880
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.803780043157973e-05,
+      "loss": 0.4104,
+      "step": 40890
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.803482401964432e-05,
+      "loss": 0.4175,
+      "step": 40900
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.8031847607708908e-05,
+      "loss": 0.4188,
+      "step": 40910
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.8028871195773498e-05,
+      "loss": 0.4208,
+      "step": 40920
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.8025894783838085e-05,
+      "loss": 0.4339,
+      "step": 40930
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.8022918371902675e-05,
+      "loss": 0.4164,
+      "step": 40940
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.8019941959967262e-05,
+      "loss": 0.4318,
+      "step": 40950
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.8016965548031852e-05,
+      "loss": 0.4453,
+      "step": 40960
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.801398913609644e-05,
+      "loss": 0.4303,
+      "step": 40970
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.8011012724161023e-05,
+      "loss": 0.4253,
+      "step": 40980
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.8008036312225613e-05,
+      "loss": 0.4343,
+      "step": 40990
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.80050599002902e-05,
+      "loss": 0.421,
+      "step": 41000
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.800208348835479e-05,
+      "loss": 0.4409,
+      "step": 41010
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7999107076419377e-05,
+      "loss": 0.4422,
+      "step": 41020
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7996130664483967e-05,
+      "loss": 0.4181,
+      "step": 41030
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7993154252548554e-05,
+      "loss": 0.4334,
+      "step": 41040
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.7990177840613144e-05,
+      "loss": 0.4358,
+      "step": 41050
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.798720142867773e-05,
+      "loss": 0.417,
+      "step": 41060
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.798422501674232e-05,
+      "loss": 0.4344,
+      "step": 41070
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.7981248604806905e-05,
+      "loss": 0.4422,
+      "step": 41080
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7978272192871495e-05,
+      "loss": 0.4331,
+      "step": 41090
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7975295780936082e-05,
+      "loss": 0.4263,
+      "step": 41100
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7972319369000673e-05,
+      "loss": 0.4243,
+      "step": 41110
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.796934295706526e-05,
+      "loss": 0.4199,
+      "step": 41120
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.796636654512985e-05,
+      "loss": 0.433,
+      "step": 41130
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7963390133194437e-05,
+      "loss": 0.4428,
+      "step": 41140
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7960413721259027e-05,
+      "loss": 0.4291,
+      "step": 41150
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7957437309323614e-05,
+      "loss": 0.4306,
+      "step": 41160
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7954460897388204e-05,
+      "loss": 0.4356,
+      "step": 41170
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.795148448545279e-05,
+      "loss": 0.432,
+      "step": 41180
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7948508073517375e-05,
+      "loss": 0.4243,
+      "step": 41190
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7945531661581965e-05,
+      "loss": 0.4366,
+      "step": 41200
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7942555249646552e-05,
+      "loss": 0.4341,
+      "step": 41210
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7939578837711142e-05,
+      "loss": 0.4238,
+      "step": 41220
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.793660242577573e-05,
+      "loss": 0.4402,
+      "step": 41230
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.793362601384032e-05,
+      "loss": 0.432,
+      "step": 41240
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7930649601904906e-05,
+      "loss": 0.4117,
+      "step": 41250
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7927673189969496e-05,
+      "loss": 0.4341,
+      "step": 41260
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7924696778034083e-05,
+      "loss": 0.4234,
+      "step": 41270
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7921720366098674e-05,
+      "loss": 0.4225,
+      "step": 41280
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7918743954163257e-05,
+      "loss": 0.4296,
+      "step": 41290
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7915767542227844e-05,
+      "loss": 0.4266,
+      "step": 41300
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7912791130292434e-05,
+      "loss": 0.4273,
+      "step": 41310
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.790981471835702e-05,
+      "loss": 0.4163,
+      "step": 41320
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.790683830642161e-05,
+      "loss": 0.437,
+      "step": 41330
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.79038618944862e-05,
+      "loss": 0.4429,
+      "step": 41340
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.790088548255079e-05,
+      "loss": 0.4163,
+      "step": 41350
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7897909070615376e-05,
+      "loss": 0.4382,
+      "step": 41360
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7894932658679966e-05,
+      "loss": 0.4123,
+      "step": 41370
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7891956246744553e-05,
+      "loss": 0.4245,
+      "step": 41380
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7888979834809136e-05,
+      "loss": 0.427,
+      "step": 41390
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7886003422873727e-05,
+      "loss": 0.4353,
+      "step": 41400
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7883027010938313e-05,
+      "loss": 0.4438,
+      "step": 41410
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7880050599002904e-05,
+      "loss": 0.4341,
+      "step": 41420
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.787707418706749e-05,
+      "loss": 0.4217,
+      "step": 41430
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.787409777513208e-05,
+      "loss": 0.4322,
+      "step": 41440
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7871121363196668e-05,
+      "loss": 0.4221,
+      "step": 41450
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7868144951261258e-05,
+      "loss": 0.4452,
+      "step": 41460
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7865168539325845e-05,
+      "loss": 0.4199,
+      "step": 41470
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7862192127390435e-05,
+      "loss": 0.4379,
+      "step": 41480
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.785921571545502e-05,
+      "loss": 0.419,
+      "step": 41490
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.785623930351961e-05,
+      "loss": 0.427,
+      "step": 41500
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7853262891584196e-05,
+      "loss": 0.431,
+      "step": 41510
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7850286479648786e-05,
+      "loss": 0.4318,
+      "step": 41520
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7847310067713373e-05,
+      "loss": 0.4395,
+      "step": 41530
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7844333655777964e-05,
+      "loss": 0.4059,
+      "step": 41540
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.784135724384255e-05,
+      "loss": 0.4249,
+      "step": 41550
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.783838083190714e-05,
+      "loss": 0.4355,
+      "step": 41560
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.7835404419971728e-05,
+      "loss": 0.4372,
+      "step": 41570
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.7832428008036318e-05,
+      "loss": 0.4226,
+      "step": 41580
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.7829451596100905e-05,
+      "loss": 0.425,
+      "step": 41590
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.7826475184165488e-05,
+      "loss": 0.4093,
+      "step": 41600
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.782349877223008e-05,
+      "loss": 0.4253,
+      "step": 41610
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.7820522360294665e-05,
+      "loss": 0.4366,
+      "step": 41620
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.7817545948359256e-05,
+      "loss": 0.4312,
+      "step": 41630
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.7814569536423843e-05,
+      "loss": 0.4477,
+      "step": 41640
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.7811593124488433e-05,
+      "loss": 0.4247,
+      "step": 41650
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.780861671255302e-05,
+      "loss": 0.4288,
+      "step": 41660
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.780564030061761e-05,
+      "loss": 0.437,
+      "step": 41670
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.7802663888682197e-05,
+      "loss": 0.4309,
+      "step": 41680
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.7799687476746787e-05,
+      "loss": 0.4208,
+      "step": 41690
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.779671106481137e-05,
+      "loss": 0.422,
+      "step": 41700
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.7793734652875958e-05,
+      "loss": 0.4363,
+      "step": 41710
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.7790758240940548e-05,
+      "loss": 0.4311,
+      "step": 41720
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.7787781829005135e-05,
+      "loss": 0.4257,
+      "step": 41730
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.7784805417069725e-05,
+      "loss": 0.4231,
+      "step": 41740
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.7781829005134312e-05,
+      "loss": 0.4183,
+      "step": 41750
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.7778852593198902e-05,
+      "loss": 0.4398,
+      "step": 41760
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.777587618126349e-05,
+      "loss": 0.4395,
+      "step": 41770
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.777289976932808e-05,
+      "loss": 0.4379,
+      "step": 41780
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.7769923357392666e-05,
+      "loss": 0.4287,
+      "step": 41790
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.776694694545725e-05,
+      "loss": 0.4452,
+      "step": 41800
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.776397053352184e-05,
+      "loss": 0.415,
+      "step": 41810
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.7760994121586427e-05,
+      "loss": 0.4265,
+      "step": 41820
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.7758017709651017e-05,
+      "loss": 0.4254,
+      "step": 41830
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.7755041297715604e-05,
+      "loss": 0.4323,
+      "step": 41840
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.7752064885780195e-05,
+      "loss": 0.4327,
+      "step": 41850
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.774908847384478e-05,
+      "loss": 0.403,
+      "step": 41860
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.7746112061909372e-05,
+      "loss": 0.4275,
+      "step": 41870
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.774313564997396e-05,
+      "loss": 0.4326,
+      "step": 41880
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.774015923803855e-05,
+      "loss": 0.4284,
+      "step": 41890
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.7737182826103133e-05,
+      "loss": 0.4061,
+      "step": 41900
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.7734206414167723e-05,
+      "loss": 0.4179,
+      "step": 41910
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.773123000223231e-05,
+      "loss": 0.4202,
+      "step": 41920
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.77282535902969e-05,
+      "loss": 0.4287,
+      "step": 41930
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.7725277178361487e-05,
+      "loss": 0.4182,
+      "step": 41940
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.7722300766426077e-05,
+      "loss": 0.4314,
+      "step": 41950
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.7719324354490664e-05,
+      "loss": 0.4503,
+      "step": 41960
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.7716347942555254e-05,
+      "loss": 0.4292,
+      "step": 41970
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.771337153061984e-05,
+      "loss": 0.399,
+      "step": 41980
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.771039511868443e-05,
+      "loss": 0.4136,
+      "step": 41990
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.7707418706749015e-05,
+      "loss": 0.4342,
+      "step": 42000
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.7704442294813602e-05,
+      "loss": 0.443,
+      "step": 42010
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.7701465882878192e-05,
+      "loss": 0.4292,
+      "step": 42020
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.769848947094278e-05,
+      "loss": 0.4149,
+      "step": 42030
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.769551305900737e-05,
+      "loss": 0.4206,
+      "step": 42040
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.7692536647071956e-05,
+      "loss": 0.4295,
+      "step": 42050
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.7689560235136547e-05,
+      "loss": 0.4215,
+      "step": 42060
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.7686583823201134e-05,
+      "loss": 0.431,
+      "step": 42070
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.7683607411265724e-05,
+      "loss": 0.4307,
+      "step": 42080
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.768063099933031e-05,
+      "loss": 0.4208,
+      "step": 42090
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.76776545873949e-05,
+      "loss": 0.4278,
+      "step": 42100
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.7674678175459485e-05,
+      "loss": 0.4312,
+      "step": 42110
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.767170176352407e-05,
+      "loss": 0.4242,
+      "step": 42120
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.766872535158866e-05,
+      "loss": 0.4341,
+      "step": 42130
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.766574893965325e-05,
+      "loss": 0.4312,
+      "step": 42140
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.766277252771784e-05,
+      "loss": 0.4175,
+      "step": 42150
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.7659796115782426e-05,
+      "loss": 0.421,
+      "step": 42160
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.7656819703847016e-05,
+      "loss": 0.4281,
+      "step": 42170
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.7653843291911603e-05,
+      "loss": 0.4285,
+      "step": 42180
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.7650866879976193e-05,
+      "loss": 0.4373,
+      "step": 42190
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.764789046804078e-05,
+      "loss": 0.4351,
+      "step": 42200
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.7644914056105364e-05,
+      "loss": 0.4123,
+      "step": 42210
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.7641937644169954e-05,
+      "loss": 0.4381,
+      "step": 42220
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.763896123223454e-05,
+      "loss": 0.4216,
+      "step": 42230
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.763598482029913e-05,
+      "loss": 0.4148,
+      "step": 42240
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.7633008408363718e-05,
+      "loss": 0.4353,
+      "step": 42250
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.763003199642831e-05,
+      "loss": 0.4215,
+      "step": 42260
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.7627055584492895e-05,
+      "loss": 0.4288,
+      "step": 42270
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.7624079172557486e-05,
+      "loss": 0.4157,
+      "step": 42280
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.7621102760622072e-05,
+      "loss": 0.4307,
+      "step": 42290
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.7618126348686663e-05,
+      "loss": 0.4319,
+      "step": 42300
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.7615149936751246e-05,
+      "loss": 0.4343,
+      "step": 42310
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.7612173524815836e-05,
+      "loss": 0.4383,
+      "step": 42320
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.7609197112880423e-05,
+      "loss": 0.4292,
+      "step": 42330
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.7606220700945014e-05,
+      "loss": 0.4308,
+      "step": 42340
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.76032442890096e-05,
+      "loss": 0.4241,
+      "step": 42350
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.760026787707419e-05,
+      "loss": 0.424,
+      "step": 42360
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.7597291465138778e-05,
+      "loss": 0.4233,
+      "step": 42370
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.7594315053203368e-05,
+      "loss": 0.4297,
+      "step": 42380
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.7591338641267955e-05,
+      "loss": 0.4253,
+      "step": 42390
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.7588362229332545e-05,
+      "loss": 0.4376,
+      "step": 42400
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.758538581739713e-05,
+      "loss": 0.4145,
+      "step": 42410
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.7582409405461716e-05,
+      "loss": 0.4132,
+      "step": 42420
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.7579432993526306e-05,
+      "loss": 0.4346,
+      "step": 42430
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.7576456581590893e-05,
+      "loss": 0.4263,
+      "step": 42440
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.7573480169655483e-05,
+      "loss": 0.4243,
+      "step": 42450
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.757050375772007e-05,
+      "loss": 0.4312,
+      "step": 42460
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.756752734578466e-05,
+      "loss": 0.4313,
+      "step": 42470
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.7564550933849247e-05,
+      "loss": 0.4202,
+      "step": 42480
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.7561574521913837e-05,
+      "loss": 0.4287,
+      "step": 42490
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.7558598109978424e-05,
+      "loss": 0.4111,
+      "step": 42500
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.7555621698043015e-05,
+      "loss": 0.4195,
+      "step": 42510
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.7552645286107598e-05,
+      "loss": 0.4343,
+      "step": 42520
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.7549668874172185e-05,
+      "loss": 0.4466,
+      "step": 42530
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.7546692462236775e-05,
+      "loss": 0.4351,
+      "step": 42540
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.7543716050301362e-05,
+      "loss": 0.4252,
+      "step": 42550
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.7540739638365953e-05,
+      "loss": 0.4212,
+      "step": 42560
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.753776322643054e-05,
+      "loss": 0.434,
+      "step": 42570
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.753478681449513e-05,
+      "loss": 0.4255,
+      "step": 42580
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.7531810402559717e-05,
+      "loss": 0.4271,
+      "step": 42590
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.7528833990624307e-05,
+      "loss": 0.4233,
+      "step": 42600
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.7525857578688894e-05,
+      "loss": 0.4217,
+      "step": 42610
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.7522881166753477e-05,
+      "loss": 0.417,
+      "step": 42620
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.7519904754818068e-05,
+      "loss": 0.4182,
+      "step": 42630
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.7516928342882655e-05,
+      "loss": 0.4225,
+      "step": 42640
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.7513951930947245e-05,
+      "loss": 0.4186,
+      "step": 42650
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.7510975519011832e-05,
+      "loss": 0.4241,
+      "step": 42660
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.7507999107076422e-05,
+      "loss": 0.4021,
+      "step": 42670
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.750502269514101e-05,
+      "loss": 0.4337,
+      "step": 42680
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.75020462832056e-05,
+      "loss": 0.4135,
+      "step": 42690
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.7499069871270186e-05,
+      "loss": 0.4107,
+      "step": 42700
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.7496093459334776e-05,
+      "loss": 0.4332,
+      "step": 42710
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.749311704739936e-05,
+      "loss": 0.4226,
+      "step": 42720
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.749014063546395e-05,
+      "loss": 0.419,
+      "step": 42730
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.7487164223528537e-05,
+      "loss": 0.4298,
+      "step": 42740
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.7484187811593127e-05,
+      "loss": 0.428,
+      "step": 42750
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.7481211399657714e-05,
+      "loss": 0.4237,
+      "step": 42760
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.7478234987722305e-05,
+      "loss": 0.4398,
+      "step": 42770
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.747525857578689e-05,
+      "loss": 0.4393,
+      "step": 42780
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.7472282163851482e-05,
+      "loss": 0.4276,
+      "step": 42790
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.746930575191607e-05,
+      "loss": 0.425,
+      "step": 42800
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.746632933998066e-05,
+      "loss": 0.4307,
+      "step": 42810
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.7463352928045242e-05,
+      "loss": 0.4166,
+      "step": 42820
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.746037651610983e-05,
+      "loss": 0.4268,
+      "step": 42830
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.745740010417442e-05,
+      "loss": 0.4257,
+      "step": 42840
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.7454423692239007e-05,
+      "loss": 0.4199,
+      "step": 42850
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.7451447280303597e-05,
+      "loss": 0.4393,
+      "step": 42860
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.7448470868368184e-05,
+      "loss": 0.4449,
+      "step": 42870
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.7445494456432774e-05,
+      "loss": 0.4264,
+      "step": 42880
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.744251804449736e-05,
+      "loss": 0.44,
+      "step": 42890
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.743954163256195e-05,
+      "loss": 0.4183,
+      "step": 42900
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.7436565220626538e-05,
+      "loss": 0.4226,
+      "step": 42910
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.743358880869113e-05,
+      "loss": 0.4192,
+      "step": 42920
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.7430612396755712e-05,
+      "loss": 0.4225,
+      "step": 42930
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.74276359848203e-05,
+      "loss": 0.4177,
+      "step": 42940
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.742465957288489e-05,
+      "loss": 0.4283,
+      "step": 42950
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.7421683160949476e-05,
+      "loss": 0.4232,
+      "step": 42960
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.7418706749014066e-05,
+      "loss": 0.4353,
+      "step": 42970
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.7415730337078653e-05,
+      "loss": 0.4323,
+      "step": 42980
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.7412753925143243e-05,
+      "loss": 0.4339,
+      "step": 42990
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.740977751320783e-05,
+      "loss": 0.4198,
+      "step": 43000
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.740680110127242e-05,
+      "loss": 0.4197,
+      "step": 43010
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.7403824689337008e-05,
+      "loss": 0.4251,
+      "step": 43020
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.740084827740159e-05,
+      "loss": 0.4307,
+      "step": 43030
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.739787186546618e-05,
+      "loss": 0.4231,
+      "step": 43040
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.7394895453530768e-05,
+      "loss": 0.4462,
+      "step": 43050
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.739191904159536e-05,
+      "loss": 0.4155,
+      "step": 43060
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.7388942629659945e-05,
+      "loss": 0.4332,
+      "step": 43070
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.7385966217724536e-05,
+      "loss": 0.4237,
+      "step": 43080
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.7382989805789123e-05,
+      "loss": 0.4203,
+      "step": 43090
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.7380013393853713e-05,
+      "loss": 0.4246,
+      "step": 43100
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.73770369819183e-05,
+      "loss": 0.4284,
+      "step": 43110
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.737406056998289e-05,
+      "loss": 0.4118,
+      "step": 43120
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.7371084158047474e-05,
+      "loss": 0.4216,
+      "step": 43130
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.7368107746112064e-05,
+      "loss": 0.4225,
+      "step": 43140
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.736513133417665e-05,
+      "loss": 0.4246,
+      "step": 43150
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.736215492224124e-05,
+      "loss": 0.4224,
+      "step": 43160
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.7359178510305828e-05,
+      "loss": 0.4443,
+      "step": 43170
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.7356202098370418e-05,
+      "loss": 0.4197,
+      "step": 43180
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.7353225686435005e-05,
+      "loss": 0.4278,
+      "step": 43190
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.7350249274499595e-05,
+      "loss": 0.4366,
+      "step": 43200
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.7347272862564182e-05,
+      "loss": 0.4268,
+      "step": 43210
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.7344296450628773e-05,
+      "loss": 0.4104,
+      "step": 43220
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.7341320038693356e-05,
+      "loss": 0.4336,
+      "step": 43230
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.7338343626757943e-05,
+      "loss": 0.4339,
+      "step": 43240
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.7335367214822533e-05,
+      "loss": 0.4204,
+      "step": 43250
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.733239080288712e-05,
+      "loss": 0.4274,
+      "step": 43260
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.732941439095171e-05,
+      "loss": 0.4274,
+      "step": 43270
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.7326437979016297e-05,
+      "loss": 0.4068,
+      "step": 43280
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.7323461567080888e-05,
+      "loss": 0.4106,
+      "step": 43290
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.7320485155145475e-05,
+      "loss": 0.4148,
+      "step": 43300
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.7317508743210065e-05,
+      "loss": 0.432,
+      "step": 43310
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.7314532331274652e-05,
+      "loss": 0.435,
+      "step": 43320
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.7311555919339242e-05,
+      "loss": 0.4302,
+      "step": 43330
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.7308579507403826e-05,
+      "loss": 0.4347,
+      "step": 43340
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.7305603095468412e-05,
+      "loss": 0.4349,
+      "step": 43350
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.7302626683533003e-05,
+      "loss": 0.4248,
+      "step": 43360
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.729965027159759e-05,
+      "loss": 0.4218,
+      "step": 43370
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.729667385966218e-05,
+      "loss": 0.4142,
+      "step": 43380
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.7293697447726767e-05,
+      "loss": 0.4116,
+      "step": 43390
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.7290721035791357e-05,
+      "loss": 0.4337,
+      "step": 43400
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.7287744623855944e-05,
+      "loss": 0.431,
+      "step": 43410
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.7284768211920534e-05,
+      "loss": 0.4356,
+      "step": 43420
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.728179179998512e-05,
+      "loss": 0.4231,
+      "step": 43430
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.7278815388049708e-05,
+      "loss": 0.418,
+      "step": 43440
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.7275838976114295e-05,
+      "loss": 0.4204,
+      "step": 43450
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.7272862564178882e-05,
+      "loss": 0.3974,
+      "step": 43460
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.7269886152243472e-05,
+      "loss": 0.425,
+      "step": 43470
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.726690974030806e-05,
+      "loss": 0.4222,
+      "step": 43480
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.726393332837265e-05,
+      "loss": 0.4246,
+      "step": 43490
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.7260956916437236e-05,
+      "loss": 0.4258,
+      "step": 43500
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.7257980504501827e-05,
+      "loss": 0.4308,
+      "step": 43510
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.7255004092566413e-05,
+      "loss": 0.4319,
+      "step": 43520
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.7252027680631004e-05,
+      "loss": 0.4384,
+      "step": 43530
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.7249051268695587e-05,
+      "loss": 0.4456,
+      "step": 43540
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.7246074856760178e-05,
+      "loss": 0.4243,
+      "step": 43550
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.7243098444824764e-05,
+      "loss": 0.4315,
+      "step": 43560
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.7240122032889355e-05,
+      "loss": 0.432,
+      "step": 43570
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.723714562095394e-05,
+      "loss": 0.412,
+      "step": 43580
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.7234169209018532e-05,
+      "loss": 0.4264,
+      "step": 43590
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.723119279708312e-05,
+      "loss": 0.4276,
+      "step": 43600
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.722821638514771e-05,
+      "loss": 0.4227,
+      "step": 43610
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.7225239973212296e-05,
+      "loss": 0.4176,
+      "step": 43620
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.7222263561276886e-05,
+      "loss": 0.4233,
+      "step": 43630
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.721928714934147e-05,
+      "loss": 0.4155,
+      "step": 43640
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.7216310737406057e-05,
+      "loss": 0.4057,
+      "step": 43650
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.7213334325470647e-05,
+      "loss": 0.4311,
+      "step": 43660
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.7210357913535234e-05,
+      "loss": 0.4201,
+      "step": 43670
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.7207381501599824e-05,
+      "loss": 0.4231,
+      "step": 43680
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.720440508966441e-05,
+      "loss": 0.4222,
+      "step": 43690
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.7201428677729e-05,
+      "loss": 0.4308,
+      "step": 43700
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.7198452265793588e-05,
+      "loss": 0.4225,
+      "step": 43710
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.719547585385818e-05,
+      "loss": 0.4277,
+      "step": 43720
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.7192499441922765e-05,
+      "loss": 0.4264,
+      "step": 43730
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.718952302998735e-05,
+      "loss": 0.4193,
+      "step": 43740
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.718654661805194e-05,
+      "loss": 0.4259,
+      "step": 43750
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7183570206116526e-05,
+      "loss": 0.4231,
+      "step": 43760
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7180593794181116e-05,
+      "loss": 0.4279,
+      "step": 43770
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7177617382245703e-05,
+      "loss": 0.4154,
+      "step": 43780
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7174640970310294e-05,
+      "loss": 0.4128,
+      "step": 43790
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.717166455837488e-05,
+      "loss": 0.418,
+      "step": 43800
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.716868814643947e-05,
+      "loss": 0.4194,
+      "step": 43810
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.7165711734504058e-05,
+      "loss": 0.4301,
+      "step": 43820
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.7162735322568648e-05,
+      "loss": 0.414,
+      "step": 43830
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.7159758910633235e-05,
+      "loss": 0.4265,
+      "step": 43840
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.7156782498697822e-05,
+      "loss": 0.4295,
+      "step": 43850
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.715380608676241e-05,
+      "loss": 0.4153,
+      "step": 43860
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.7150829674827e-05,
+      "loss": 0.4325,
+      "step": 43870
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.7147853262891586e-05,
+      "loss": 0.4353,
+      "step": 43880
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.7144876850956173e-05,
+      "loss": 0.4522,
+      "step": 43890
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.7141900439020763e-05,
+      "loss": 0.4254,
+      "step": 43900
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.713892402708535e-05,
+      "loss": 0.4362,
+      "step": 43910
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.713594761514994e-05,
+      "loss": 0.4064,
+      "step": 43920
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.7132971203214527e-05,
+      "loss": 0.4244,
+      "step": 43930
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.7129994791279117e-05,
+      "loss": 0.4291,
+      "step": 43940
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.71270183793437e-05,
+      "loss": 0.4202,
+      "step": 43950
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.712404196740829e-05,
+      "loss": 0.4035,
+      "step": 43960
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.7121065555472878e-05,
+      "loss": 0.4198,
+      "step": 43970
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.711808914353747e-05,
+      "loss": 0.4238,
+      "step": 43980
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.7115112731602055e-05,
+      "loss": 0.4226,
+      "step": 43990
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.7112136319666646e-05,
+      "loss": 0.4372,
+      "step": 44000
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.7109159907731232e-05,
+      "loss": 0.4175,
+      "step": 44010
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.7106183495795823e-05,
+      "loss": 0.4059,
+      "step": 44020
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.710320708386041e-05,
+      "loss": 0.4213,
+      "step": 44030
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.7100230671925e-05,
+      "loss": 0.419,
+      "step": 44040
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.7097254259989583e-05,
+      "loss": 0.4154,
+      "step": 44050
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.709427784805417e-05,
+      "loss": 0.415,
+      "step": 44060
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.709130143611876e-05,
+      "loss": 0.422,
+      "step": 44070
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.7088325024183348e-05,
+      "loss": 0.4308,
+      "step": 44080
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.7085348612247938e-05,
+      "loss": 0.4416,
+      "step": 44090
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.7082372200312525e-05,
+      "loss": 0.4388,
+      "step": 44100
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.7079395788377115e-05,
+      "loss": 0.434,
+      "step": 44110
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.7076419376441702e-05,
+      "loss": 0.4208,
+      "step": 44120
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.7073442964506292e-05,
+      "loss": 0.4195,
+      "step": 44130
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.707046655257088e-05,
+      "loss": 0.4176,
+      "step": 44140
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.7067490140635463e-05,
+      "loss": 0.4163,
+      "step": 44150
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.7064513728700053e-05,
+      "loss": 0.4274,
+      "step": 44160
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.706153731676464e-05,
+      "loss": 0.4278,
+      "step": 44170
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.705856090482923e-05,
+      "loss": 0.4232,
+      "step": 44180
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.7055584492893817e-05,
+      "loss": 0.4055,
+      "step": 44190
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.7052608080958407e-05,
+      "loss": 0.4117,
+      "step": 44200
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.7049631669022994e-05,
+      "loss": 0.4274,
+      "step": 44210
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.7046655257087584e-05,
+      "loss": 0.4213,
+      "step": 44220
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.704367884515217e-05,
+      "loss": 0.429,
+      "step": 44230
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.704070243321676e-05,
+      "loss": 0.4284,
+      "step": 44240
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.703772602128135e-05,
+      "loss": 0.4019,
+      "step": 44250
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.7034749609345935e-05,
+      "loss": 0.4101,
+      "step": 44260
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.7031773197410522e-05,
+      "loss": 0.4206,
+      "step": 44270
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.7028796785475113e-05,
+      "loss": 0.409,
+      "step": 44280
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.70258203735397e-05,
+      "loss": 0.4183,
+      "step": 44290
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.702284396160429e-05,
+      "loss": 0.4158,
+      "step": 44300
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.7019867549668877e-05,
+      "loss": 0.421,
+      "step": 44310
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.7016891137733464e-05,
+      "loss": 0.4237,
+      "step": 44320
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.7013914725798054e-05,
+      "loss": 0.4226,
+      "step": 44330
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.701093831386264e-05,
+      "loss": 0.4343,
+      "step": 44340
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.700796190192723e-05,
+      "loss": 0.4156,
+      "step": 44350
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.7004985489991815e-05,
+      "loss": 0.4269,
+      "step": 44360
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.7002009078056405e-05,
+      "loss": 0.4053,
+      "step": 44370
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.6999032666120992e-05,
+      "loss": 0.4225,
+      "step": 44380
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.6996056254185582e-05,
+      "loss": 0.4339,
+      "step": 44390
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.699307984225017e-05,
+      "loss": 0.4184,
+      "step": 44400
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.699010343031476e-05,
+      "loss": 0.4302,
+      "step": 44410
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.6987127018379346e-05,
+      "loss": 0.4337,
+      "step": 44420
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.6984150606443936e-05,
+      "loss": 0.4339,
+      "step": 44430
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.6981174194508523e-05,
+      "loss": 0.4169,
+      "step": 44440
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.6978197782573114e-05,
+      "loss": 0.412,
+      "step": 44450
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.6975221370637697e-05,
+      "loss": 0.421,
+      "step": 44460
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.6972244958702284e-05,
+      "loss": 0.403,
+      "step": 44470
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.6969268546766874e-05,
+      "loss": 0.4084,
+      "step": 44480
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.696629213483146e-05,
+      "loss": 0.4267,
+      "step": 44490
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.696331572289605e-05,
+      "loss": 0.4135,
+      "step": 44500
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.696033931096064e-05,
+      "loss": 0.4362,
+      "step": 44510
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.695736289902523e-05,
+      "loss": 0.4226,
+      "step": 44520
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.6954386487089816e-05,
+      "loss": 0.4032,
+      "step": 44530
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.6951410075154406e-05,
+      "loss": 0.4205,
+      "step": 44540
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.6948433663218993e-05,
+      "loss": 0.4058,
+      "step": 44550
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.6945457251283576e-05,
+      "loss": 0.4079,
+      "step": 44560
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.6942480839348167e-05,
+      "loss": 0.4168,
+      "step": 44570
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.6939504427412753e-05,
+      "loss": 0.4142,
+      "step": 44580
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.6936528015477344e-05,
+      "loss": 0.4225,
+      "step": 44590
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.693355160354193e-05,
+      "loss": 0.437,
+      "step": 44600
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.693057519160652e-05,
+      "loss": 0.4152,
+      "step": 44610
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6927598779671108e-05,
+      "loss": 0.4235,
+      "step": 44620
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6924622367735698e-05,
+      "loss": 0.4156,
+      "step": 44630
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.6921645955800285e-05,
+      "loss": 0.4373,
+      "step": 44640
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.6918669543864875e-05,
+      "loss": 0.4078,
+      "step": 44650
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.6915693131929462e-05,
+      "loss": 0.4049,
+      "step": 44660
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.691271671999405e-05,
+      "loss": 0.4168,
+      "step": 44670
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.6909740308058636e-05,
+      "loss": 0.4304,
+      "step": 44680
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.6906763896123226e-05,
+      "loss": 0.436,
+      "step": 44690
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.6903787484187813e-05,
+      "loss": 0.4211,
+      "step": 44700
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.6900811072252403e-05,
+      "loss": 0.4308,
+      "step": 44710
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.689783466031699e-05,
+      "loss": 0.4276,
+      "step": 44720
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.689485824838158e-05,
+      "loss": 0.426,
+      "step": 44730
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.6891881836446168e-05,
+      "loss": 0.4277,
+      "step": 44740
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.6888905424510754e-05,
+      "loss": 0.4218,
+      "step": 44750
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.6885929012575345e-05,
+      "loss": 0.4164,
+      "step": 44760
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.6882952600639928e-05,
+      "loss": 0.4145,
+      "step": 44770
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.687997618870452e-05,
+      "loss": 0.4211,
+      "step": 44780
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.6876999776769105e-05,
+      "loss": 0.4055,
+      "step": 44790
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.6874023364833696e-05,
+      "loss": 0.4289,
+      "step": 44800
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.6871046952898283e-05,
+      "loss": 0.4145,
+      "step": 44810
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.6868070540962873e-05,
+      "loss": 0.4482,
+      "step": 44820
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.686509412902746e-05,
+      "loss": 0.4139,
+      "step": 44830
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.686211771709205e-05,
+      "loss": 0.4099,
+      "step": 44840
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.6859141305156637e-05,
+      "loss": 0.4239,
+      "step": 44850
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.6856164893221227e-05,
+      "loss": 0.4258,
+      "step": 44860
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.685318848128581e-05,
+      "loss": 0.4303,
+      "step": 44870
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.6850212069350398e-05,
+      "loss": 0.4278,
+      "step": 44880
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.6847235657414988e-05,
+      "loss": 0.4174,
+      "step": 44890
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.6844259245479575e-05,
+      "loss": 0.4264,
+      "step": 44900
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.6841282833544165e-05,
+      "loss": 0.4321,
+      "step": 44910
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.6838306421608752e-05,
+      "loss": 0.4186,
+      "step": 44920
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.6835330009673342e-05,
+      "loss": 0.4186,
+      "step": 44930
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.683235359773793e-05,
+      "loss": 0.4325,
+      "step": 44940
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.682937718580252e-05,
+      "loss": 0.4265,
+      "step": 44950
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.6826400773867106e-05,
+      "loss": 0.4212,
+      "step": 44960
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.682342436193169e-05,
+      "loss": 0.4119,
+      "step": 44970
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.682044794999628e-05,
+      "loss": 0.4263,
+      "step": 44980
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.6817471538060867e-05,
+      "loss": 0.4171,
+      "step": 44990
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.6814495126125457e-05,
+      "loss": 0.412,
+      "step": 45000
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.6811518714190044e-05,
+      "loss": 0.4107,
+      "step": 45010
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.6808542302254635e-05,
+      "loss": 0.4143,
+      "step": 45020
+    },
+    {
+      "epoch": 10.0,
+      "eval_cer": 4.564102564102564,
+      "eval_loss": 0.42040038108825684,
+      "eval_runtime": 1.3145,
+      "eval_samples_per_second": 7.608,
+      "eval_steps_per_second": 0.761,
+      "eval_wer": 0.5324675324675324,
+      "step": 45020
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.680556589031922e-05,
+      "loss": 0.4206,
+      "step": 45030
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.6802589478383812e-05,
+      "loss": 0.4085,
+      "step": 45040
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.67996130664484e-05,
+      "loss": 0.4186,
+      "step": 45050
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.679663665451299e-05,
+      "loss": 0.4178,
+      "step": 45060
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.6793660242577576e-05,
+      "loss": 0.414,
+      "step": 45070
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.6790683830642163e-05,
+      "loss": 0.4069,
+      "step": 45080
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.678770741870675e-05,
+      "loss": 0.4115,
+      "step": 45090
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.678473100677134e-05,
+      "loss": 0.3994,
+      "step": 45100
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.6781754594835927e-05,
+      "loss": 0.4027,
+      "step": 45110
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.6778778182900517e-05,
+      "loss": 0.4159,
+      "step": 45120
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.6775801770965104e-05,
+      "loss": 0.4116,
+      "step": 45130
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.6772825359029694e-05,
+      "loss": 0.4153,
+      "step": 45140
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.676984894709428e-05,
+      "loss": 0.4104,
+      "step": 45150
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.676687253515887e-05,
+      "loss": 0.3891,
+      "step": 45160
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.676389612322346e-05,
+      "loss": 0.412,
+      "step": 45170
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.6760919711288042e-05,
+      "loss": 0.4126,
+      "step": 45180
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.6757943299352632e-05,
+      "loss": 0.3996,
+      "step": 45190
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.675496688741722e-05,
+      "loss": 0.3975,
+      "step": 45200
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.675199047548181e-05,
+      "loss": 0.423,
+      "step": 45210
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.6749014063546396e-05,
+      "loss": 0.4022,
+      "step": 45220
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.6746037651610987e-05,
+      "loss": 0.4167,
+      "step": 45230
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.6743061239675573e-05,
+      "loss": 0.4057,
+      "step": 45240
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.6740084827740164e-05,
+      "loss": 0.4331,
+      "step": 45250
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.673710841580475e-05,
+      "loss": 0.4135,
+      "step": 45260
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.673413200386934e-05,
+      "loss": 0.402,
+      "step": 45270
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.6731155591933924e-05,
+      "loss": 0.4059,
+      "step": 45280
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.672817917999851e-05,
+      "loss": 0.4135,
+      "step": 45290
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.67252027680631e-05,
+      "loss": 0.394,
+      "step": 45300
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.672222635612769e-05,
+      "loss": 0.4116,
+      "step": 45310
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.671924994419228e-05,
+      "loss": 0.4058,
+      "step": 45320
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.6716273532256866e-05,
+      "loss": 0.402,
+      "step": 45330
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.6713297120321456e-05,
+      "loss": 0.4189,
+      "step": 45340
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.6710320708386043e-05,
+      "loss": 0.3995,
+      "step": 45350
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.6707344296450633e-05,
+      "loss": 0.406,
+      "step": 45360
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.670436788451522e-05,
+      "loss": 0.4095,
+      "step": 45370
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.6701391472579804e-05,
+      "loss": 0.4117,
+      "step": 45380
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.6698415060644394e-05,
+      "loss": 0.4168,
+      "step": 45390
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.669543864870898e-05,
+      "loss": 0.4029,
+      "step": 45400
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.669246223677357e-05,
+      "loss": 0.4116,
+      "step": 45410
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.6689485824838158e-05,
+      "loss": 0.4064,
+      "step": 45420
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.6686509412902748e-05,
+      "loss": 0.3983,
+      "step": 45430
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.6683533000967335e-05,
+      "loss": 0.4023,
+      "step": 45440
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.6680556589031925e-05,
+      "loss": 0.4139,
+      "step": 45450
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.6677580177096512e-05,
+      "loss": 0.4133,
+      "step": 45460
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.6674603765161103e-05,
+      "loss": 0.3984,
+      "step": 45470
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.6671627353225686e-05,
+      "loss": 0.4002,
+      "step": 45480
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.6668650941290276e-05,
+      "loss": 0.4061,
+      "step": 45490
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.6665674529354863e-05,
+      "loss": 0.408,
+      "step": 45500
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.6662698117419454e-05,
+      "loss": 0.4069,
+      "step": 45510
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.665972170548404e-05,
+      "loss": 0.4153,
+      "step": 45520
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.665674529354863e-05,
+      "loss": 0.4055,
+      "step": 45530
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.6653768881613218e-05,
+      "loss": 0.4289,
+      "step": 45540
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.6650792469677808e-05,
+      "loss": 0.4086,
+      "step": 45550
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.6647816057742395e-05,
+      "loss": 0.4119,
+      "step": 45560
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.6644839645806985e-05,
+      "loss": 0.4153,
+      "step": 45570
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.6641863233871572e-05,
+      "loss": 0.412,
+      "step": 45580
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.6638886821936156e-05,
+      "loss": 0.4294,
+      "step": 45590
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.6635910410000746e-05,
+      "loss": 0.4049,
+      "step": 45600
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.6632933998065333e-05,
+      "loss": 0.4211,
+      "step": 45610
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.6629957586129923e-05,
+      "loss": 0.423,
+      "step": 45620
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.662698117419451e-05,
+      "loss": 0.4033,
+      "step": 45630
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.66240047622591e-05,
+      "loss": 0.4196,
+      "step": 45640
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.6621028350323687e-05,
+      "loss": 0.421,
+      "step": 45650
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.6618051938388277e-05,
+      "loss": 0.4125,
+      "step": 45660
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.6615075526452864e-05,
+      "loss": 0.4118,
+      "step": 45670
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.6612099114517455e-05,
+      "loss": 0.4164,
+      "step": 45680
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.6609122702582038e-05,
+      "loss": 0.3959,
+      "step": 45690
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.6606146290646625e-05,
+      "loss": 0.411,
+      "step": 45700
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.6603169878711215e-05,
+      "loss": 0.4075,
+      "step": 45710
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.6600193466775802e-05,
+      "loss": 0.3886,
+      "step": 45720
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.6597217054840393e-05,
+      "loss": 0.4155,
+      "step": 45730
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.659424064290498e-05,
+      "loss": 0.4149,
+      "step": 45740
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.659126423096957e-05,
+      "loss": 0.4032,
+      "step": 45750
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.6588287819034157e-05,
+      "loss": 0.405,
+      "step": 45760
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.6585311407098747e-05,
+      "loss": 0.4086,
+      "step": 45770
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.6582334995163334e-05,
+      "loss": 0.4202,
+      "step": 45780
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.6579358583227917e-05,
+      "loss": 0.4097,
+      "step": 45790
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.6576382171292508e-05,
+      "loss": 0.4257,
+      "step": 45800
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.6573405759357094e-05,
+      "loss": 0.4231,
+      "step": 45810
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.6570429347421685e-05,
+      "loss": 0.4061,
+      "step": 45820
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.656745293548627e-05,
+      "loss": 0.4049,
+      "step": 45830
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.6564476523550862e-05,
+      "loss": 0.4133,
+      "step": 45840
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.656150011161545e-05,
+      "loss": 0.4049,
+      "step": 45850
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.655852369968004e-05,
+      "loss": 0.4156,
+      "step": 45860
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.6555547287744626e-05,
+      "loss": 0.4297,
+      "step": 45870
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.6552570875809216e-05,
+      "loss": 0.4174,
+      "step": 45880
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.65495944638738e-05,
+      "loss": 0.4129,
+      "step": 45890
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.654661805193839e-05,
+      "loss": 0.4115,
+      "step": 45900
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.6543641640002977e-05,
+      "loss": 0.3879,
+      "step": 45910
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.6540665228067567e-05,
+      "loss": 0.4015,
+      "step": 45920
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.6537688816132154e-05,
+      "loss": 0.3972,
+      "step": 45930
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.6534712404196745e-05,
+      "loss": 0.4073,
+      "step": 45940
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.653173599226133e-05,
+      "loss": 0.4196,
+      "step": 45950
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.652875958032592e-05,
+      "loss": 0.4012,
+      "step": 45960
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.652578316839051e-05,
+      "loss": 0.396,
+      "step": 45970
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.65228067564551e-05,
+      "loss": 0.4051,
+      "step": 45980
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.6519830344519686e-05,
+      "loss": 0.4339,
+      "step": 45990
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.651685393258427e-05,
+      "loss": 0.4087,
+      "step": 46000
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.651387752064886e-05,
+      "loss": 0.4018,
+      "step": 46010
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.6510901108713446e-05,
+      "loss": 0.4031,
+      "step": 46020
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.6507924696778037e-05,
+      "loss": 0.4103,
+      "step": 46030
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.6504948284842624e-05,
+      "loss": 0.4016,
+      "step": 46040
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.6501971872907214e-05,
+      "loss": 0.4082,
+      "step": 46050
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.64989954609718e-05,
+      "loss": 0.4134,
+      "step": 46060
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.649601904903639e-05,
+      "loss": 0.4111,
+      "step": 46070
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.6493042637100978e-05,
+      "loss": 0.3895,
+      "step": 46080
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.649006622516557e-05,
+      "loss": 0.4041,
+      "step": 46090
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.6487089813230152e-05,
+      "loss": 0.4187,
+      "step": 46100
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.648411340129474e-05,
+      "loss": 0.4104,
+      "step": 46110
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.648113698935933e-05,
+      "loss": 0.4116,
+      "step": 46120
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.6478160577423916e-05,
+      "loss": 0.4073,
+      "step": 46130
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.6475184165488506e-05,
+      "loss": 0.4162,
+      "step": 46140
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.6472207753553093e-05,
+      "loss": 0.4031,
+      "step": 46150
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.6469231341617683e-05,
+      "loss": 0.4102,
+      "step": 46160
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.646625492968227e-05,
+      "loss": 0.4132,
+      "step": 46170
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.646327851774686e-05,
+      "loss": 0.4063,
+      "step": 46180
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.6460302105811447e-05,
+      "loss": 0.413,
+      "step": 46190
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.645732569387603e-05,
+      "loss": 0.4191,
+      "step": 46200
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.645434928194062e-05,
+      "loss": 0.4189,
+      "step": 46210
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.6451372870005208e-05,
+      "loss": 0.404,
+      "step": 46220
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.64483964580698e-05,
+      "loss": 0.4117,
+      "step": 46230
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.6445420046134385e-05,
+      "loss": 0.3962,
+      "step": 46240
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.6442443634198976e-05,
+      "loss": 0.402,
+      "step": 46250
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.6439467222263563e-05,
+      "loss": 0.4083,
+      "step": 46260
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.6436490810328153e-05,
+      "loss": 0.4189,
+      "step": 46270
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.643351439839274e-05,
+      "loss": 0.4067,
+      "step": 46280
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.643053798645733e-05,
+      "loss": 0.4078,
+      "step": 46290
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.6427561574521914e-05,
+      "loss": 0.4071,
+      "step": 46300
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.6424585162586504e-05,
+      "loss": 0.4044,
+      "step": 46310
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.642160875065109e-05,
+      "loss": 0.4093,
+      "step": 46320
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.641863233871568e-05,
+      "loss": 0.3998,
+      "step": 46330
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.6415655926780268e-05,
+      "loss": 0.3991,
+      "step": 46340
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.6412679514844858e-05,
+      "loss": 0.4126,
+      "step": 46350
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.6409703102909445e-05,
+      "loss": 0.4154,
+      "step": 46360
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.6406726690974035e-05,
+      "loss": 0.4129,
+      "step": 46370
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.6403750279038622e-05,
+      "loss": 0.4156,
+      "step": 46380
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.6400773867103213e-05,
+      "loss": 0.4207,
+      "step": 46390
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.63977974551678e-05,
+      "loss": 0.4047,
+      "step": 46400
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.6394821043232383e-05,
+      "loss": 0.4006,
+      "step": 46410
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.6391844631296973e-05,
+      "loss": 0.4093,
+      "step": 46420
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.638886821936156e-05,
+      "loss": 0.4079,
+      "step": 46430
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.638589180742615e-05,
+      "loss": 0.406,
+      "step": 46440
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.6382915395490737e-05,
+      "loss": 0.403,
+      "step": 46450
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.6379938983555328e-05,
+      "loss": 0.4038,
+      "step": 46460
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.6376962571619915e-05,
+      "loss": 0.4177,
+      "step": 46470
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.6373986159684505e-05,
+      "loss": 0.4196,
+      "step": 46480
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.6371009747749092e-05,
+      "loss": 0.4008,
+      "step": 46490
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.6368033335813682e-05,
+      "loss": 0.4128,
+      "step": 46500
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.6365056923878265e-05,
+      "loss": 0.4134,
+      "step": 46510
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.6362080511942852e-05,
+      "loss": 0.4161,
+      "step": 46520
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.6359104100007443e-05,
+      "loss": 0.3984,
+      "step": 46530
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.635612768807203e-05,
+      "loss": 0.4162,
+      "step": 46540
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.635315127613662e-05,
+      "loss": 0.4044,
+      "step": 46550
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.6350174864201207e-05,
+      "loss": 0.4197,
+      "step": 46560
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.6347198452265797e-05,
+      "loss": 0.3835,
+      "step": 46570
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.6344222040330384e-05,
+      "loss": 0.4188,
+      "step": 46580
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.6341245628394974e-05,
+      "loss": 0.4102,
+      "step": 46590
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.633826921645956e-05,
+      "loss": 0.4055,
+      "step": 46600
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.6335292804524145e-05,
+      "loss": 0.4119,
+      "step": 46610
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.6332316392588735e-05,
+      "loss": 0.4099,
+      "step": 46620
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.6329339980653322e-05,
+      "loss": 0.4238,
+      "step": 46630
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.6326363568717912e-05,
+      "loss": 0.4057,
+      "step": 46640
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.63233871567825e-05,
+      "loss": 0.4152,
+      "step": 46650
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.632041074484709e-05,
+      "loss": 0.4144,
+      "step": 46660
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.6317434332911676e-05,
+      "loss": 0.4132,
+      "step": 46670
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.6314457920976266e-05,
+      "loss": 0.4011,
+      "step": 46680
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.6311481509040853e-05,
+      "loss": 0.3992,
+      "step": 46690
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.6308505097105444e-05,
+      "loss": 0.4144,
+      "step": 46700
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.6305528685170027e-05,
+      "loss": 0.4184,
+      "step": 46710
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.6302552273234617e-05,
+      "loss": 0.4029,
+      "step": 46720
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.6299575861299204e-05,
+      "loss": 0.4043,
+      "step": 46730
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.6296599449363795e-05,
+      "loss": 0.4111,
+      "step": 46740
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.629362303742838e-05,
+      "loss": 0.4228,
+      "step": 46750
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.6290646625492972e-05,
+      "loss": 0.4094,
+      "step": 46760
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.628767021355756e-05,
+      "loss": 0.415,
+      "step": 46770
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.628469380162215e-05,
+      "loss": 0.3919,
+      "step": 46780
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.6281717389686736e-05,
+      "loss": 0.397,
+      "step": 46790
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.6278740977751326e-05,
+      "loss": 0.4018,
+      "step": 46800
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.6275764565815913e-05,
+      "loss": 0.4032,
+      "step": 46810
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.6272788153880497e-05,
+      "loss": 0.4033,
+      "step": 46820
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.6269811741945087e-05,
+      "loss": 0.4045,
+      "step": 46830
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.6266835330009674e-05,
+      "loss": 0.3996,
+      "step": 46840
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.6263858918074264e-05,
+      "loss": 0.4105,
+      "step": 46850
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.626088250613885e-05,
+      "loss": 0.4127,
+      "step": 46860
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.625790609420344e-05,
+      "loss": 0.4053,
+      "step": 46870
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.6254929682268028e-05,
+      "loss": 0.4106,
+      "step": 46880
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.625195327033262e-05,
+      "loss": 0.4046,
+      "step": 46890
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.6248976858397205e-05,
+      "loss": 0.4163,
+      "step": 46900
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.6246000446461796e-05,
+      "loss": 0.3922,
+      "step": 46910
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.624302403452638e-05,
+      "loss": 0.4091,
+      "step": 46920
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.6240047622590966e-05,
+      "loss": 0.3861,
+      "step": 46930
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.6237071210655556e-05,
+      "loss": 0.4202,
+      "step": 46940
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.6234094798720143e-05,
+      "loss": 0.4028,
+      "step": 46950
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.6231118386784734e-05,
+      "loss": 0.4023,
+      "step": 46960
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.622814197484932e-05,
+      "loss": 0.409,
+      "step": 46970
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.622516556291391e-05,
+      "loss": 0.398,
+      "step": 46980
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.6222189150978498e-05,
+      "loss": 0.4123,
+      "step": 46990
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.6219212739043088e-05,
+      "loss": 0.422,
+      "step": 47000
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.6216236327107675e-05,
+      "loss": 0.4206,
+      "step": 47010
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.621325991517226e-05,
+      "loss": 0.4137,
+      "step": 47020
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.621028350323685e-05,
+      "loss": 0.4088,
+      "step": 47030
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.6207307091301436e-05,
+      "loss": 0.4168,
+      "step": 47040
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.6204330679366026e-05,
+      "loss": 0.4085,
+      "step": 47050
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.6201354267430613e-05,
+      "loss": 0.3998,
+      "step": 47060
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.6198377855495203e-05,
+      "loss": 0.3967,
+      "step": 47070
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.619540144355979e-05,
+      "loss": 0.4185,
+      "step": 47080
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.619242503162438e-05,
+      "loss": 0.4069,
+      "step": 47090
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.6189448619688967e-05,
+      "loss": 0.4121,
+      "step": 47100
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.6186472207753557e-05,
+      "loss": 0.4139,
+      "step": 47110
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.618349579581814e-05,
+      "loss": 0.4069,
+      "step": 47120
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.618051938388273e-05,
+      "loss": 0.4052,
+      "step": 47130
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.6177542971947318e-05,
+      "loss": 0.4089,
+      "step": 47140
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.617456656001191e-05,
+      "loss": 0.409,
+      "step": 47150
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.6171590148076495e-05,
+      "loss": 0.417,
+      "step": 47160
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.6168613736141086e-05,
+      "loss": 0.4195,
+      "step": 47170
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.6165637324205672e-05,
+      "loss": 0.4206,
+      "step": 47180
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.6162660912270263e-05,
+      "loss": 0.3999,
+      "step": 47190
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.615968450033485e-05,
+      "loss": 0.3966,
+      "step": 47200
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.615670808839944e-05,
+      "loss": 0.4321,
+      "step": 47210
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.6153731676464023e-05,
+      "loss": 0.4362,
+      "step": 47220
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.615075526452861e-05,
+      "loss": 0.4093,
+      "step": 47230
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.61477788525932e-05,
+      "loss": 0.4255,
+      "step": 47240
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.6144802440657787e-05,
+      "loss": 0.4174,
+      "step": 47250
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.6141826028722378e-05,
+      "loss": 0.4258,
+      "step": 47260
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.6138849616786965e-05,
+      "loss": 0.4144,
+      "step": 47270
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.6135873204851555e-05,
+      "loss": 0.3964,
+      "step": 47280
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.6132896792916142e-05,
+      "loss": 0.4015,
+      "step": 47290
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.6129920380980732e-05,
+      "loss": 0.4137,
+      "step": 47300
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.612694396904532e-05,
+      "loss": 0.3864,
+      "step": 47310
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.612396755710991e-05,
+      "loss": 0.4057,
+      "step": 47320
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.6120991145174493e-05,
+      "loss": 0.4041,
+      "step": 47330
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.611801473323908e-05,
+      "loss": 0.396,
+      "step": 47340
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.611503832130367e-05,
+      "loss": 0.4022,
+      "step": 47350
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.6112061909368257e-05,
+      "loss": 0.4101,
+      "step": 47360
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.6109085497432847e-05,
+      "loss": 0.4051,
+      "step": 47370
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.6106109085497434e-05,
+      "loss": 0.4086,
+      "step": 47380
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.6103132673562024e-05,
+      "loss": 0.4115,
+      "step": 47390
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.610015626162661e-05,
+      "loss": 0.3991,
+      "step": 47400
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.60971798496912e-05,
+      "loss": 0.4137,
+      "step": 47410
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.609420343775579e-05,
+      "loss": 0.4003,
+      "step": 47420
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.6091227025820375e-05,
+      "loss": 0.4061,
+      "step": 47430
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.6088250613884962e-05,
+      "loss": 0.4007,
+      "step": 47440
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.608527420194955e-05,
+      "loss": 0.3889,
+      "step": 47450
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.608229779001414e-05,
+      "loss": 0.3984,
+      "step": 47460
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.6079321378078726e-05,
+      "loss": 0.416,
+      "step": 47470
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.6076344966143317e-05,
+      "loss": 0.4058,
+      "step": 47480
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.6073368554207904e-05,
+      "loss": 0.4019,
+      "step": 47490
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.6070392142272494e-05,
+      "loss": 0.4216,
+      "step": 47500
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.606741573033708e-05,
+      "loss": 0.4151,
+      "step": 47510
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.606443931840167e-05,
+      "loss": 0.4031,
+      "step": 47520
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.6061462906466255e-05,
+      "loss": 0.3975,
+      "step": 47530
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.6058486494530845e-05,
+      "loss": 0.4165,
+      "step": 47540
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.6055510082595432e-05,
+      "loss": 0.4171,
+      "step": 47550
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.6052533670660022e-05,
+      "loss": 0.4107,
+      "step": 47560
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.604955725872461e-05,
+      "loss": 0.4027,
+      "step": 47570
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.60465808467892e-05,
+      "loss": 0.4155,
+      "step": 47580
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.6043604434853786e-05,
+      "loss": 0.4113,
+      "step": 47590
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.6040628022918376e-05,
+      "loss": 0.4134,
+      "step": 47600
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.6037651610982963e-05,
+      "loss": 0.4127,
+      "step": 47610
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.6034675199047554e-05,
+      "loss": 0.4173,
+      "step": 47620
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.6031698787112137e-05,
+      "loss": 0.3998,
+      "step": 47630
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.6028722375176724e-05,
+      "loss": 0.4091,
+      "step": 47640
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.6025745963241314e-05,
+      "loss": 0.411,
+      "step": 47650
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.60227695513059e-05,
+      "loss": 0.4089,
+      "step": 47660
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.601979313937049e-05,
+      "loss": 0.405,
+      "step": 47670
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.601681672743508e-05,
+      "loss": 0.4108,
+      "step": 47680
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.601384031549967e-05,
+      "loss": 0.3964,
+      "step": 47690
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.6010863903564256e-05,
+      "loss": 0.4061,
+      "step": 47700
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.6007887491628846e-05,
+      "loss": 0.4091,
+      "step": 47710
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.6004911079693433e-05,
+      "loss": 0.3882,
+      "step": 47720
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.6001934667758023e-05,
+      "loss": 0.3983,
+      "step": 47730
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.5998958255822607e-05,
+      "loss": 0.4169,
+      "step": 47740
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.5995981843887193e-05,
+      "loss": 0.3979,
+      "step": 47750
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.5993005431951784e-05,
+      "loss": 0.398,
+      "step": 47760
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.599002902001637e-05,
+      "loss": 0.4065,
+      "step": 47770
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.598705260808096e-05,
+      "loss": 0.4251,
+      "step": 47780
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.5984076196145548e-05,
+      "loss": 0.4026,
+      "step": 47790
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.5981099784210138e-05,
+      "loss": 0.4072,
+      "step": 47800
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.5978123372274725e-05,
+      "loss": 0.4152,
+      "step": 47810
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.5975146960339315e-05,
+      "loss": 0.4201,
+      "step": 47820
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.5972170548403902e-05,
+      "loss": 0.3889,
+      "step": 47830
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.596919413646849e-05,
+      "loss": 0.4063,
+      "step": 47840
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.5966217724533076e-05,
+      "loss": 0.4047,
+      "step": 47850
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.5963241312597666e-05,
+      "loss": 0.4096,
+      "step": 47860
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.5960264900662253e-05,
+      "loss": 0.402,
+      "step": 47870
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.595728848872684e-05,
+      "loss": 0.4125,
+      "step": 47880
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.595431207679143e-05,
+      "loss": 0.3972,
+      "step": 47890
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.5951335664856017e-05,
+      "loss": 0.4063,
+      "step": 47900
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.5948359252920608e-05,
+      "loss": 0.4038,
+      "step": 47910
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.5945382840985194e-05,
+      "loss": 0.3942,
+      "step": 47920
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.5942406429049785e-05,
+      "loss": 0.3892,
+      "step": 47930
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.5939430017114368e-05,
+      "loss": 0.3994,
+      "step": 47940
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.593645360517896e-05,
+      "loss": 0.3979,
+      "step": 47950
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.5933477193243545e-05,
+      "loss": 0.4103,
+      "step": 47960
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.5930500781308136e-05,
+      "loss": 0.404,
+      "step": 47970
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.5927524369372723e-05,
+      "loss": 0.4146,
+      "step": 47980
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.5924547957437313e-05,
+      "loss": 0.4113,
+      "step": 47990
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.59215715455019e-05,
+      "loss": 0.4035,
+      "step": 48000
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.591859513356649e-05,
+      "loss": 0.4129,
+      "step": 48010
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.5915618721631077e-05,
+      "loss": 0.417,
+      "step": 48020
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.5912642309695667e-05,
+      "loss": 0.4079,
+      "step": 48030
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.590966589776025e-05,
+      "loss": 0.4062,
+      "step": 48040
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.5906689485824838e-05,
+      "loss": 0.4222,
+      "step": 48050
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.5903713073889428e-05,
+      "loss": 0.4155,
+      "step": 48060
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.5900736661954015e-05,
+      "loss": 0.4008,
+      "step": 48070
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.5897760250018605e-05,
+      "loss": 0.396,
+      "step": 48080
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.5894783838083192e-05,
+      "loss": 0.4064,
+      "step": 48090
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.5891807426147782e-05,
+      "loss": 0.405,
+      "step": 48100
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.588883101421237e-05,
+      "loss": 0.3901,
+      "step": 48110
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.588585460227696e-05,
+      "loss": 0.4095,
+      "step": 48120
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.5882878190341546e-05,
+      "loss": 0.4106,
+      "step": 48130
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.5879901778406137e-05,
+      "loss": 0.391,
+      "step": 48140
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.587692536647072e-05,
+      "loss": 0.3912,
+      "step": 48150
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.5873948954535307e-05,
+      "loss": 0.4124,
+      "step": 48160
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.5870972542599897e-05,
+      "loss": 0.4002,
+      "step": 48170
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.5867996130664484e-05,
+      "loss": 0.4097,
+      "step": 48180
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.5865019718729075e-05,
+      "loss": 0.4081,
+      "step": 48190
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.586204330679366e-05,
+      "loss": 0.4056,
+      "step": 48200
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.5859066894858252e-05,
+      "loss": 0.4049,
+      "step": 48210
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.585609048292284e-05,
+      "loss": 0.4114,
+      "step": 48220
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.585311407098743e-05,
+      "loss": 0.4047,
+      "step": 48230
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.5850137659052016e-05,
+      "loss": 0.418,
+      "step": 48240
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.5847161247116603e-05,
+      "loss": 0.4141,
+      "step": 48250
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.584418483518119e-05,
+      "loss": 0.4079,
+      "step": 48260
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.584120842324578e-05,
+      "loss": 0.3946,
+      "step": 48270
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.5838232011310367e-05,
+      "loss": 0.4108,
+      "step": 48280
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.5835255599374954e-05,
+      "loss": 0.4037,
+      "step": 48290
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.5832279187439544e-05,
+      "loss": 0.4094,
+      "step": 48300
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.582930277550413e-05,
+      "loss": 0.4133,
+      "step": 48310
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.582632636356872e-05,
+      "loss": 0.405,
+      "step": 48320
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.5823349951633308e-05,
+      "loss": 0.4107,
+      "step": 48330
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.58203735396979e-05,
+      "loss": 0.419,
+      "step": 48340
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.5817397127762482e-05,
+      "loss": 0.4014,
+      "step": 48350
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.5814420715827072e-05,
+      "loss": 0.4192,
+      "step": 48360
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.581144430389166e-05,
+      "loss": 0.4029,
+      "step": 48370
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.580846789195625e-05,
+      "loss": 0.395,
+      "step": 48380
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.5805491480020836e-05,
+      "loss": 0.4178,
+      "step": 48390
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.5802515068085427e-05,
+      "loss": 0.4102,
+      "step": 48400
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.5799538656150013e-05,
+      "loss": 0.3935,
+      "step": 48410
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.5796562244214604e-05,
+      "loss": 0.395,
+      "step": 48420
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.579358583227919e-05,
+      "loss": 0.4145,
+      "step": 48430
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.579060942034378e-05,
+      "loss": 0.4188,
+      "step": 48440
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.5787633008408364e-05,
+      "loss": 0.4015,
+      "step": 48450
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.578465659647295e-05,
+      "loss": 0.4025,
+      "step": 48460
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.578168018453754e-05,
+      "loss": 0.405,
+      "step": 48470
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.577870377260213e-05,
+      "loss": 0.4002,
+      "step": 48480
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.577572736066672e-05,
+      "loss": 0.4041,
+      "step": 48490
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.5772750948731306e-05,
+      "loss": 0.4145,
+      "step": 48500
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.5769774536795896e-05,
+      "loss": 0.41,
+      "step": 48510
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.5766798124860483e-05,
+      "loss": 0.3995,
+      "step": 48520
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.5763821712925073e-05,
+      "loss": 0.4149,
+      "step": 48530
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.576084530098966e-05,
+      "loss": 0.401,
+      "step": 48540
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.575786888905425e-05,
+      "loss": 0.3971,
+      "step": 48550
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.5754892477118834e-05,
+      "loss": 0.3979,
+      "step": 48560
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.575191606518342e-05,
+      "loss": 0.4083,
+      "step": 48570
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.574893965324801e-05,
+      "loss": 0.3951,
+      "step": 48580
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.5745963241312598e-05,
+      "loss": 0.3919,
+      "step": 48590
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.5742986829377188e-05,
+      "loss": 0.4049,
+      "step": 48600
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.5740010417441775e-05,
+      "loss": 0.3969,
+      "step": 48610
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.5737034005506365e-05,
+      "loss": 0.4107,
+      "step": 48620
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.5734057593570952e-05,
+      "loss": 0.4072,
+      "step": 48630
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.5731081181635543e-05,
+      "loss": 0.3974,
+      "step": 48640
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.572810476970013e-05,
+      "loss": 0.3983,
+      "step": 48650
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.5725128357764716e-05,
+      "loss": 0.4184,
+      "step": 48660
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.5722151945829303e-05,
+      "loss": 0.4181,
+      "step": 48670
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.5719175533893894e-05,
+      "loss": 0.4016,
+      "step": 48680
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.571619912195848e-05,
+      "loss": 0.4015,
+      "step": 48690
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.571322271002307e-05,
+      "loss": 0.4069,
+      "step": 48700
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.5710246298087658e-05,
+      "loss": 0.4107,
+      "step": 48710
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.5707269886152245e-05,
+      "loss": 0.3945,
+      "step": 48720
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.5704293474216835e-05,
+      "loss": 0.4012,
+      "step": 48730
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.5701317062281422e-05,
+      "loss": 0.4163,
+      "step": 48740
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.5698340650346012e-05,
+      "loss": 0.4108,
+      "step": 48750
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.5695364238410596e-05,
+      "loss": 0.3999,
+      "step": 48760
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.5692387826475186e-05,
+      "loss": 0.4174,
+      "step": 48770
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.5689411414539773e-05,
+      "loss": 0.3912,
+      "step": 48780
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.5686435002604363e-05,
+      "loss": 0.4158,
+      "step": 48790
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.568345859066895e-05,
+      "loss": 0.4031,
+      "step": 48800
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.568048217873354e-05,
+      "loss": 0.3963,
+      "step": 48810
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.5677505766798127e-05,
+      "loss": 0.4174,
+      "step": 48820
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.5674529354862717e-05,
+      "loss": 0.3993,
+      "step": 48830
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.5671552942927304e-05,
+      "loss": 0.3828,
+      "step": 48840
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.5668576530991895e-05,
+      "loss": 0.4136,
+      "step": 48850
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.5665600119056478e-05,
+      "loss": 0.3977,
+      "step": 48860
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.5662623707121065e-05,
+      "loss": 0.4214,
+      "step": 48870
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.5659647295185655e-05,
+      "loss": 0.4067,
+      "step": 48880
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.5656670883250242e-05,
+      "loss": 0.4058,
+      "step": 48890
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.5653694471314832e-05,
+      "loss": 0.3916,
+      "step": 48900
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.565071805937942e-05,
+      "loss": 0.3959,
+      "step": 48910
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.564774164744401e-05,
+      "loss": 0.398,
+      "step": 48920
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.5644765235508597e-05,
+      "loss": 0.4015,
+      "step": 48930
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.5641788823573187e-05,
+      "loss": 0.4106,
+      "step": 48940
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.5638812411637774e-05,
+      "loss": 0.4008,
+      "step": 48950
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.5635835999702357e-05,
+      "loss": 0.4188,
+      "step": 48960
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.5632859587766948e-05,
+      "loss": 0.3931,
+      "step": 48970
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.5629883175831534e-05,
+      "loss": 0.4038,
+      "step": 48980
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.5626906763896125e-05,
+      "loss": 0.3953,
+      "step": 48990
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.562393035196071e-05,
+      "loss": 0.4045,
+      "step": 49000
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.5620953940025302e-05,
+      "loss": 0.3971,
+      "step": 49010
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.561797752808989e-05,
+      "loss": 0.4143,
+      "step": 49020
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.561500111615448e-05,
+      "loss": 0.4121,
+      "step": 49030
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.5612024704219066e-05,
+      "loss": 0.4078,
+      "step": 49040
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.5609048292283656e-05,
+      "loss": 0.4075,
+      "step": 49050
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.5606071880348243e-05,
+      "loss": 0.4002,
+      "step": 49060
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.560309546841283e-05,
+      "loss": 0.4151,
+      "step": 49070
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.5600119056477417e-05,
+      "loss": 0.4187,
+      "step": 49080
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.5597142644542007e-05,
+      "loss": 0.41,
+      "step": 49090
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.5594166232606594e-05,
+      "loss": 0.4097,
+      "step": 49100
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.5591189820671184e-05,
+      "loss": 0.3981,
+      "step": 49110
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.558821340873577e-05,
+      "loss": 0.4176,
+      "step": 49120
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.558523699680036e-05,
+      "loss": 0.407,
+      "step": 49130
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.558226058486495e-05,
+      "loss": 0.408,
+      "step": 49140
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.5579284172929535e-05,
+      "loss": 0.3948,
+      "step": 49150
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.5576307760994126e-05,
+      "loss": 0.405,
+      "step": 49160
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.557333134905871e-05,
+      "loss": 0.4111,
+      "step": 49170
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.55703549371233e-05,
+      "loss": 0.4243,
+      "step": 49180
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.5567378525187886e-05,
+      "loss": 0.4021,
+      "step": 49190
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.5564402113252477e-05,
+      "loss": 0.4091,
+      "step": 49200
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.5561425701317064e-05,
+      "loss": 0.3978,
+      "step": 49210
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.5558449289381654e-05,
+      "loss": 0.3925,
+      "step": 49220
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.555547287744624e-05,
+      "loss": 0.4046,
+      "step": 49230
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.555249646551083e-05,
+      "loss": 0.4053,
+      "step": 49240
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.5549520053575418e-05,
+      "loss": 0.3943,
+      "step": 49250
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.5546543641640008e-05,
+      "loss": 0.4066,
+      "step": 49260
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.5543567229704592e-05,
+      "loss": 0.3915,
+      "step": 49270
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.554059081776918e-05,
+      "loss": 0.4167,
+      "step": 49280
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.553761440583377e-05,
+      "loss": 0.3973,
+      "step": 49290
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.5534637993898356e-05,
+      "loss": 0.4013,
+      "step": 49300
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.5531661581962946e-05,
+      "loss": 0.4008,
+      "step": 49310
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.5528685170027533e-05,
+      "loss": 0.4137,
+      "step": 49320
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.5525708758092123e-05,
+      "loss": 0.4128,
+      "step": 49330
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.552273234615671e-05,
+      "loss": 0.4018,
+      "step": 49340
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.55197559342213e-05,
+      "loss": 0.4051,
+      "step": 49350
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.5516779522285887e-05,
+      "loss": 0.3947,
+      "step": 49360
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.551380311035047e-05,
+      "loss": 0.3973,
+      "step": 49370
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.551082669841506e-05,
+      "loss": 0.3864,
+      "step": 49380
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.5507850286479648e-05,
+      "loss": 0.3969,
+      "step": 49390
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.550487387454424e-05,
+      "loss": 0.4008,
+      "step": 49400
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.5501897462608825e-05,
+      "loss": 0.4023,
+      "step": 49410
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.5498921050673416e-05,
+      "loss": 0.4096,
+      "step": 49420
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.5495944638738002e-05,
+      "loss": 0.41,
+      "step": 49430
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.5492968226802593e-05,
+      "loss": 0.395,
+      "step": 49440
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.548999181486718e-05,
+      "loss": 0.3998,
+      "step": 49450
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.548701540293177e-05,
+      "loss": 0.4028,
+      "step": 49460
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.5484038990996357e-05,
+      "loss": 0.3952,
+      "step": 49470
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.5481062579060944e-05,
+      "loss": 0.395,
+      "step": 49480
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.547808616712553e-05,
+      "loss": 0.3973,
+      "step": 49490
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.547510975519012e-05,
+      "loss": 0.3985,
+      "step": 49500
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.5472133343254708e-05,
+      "loss": 0.3967,
+      "step": 49510
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.5469156931319298e-05,
+      "loss": 0.3847,
+      "step": 49520
+    },
+    {
+      "epoch": 11.0,
+      "eval_cer": 4.555555555555555,
+      "eval_loss": 0.4103516936302185,
+      "eval_runtime": 1.2867,
+      "eval_samples_per_second": 7.772,
+      "eval_steps_per_second": 0.777,
+      "eval_wer": 0.6233766233766234,
+      "step": 49522
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.5466180519383885e-05,
+      "loss": 0.4063,
+      "step": 49530
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.5463204107448475e-05,
+      "loss": 0.404,
+      "step": 49540
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.5460227695513062e-05,
+      "loss": 0.3999,
+      "step": 49550
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.5457251283577653e-05,
+      "loss": 0.3721,
+      "step": 49560
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.545427487164224e-05,
+      "loss": 0.3869,
+      "step": 49570
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.5451298459706823e-05,
+      "loss": 0.4015,
+      "step": 49580
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.5448322047771413e-05,
+      "loss": 0.3896,
+      "step": 49590
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.5445345635836e-05,
+      "loss": 0.3979,
+      "step": 49600
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.544236922390059e-05,
+      "loss": 0.393,
+      "step": 49610
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.5439392811965177e-05,
+      "loss": 0.3977,
+      "step": 49620
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.5436416400029768e-05,
+      "loss": 0.3992,
+      "step": 49630
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.5433439988094354e-05,
+      "loss": 0.3951,
+      "step": 49640
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.5430463576158945e-05,
+      "loss": 0.391,
+      "step": 49650
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.542748716422353e-05,
+      "loss": 0.3879,
+      "step": 49660
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.5424510752288122e-05,
+      "loss": 0.3877,
+      "step": 49670
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.5421534340352705e-05,
+      "loss": 0.3889,
+      "step": 49680
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.5418557928417292e-05,
+      "loss": 0.3911,
+      "step": 49690
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.5415581516481883e-05,
+      "loss": 0.3791,
+      "step": 49700
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.541260510454647e-05,
+      "loss": 0.3979,
+      "step": 49710
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.540962869261106e-05,
+      "loss": 0.3953,
+      "step": 49720
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.5406652280675647e-05,
+      "loss": 0.3957,
+      "step": 49730
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.5403675868740237e-05,
+      "loss": 0.3874,
+      "step": 49740
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.5400699456804824e-05,
+      "loss": 0.4037,
+      "step": 49750
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.5397723044869414e-05,
+      "loss": 0.3904,
+      "step": 49760
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.5394746632934e-05,
+      "loss": 0.3989,
+      "step": 49770
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.5391770220998585e-05,
+      "loss": 0.398,
+      "step": 49780
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.5388793809063175e-05,
+      "loss": 0.4055,
+      "step": 49790
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.5385817397127762e-05,
+      "loss": 0.4009,
+      "step": 49800
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.5382840985192352e-05,
+      "loss": 0.3797,
+      "step": 49810
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.537986457325694e-05,
+      "loss": 0.3867,
+      "step": 49820
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.537688816132153e-05,
+      "loss": 0.3958,
+      "step": 49830
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.5373911749386116e-05,
+      "loss": 0.3891,
+      "step": 49840
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.5370935337450706e-05,
+      "loss": 0.4066,
+      "step": 49850
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.5367958925515293e-05,
+      "loss": 0.3862,
+      "step": 49860
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.5364982513579884e-05,
+      "loss": 0.3885,
+      "step": 49870
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.536200610164447e-05,
+      "loss": 0.4106,
+      "step": 49880
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.5359029689709057e-05,
+      "loss": 0.4055,
+      "step": 49890
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.5356053277773644e-05,
+      "loss": 0.3783,
+      "step": 49900
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.5353076865838235e-05,
+      "loss": 0.4102,
+      "step": 49910
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.535010045390282e-05,
+      "loss": 0.3999,
+      "step": 49920
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.5347124041967412e-05,
+      "loss": 0.3841,
+      "step": 49930
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.5344147630032e-05,
+      "loss": 0.3912,
+      "step": 49940
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.534117121809659e-05,
+      "loss": 0.3797,
+      "step": 49950
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.5338194806161176e-05,
+      "loss": 0.4146,
+      "step": 49960
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.5335218394225766e-05,
+      "loss": 0.4042,
+      "step": 49970
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.5332241982290353e-05,
+      "loss": 0.395,
+      "step": 49980
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.5329265570354937e-05,
+      "loss": 0.4021,
+      "step": 49990
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.5326289158419527e-05,
+      "loss": 0.386,
+      "step": 50000
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.5323312746484114e-05,
+      "loss": 0.396,
+      "step": 50010
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.5320336334548704e-05,
+      "loss": 0.3863,
+      "step": 50020
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.531735992261329e-05,
+      "loss": 0.3963,
+      "step": 50030
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.531438351067788e-05,
+      "loss": 0.3844,
+      "step": 50040
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.5311407098742468e-05,
+      "loss": 0.3815,
+      "step": 50050
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.530843068680706e-05,
+      "loss": 0.3903,
+      "step": 50060
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.5305454274871645e-05,
+      "loss": 0.3924,
+      "step": 50070
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.5302477862936236e-05,
+      "loss": 0.3792,
+      "step": 50080
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.529950145100082e-05,
+      "loss": 0.3993,
+      "step": 50090
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.5296525039065406e-05,
+      "loss": 0.3884,
+      "step": 50100
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.5293548627129996e-05,
+      "loss": 0.3905,
+      "step": 50110
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.5290572215194583e-05,
+      "loss": 0.4006,
+      "step": 50120
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.5287595803259174e-05,
+      "loss": 0.3974,
+      "step": 50130
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.528461939132376e-05,
+      "loss": 0.4015,
+      "step": 50140
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.528164297938835e-05,
+      "loss": 0.3786,
+      "step": 50150
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.5278666567452938e-05,
+      "loss": 0.3969,
+      "step": 50160
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.5275690155517528e-05,
+      "loss": 0.3922,
+      "step": 50170
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.5272713743582115e-05,
+      "loss": 0.3706,
+      "step": 50180
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.5269737331646698e-05,
+      "loss": 0.3809,
+      "step": 50190
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.526676091971129e-05,
+      "loss": 0.3949,
+      "step": 50200
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.5263784507775875e-05,
+      "loss": 0.377,
+      "step": 50210
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.5260808095840466e-05,
+      "loss": 0.3978,
+      "step": 50220
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.5257831683905053e-05,
+      "loss": 0.3881,
+      "step": 50230
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.5254855271969643e-05,
+      "loss": 0.3842,
+      "step": 50240
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.525187886003423e-05,
+      "loss": 0.3841,
+      "step": 50250
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.524890244809882e-05,
+      "loss": 0.3994,
+      "step": 50260
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.5245926036163407e-05,
+      "loss": 0.383,
+      "step": 50270
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.5242949624227997e-05,
+      "loss": 0.388,
+      "step": 50280
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.5239973212292584e-05,
+      "loss": 0.3903,
+      "step": 50290
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.523699680035717e-05,
+      "loss": 0.4109,
+      "step": 50300
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.5234020388421758e-05,
+      "loss": 0.4053,
+      "step": 50310
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.5231043976486348e-05,
+      "loss": 0.3984,
+      "step": 50320
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.5228067564550935e-05,
+      "loss": 0.3857,
+      "step": 50330
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.5225091152615525e-05,
+      "loss": 0.3865,
+      "step": 50340
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.5222114740680112e-05,
+      "loss": 0.3861,
+      "step": 50350
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.5219138328744703e-05,
+      "loss": 0.3775,
+      "step": 50360
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.521616191680929e-05,
+      "loss": 0.3965,
+      "step": 50370
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.521318550487388e-05,
+      "loss": 0.3912,
+      "step": 50380
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.5210209092938467e-05,
+      "loss": 0.3763,
+      "step": 50390
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.520723268100305e-05,
+      "loss": 0.3904,
+      "step": 50400
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.520425626906764e-05,
+      "loss": 0.3931,
+      "step": 50410
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.5201279857132227e-05,
+      "loss": 0.4009,
+      "step": 50420
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.5198303445196818e-05,
+      "loss": 0.3877,
+      "step": 50430
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.5195327033261405e-05,
+      "loss": 0.3801,
+      "step": 50440
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.5192350621325995e-05,
+      "loss": 0.3915,
+      "step": 50450
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.5189374209390582e-05,
+      "loss": 0.4018,
+      "step": 50460
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.5186397797455172e-05,
+      "loss": 0.4076,
+      "step": 50470
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.518342138551976e-05,
+      "loss": 0.3935,
+      "step": 50480
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.518044497358435e-05,
+      "loss": 0.3859,
+      "step": 50490
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.5177468561648933e-05,
+      "loss": 0.3905,
+      "step": 50500
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.517449214971352e-05,
+      "loss": 0.393,
+      "step": 50510
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.517151573777811e-05,
+      "loss": 0.3909,
+      "step": 50520
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.5168539325842697e-05,
+      "loss": 0.3928,
+      "step": 50530
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.5165562913907287e-05,
+      "loss": 0.3999,
+      "step": 50540
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.5162586501971874e-05,
+      "loss": 0.3824,
+      "step": 50550
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.5159610090036464e-05,
+      "loss": 0.3974,
+      "step": 50560
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.515663367810105e-05,
+      "loss": 0.3879,
+      "step": 50570
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.515365726616564e-05,
+      "loss": 0.3856,
+      "step": 50580
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.515068085423023e-05,
+      "loss": 0.4008,
+      "step": 50590
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.5147704442294812e-05,
+      "loss": 0.4042,
+      "step": 50600
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.5144728030359402e-05,
+      "loss": 0.3846,
+      "step": 50610
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.514175161842399e-05,
+      "loss": 0.3902,
+      "step": 50620
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.513877520648858e-05,
+      "loss": 0.3743,
+      "step": 50630
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.5135798794553166e-05,
+      "loss": 0.4015,
+      "step": 50640
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.5132822382617757e-05,
+      "loss": 0.3936,
+      "step": 50650
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.5129845970682344e-05,
+      "loss": 0.3923,
+      "step": 50660
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.5126869558746934e-05,
+      "loss": 0.3823,
+      "step": 50670
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.512389314681152e-05,
+      "loss": 0.382,
+      "step": 50680
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.512091673487611e-05,
+      "loss": 0.401,
+      "step": 50690
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.5117940322940695e-05,
+      "loss": 0.395,
+      "step": 50700
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.5114963911005285e-05,
+      "loss": 0.4084,
+      "step": 50710
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.5111987499069872e-05,
+      "loss": 0.385,
+      "step": 50720
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.5109011087134462e-05,
+      "loss": 0.3885,
+      "step": 50730
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.510603467519905e-05,
+      "loss": 0.3902,
+      "step": 50740
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.510305826326364e-05,
+      "loss": 0.3895,
+      "step": 50750
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.5100081851328226e-05,
+      "loss": 0.4005,
+      "step": 50760
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.5097105439392816e-05,
+      "loss": 0.3807,
+      "step": 50770
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.5094129027457403e-05,
+      "loss": 0.3787,
+      "step": 50780
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.5091152615521994e-05,
+      "loss": 0.3828,
+      "step": 50790
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.508817620358658e-05,
+      "loss": 0.3937,
+      "step": 50800
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.5085199791651164e-05,
+      "loss": 0.3766,
+      "step": 50810
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.5082223379715754e-05,
+      "loss": 0.3953,
+      "step": 50820
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.507924696778034e-05,
+      "loss": 0.3955,
+      "step": 50830
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.507627055584493e-05,
+      "loss": 0.3862,
+      "step": 50840
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.507329414390952e-05,
+      "loss": 0.403,
+      "step": 50850
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.507031773197411e-05,
+      "loss": 0.3897,
+      "step": 50860
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.5067341320038696e-05,
+      "loss": 0.39,
+      "step": 50870
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.5064364908103286e-05,
+      "loss": 0.3834,
+      "step": 50880
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.5061388496167873e-05,
+      "loss": 0.3865,
+      "step": 50890
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.5058412084232463e-05,
+      "loss": 0.4061,
+      "step": 50900
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.5055435672297046e-05,
+      "loss": 0.3795,
+      "step": 50910
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.5052459260361633e-05,
+      "loss": 0.3886,
+      "step": 50920
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.5049482848426224e-05,
+      "loss": 0.4084,
+      "step": 50930
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.504650643649081e-05,
+      "loss": 0.3855,
+      "step": 50940
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.50435300245554e-05,
+      "loss": 0.4088,
+      "step": 50950
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.5040553612619988e-05,
+      "loss": 0.3903,
+      "step": 50960
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.5037577200684578e-05,
+      "loss": 0.3932,
+      "step": 50970
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.5034600788749165e-05,
+      "loss": 0.3784,
+      "step": 50980
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.5031624376813755e-05,
+      "loss": 0.3966,
+      "step": 50990
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.5028647964878342e-05,
+      "loss": 0.3915,
+      "step": 51000
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.5025671552942926e-05,
+      "loss": 0.4,
+      "step": 51010
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.5022695141007516e-05,
+      "loss": 0.3932,
+      "step": 51020
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.5019718729072103e-05,
+      "loss": 0.3983,
+      "step": 51030
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.5016742317136693e-05,
+      "loss": 0.4047,
+      "step": 51040
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.501376590520128e-05,
+      "loss": 0.3824,
+      "step": 51050
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.501078949326587e-05,
+      "loss": 0.3863,
+      "step": 51060
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.5007813081330457e-05,
+      "loss": 0.3875,
+      "step": 51070
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.5004836669395047e-05,
+      "loss": 0.3952,
+      "step": 51080
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.5001860257459634e-05,
+      "loss": 0.3873,
+      "step": 51090
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.4998883845524225e-05,
+      "loss": 0.3758,
+      "step": 51100
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.4995907433588808e-05,
+      "loss": 0.392,
+      "step": 51110
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.49929310216534e-05,
+      "loss": 0.4078,
+      "step": 51120
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.4989954609717985e-05,
+      "loss": 0.3962,
+      "step": 51130
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.4986978197782576e-05,
+      "loss": 0.401,
+      "step": 51140
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.4984001785847163e-05,
+      "loss": 0.3982,
+      "step": 51150
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.4981025373911753e-05,
+      "loss": 0.3721,
+      "step": 51160
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.497804896197634e-05,
+      "loss": 0.4111,
+      "step": 51170
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.497507255004093e-05,
+      "loss": 0.4004,
+      "step": 51180
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.4972096138105517e-05,
+      "loss": 0.4076,
+      "step": 51190
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.4969119726170107e-05,
+      "loss": 0.3882,
+      "step": 51200
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.4966143314234694e-05,
+      "loss": 0.3911,
+      "step": 51210
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.4963166902299278e-05,
+      "loss": 0.3899,
+      "step": 51220
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.4960190490363868e-05,
+      "loss": 0.3994,
+      "step": 51230
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.4957214078428455e-05,
+      "loss": 0.3772,
+      "step": 51240
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.4954237666493045e-05,
+      "loss": 0.3922,
+      "step": 51250
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.4951261254557632e-05,
+      "loss": 0.4152,
+      "step": 51260
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.4948284842622222e-05,
+      "loss": 0.373,
+      "step": 51270
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.494530843068681e-05,
+      "loss": 0.3925,
+      "step": 51280
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.49423320187514e-05,
+      "loss": 0.3894,
+      "step": 51290
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.4939355606815986e-05,
+      "loss": 0.3987,
+      "step": 51300
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.4936379194880577e-05,
+      "loss": 0.3815,
+      "step": 51310
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.493340278294516e-05,
+      "loss": 0.3866,
+      "step": 51320
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.4930426371009747e-05,
+      "loss": 0.3959,
+      "step": 51330
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.4927449959074337e-05,
+      "loss": 0.3854,
+      "step": 51340
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.4924473547138924e-05,
+      "loss": 0.4014,
+      "step": 51350
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.4921497135203515e-05,
+      "loss": 0.3922,
+      "step": 51360
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.49185207232681e-05,
+      "loss": 0.3954,
+      "step": 51370
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.4915544311332692e-05,
+      "loss": 0.3973,
+      "step": 51380
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.491256789939728e-05,
+      "loss": 0.4084,
+      "step": 51390
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.490959148746187e-05,
+      "loss": 0.3983,
+      "step": 51400
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.4906615075526456e-05,
+      "loss": 0.395,
+      "step": 51410
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.490363866359104e-05,
+      "loss": 0.4144,
+      "step": 51420
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.490066225165563e-05,
+      "loss": 0.3914,
+      "step": 51430
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.4897685839720217e-05,
+      "loss": 0.3898,
+      "step": 51440
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.4894709427784807e-05,
+      "loss": 0.3978,
+      "step": 51450
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.4891733015849394e-05,
+      "loss": 0.3847,
+      "step": 51460
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.4888756603913984e-05,
+      "loss": 0.3843,
+      "step": 51470
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.488578019197857e-05,
+      "loss": 0.3975,
+      "step": 51480
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.488280378004316e-05,
+      "loss": 0.3767,
+      "step": 51490
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.4879827368107748e-05,
+      "loss": 0.3854,
+      "step": 51500
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.487685095617234e-05,
+      "loss": 0.4142,
+      "step": 51510
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.4873874544236922e-05,
+      "loss": 0.3853,
+      "step": 51520
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.4870898132301512e-05,
+      "loss": 0.3911,
+      "step": 51530
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.48679217203661e-05,
+      "loss": 0.3862,
+      "step": 51540
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.486494530843069e-05,
+      "loss": 0.3899,
+      "step": 51550
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.4861968896495276e-05,
+      "loss": 0.3812,
+      "step": 51560
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.4858992484559867e-05,
+      "loss": 0.4017,
+      "step": 51570
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.4856016072624453e-05,
+      "loss": 0.3959,
+      "step": 51580
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.4853039660689044e-05,
+      "loss": 0.3882,
+      "step": 51590
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.485006324875363e-05,
+      "loss": 0.3842,
+      "step": 51600
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.484708683681822e-05,
+      "loss": 0.4002,
+      "step": 51610
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.4844110424882808e-05,
+      "loss": 0.3917,
+      "step": 51620
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.484113401294739e-05,
+      "loss": 0.3984,
+      "step": 51630
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.483815760101198e-05,
+      "loss": 0.3871,
+      "step": 51640
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.483518118907657e-05,
+      "loss": 0.3873,
+      "step": 51650
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.483220477714116e-05,
+      "loss": 0.378,
+      "step": 51660
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.4829228365205746e-05,
+      "loss": 0.4006,
+      "step": 51670
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.4826251953270336e-05,
+      "loss": 0.3757,
+      "step": 51680
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.4823275541334923e-05,
+      "loss": 0.3814,
+      "step": 51690
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.4820299129399513e-05,
+      "loss": 0.3905,
+      "step": 51700
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.48173227174641e-05,
+      "loss": 0.3967,
+      "step": 51710
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.481434630552869e-05,
+      "loss": 0.3983,
+      "step": 51720
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.4811369893593274e-05,
+      "loss": 0.4058,
+      "step": 51730
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.480839348165786e-05,
+      "loss": 0.3878,
+      "step": 51740
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.480541706972245e-05,
+      "loss": 0.3826,
+      "step": 51750
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.4802440657787038e-05,
+      "loss": 0.3983,
+      "step": 51760
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.4799464245851628e-05,
+      "loss": 0.3804,
+      "step": 51770
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.4796487833916215e-05,
+      "loss": 0.396,
+      "step": 51780
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.4793511421980805e-05,
+      "loss": 0.3983,
+      "step": 51790
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.4790535010045392e-05,
+      "loss": 0.368,
+      "step": 51800
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.4787558598109983e-05,
+      "loss": 0.3943,
+      "step": 51810
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.478458218617457e-05,
+      "loss": 0.3902,
+      "step": 51820
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.4781605774239156e-05,
+      "loss": 0.4069,
+      "step": 51830
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.4778629362303743e-05,
+      "loss": 0.3808,
+      "step": 51840
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.477565295036833e-05,
+      "loss": 0.375,
+      "step": 51850
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.477267653843292e-05,
+      "loss": 0.3857,
+      "step": 51860
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.4769700126497507e-05,
+      "loss": 0.3757,
+      "step": 51870
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.4766723714562098e-05,
+      "loss": 0.3943,
+      "step": 51880
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.4763747302626685e-05,
+      "loss": 0.4176,
+      "step": 51890
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.4760770890691275e-05,
+      "loss": 0.3922,
+      "step": 51900
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.4757794478755862e-05,
+      "loss": 0.3945,
+      "step": 51910
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.4754818066820452e-05,
+      "loss": 0.3855,
+      "step": 51920
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.4751841654885036e-05,
+      "loss": 0.3882,
+      "step": 51930
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.4748865242949626e-05,
+      "loss": 0.3916,
+      "step": 51940
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.4745888831014213e-05,
+      "loss": 0.383,
+      "step": 51950
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.4742912419078803e-05,
+      "loss": 0.3919,
+      "step": 51960
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.473993600714339e-05,
+      "loss": 0.3833,
+      "step": 51970
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.473695959520798e-05,
+      "loss": 0.393,
+      "step": 51980
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.4733983183272567e-05,
+      "loss": 0.3748,
+      "step": 51990
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.4731006771337157e-05,
+      "loss": 0.3888,
+      "step": 52000
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.4728030359401744e-05,
+      "loss": 0.3935,
+      "step": 52010
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.4725053947466335e-05,
+      "loss": 0.3987,
+      "step": 52020
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.472207753553092e-05,
+      "loss": 0.3966,
+      "step": 52030
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.4719101123595505e-05,
+      "loss": 0.3969,
+      "step": 52040
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.4716124711660095e-05,
+      "loss": 0.395,
+      "step": 52050
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.4713148299724682e-05,
+      "loss": 0.3973,
+      "step": 52060
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.4710171887789272e-05,
+      "loss": 0.3834,
+      "step": 52070
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.470719547585386e-05,
+      "loss": 0.3837,
+      "step": 52080
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.470421906391845e-05,
+      "loss": 0.4031,
+      "step": 52090
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.4701242651983037e-05,
+      "loss": 0.3827,
+      "step": 52100
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.4698266240047627e-05,
+      "loss": 0.3857,
+      "step": 52110
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.4695289828112214e-05,
+      "loss": 0.406,
+      "step": 52120
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.4692313416176804e-05,
+      "loss": 0.3774,
+      "step": 52130
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.4689337004241388e-05,
+      "loss": 0.3897,
+      "step": 52140
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.4686360592305974e-05,
+      "loss": 0.3846,
+      "step": 52150
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.4683384180370565e-05,
+      "loss": 0.3916,
+      "step": 52160
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.468040776843515e-05,
+      "loss": 0.39,
+      "step": 52170
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.4677431356499742e-05,
+      "loss": 0.4062,
+      "step": 52180
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.467445494456433e-05,
+      "loss": 0.3854,
+      "step": 52190
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.467147853262892e-05,
+      "loss": 0.3969,
+      "step": 52200
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.4668502120693506e-05,
+      "loss": 0.3826,
+      "step": 52210
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.4665525708758096e-05,
+      "loss": 0.3895,
+      "step": 52220
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.4662549296822683e-05,
+      "loss": 0.3851,
+      "step": 52230
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.465957288488727e-05,
+      "loss": 0.3801,
+      "step": 52240
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.4656596472951857e-05,
+      "loss": 0.3689,
+      "step": 52250
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.4653620061016447e-05,
+      "loss": 0.3877,
+      "step": 52260
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.4650643649081034e-05,
+      "loss": 0.3877,
+      "step": 52270
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.464766723714562e-05,
+      "loss": 0.3946,
+      "step": 52280
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.464469082521021e-05,
+      "loss": 0.4086,
+      "step": 52290
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.4641714413274798e-05,
+      "loss": 0.3844,
+      "step": 52300
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.463873800133939e-05,
+      "loss": 0.3737,
+      "step": 52310
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.4635761589403975e-05,
+      "loss": 0.397,
+      "step": 52320
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.4632785177468566e-05,
+      "loss": 0.3885,
+      "step": 52330
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.462980876553315e-05,
+      "loss": 0.3898,
+      "step": 52340
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.462683235359774e-05,
+      "loss": 0.3904,
+      "step": 52350
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.4623855941662326e-05,
+      "loss": 0.3838,
+      "step": 52360
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.4620879529726917e-05,
+      "loss": 0.3902,
+      "step": 52370
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.4617903117791504e-05,
+      "loss": 0.3917,
+      "step": 52380
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.4614926705856094e-05,
+      "loss": 0.3796,
+      "step": 52390
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.461195029392068e-05,
+      "loss": 0.3862,
+      "step": 52400
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.460897388198527e-05,
+      "loss": 0.392,
+      "step": 52410
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.4605997470049858e-05,
+      "loss": 0.3934,
+      "step": 52420
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.4603021058114448e-05,
+      "loss": 0.4046,
+      "step": 52430
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.4600044646179032e-05,
+      "loss": 0.3897,
+      "step": 52440
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.459706823424362e-05,
+      "loss": 0.3857,
+      "step": 52450
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.459409182230821e-05,
+      "loss": 0.3872,
+      "step": 52460
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.4591115410372796e-05,
+      "loss": 0.3973,
+      "step": 52470
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.4588138998437386e-05,
+      "loss": 0.3851,
+      "step": 52480
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.4585162586501973e-05,
+      "loss": 0.3897,
+      "step": 52490
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.4582186174566563e-05,
+      "loss": 0.3928,
+      "step": 52500
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.457920976263115e-05,
+      "loss": 0.3828,
+      "step": 52510
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.457623335069574e-05,
+      "loss": 0.3847,
+      "step": 52520
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.4573256938760327e-05,
+      "loss": 0.3783,
+      "step": 52530
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.4570280526824918e-05,
+      "loss": 0.3869,
+      "step": 52540
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.45673041148895e-05,
+      "loss": 0.384,
+      "step": 52550
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.4564327702954088e-05,
+      "loss": 0.3739,
+      "step": 52560
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.456135129101868e-05,
+      "loss": 0.411,
+      "step": 52570
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.4558374879083265e-05,
+      "loss": 0.372,
+      "step": 52580
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.4555398467147856e-05,
+      "loss": 0.404,
+      "step": 52590
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.4552422055212442e-05,
+      "loss": 0.3808,
+      "step": 52600
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.4549445643277033e-05,
+      "loss": 0.3936,
+      "step": 52610
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.454646923134162e-05,
+      "loss": 0.3754,
+      "step": 52620
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.454349281940621e-05,
+      "loss": 0.3896,
+      "step": 52630
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.4540516407470797e-05,
+      "loss": 0.3758,
+      "step": 52640
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.4537539995535384e-05,
+      "loss": 0.3992,
+      "step": 52650
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.453456358359997e-05,
+      "loss": 0.3921,
+      "step": 52660
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.453158717166456e-05,
+      "loss": 0.3917,
+      "step": 52670
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.4528610759729148e-05,
+      "loss": 0.3816,
+      "step": 52680
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.4525634347793738e-05,
+      "loss": 0.3784,
+      "step": 52690
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.4522657935858325e-05,
+      "loss": 0.3903,
+      "step": 52700
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.4519681523922912e-05,
+      "loss": 0.3915,
+      "step": 52710
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.4516705111987502e-05,
+      "loss": 0.3911,
+      "step": 52720
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.451372870005209e-05,
+      "loss": 0.4021,
+      "step": 52730
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.451075228811668e-05,
+      "loss": 0.3944,
+      "step": 52740
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.4507775876181263e-05,
+      "loss": 0.396,
+      "step": 52750
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.4504799464245853e-05,
+      "loss": 0.3994,
+      "step": 52760
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.450182305231044e-05,
+      "loss": 0.3869,
+      "step": 52770
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.449884664037503e-05,
+      "loss": 0.3823,
+      "step": 52780
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.4495870228439617e-05,
+      "loss": 0.383,
+      "step": 52790
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.4492893816504208e-05,
+      "loss": 0.39,
+      "step": 52800
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.4489917404568794e-05,
+      "loss": 0.388,
+      "step": 52810
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.4486940992633385e-05,
+      "loss": 0.3977,
+      "step": 52820
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.448396458069797e-05,
+      "loss": 0.3891,
+      "step": 52830
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.4480988168762562e-05,
+      "loss": 0.3979,
+      "step": 52840
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.4478011756827145e-05,
+      "loss": 0.3817,
+      "step": 52850
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.4475035344891732e-05,
+      "loss": 0.3854,
+      "step": 52860
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.4472058932956323e-05,
+      "loss": 0.4035,
+      "step": 52870
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.446908252102091e-05,
+      "loss": 0.3971,
+      "step": 52880
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.44661061090855e-05,
+      "loss": 0.3872,
+      "step": 52890
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.4463129697150087e-05,
+      "loss": 0.3926,
+      "step": 52900
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.4460153285214677e-05,
+      "loss": 0.3944,
+      "step": 52910
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.4457176873279264e-05,
+      "loss": 0.3934,
+      "step": 52920
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.4454200461343854e-05,
+      "loss": 0.3972,
+      "step": 52930
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.445122404940844e-05,
+      "loss": 0.3921,
+      "step": 52940
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.444824763747303e-05,
+      "loss": 0.3923,
+      "step": 52950
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.4445271225537615e-05,
+      "loss": 0.3851,
+      "step": 52960
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.4442294813602202e-05,
+      "loss": 0.3893,
+      "step": 52970
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.4439318401666792e-05,
+      "loss": 0.3942,
+      "step": 52980
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.443634198973138e-05,
+      "loss": 0.3864,
+      "step": 52990
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.443336557779597e-05,
+      "loss": 0.3964,
+      "step": 53000
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.4430389165860556e-05,
+      "loss": 0.3878,
+      "step": 53010
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.4427412753925146e-05,
+      "loss": 0.3984,
+      "step": 53020
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.4424436341989733e-05,
+      "loss": 0.3824,
+      "step": 53030
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.4421459930054324e-05,
+      "loss": 0.3883,
+      "step": 53040
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.441848351811891e-05,
+      "loss": 0.3844,
+      "step": 53050
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.4415507106183497e-05,
+      "loss": 0.3917,
+      "step": 53060
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.4412530694248084e-05,
+      "loss": 0.3874,
+      "step": 53070
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.4409554282312675e-05,
+      "loss": 0.3857,
+      "step": 53080
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.440657787037726e-05,
+      "loss": 0.3953,
+      "step": 53090
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.4403601458441852e-05,
+      "loss": 0.3902,
+      "step": 53100
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.440062504650644e-05,
+      "loss": 0.3787,
+      "step": 53110
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.439764863457103e-05,
+      "loss": 0.3912,
+      "step": 53120
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.4394672222635616e-05,
+      "loss": 0.4022,
+      "step": 53130
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.4391695810700203e-05,
+      "loss": 0.3878,
+      "step": 53140
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.4388719398764793e-05,
+      "loss": 0.3882,
+      "step": 53150
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.4385742986829377e-05,
+      "loss": 0.379,
+      "step": 53160
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.4382766574893967e-05,
+      "loss": 0.3926,
+      "step": 53170
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.4379790162958554e-05,
+      "loss": 0.412,
+      "step": 53180
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.4376813751023144e-05,
+      "loss": 0.3857,
+      "step": 53190
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.437383733908773e-05,
+      "loss": 0.3859,
+      "step": 53200
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.437086092715232e-05,
+      "loss": 0.3839,
+      "step": 53210
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.4367884515216908e-05,
+      "loss": 0.3817,
+      "step": 53220
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.43649081032815e-05,
+      "loss": 0.402,
+      "step": 53230
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.4361931691346085e-05,
+      "loss": 0.4093,
+      "step": 53240
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.4358955279410676e-05,
+      "loss": 0.4043,
+      "step": 53250
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.435597886747526e-05,
+      "loss": 0.3972,
+      "step": 53260
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.4353002455539846e-05,
+      "loss": 0.3918,
+      "step": 53270
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.4350026043604436e-05,
+      "loss": 0.374,
+      "step": 53280
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.4347049631669023e-05,
+      "loss": 0.3808,
+      "step": 53290
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.4344073219733613e-05,
+      "loss": 0.3912,
+      "step": 53300
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.43410968077982e-05,
+      "loss": 0.3904,
+      "step": 53310
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.433812039586279e-05,
+      "loss": 0.3782,
+      "step": 53320
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.4335143983927378e-05,
+      "loss": 0.3761,
+      "step": 53330
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.4332167571991968e-05,
+      "loss": 0.3866,
+      "step": 53340
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.4329191160056555e-05,
+      "loss": 0.3886,
+      "step": 53350
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.4326214748121145e-05,
+      "loss": 0.3929,
+      "step": 53360
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.432323833618573e-05,
+      "loss": 0.4007,
+      "step": 53370
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.4320261924250315e-05,
+      "loss": 0.3911,
+      "step": 53380
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.4317285512314906e-05,
+      "loss": 0.4006,
+      "step": 53390
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.4314309100379493e-05,
+      "loss": 0.3873,
+      "step": 53400
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.4311332688444083e-05,
+      "loss": 0.3932,
+      "step": 53410
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.430835627650867e-05,
+      "loss": 0.3956,
+      "step": 53420
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.430537986457326e-05,
+      "loss": 0.3938,
+      "step": 53430
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.4302403452637847e-05,
+      "loss": 0.3838,
+      "step": 53440
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.4299427040702437e-05,
+      "loss": 0.3836,
+      "step": 53450
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.4296450628767024e-05,
+      "loss": 0.3737,
+      "step": 53460
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.429347421683161e-05,
+      "loss": 0.4104,
+      "step": 53470
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.4290497804896198e-05,
+      "loss": 0.3857,
+      "step": 53480
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.4287521392960788e-05,
+      "loss": 0.3955,
+      "step": 53490
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.4284544981025375e-05,
+      "loss": 0.3782,
+      "step": 53500
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.4281568569089965e-05,
+      "loss": 0.381,
+      "step": 53510
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.4278592157154552e-05,
+      "loss": 0.396,
+      "step": 53520
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.4275615745219143e-05,
+      "loss": 0.4062,
+      "step": 53530
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.427263933328373e-05,
+      "loss": 0.3893,
+      "step": 53540
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.426966292134832e-05,
+      "loss": 0.3893,
+      "step": 53550
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.4266686509412907e-05,
+      "loss": 0.3898,
+      "step": 53560
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.426371009747749e-05,
+      "loss": 0.398,
+      "step": 53570
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.426073368554208e-05,
+      "loss": 0.3942,
+      "step": 53580
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.4257757273606667e-05,
+      "loss": 0.3908,
+      "step": 53590
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.4254780861671258e-05,
+      "loss": 0.3923,
+      "step": 53600
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.4251804449735845e-05,
+      "loss": 0.3889,
+      "step": 53610
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.4248828037800435e-05,
+      "loss": 0.3929,
+      "step": 53620
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.4245851625865022e-05,
+      "loss": 0.3684,
+      "step": 53630
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.4242875213929612e-05,
+      "loss": 0.3951,
+      "step": 53640
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.42398988019942e-05,
+      "loss": 0.3798,
+      "step": 53650
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.423692239005879e-05,
+      "loss": 0.3861,
+      "step": 53660
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.4233945978123373e-05,
+      "loss": 0.3704,
+      "step": 53670
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.423096956618796e-05,
+      "loss": 0.3935,
+      "step": 53680
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.422799315425255e-05,
+      "loss": 0.3856,
+      "step": 53690
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.4225016742317137e-05,
+      "loss": 0.4015,
+      "step": 53700
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.4222040330381727e-05,
+      "loss": 0.3845,
+      "step": 53710
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.4219063918446314e-05,
+      "loss": 0.3948,
+      "step": 53720
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.4216087506510904e-05,
+      "loss": 0.3962,
+      "step": 53730
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.421311109457549e-05,
+      "loss": 0.3988,
+      "step": 53740
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.421013468264008e-05,
+      "loss": 0.3884,
+      "step": 53750
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.420715827070467e-05,
+      "loss": 0.3952,
+      "step": 53760
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.420418185876926e-05,
+      "loss": 0.3871,
+      "step": 53770
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.4201205446833842e-05,
+      "loss": 0.4001,
+      "step": 53780
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.419822903489843e-05,
+      "loss": 0.3887,
+      "step": 53790
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.419525262296302e-05,
+      "loss": 0.3807,
+      "step": 53800
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.4192276211027606e-05,
+      "loss": 0.3992,
+      "step": 53810
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.4189299799092197e-05,
+      "loss": 0.3878,
+      "step": 53820
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.4186323387156783e-05,
+      "loss": 0.3735,
+      "step": 53830
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.4183346975221374e-05,
+      "loss": 0.3747,
+      "step": 53840
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.418037056328596e-05,
+      "loss": 0.3939,
+      "step": 53850
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.417739415135055e-05,
+      "loss": 0.3913,
+      "step": 53860
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.4174417739415138e-05,
+      "loss": 0.3879,
+      "step": 53870
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.4171441327479725e-05,
+      "loss": 0.3758,
+      "step": 53880
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.416846491554431e-05,
+      "loss": 0.3899,
+      "step": 53890
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.4165488503608902e-05,
+      "loss": 0.4025,
+      "step": 53900
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.416251209167349e-05,
+      "loss": 0.3739,
+      "step": 53910
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.415953567973808e-05,
+      "loss": 0.3939,
+      "step": 53920
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.4156559267802666e-05,
+      "loss": 0.3776,
+      "step": 53930
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.4153582855867256e-05,
+      "loss": 0.3835,
+      "step": 53940
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.4150606443931843e-05,
+      "loss": 0.3781,
+      "step": 53950
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.4147630031996434e-05,
+      "loss": 0.393,
+      "step": 53960
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.414465362006102e-05,
+      "loss": 0.4004,
+      "step": 53970
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.4141677208125604e-05,
+      "loss": 0.3797,
+      "step": 53980
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.4138700796190194e-05,
+      "loss": 0.3898,
+      "step": 53990
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.413572438425478e-05,
+      "loss": 0.3862,
+      "step": 54000
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.413274797231937e-05,
+      "loss": 0.3929,
+      "step": 54010
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.4129771560383958e-05,
+      "loss": 0.3849,
+      "step": 54020
+    },
+    {
+      "epoch": 12.0,
+      "eval_cer": 4.549145299145299,
+      "eval_loss": 0.4053784906864166,
+      "eval_runtime": 1.3105,
+      "eval_samples_per_second": 7.631,
+      "eval_steps_per_second": 0.763,
+      "eval_wer": 0.5584415584415584,
+      "step": 54024
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.412679514844855e-05,
+      "loss": 0.3838,
+      "step": 54030
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.4123818736513135e-05,
+      "loss": 0.3928,
+      "step": 54040
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.4120842324577726e-05,
+      "loss": 0.3904,
+      "step": 54050
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4117865912642313e-05,
+      "loss": 0.3855,
+      "step": 54060
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4114889500706903e-05,
+      "loss": 0.3818,
+      "step": 54070
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4111913088771486e-05,
+      "loss": 0.3724,
+      "step": 54080
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4108936676836073e-05,
+      "loss": 0.3641,
+      "step": 54090
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4105960264900664e-05,
+      "loss": 0.3819,
+      "step": 54100
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.410298385296525e-05,
+      "loss": 0.3724,
+      "step": 54110
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.410000744102984e-05,
+      "loss": 0.378,
+      "step": 54120
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.4097031029094428e-05,
+      "loss": 0.3709,
+      "step": 54130
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.4094054617159018e-05,
+      "loss": 0.3764,
+      "step": 54140
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.4091078205223605e-05,
+      "loss": 0.3831,
+      "step": 54150
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.4088101793288195e-05,
+      "loss": 0.3823,
+      "step": 54160
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.4085125381352782e-05,
+      "loss": 0.384,
+      "step": 54170
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.4082148969417366e-05,
+      "loss": 0.3716,
+      "step": 54180
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.4079172557481956e-05,
+      "loss": 0.3785,
+      "step": 54190
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.4076196145546543e-05,
+      "loss": 0.3772,
+      "step": 54200
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.4073219733611133e-05,
+      "loss": 0.364,
+      "step": 54210
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.407024332167572e-05,
+      "loss": 0.3805,
+      "step": 54220
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.406726690974031e-05,
+      "loss": 0.3962,
+      "step": 54230
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.4064290497804897e-05,
+      "loss": 0.3709,
+      "step": 54240
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.4061314085869487e-05,
+      "loss": 0.3769,
+      "step": 54250
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.4058337673934074e-05,
+      "loss": 0.3598,
+      "step": 54260
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.4055361261998665e-05,
+      "loss": 0.3701,
+      "step": 54270
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.405238485006325e-05,
+      "loss": 0.3731,
+      "step": 54280
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.404940843812784e-05,
+      "loss": 0.3795,
+      "step": 54290
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.4046432026192425e-05,
+      "loss": 0.3781,
+      "step": 54300
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.4043455614257016e-05,
+      "loss": 0.3912,
+      "step": 54310
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.4040479202321603e-05,
+      "loss": 0.3754,
+      "step": 54320
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.4037502790386193e-05,
+      "loss": 0.3843,
+      "step": 54330
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.403452637845078e-05,
+      "loss": 0.3684,
+      "step": 54340
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.403154996651537e-05,
+      "loss": 0.3746,
+      "step": 54350
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.4028573554579957e-05,
+      "loss": 0.3766,
+      "step": 54360
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.4025597142644547e-05,
+      "loss": 0.3748,
+      "step": 54370
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.4022620730709134e-05,
+      "loss": 0.3912,
+      "step": 54380
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.4019644318773718e-05,
+      "loss": 0.386,
+      "step": 54390
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.4016667906838308e-05,
+      "loss": 0.3753,
+      "step": 54400
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.4013691494902895e-05,
+      "loss": 0.3647,
+      "step": 54410
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.4010715082967485e-05,
+      "loss": 0.3791,
+      "step": 54420
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.4007738671032072e-05,
+      "loss": 0.3857,
+      "step": 54430
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.4004762259096662e-05,
+      "loss": 0.3699,
+      "step": 54440
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.400178584716125e-05,
+      "loss": 0.3675,
+      "step": 54450
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.399880943522584e-05,
+      "loss": 0.3854,
+      "step": 54460
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.3995833023290426e-05,
+      "loss": 0.3734,
+      "step": 54470
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.3992856611355017e-05,
+      "loss": 0.3632,
+      "step": 54480
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.39898801994196e-05,
+      "loss": 0.3735,
+      "step": 54490
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.3986903787484187e-05,
+      "loss": 0.3685,
+      "step": 54500
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.3983927375548777e-05,
+      "loss": 0.385,
+      "step": 54510
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.3980950963613364e-05,
+      "loss": 0.3666,
+      "step": 54520
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.3977974551677955e-05,
+      "loss": 0.3754,
+      "step": 54530
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.397499813974254e-05,
+      "loss": 0.3724,
+      "step": 54540
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.397202172780713e-05,
+      "loss": 0.3751,
+      "step": 54550
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.396904531587172e-05,
+      "loss": 0.3724,
+      "step": 54560
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.396606890393631e-05,
+      "loss": 0.3808,
+      "step": 54570
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.3963092492000896e-05,
+      "loss": 0.3746,
+      "step": 54580
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.396011608006548e-05,
+      "loss": 0.3842,
+      "step": 54590
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.395713966813007e-05,
+      "loss": 0.364,
+      "step": 54600
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.3954163256194656e-05,
+      "loss": 0.3608,
+      "step": 54610
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.3951186844259247e-05,
+      "loss": 0.3845,
+      "step": 54620
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.3948210432323834e-05,
+      "loss": 0.3725,
+      "step": 54630
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.3945234020388424e-05,
+      "loss": 0.3888,
+      "step": 54640
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.394225760845301e-05,
+      "loss": 0.4032,
+      "step": 54650
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.39392811965176e-05,
+      "loss": 0.3698,
+      "step": 54660
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.3936304784582188e-05,
+      "loss": 0.3814,
+      "step": 54670
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.393332837264678e-05,
+      "loss": 0.3858,
+      "step": 54680
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.3930351960711365e-05,
+      "loss": 0.3714,
+      "step": 54690
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.3927375548775952e-05,
+      "loss": 0.3803,
+      "step": 54700
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.392439913684054e-05,
+      "loss": 0.3704,
+      "step": 54710
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.392142272490513e-05,
+      "loss": 0.3696,
+      "step": 54720
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.3918446312969716e-05,
+      "loss": 0.3676,
+      "step": 54730
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.3915469901034306e-05,
+      "loss": 0.3761,
+      "step": 54740
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.3912493489098893e-05,
+      "loss": 0.3804,
+      "step": 54750
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.3909517077163484e-05,
+      "loss": 0.377,
+      "step": 54760
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.390654066522807e-05,
+      "loss": 0.3848,
+      "step": 54770
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.390356425329266e-05,
+      "loss": 0.3873,
+      "step": 54780
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.3900587841357248e-05,
+      "loss": 0.3717,
+      "step": 54790
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.389761142942183e-05,
+      "loss": 0.3716,
+      "step": 54800
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.389463501748642e-05,
+      "loss": 0.3624,
+      "step": 54810
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.389165860555101e-05,
+      "loss": 0.367,
+      "step": 54820
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.38886821936156e-05,
+      "loss": 0.3672,
+      "step": 54830
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.3885705781680186e-05,
+      "loss": 0.3768,
+      "step": 54840
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.3882729369744776e-05,
+      "loss": 0.3941,
+      "step": 54850
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.3879752957809363e-05,
+      "loss": 0.3611,
+      "step": 54860
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.3876776545873953e-05,
+      "loss": 0.3742,
+      "step": 54870
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.387380013393854e-05,
+      "loss": 0.3567,
+      "step": 54880
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.387082372200313e-05,
+      "loss": 0.3848,
+      "step": 54890
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.3867847310067714e-05,
+      "loss": 0.3539,
+      "step": 54900
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.38648708981323e-05,
+      "loss": 0.3834,
+      "step": 54910
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.386189448619689e-05,
+      "loss": 0.3965,
+      "step": 54920
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.3858918074261478e-05,
+      "loss": 0.3675,
+      "step": 54930
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.3855941662326068e-05,
+      "loss": 0.3792,
+      "step": 54940
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.3852965250390655e-05,
+      "loss": 0.3759,
+      "step": 54950
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.3849988838455245e-05,
+      "loss": 0.3696,
+      "step": 54960
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.3847012426519832e-05,
+      "loss": 0.3743,
+      "step": 54970
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.3844036014584423e-05,
+      "loss": 0.3729,
+      "step": 54980
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.384105960264901e-05,
+      "loss": 0.3739,
+      "step": 54990
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.3838083190713593e-05,
+      "loss": 0.3855,
+      "step": 55000
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.3835106778778183e-05,
+      "loss": 0.3826,
+      "step": 55010
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.383213036684277e-05,
+      "loss": 0.3815,
+      "step": 55020
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.382915395490736e-05,
+      "loss": 0.3877,
+      "step": 55030
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.3826177542971947e-05,
+      "loss": 0.3917,
+      "step": 55040
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.3823201131036538e-05,
+      "loss": 0.3796,
+      "step": 55050
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.3820224719101125e-05,
+      "loss": 0.3605,
+      "step": 55060
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.3817248307165715e-05,
+      "loss": 0.3628,
+      "step": 55070
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.3814271895230302e-05,
+      "loss": 0.3853,
+      "step": 55080
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.3811295483294892e-05,
+      "loss": 0.3862,
+      "step": 55090
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.380831907135948e-05,
+      "loss": 0.3809,
+      "step": 55100
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.3805342659424066e-05,
+      "loss": 0.3716,
+      "step": 55110
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.3802366247488653e-05,
+      "loss": 0.3603,
+      "step": 55120
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.3799389835553243e-05,
+      "loss": 0.3897,
+      "step": 55130
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.379641342361783e-05,
+      "loss": 0.3954,
+      "step": 55140
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.379343701168242e-05,
+      "loss": 0.383,
+      "step": 55150
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.3790460599747007e-05,
+      "loss": 0.3874,
+      "step": 55160
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.3787484187811597e-05,
+      "loss": 0.3754,
+      "step": 55170
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.3784507775876184e-05,
+      "loss": 0.3761,
+      "step": 55180
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.3781531363940775e-05,
+      "loss": 0.372,
+      "step": 55190
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.377855495200536e-05,
+      "loss": 0.3747,
+      "step": 55200
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.3775578540069945e-05,
+      "loss": 0.3924,
+      "step": 55210
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.3772602128134535e-05,
+      "loss": 0.3715,
+      "step": 55220
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.3769625716199122e-05,
+      "loss": 0.3773,
+      "step": 55230
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.3766649304263712e-05,
+      "loss": 0.358,
+      "step": 55240
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.37636728923283e-05,
+      "loss": 0.3915,
+      "step": 55250
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.376069648039289e-05,
+      "loss": 0.3779,
+      "step": 55260
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.3757720068457476e-05,
+      "loss": 0.3701,
+      "step": 55270
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.3754743656522067e-05,
+      "loss": 0.371,
+      "step": 55280
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.3751767244586654e-05,
+      "loss": 0.3717,
+      "step": 55290
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.3748790832651244e-05,
+      "loss": 0.3794,
+      "step": 55300
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.3745814420715827e-05,
+      "loss": 0.3803,
+      "step": 55310
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.3742838008780414e-05,
+      "loss": 0.3808,
+      "step": 55320
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.3739861596845005e-05,
+      "loss": 0.3965,
+      "step": 55330
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.373688518490959e-05,
+      "loss": 0.3673,
+      "step": 55340
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.3733908772974182e-05,
+      "loss": 0.3707,
+      "step": 55350
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.373093236103877e-05,
+      "loss": 0.3846,
+      "step": 55360
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.372795594910336e-05,
+      "loss": 0.3938,
+      "step": 55370
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.3724979537167946e-05,
+      "loss": 0.39,
+      "step": 55380
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.3722003125232536e-05,
+      "loss": 0.3682,
+      "step": 55390
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.3719026713297123e-05,
+      "loss": 0.379,
+      "step": 55400
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.3716050301361707e-05,
+      "loss": 0.3733,
+      "step": 55410
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.3713073889426297e-05,
+      "loss": 0.3882,
+      "step": 55420
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.3710097477490884e-05,
+      "loss": 0.3585,
+      "step": 55430
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.3707121065555474e-05,
+      "loss": 0.3781,
+      "step": 55440
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.370414465362006e-05,
+      "loss": 0.3758,
+      "step": 55450
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.370116824168465e-05,
+      "loss": 0.3758,
+      "step": 55460
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.3698191829749238e-05,
+      "loss": 0.3645,
+      "step": 55470
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.369521541781383e-05,
+      "loss": 0.395,
+      "step": 55480
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.3692239005878415e-05,
+      "loss": 0.3888,
+      "step": 55490
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3689262593943006e-05,
+      "loss": 0.3865,
+      "step": 55500
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3686286182007593e-05,
+      "loss": 0.3778,
+      "step": 55510
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.368330977007218e-05,
+      "loss": 0.3683,
+      "step": 55520
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3680333358136766e-05,
+      "loss": 0.3826,
+      "step": 55530
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3677356946201357e-05,
+      "loss": 0.3811,
+      "step": 55540
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.3674380534265944e-05,
+      "loss": 0.3661,
+      "step": 55550
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.3671404122330534e-05,
+      "loss": 0.3786,
+      "step": 55560
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.366842771039512e-05,
+      "loss": 0.3928,
+      "step": 55570
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.366545129845971e-05,
+      "loss": 0.387,
+      "step": 55580
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.3662474886524298e-05,
+      "loss": 0.3776,
+      "step": 55590
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.3659498474588888e-05,
+      "loss": 0.3853,
+      "step": 55600
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.3656522062653475e-05,
+      "loss": 0.399,
+      "step": 55610
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.365354565071806e-05,
+      "loss": 0.3694,
+      "step": 55620
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.365056923878265e-05,
+      "loss": 0.371,
+      "step": 55630
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.3647592826847236e-05,
+      "loss": 0.3791,
+      "step": 55640
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.3644616414911826e-05,
+      "loss": 0.3855,
+      "step": 55650
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.3641640002976413e-05,
+      "loss": 0.382,
+      "step": 55660
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.3638663591041003e-05,
+      "loss": 0.3781,
+      "step": 55670
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.363568717910559e-05,
+      "loss": 0.3741,
+      "step": 55680
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.363271076717018e-05,
+      "loss": 0.3792,
+      "step": 55690
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.3629734355234767e-05,
+      "loss": 0.3705,
+      "step": 55700
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.3626757943299358e-05,
+      "loss": 0.3856,
+      "step": 55710
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.362378153136394e-05,
+      "loss": 0.3688,
+      "step": 55720
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.3620805119428528e-05,
+      "loss": 0.3929,
+      "step": 55730
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.361782870749312e-05,
+      "loss": 0.3886,
+      "step": 55740
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.3614852295557705e-05,
+      "loss": 0.378,
+      "step": 55750
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.3611875883622296e-05,
+      "loss": 0.39,
+      "step": 55760
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.3608899471686882e-05,
+      "loss": 0.3891,
+      "step": 55770
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.3605923059751473e-05,
+      "loss": 0.3735,
+      "step": 55780
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.360294664781606e-05,
+      "loss": 0.3752,
+      "step": 55790
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.359997023588065e-05,
+      "loss": 0.3744,
+      "step": 55800
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.3596993823945237e-05,
+      "loss": 0.3786,
+      "step": 55810
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.359401741200982e-05,
+      "loss": 0.3867,
+      "step": 55820
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.359104100007441e-05,
+      "loss": 0.3857,
+      "step": 55830
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.3588064588138997e-05,
+      "loss": 0.3877,
+      "step": 55840
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.3585088176203588e-05,
+      "loss": 0.3803,
+      "step": 55850
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.3582111764268175e-05,
+      "loss": 0.3667,
+      "step": 55860
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.3579135352332765e-05,
+      "loss": 0.3632,
+      "step": 55870
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.3576158940397352e-05,
+      "loss": 0.3758,
+      "step": 55880
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.3573182528461942e-05,
+      "loss": 0.388,
+      "step": 55890
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.357020611652653e-05,
+      "loss": 0.3717,
+      "step": 55900
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.356722970459112e-05,
+      "loss": 0.3817,
+      "step": 55910
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.3564253292655703e-05,
+      "loss": 0.3714,
+      "step": 55920
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.3561276880720293e-05,
+      "loss": 0.3749,
+      "step": 55930
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.355830046878488e-05,
+      "loss": 0.3634,
+      "step": 55940
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.355532405684947e-05,
+      "loss": 0.373,
+      "step": 55950
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.3552347644914057e-05,
+      "loss": 0.3763,
+      "step": 55960
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.3549371232978648e-05,
+      "loss": 0.367,
+      "step": 55970
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.3546394821043234e-05,
+      "loss": 0.3749,
+      "step": 55980
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.3543418409107825e-05,
+      "loss": 0.3745,
+      "step": 55990
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.354044199717241e-05,
+      "loss": 0.3607,
+      "step": 56000
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.3537465585237002e-05,
+      "loss": 0.3912,
+      "step": 56010
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.353448917330159e-05,
+      "loss": 0.3732,
+      "step": 56020
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.3531512761366172e-05,
+      "loss": 0.38,
+      "step": 56030
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.3528536349430763e-05,
+      "loss": 0.3705,
+      "step": 56040
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.352555993749535e-05,
+      "loss": 0.3839,
+      "step": 56050
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.352258352555994e-05,
+      "loss": 0.3844,
+      "step": 56060
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.3519607113624527e-05,
+      "loss": 0.3821,
+      "step": 56070
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.3516630701689117e-05,
+      "loss": 0.3836,
+      "step": 56080
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.3513654289753704e-05,
+      "loss": 0.384,
+      "step": 56090
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.3510677877818294e-05,
+      "loss": 0.3974,
+      "step": 56100
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.350770146588288e-05,
+      "loss": 0.3837,
+      "step": 56110
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.350472505394747e-05,
+      "loss": 0.3945,
+      "step": 56120
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.3501748642012055e-05,
+      "loss": 0.3744,
+      "step": 56130
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.3498772230076642e-05,
+      "loss": 0.3706,
+      "step": 56140
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.3495795818141232e-05,
+      "loss": 0.3872,
+      "step": 56150
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.349281940620582e-05,
+      "loss": 0.3817,
+      "step": 56160
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.348984299427041e-05,
+      "loss": 0.3742,
+      "step": 56170
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.3486866582334996e-05,
+      "loss": 0.3661,
+      "step": 56180
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.3483890170399586e-05,
+      "loss": 0.3867,
+      "step": 56190
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.3480913758464173e-05,
+      "loss": 0.364,
+      "step": 56200
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.3477937346528764e-05,
+      "loss": 0.379,
+      "step": 56210
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.347496093459335e-05,
+      "loss": 0.385,
+      "step": 56220
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.3471984522657937e-05,
+      "loss": 0.3785,
+      "step": 56230
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.3469008110722524e-05,
+      "loss": 0.3654,
+      "step": 56240
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.346603169878711e-05,
+      "loss": 0.3838,
+      "step": 56250
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.34630552868517e-05,
+      "loss": 0.3882,
+      "step": 56260
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.346007887491629e-05,
+      "loss": 0.3647,
+      "step": 56270
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.345710246298088e-05,
+      "loss": 0.3811,
+      "step": 56280
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.3454126051045466e-05,
+      "loss": 0.3785,
+      "step": 56290
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.3451149639110056e-05,
+      "loss": 0.3692,
+      "step": 56300
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.3448173227174643e-05,
+      "loss": 0.3792,
+      "step": 56310
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.3445196815239233e-05,
+      "loss": 0.3874,
+      "step": 56320
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.3442220403303817e-05,
+      "loss": 0.3682,
+      "step": 56330
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.3439243991368407e-05,
+      "loss": 0.3742,
+      "step": 56340
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.3436267579432994e-05,
+      "loss": 0.3722,
+      "step": 56350
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.3433291167497584e-05,
+      "loss": 0.3717,
+      "step": 56360
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.343031475556217e-05,
+      "loss": 0.3746,
+      "step": 56370
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.342733834362676e-05,
+      "loss": 0.3924,
+      "step": 56380
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.3424361931691348e-05,
+      "loss": 0.3766,
+      "step": 56390
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.342138551975594e-05,
+      "loss": 0.3761,
+      "step": 56400
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.3418409107820525e-05,
+      "loss": 0.3658,
+      "step": 56410
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.3415432695885116e-05,
+      "loss": 0.3878,
+      "step": 56420
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.3412456283949702e-05,
+      "loss": 0.3599,
+      "step": 56430
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.3409479872014286e-05,
+      "loss": 0.3828,
+      "step": 56440
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.3406503460078876e-05,
+      "loss": 0.3757,
+      "step": 56450
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.3403527048143463e-05,
+      "loss": 0.3885,
+      "step": 56460
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.3400550636208053e-05,
+      "loss": 0.3636,
+      "step": 56470
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.339757422427264e-05,
+      "loss": 0.3839,
+      "step": 56480
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.339459781233723e-05,
+      "loss": 0.3652,
+      "step": 56490
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.3391621400401818e-05,
+      "loss": 0.3866,
+      "step": 56500
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.3388644988466408e-05,
+      "loss": 0.3598,
+      "step": 56510
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.3385668576530995e-05,
+      "loss": 0.39,
+      "step": 56520
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.3382692164595585e-05,
+      "loss": 0.3715,
+      "step": 56530
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.337971575266017e-05,
+      "loss": 0.3694,
+      "step": 56540
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.3376739340724755e-05,
+      "loss": 0.3876,
+      "step": 56550
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.3373762928789346e-05,
+      "loss": 0.3757,
+      "step": 56560
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.3370786516853933e-05,
+      "loss": 0.3836,
+      "step": 56570
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.3367810104918523e-05,
+      "loss": 0.3644,
+      "step": 56580
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.336483369298311e-05,
+      "loss": 0.3799,
+      "step": 56590
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.33618572810477e-05,
+      "loss": 0.3936,
+      "step": 56600
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.3358880869112287e-05,
+      "loss": 0.3685,
+      "step": 56610
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.3355904457176877e-05,
+      "loss": 0.3825,
+      "step": 56620
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.3352928045241464e-05,
+      "loss": 0.3732,
+      "step": 56630
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.334995163330605e-05,
+      "loss": 0.3809,
+      "step": 56640
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.3346975221370638e-05,
+      "loss": 0.3828,
+      "step": 56650
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.3343998809435228e-05,
+      "loss": 0.3792,
+      "step": 56660
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.3341022397499815e-05,
+      "loss": 0.3673,
+      "step": 56670
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.3338045985564402e-05,
+      "loss": 0.3717,
+      "step": 56680
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.3335069573628992e-05,
+      "loss": 0.3746,
+      "step": 56690
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.333209316169358e-05,
+      "loss": 0.3737,
+      "step": 56700
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.332911674975817e-05,
+      "loss": 0.3603,
+      "step": 56710
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.3326140337822756e-05,
+      "loss": 0.3865,
+      "step": 56720
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.3323163925887347e-05,
+      "loss": 0.3777,
+      "step": 56730
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.332018751395193e-05,
+      "loss": 0.3663,
+      "step": 56740
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.331721110201652e-05,
+      "loss": 0.3727,
+      "step": 56750
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.3314234690081107e-05,
+      "loss": 0.3644,
+      "step": 56760
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.3311258278145698e-05,
+      "loss": 0.3713,
+      "step": 56770
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.3308281866210285e-05,
+      "loss": 0.3643,
+      "step": 56780
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.3305305454274875e-05,
+      "loss": 0.3767,
+      "step": 56790
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.3302329042339462e-05,
+      "loss": 0.3667,
+      "step": 56800
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.3299352630404052e-05,
+      "loss": 0.3914,
+      "step": 56810
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.329637621846864e-05,
+      "loss": 0.3571,
+      "step": 56820
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.329339980653323e-05,
+      "loss": 0.3675,
+      "step": 56830
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.3290423394597816e-05,
+      "loss": 0.3874,
+      "step": 56840
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.32874469826624e-05,
+      "loss": 0.3826,
+      "step": 56850
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.328447057072699e-05,
+      "loss": 0.3681,
+      "step": 56860
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.3281494158791577e-05,
+      "loss": 0.3833,
+      "step": 56870
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.3278517746856167e-05,
+      "loss": 0.3776,
+      "step": 56880
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.3275541334920754e-05,
+      "loss": 0.383,
+      "step": 56890
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.3272564922985344e-05,
+      "loss": 0.3777,
+      "step": 56900
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.326958851104993e-05,
+      "loss": 0.3788,
+      "step": 56910
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.326661209911452e-05,
+      "loss": 0.3821,
+      "step": 56920
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.326363568717911e-05,
+      "loss": 0.3737,
+      "step": 56930
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.32606592752437e-05,
+      "loss": 0.3613,
+      "step": 56940
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.3257682863308282e-05,
+      "loss": 0.385,
+      "step": 56950
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.325470645137287e-05,
+      "loss": 0.365,
+      "step": 56960
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.325173003943746e-05,
+      "loss": 0.3855,
+      "step": 56970
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.3248753627502046e-05,
+      "loss": 0.361,
+      "step": 56980
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.3245777215566637e-05,
+      "loss": 0.3741,
+      "step": 56990
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.3242800803631223e-05,
+      "loss": 0.369,
+      "step": 57000
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.3239824391695814e-05,
+      "loss": 0.3891,
+      "step": 57010
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.32368479797604e-05,
+      "loss": 0.3735,
+      "step": 57020
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.323387156782499e-05,
+      "loss": 0.3731,
+      "step": 57030
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.3230895155889578e-05,
+      "loss": 0.3819,
+      "step": 57040
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.3227918743954165e-05,
+      "loss": 0.3772,
+      "step": 57050
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.322494233201875e-05,
+      "loss": 0.3711,
+      "step": 57060
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.3221965920083342e-05,
+      "loss": 0.3796,
+      "step": 57070
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.321898950814793e-05,
+      "loss": 0.3911,
+      "step": 57080
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.321601309621252e-05,
+      "loss": 0.3987,
+      "step": 57090
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.3213036684277106e-05,
+      "loss": 0.3918,
+      "step": 57100
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.3210060272341693e-05,
+      "loss": 0.3717,
+      "step": 57110
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.3207083860406283e-05,
+      "loss": 0.3798,
+      "step": 57120
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.320410744847087e-05,
+      "loss": 0.3562,
+      "step": 57130
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.320113103653546e-05,
+      "loss": 0.3797,
+      "step": 57140
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.3198154624600044e-05,
+      "loss": 0.3697,
+      "step": 57150
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.3195178212664634e-05,
+      "loss": 0.3754,
+      "step": 57160
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.319220180072922e-05,
+      "loss": 0.3806,
+      "step": 57170
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.318922538879381e-05,
+      "loss": 0.3644,
+      "step": 57180
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.3186248976858398e-05,
+      "loss": 0.3719,
+      "step": 57190
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.318327256492299e-05,
+      "loss": 0.3682,
+      "step": 57200
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.3180296152987575e-05,
+      "loss": 0.3827,
+      "step": 57210
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.3177319741052166e-05,
+      "loss": 0.3712,
+      "step": 57220
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.3174343329116753e-05,
+      "loss": 0.3666,
+      "step": 57230
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.3171366917181343e-05,
+      "loss": 0.3865,
+      "step": 57240
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.316839050524593e-05,
+      "loss": 0.3768,
+      "step": 57250
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.3165414093310513e-05,
+      "loss": 0.3662,
+      "step": 57260
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.3162437681375104e-05,
+      "loss": 0.3858,
+      "step": 57270
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.315946126943969e-05,
+      "loss": 0.3663,
+      "step": 57280
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.315648485750428e-05,
+      "loss": 0.3878,
+      "step": 57290
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.3153508445568868e-05,
+      "loss": 0.3693,
+      "step": 57300
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.3150532033633458e-05,
+      "loss": 0.382,
+      "step": 57310
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.3147555621698045e-05,
+      "loss": 0.3731,
+      "step": 57320
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.3144579209762635e-05,
+      "loss": 0.3608,
+      "step": 57330
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.3141602797827222e-05,
+      "loss": 0.3758,
+      "step": 57340
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.3138626385891812e-05,
+      "loss": 0.3736,
+      "step": 57350
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.3135649973956396e-05,
+      "loss": 0.3738,
+      "step": 57360
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.3132673562020983e-05,
+      "loss": 0.3714,
+      "step": 57370
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.3129697150085573e-05,
+      "loss": 0.3685,
+      "step": 57380
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.312672073815016e-05,
+      "loss": 0.3753,
+      "step": 57390
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.312374432621475e-05,
+      "loss": 0.3689,
+      "step": 57400
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.3120767914279337e-05,
+      "loss": 0.3919,
+      "step": 57410
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.3117791502343927e-05,
+      "loss": 0.3703,
+      "step": 57420
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.3114815090408514e-05,
+      "loss": 0.3839,
+      "step": 57430
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.3111838678473105e-05,
+      "loss": 0.3769,
+      "step": 57440
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.310886226653769e-05,
+      "loss": 0.374,
+      "step": 57450
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.310588585460228e-05,
+      "loss": 0.3539,
+      "step": 57460
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.3102909442666865e-05,
+      "loss": 0.3892,
+      "step": 57470
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.3099933030731456e-05,
+      "loss": 0.3823,
+      "step": 57480
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.3096956618796042e-05,
+      "loss": 0.3749,
+      "step": 57490
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.3093980206860633e-05,
+      "loss": 0.3727,
+      "step": 57500
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.309100379492522e-05,
+      "loss": 0.3815,
+      "step": 57510
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.308802738298981e-05,
+      "loss": 0.3707,
+      "step": 57520
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.3085050971054397e-05,
+      "loss": 0.3694,
+      "step": 57530
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.3082074559118984e-05,
+      "loss": 0.3828,
+      "step": 57540
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.3079098147183574e-05,
+      "loss": 0.3786,
+      "step": 57550
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.3076121735248158e-05,
+      "loss": 0.3775,
+      "step": 57560
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.3073145323312748e-05,
+      "loss": 0.3864,
+      "step": 57570
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.3070168911377335e-05,
+      "loss": 0.364,
+      "step": 57580
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.3067192499441925e-05,
+      "loss": 0.3611,
+      "step": 57590
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.3064216087506512e-05,
+      "loss": 0.3696,
+      "step": 57600
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.3061239675571102e-05,
+      "loss": 0.3575,
+      "step": 57610
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.305826326363569e-05,
+      "loss": 0.3811,
+      "step": 57620
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.305528685170028e-05,
+      "loss": 0.3857,
+      "step": 57630
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.3052310439764866e-05,
+      "loss": 0.3807,
+      "step": 57640
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.3049334027829457e-05,
+      "loss": 0.3765,
+      "step": 57650
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.304635761589404e-05,
+      "loss": 0.3716,
+      "step": 57660
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.3043381203958627e-05,
+      "loss": 0.3778,
+      "step": 57670
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.3040404792023217e-05,
+      "loss": 0.3768,
+      "step": 57680
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.3037428380087804e-05,
+      "loss": 0.3858,
+      "step": 57690
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.3034451968152394e-05,
+      "loss": 0.372,
+      "step": 57700
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.303147555621698e-05,
+      "loss": 0.3651,
+      "step": 57710
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.302849914428157e-05,
+      "loss": 0.3857,
+      "step": 57720
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.302552273234616e-05,
+      "loss": 0.3754,
+      "step": 57730
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.302254632041075e-05,
+      "loss": 0.3739,
+      "step": 57740
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.3019569908475336e-05,
+      "loss": 0.3782,
+      "step": 57750
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.3016593496539926e-05,
+      "loss": 0.382,
+      "step": 57760
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.301361708460451e-05,
+      "loss": 0.383,
+      "step": 57770
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.3010640672669096e-05,
+      "loss": 0.3731,
+      "step": 57780
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.3007664260733687e-05,
+      "loss": 0.3806,
+      "step": 57790
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.3004687848798274e-05,
+      "loss": 0.3681,
+      "step": 57800
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.3001711436862864e-05,
+      "loss": 0.3868,
+      "step": 57810
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.299873502492745e-05,
+      "loss": 0.3688,
+      "step": 57820
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.299575861299204e-05,
+      "loss": 0.3754,
+      "step": 57830
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.2992782201056628e-05,
+      "loss": 0.3734,
+      "step": 57840
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2989805789121218e-05,
+      "loss": 0.3864,
+      "step": 57850
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2986829377185805e-05,
+      "loss": 0.3844,
+      "step": 57860
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.2983852965250392e-05,
+      "loss": 0.3814,
+      "step": 57870
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.298087655331498e-05,
+      "loss": 0.3757,
+      "step": 57880
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.297790014137957e-05,
+      "loss": 0.3748,
+      "step": 57890
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2974923729444156e-05,
+      "loss": 0.3918,
+      "step": 57900
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2971947317508746e-05,
+      "loss": 0.3677,
+      "step": 57910
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2968970905573333e-05,
+      "loss": 0.3754,
+      "step": 57920
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.2965994493637924e-05,
+      "loss": 0.3784,
+      "step": 57930
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.296301808170251e-05,
+      "loss": 0.3804,
+      "step": 57940
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.29600416697671e-05,
+      "loss": 0.3842,
+      "step": 57950
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.2957065257831688e-05,
+      "loss": 0.3795,
+      "step": 57960
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.295408884589627e-05,
+      "loss": 0.3805,
+      "step": 57970
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.295111243396086e-05,
+      "loss": 0.357,
+      "step": 57980
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.294813602202545e-05,
+      "loss": 0.3638,
+      "step": 57990
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.294515961009004e-05,
+      "loss": 0.3881,
+      "step": 58000
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2942183198154626e-05,
+      "loss": 0.3937,
+      "step": 58010
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.2939206786219216e-05,
+      "loss": 0.3848,
+      "step": 58020
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2936230374283803e-05,
+      "loss": 0.3685,
+      "step": 58030
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.2933253962348393e-05,
+      "loss": 0.3826,
+      "step": 58040
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.293027755041298e-05,
+      "loss": 0.3762,
+      "step": 58050
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.292730113847757e-05,
+      "loss": 0.3865,
+      "step": 58060
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2924324726542154e-05,
+      "loss": 0.3751,
+      "step": 58070
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.292134831460674e-05,
+      "loss": 0.3694,
+      "step": 58080
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.291837190267133e-05,
+      "loss": 0.3578,
+      "step": 58090
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2915395490735918e-05,
+      "loss": 0.38,
+      "step": 58100
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.2912419078800508e-05,
+      "loss": 0.3714,
+      "step": 58110
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2909442666865095e-05,
+      "loss": 0.3807,
+      "step": 58120
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2906466254929685e-05,
+      "loss": 0.3727,
+      "step": 58130
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2903489842994272e-05,
+      "loss": 0.37,
+      "step": 58140
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.2900513431058863e-05,
+      "loss": 0.3717,
+      "step": 58150
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.289753701912345e-05,
+      "loss": 0.3825,
+      "step": 58160
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.289456060718804e-05,
+      "loss": 0.3692,
+      "step": 58170
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.2891584195252623e-05,
+      "loss": 0.3634,
+      "step": 58180
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.288860778331721e-05,
+      "loss": 0.3664,
+      "step": 58190
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.28856313713818e-05,
+      "loss": 0.3809,
+      "step": 58200
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2882654959446387e-05,
+      "loss": 0.3838,
+      "step": 58210
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2879678547510978e-05,
+      "loss": 0.3703,
+      "step": 58220
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2876702135575564e-05,
+      "loss": 0.367,
+      "step": 58230
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.2873725723640155e-05,
+      "loss": 0.3733,
+      "step": 58240
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.287074931170474e-05,
+      "loss": 0.3726,
+      "step": 58250
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2867772899769332e-05,
+      "loss": 0.3697,
+      "step": 58260
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.286479648783392e-05,
+      "loss": 0.3626,
+      "step": 58270
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2861820075898506e-05,
+      "loss": 0.3861,
+      "step": 58280
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.2858843663963093e-05,
+      "loss": 0.386,
+      "step": 58290
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2855867252027683e-05,
+      "loss": 0.3683,
+      "step": 58300
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.285289084009227e-05,
+      "loss": 0.381,
+      "step": 58310
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.284991442815686e-05,
+      "loss": 0.3779,
+      "step": 58320
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.2846938016221447e-05,
+      "loss": 0.3775,
+      "step": 58330
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2843961604286037e-05,
+      "loss": 0.3838,
+      "step": 58340
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2840985192350624e-05,
+      "loss": 0.3754,
+      "step": 58350
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2838008780415214e-05,
+      "loss": 0.3829,
+      "step": 58360
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.28350323684798e-05,
+      "loss": 0.3743,
+      "step": 58370
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.2832055956544385e-05,
+      "loss": 0.3721,
+      "step": 58380
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2829079544608975e-05,
+      "loss": 0.3864,
+      "step": 58390
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2826103132673562e-05,
+      "loss": 0.3714,
+      "step": 58400
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.2823126720738152e-05,
+      "loss": 0.3709,
+      "step": 58410
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.282015030880274e-05,
+      "loss": 0.3752,
+      "step": 58420
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.281717389686733e-05,
+      "loss": 0.3724,
+      "step": 58430
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2814197484931916e-05,
+      "loss": 0.3699,
+      "step": 58440
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2811221072996507e-05,
+      "loss": 0.3609,
+      "step": 58450
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2808244661061094e-05,
+      "loss": 0.3976,
+      "step": 58460
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.2805268249125684e-05,
+      "loss": 0.3728,
+      "step": 58470
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2802291837190267e-05,
+      "loss": 0.3752,
+      "step": 58480
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2799315425254854e-05,
+      "loss": 0.3779,
+      "step": 58490
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.2796339013319445e-05,
+      "loss": 0.3773,
+      "step": 58500
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.279336260138403e-05,
+      "loss": 0.3789,
+      "step": 58510
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2790386189448622e-05,
+      "loss": 0.3854,
+      "step": 58520
+    },
+    {
+      "epoch": 13.0,
+      "eval_cer": 4.544871794871795,
+      "eval_loss": 0.4418568015098572,
+      "eval_runtime": 1.2784,
+      "eval_samples_per_second": 7.822,
+      "eval_steps_per_second": 0.782,
+      "eval_wer": 0.5974025974025974,
+      "step": 58526
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.278740977751321e-05,
+      "loss": 0.3701,
+      "step": 58530
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.27844333655778e-05,
+      "loss": 0.3662,
+      "step": 58540
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2781456953642386e-05,
+      "loss": 0.379,
+      "step": 58550
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.2778480541706976e-05,
+      "loss": 0.3717,
+      "step": 58560
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2775504129771563e-05,
+      "loss": 0.3833,
+      "step": 58570
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2772527717836153e-05,
+      "loss": 0.3577,
+      "step": 58580
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2769551305900737e-05,
+      "loss": 0.3634,
+      "step": 58590
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.2766574893965324e-05,
+      "loss": 0.357,
+      "step": 58600
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.2763598482029914e-05,
+      "loss": 0.3632,
+      "step": 58610
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.27606220700945e-05,
+      "loss": 0.3531,
+      "step": 58620
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.275764565815909e-05,
+      "loss": 0.3486,
+      "step": 58630
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.2754669246223678e-05,
+      "loss": 0.3648,
+      "step": 58640
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.275169283428827e-05,
+      "loss": 0.3617,
+      "step": 58650
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2748716422352855e-05,
+      "loss": 0.3726,
+      "step": 58660
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2745740010417446e-05,
+      "loss": 0.3528,
+      "step": 58670
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.2742763598482033e-05,
+      "loss": 0.3671,
+      "step": 58680
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 2.273978718654662e-05,
+      "loss": 0.3641,
+      "step": 58690
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2736810774611206e-05,
+      "loss": 0.3609,
+      "step": 58700
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2733834362675797e-05,
+      "loss": 0.3709,
+      "step": 58710
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2730857950740384e-05,
+      "loss": 0.3779,
+      "step": 58720
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.2727881538804974e-05,
+      "loss": 0.3495,
+      "step": 58730
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 2.272490512686956e-05,
+      "loss": 0.3638,
+      "step": 58740
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.272192871493415e-05,
+      "loss": 0.3672,
+      "step": 58750
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2718952302998738e-05,
+      "loss": 0.3593,
+      "step": 58760
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2715975891063328e-05,
+      "loss": 0.3641,
+      "step": 58770
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 2.2712999479127915e-05,
+      "loss": 0.3676,
+      "step": 58780
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.27100230671925e-05,
+      "loss": 0.3782,
+      "step": 58790
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.270704665525709e-05,
+      "loss": 0.3673,
+      "step": 58800
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2704070243321676e-05,
+      "loss": 0.3525,
+      "step": 58810
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2701093831386266e-05,
+      "loss": 0.3737,
+      "step": 58820
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 2.2698117419450853e-05,
+      "loss": 0.3849,
+      "step": 58830
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2695141007515443e-05,
+      "loss": 0.3502,
+      "step": 58840
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.269216459558003e-05,
+      "loss": 0.3793,
+      "step": 58850
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.268918818364462e-05,
+      "loss": 0.3653,
+      "step": 58860
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 2.2686211771709207e-05,
+      "loss": 0.3628,
+      "step": 58870
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2683235359773798e-05,
+      "loss": 0.3581,
+      "step": 58880
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.268025894783838e-05,
+      "loss": 0.3548,
+      "step": 58890
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2677282535902968e-05,
+      "loss": 0.3548,
+      "step": 58900
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.267430612396756e-05,
+      "loss": 0.3712,
+      "step": 58910
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 2.2671329712032145e-05,
+      "loss": 0.353,
+      "step": 58920
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2668353300096735e-05,
+      "loss": 0.3425,
+      "step": 58930
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2665376888161322e-05,
+      "loss": 0.3572,
+      "step": 58940
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.2662400476225913e-05,
+      "loss": 0.3782,
+      "step": 58950
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 2.26594240642905e-05,
+      "loss": 0.3694,
+      "step": 58960
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.265644765235509e-05,
+      "loss": 0.3659,
+      "step": 58970
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2653471240419677e-05,
+      "loss": 0.3598,
+      "step": 58980
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2650494828484267e-05,
+      "loss": 0.3605,
+      "step": 58990
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.264751841654885e-05,
+      "loss": 0.3781,
+      "step": 59000
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 2.2644542004613437e-05,
+      "loss": 0.3588,
+      "step": 59010
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2641565592678028e-05,
+      "loss": 0.3591,
+      "step": 59020
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2638589180742615e-05,
+      "loss": 0.3704,
+      "step": 59030
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2635612768807205e-05,
+      "loss": 0.3603,
+      "step": 59040
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 2.2632636356871792e-05,
+      "loss": 0.345,
+      "step": 59050
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2629659944936382e-05,
+      "loss": 0.3639,
+      "step": 59060
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.262668353300097e-05,
+      "loss": 0.3577,
+      "step": 59070
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.262370712106556e-05,
+      "loss": 0.3628,
+      "step": 59080
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2620730709130146e-05,
+      "loss": 0.3769,
+      "step": 59090
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 2.2617754297194733e-05,
+      "loss": 0.367,
+      "step": 59100
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.261477788525932e-05,
+      "loss": 0.3683,
+      "step": 59110
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.261180147332391e-05,
+      "loss": 0.3914,
+      "step": 59120
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.2608825061388497e-05,
+      "loss": 0.3635,
+      "step": 59130
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 2.2605848649453087e-05,
+      "loss": 0.3444,
+      "step": 59140
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2602872237517674e-05,
+      "loss": 0.3612,
+      "step": 59150
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2599895825582265e-05,
+      "loss": 0.38,
+      "step": 59160
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.259691941364685e-05,
+      "loss": 0.3667,
+      "step": 59170
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.2593943001711442e-05,
+      "loss": 0.351,
+      "step": 59180
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 2.259096658977603e-05,
+      "loss": 0.3455,
+      "step": 59190
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2587990177840612e-05,
+      "loss": 0.3853,
+      "step": 59200
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.2585013765905203e-05,
+      "loss": 0.3668,
+      "step": 59210
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.258203735396979e-05,
+      "loss": 0.3735,
+      "step": 59220
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 2.257906094203438e-05,
+      "loss": 0.37,
+      "step": 59230
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2576084530098967e-05,
+      "loss": 0.3666,
+      "step": 59240
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2573108118163557e-05,
+      "loss": 0.3622,
+      "step": 59250
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2570131706228144e-05,
+      "loss": 0.3659,
+      "step": 59260
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.2567155294292734e-05,
+      "loss": 0.3719,
+      "step": 59270
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 2.256417888235732e-05,
+      "loss": 0.3647,
+      "step": 59280
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.256120247042191e-05,
+      "loss": 0.3782,
+      "step": 59290
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2558226058486495e-05,
+      "loss": 0.3701,
+      "step": 59300
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2555249646551082e-05,
+      "loss": 0.3619,
+      "step": 59310
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 2.2552273234615672e-05,
+      "loss": 0.3578,
+      "step": 59320
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.254929682268026e-05,
+      "loss": 0.3588,
+      "step": 59330
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.254632041074485e-05,
+      "loss": 0.3647,
+      "step": 59340
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2543343998809436e-05,
+      "loss": 0.3716,
+      "step": 59350
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2540367586874026e-05,
+      "loss": 0.3553,
+      "step": 59360
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 2.2537391174938613e-05,
+      "loss": 0.3742,
+      "step": 59370
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2534414763003204e-05,
+      "loss": 0.3543,
+      "step": 59380
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.253143835106779e-05,
+      "loss": 0.36,
+      "step": 59390
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2528461939132374e-05,
+      "loss": 0.3663,
+      "step": 59400
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 2.2525485527196964e-05,
+      "loss": 0.3608,
+      "step": 59410
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.252250911526155e-05,
+      "loss": 0.3703,
+      "step": 59420
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.251953270332614e-05,
+      "loss": 0.3716,
+      "step": 59430
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.251655629139073e-05,
+      "loss": 0.3556,
+      "step": 59440
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.251357987945532e-05,
+      "loss": 0.3697,
+      "step": 59450
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 2.2510603467519906e-05,
+      "loss": 0.3628,
+      "step": 59460
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2507627055584496e-05,
+      "loss": 0.3751,
+      "step": 59470
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2504650643649083e-05,
+      "loss": 0.3463,
+      "step": 59480
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.2501674231713673e-05,
+      "loss": 0.3846,
+      "step": 59490
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 2.249869781977826e-05,
+      "loss": 0.3598,
+      "step": 59500
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2495721407842847e-05,
+      "loss": 0.3674,
+      "step": 59510
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2492744995907434e-05,
+      "loss": 0.3549,
+      "step": 59520
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.2489768583972024e-05,
+      "loss": 0.3475,
+      "step": 59530
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.248679217203661e-05,
+      "loss": 0.3664,
+      "step": 59540
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 2.24838157601012e-05,
+      "loss": 0.3816,
+      "step": 59550
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2480839348165788e-05,
+      "loss": 0.3685,
+      "step": 59560
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.247786293623038e-05,
+      "loss": 0.3745,
+      "step": 59570
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2474886524294965e-05,
+      "loss": 0.3601,
+      "step": 59580
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 2.2471910112359556e-05,
+      "loss": 0.3683,
+      "step": 59590
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2468933700424142e-05,
+      "loss": 0.3595,
+      "step": 59600
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2465957288488726e-05,
+      "loss": 0.3651,
+      "step": 59610
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2462980876553316e-05,
+      "loss": 0.3605,
+      "step": 59620
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2460004464617903e-05,
+      "loss": 0.3724,
+      "step": 59630
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 2.2457028052682493e-05,
+      "loss": 0.3723,
+      "step": 59640
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.245405164074708e-05,
+      "loss": 0.3707,
+      "step": 59650
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.245107522881167e-05,
+      "loss": 0.3521,
+      "step": 59660
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2448098816876257e-05,
+      "loss": 0.3721,
+      "step": 59670
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 2.2445122404940848e-05,
+      "loss": 0.3598,
+      "step": 59680
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2442145993005435e-05,
+      "loss": 0.3622,
+      "step": 59690
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2439169581070025e-05,
+      "loss": 0.3683,
+      "step": 59700
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.243619316913461e-05,
+      "loss": 0.3478,
+      "step": 59710
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2433216757199195e-05,
+      "loss": 0.3559,
+      "step": 59720
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 2.2430240345263786e-05,
+      "loss": 0.3783,
+      "step": 59730
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2427263933328373e-05,
+      "loss": 0.3638,
+      "step": 59740
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.2424287521392963e-05,
+      "loss": 0.3677,
+      "step": 59750
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.242131110945755e-05,
+      "loss": 0.371,
+      "step": 59760
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 2.241833469752214e-05,
+      "loss": 0.3564,
+      "step": 59770
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2415358285586727e-05,
+      "loss": 0.3587,
+      "step": 59780
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2412381873651317e-05,
+      "loss": 0.3573,
+      "step": 59790
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2409405461715904e-05,
+      "loss": 0.3533,
+      "step": 59800
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2406429049780488e-05,
+      "loss": 0.3558,
+      "step": 59810
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 2.2403452637845078e-05,
+      "loss": 0.3662,
+      "step": 59820
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2400476225909665e-05,
+      "loss": 0.3702,
+      "step": 59830
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2397499813974255e-05,
+      "loss": 0.3624,
+      "step": 59840
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2394523402038842e-05,
+      "loss": 0.3521,
+      "step": 59850
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 2.2391546990103432e-05,
+      "loss": 0.3645,
+      "step": 59860
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.238857057816802e-05,
+      "loss": 0.3709,
+      "step": 59870
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.238559416623261e-05,
+      "loss": 0.3543,
+      "step": 59880
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2382617754297196e-05,
+      "loss": 0.3709,
+      "step": 59890
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2379641342361787e-05,
+      "loss": 0.3637,
+      "step": 59900
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 2.2376664930426374e-05,
+      "loss": 0.355,
+      "step": 59910
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.237368851849096e-05,
+      "loss": 0.3644,
+      "step": 59920
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2370712106555547e-05,
+      "loss": 0.3557,
+      "step": 59930
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2367735694620138e-05,
+      "loss": 0.3681,
+      "step": 59940
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 2.2364759282684725e-05,
+      "loss": 0.3499,
+      "step": 59950
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2361782870749315e-05,
+      "loss": 0.3712,
+      "step": 59960
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2358806458813902e-05,
+      "loss": 0.3442,
+      "step": 59970
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.2355830046878492e-05,
+      "loss": 0.3583,
+      "step": 59980
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.235285363494308e-05,
+      "loss": 0.3621,
+      "step": 59990
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 2.234987722300767e-05,
+      "loss": 0.3637,
+      "step": 60000
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2346900811072256e-05,
+      "loss": 0.3609,
+      "step": 60010
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.234392439913684e-05,
+      "loss": 0.37,
+      "step": 60020
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.234094798720143e-05,
+      "loss": 0.3632,
+      "step": 60030
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 2.2337971575266017e-05,
+      "loss": 0.3673,
+      "step": 60040
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2334995163330607e-05,
+      "loss": 0.3509,
+      "step": 60050
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2332018751395194e-05,
+      "loss": 0.3733,
+      "step": 60060
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.2329042339459784e-05,
+      "loss": 0.3656,
+      "step": 60070
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.232606592752437e-05,
+      "loss": 0.3545,
+      "step": 60080
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 2.232308951558896e-05,
+      "loss": 0.3737,
+      "step": 60090
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.232011310365355e-05,
+      "loss": 0.3591,
+      "step": 60100
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.231713669171814e-05,
+      "loss": 0.3472,
+      "step": 60110
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.2314160279782722e-05,
+      "loss": 0.3668,
+      "step": 60120
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 2.231118386784731e-05,
+      "loss": 0.3666,
+      "step": 60130
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.23082074559119e-05,
+      "loss": 0.3916,
+      "step": 60140
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2305231043976486e-05,
+      "loss": 0.3535,
+      "step": 60150
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2302254632041077e-05,
+      "loss": 0.3602,
+      "step": 60160
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2299278220105663e-05,
+      "loss": 0.3638,
+      "step": 60170
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 2.2296301808170254e-05,
+      "loss": 0.3571,
+      "step": 60180
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.229332539623484e-05,
+      "loss": 0.3621,
+      "step": 60190
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.229034898429943e-05,
+      "loss": 0.357,
+      "step": 60200
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.2287372572364018e-05,
+      "loss": 0.3509,
+      "step": 60210
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 2.22843961604286e-05,
+      "loss": 0.3571,
+      "step": 60220
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.228141974849319e-05,
+      "loss": 0.375,
+      "step": 60230
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.227844333655778e-05,
+      "loss": 0.3663,
+      "step": 60240
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.227546692462237e-05,
+      "loss": 0.3816,
+      "step": 60250
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2272490512686956e-05,
+      "loss": 0.3684,
+      "step": 60260
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 2.2269514100751546e-05,
+      "loss": 0.3512,
+      "step": 60270
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.2266537688816133e-05,
+      "loss": 0.3744,
+      "step": 60280
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.2263561276880723e-05,
+      "loss": 0.3729,
+      "step": 60290
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.226058486494531e-05,
+      "loss": 0.3694,
+      "step": 60300
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 2.22576084530099e-05,
+      "loss": 0.3624,
+      "step": 60310
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.2254632041074487e-05,
+      "loss": 0.3835,
+      "step": 60320
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.2251655629139074e-05,
+      "loss": 0.3682,
+      "step": 60330
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.224867921720366e-05,
+      "loss": 0.3668,
+      "step": 60340
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.224570280526825e-05,
+      "loss": 0.3703,
+      "step": 60350
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 2.2242726393332838e-05,
+      "loss": 0.3709,
+      "step": 60360
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.223974998139743e-05,
+      "loss": 0.3707,
+      "step": 60370
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.2236773569462015e-05,
+      "loss": 0.3787,
+      "step": 60380
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.2233797157526606e-05,
+      "loss": 0.3572,
+      "step": 60390
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 2.2230820745591193e-05,
+      "loss": 0.3641,
+      "step": 60400
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.2227844333655783e-05,
+      "loss": 0.3587,
+      "step": 60410
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.222486792172037e-05,
+      "loss": 0.3758,
+      "step": 60420
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.2221891509784953e-05,
+      "loss": 0.3598,
+      "step": 60430
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.2218915097849544e-05,
+      "loss": 0.3614,
+      "step": 60440
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 2.221593868591413e-05,
+      "loss": 0.3819,
+      "step": 60450
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.221296227397872e-05,
+      "loss": 0.3628,
+      "step": 60460
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.2209985862043308e-05,
+      "loss": 0.3561,
+      "step": 60470
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.2207009450107898e-05,
+      "loss": 0.3665,
+      "step": 60480
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 2.2204033038172485e-05,
+      "loss": 0.3694,
+      "step": 60490
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.2201056626237075e-05,
+      "loss": 0.3676,
+      "step": 60500
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.2198080214301662e-05,
+      "loss": 0.3562,
+      "step": 60510
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.2195103802366252e-05,
+      "loss": 0.3427,
+      "step": 60520
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.2192127390430836e-05,
+      "loss": 0.3512,
+      "step": 60530
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 2.2189150978495423e-05,
+      "loss": 0.3694,
+      "step": 60540
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.2186174566560013e-05,
+      "loss": 0.3716,
+      "step": 60550
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.21831981546246e-05,
+      "loss": 0.37,
+      "step": 60560
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.218022174268919e-05,
+      "loss": 0.3752,
+      "step": 60570
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 2.2177245330753777e-05,
+      "loss": 0.361,
+      "step": 60580
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.2174268918818367e-05,
+      "loss": 0.3538,
+      "step": 60590
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.2171292506882954e-05,
+      "loss": 0.3766,
+      "step": 60600
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.2168316094947545e-05,
+      "loss": 0.3647,
+      "step": 60610
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.216533968301213e-05,
+      "loss": 0.3546,
+      "step": 60620
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 2.216236327107672e-05,
+      "loss": 0.3837,
+      "step": 60630
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.2159386859141305e-05,
+      "loss": 0.3575,
+      "step": 60640
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.2156410447205892e-05,
+      "loss": 0.3681,
+      "step": 60650
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.2153434035270482e-05,
+      "loss": 0.3768,
+      "step": 60660
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 2.215045762333507e-05,
+      "loss": 0.3674,
+      "step": 60670
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.214748121139966e-05,
+      "loss": 0.3617,
+      "step": 60680
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.2144504799464247e-05,
+      "loss": 0.3731,
+      "step": 60690
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.2141528387528837e-05,
+      "loss": 0.365,
+      "step": 60700
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.2138551975593424e-05,
+      "loss": 0.3665,
+      "step": 60710
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 2.2135575563658014e-05,
+      "loss": 0.379,
+      "step": 60720
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.21325991517226e-05,
+      "loss": 0.3568,
+      "step": 60730
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.2129622739787188e-05,
+      "loss": 0.3619,
+      "step": 60740
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.2126646327851775e-05,
+      "loss": 0.3636,
+      "step": 60750
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 2.2123669915916365e-05,
+      "loss": 0.3538,
+      "step": 60760
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.2120693503980952e-05,
+      "loss": 0.3568,
+      "step": 60770
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.2117717092045542e-05,
+      "loss": 0.3547,
+      "step": 60780
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.211474068011013e-05,
+      "loss": 0.3642,
+      "step": 60790
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.211176426817472e-05,
+      "loss": 0.3658,
+      "step": 60800
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 2.2108787856239306e-05,
+      "loss": 0.3739,
+      "step": 60810
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.2105811444303897e-05,
+      "loss": 0.3647,
+      "step": 60820
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.2102835032368483e-05,
+      "loss": 0.3507,
+      "step": 60830
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.2099858620433067e-05,
+      "loss": 0.3713,
+      "step": 60840
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 2.2096882208497657e-05,
+      "loss": 0.3623,
+      "step": 60850
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.2093905796562244e-05,
+      "loss": 0.3719,
+      "step": 60860
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.2090929384626834e-05,
+      "loss": 0.3769,
+      "step": 60870
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.208795297269142e-05,
+      "loss": 0.3524,
+      "step": 60880
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.208497656075601e-05,
+      "loss": 0.3693,
+      "step": 60890
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 2.20820001488206e-05,
+      "loss": 0.3676,
+      "step": 60900
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.207902373688519e-05,
+      "loss": 0.366,
+      "step": 60910
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.2076047324949776e-05,
+      "loss": 0.3646,
+      "step": 60920
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.2073070913014366e-05,
+      "loss": 0.3615,
+      "step": 60930
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 2.207009450107895e-05,
+      "loss": 0.3526,
+      "step": 60940
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.2067118089143536e-05,
+      "loss": 0.3773,
+      "step": 60950
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.2064141677208127e-05,
+      "loss": 0.3519,
+      "step": 60960
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.2061165265272714e-05,
+      "loss": 0.3581,
+      "step": 60970
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.2058188853337304e-05,
+      "loss": 0.3786,
+      "step": 60980
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.205521244140189e-05,
+      "loss": 0.3619,
+      "step": 60990
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.205223602946648e-05,
+      "loss": 0.3675,
+      "step": 61000
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.2049259617531068e-05,
+      "loss": 0.3649,
+      "step": 61010
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.2046283205595658e-05,
+      "loss": 0.3569,
+      "step": 61020
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 2.2043306793660245e-05,
+      "loss": 0.3804,
+      "step": 61030
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.2040330381724832e-05,
+      "loss": 0.3839,
+      "step": 61040
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.203735396978942e-05,
+      "loss": 0.3586,
+      "step": 61050
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.203437755785401e-05,
+      "loss": 0.3713,
+      "step": 61060
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.2031401145918596e-05,
+      "loss": 0.3695,
+      "step": 61070
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 2.2028424733983183e-05,
+      "loss": 0.3747,
+      "step": 61080
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.2025448322047773e-05,
+      "loss": 0.365,
+      "step": 61090
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.202247191011236e-05,
+      "loss": 0.3677,
+      "step": 61100
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.201949549817695e-05,
+      "loss": 0.3665,
+      "step": 61110
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 2.2016519086241537e-05,
+      "loss": 0.3704,
+      "step": 61120
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.2013542674306128e-05,
+      "loss": 0.3567,
+      "step": 61130
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.201056626237071e-05,
+      "loss": 0.3706,
+      "step": 61140
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.20075898504353e-05,
+      "loss": 0.3621,
+      "step": 61150
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.200461343849989e-05,
+      "loss": 0.3642,
+      "step": 61160
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 2.200163702656448e-05,
+      "loss": 0.3718,
+      "step": 61170
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1998660614629066e-05,
+      "loss": 0.3737,
+      "step": 61180
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1995684202693656e-05,
+      "loss": 0.3582,
+      "step": 61190
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1992707790758243e-05,
+      "loss": 0.3636,
+      "step": 61200
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 2.1989731378822833e-05,
+      "loss": 0.372,
+      "step": 61210
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.198675496688742e-05,
+      "loss": 0.3736,
+      "step": 61220
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.198377855495201e-05,
+      "loss": 0.3521,
+      "step": 61230
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.1980802143016597e-05,
+      "loss": 0.375,
+      "step": 61240
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.197782573108118e-05,
+      "loss": 0.3736,
+      "step": 61250
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.197484931914577e-05,
+      "loss": 0.3646,
+      "step": 61260
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1971872907210358e-05,
+      "loss": 0.3775,
+      "step": 61270
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1968896495274948e-05,
+      "loss": 0.3713,
+      "step": 61280
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1965920083339535e-05,
+      "loss": 0.3582,
+      "step": 61290
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 2.1962943671404125e-05,
+      "loss": 0.3613,
+      "step": 61300
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1959967259468712e-05,
+      "loss": 0.3482,
+      "step": 61310
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1956990847533302e-05,
+      "loss": 0.3619,
+      "step": 61320
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.195401443559789e-05,
+      "loss": 0.3586,
+      "step": 61330
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.195103802366248e-05,
+      "loss": 0.3644,
+      "step": 61340
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 2.1948061611727063e-05,
+      "loss": 0.3508,
+      "step": 61350
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.194508519979165e-05,
+      "loss": 0.3731,
+      "step": 61360
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.194210878785624e-05,
+      "loss": 0.3725,
+      "step": 61370
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1939132375920827e-05,
+      "loss": 0.3598,
+      "step": 61380
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 2.1936155963985418e-05,
+      "loss": 0.3676,
+      "step": 61390
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1933179552050004e-05,
+      "loss": 0.3804,
+      "step": 61400
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1930203140114595e-05,
+      "loss": 0.3649,
+      "step": 61410
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.192722672817918e-05,
+      "loss": 0.3646,
+      "step": 61420
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.1924250316243772e-05,
+      "loss": 0.3726,
+      "step": 61430
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 2.192127390430836e-05,
+      "loss": 0.3672,
+      "step": 61440
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1918297492372946e-05,
+      "loss": 0.3592,
+      "step": 61450
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1915321080437533e-05,
+      "loss": 0.3557,
+      "step": 61460
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.1912344668502123e-05,
+      "loss": 0.3746,
+      "step": 61470
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.190936825656671e-05,
+      "loss": 0.3588,
+      "step": 61480
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.19063918446313e-05,
+      "loss": 0.3567,
+      "step": 61490
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1903415432695887e-05,
+      "loss": 0.3717,
+      "step": 61500
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1900439020760474e-05,
+      "loss": 0.3795,
+      "step": 61510
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.1897462608825064e-05,
+      "loss": 0.3574,
+      "step": 61520
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 2.189448619688965e-05,
+      "loss": 0.3576,
+      "step": 61530
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.189150978495424e-05,
+      "loss": 0.3659,
+      "step": 61540
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1888533373018825e-05,
+      "loss": 0.3746,
+      "step": 61550
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1885556961083415e-05,
+      "loss": 0.3717,
+      "step": 61560
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 2.1882580549148002e-05,
+      "loss": 0.3678,
+      "step": 61570
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1879604137212592e-05,
+      "loss": 0.3786,
+      "step": 61580
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.187662772527718e-05,
+      "loss": 0.3573,
+      "step": 61590
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.187365131334177e-05,
+      "loss": 0.3675,
+      "step": 61600
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1870674901406356e-05,
+      "loss": 0.388,
+      "step": 61610
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 2.1867698489470947e-05,
+      "loss": 0.3562,
+      "step": 61620
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1864722077535534e-05,
+      "loss": 0.3652,
+      "step": 61630
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1861745665600124e-05,
+      "loss": 0.3629,
+      "step": 61640
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.185876925366471e-05,
+      "loss": 0.3552,
+      "step": 61650
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 2.1855792841729294e-05,
+      "loss": 0.361,
+      "step": 61660
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1852816429793885e-05,
+      "loss": 0.3722,
+      "step": 61670
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.184984001785847e-05,
+      "loss": 0.3618,
+      "step": 61680
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.1846863605923062e-05,
+      "loss": 0.3598,
+      "step": 61690
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.184388719398765e-05,
+      "loss": 0.3531,
+      "step": 61700
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.184091078205224e-05,
+      "loss": 0.357,
+      "step": 61710
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1837934370116826e-05,
+      "loss": 0.3679,
+      "step": 61720
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1834957958181416e-05,
+      "loss": 0.3619,
+      "step": 61730
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1831981546246003e-05,
+      "loss": 0.3785,
+      "step": 61740
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 2.1829005134310593e-05,
+      "loss": 0.3653,
+      "step": 61750
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1826028722375177e-05,
+      "loss": 0.3534,
+      "step": 61760
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1823052310439764e-05,
+      "loss": 0.3555,
+      "step": 61770
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.1820075898504354e-05,
+      "loss": 0.3616,
+      "step": 61780
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.181709948656894e-05,
+      "loss": 0.3595,
+      "step": 61790
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 2.181412307463353e-05,
+      "loss": 0.3591,
+      "step": 61800
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1811146662698118e-05,
+      "loss": 0.378,
+      "step": 61810
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.180817025076271e-05,
+      "loss": 0.3536,
+      "step": 61820
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1805193838827295e-05,
+      "loss": 0.3631,
+      "step": 61830
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 2.1802217426891886e-05,
+      "loss": 0.3761,
+      "step": 61840
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1799241014956472e-05,
+      "loss": 0.3746,
+      "step": 61850
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.179626460302106e-05,
+      "loss": 0.3639,
+      "step": 61860
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1793288191085646e-05,
+      "loss": 0.3619,
+      "step": 61870
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1790311779150237e-05,
+      "loss": 0.3541,
+      "step": 61880
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 2.1787335367214823e-05,
+      "loss": 0.3579,
+      "step": 61890
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.1784358955279414e-05,
+      "loss": 0.3725,
+      "step": 61900
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.1781382543344e-05,
+      "loss": 0.3456,
+      "step": 61910
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.177840613140859e-05,
+      "loss": 0.3739,
+      "step": 61920
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 2.1775429719473178e-05,
+      "loss": 0.3662,
+      "step": 61930
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1772453307537765e-05,
+      "loss": 0.3644,
+      "step": 61940
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1769476895602355e-05,
+      "loss": 0.3663,
+      "step": 61950
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.176650048366694e-05,
+      "loss": 0.3661,
+      "step": 61960
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.176352407173153e-05,
+      "loss": 0.3575,
+      "step": 61970
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.1760547659796116e-05,
+      "loss": 0.3718,
+      "step": 61980
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.1757571247860706e-05,
+      "loss": 0.3653,
+      "step": 61990
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.1754594835925293e-05,
+      "loss": 0.3456,
+      "step": 62000
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.1751618423989883e-05,
+      "loss": 0.3597,
+      "step": 62010
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 2.174864201205447e-05,
+      "loss": 0.3555,
+      "step": 62020
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.174566560011906e-05,
+      "loss": 0.3728,
+      "step": 62030
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1742689188183647e-05,
+      "loss": 0.3674,
+      "step": 62040
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1739712776248238e-05,
+      "loss": 0.3646,
+      "step": 62050
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1736736364312824e-05,
+      "loss": 0.361,
+      "step": 62060
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 2.1733759952377408e-05,
+      "loss": 0.3524,
+      "step": 62070
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1730783540441998e-05,
+      "loss": 0.3574,
+      "step": 62080
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1727807128506585e-05,
+      "loss": 0.353,
+      "step": 62090
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1724830716571175e-05,
+      "loss": 0.3666,
+      "step": 62100
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 2.1721854304635762e-05,
+      "loss": 0.3543,
+      "step": 62110
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1718877892700353e-05,
+      "loss": 0.357,
+      "step": 62120
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.171590148076494e-05,
+      "loss": 0.355,
+      "step": 62130
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.171292506882953e-05,
+      "loss": 0.3609,
+      "step": 62140
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1709948656894117e-05,
+      "loss": 0.3636,
+      "step": 62150
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 2.1706972244958707e-05,
+      "loss": 0.3569,
+      "step": 62160
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.170399583302329e-05,
+      "loss": 0.3469,
+      "step": 62170
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.1701019421087877e-05,
+      "loss": 0.3532,
+      "step": 62180
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.1698043009152468e-05,
+      "loss": 0.364,
+      "step": 62190
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.1695066597217055e-05,
+      "loss": 0.3641,
+      "step": 62200
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1692090185281645e-05,
+      "loss": 0.3662,
+      "step": 62210
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1689113773346232e-05,
+      "loss": 0.3604,
+      "step": 62220
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.1686137361410822e-05,
+      "loss": 0.371,
+      "step": 62230
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.168316094947541e-05,
+      "loss": 0.3593,
+      "step": 62240
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 2.168018453754e-05,
+      "loss": 0.3618,
+      "step": 62250
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.1677208125604586e-05,
+      "loss": 0.3648,
+      "step": 62260
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.1674231713669173e-05,
+      "loss": 0.3723,
+      "step": 62270
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.167125530173376e-05,
+      "loss": 0.3558,
+      "step": 62280
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 2.166827888979835e-05,
+      "loss": 0.3549,
+      "step": 62290
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1665302477862937e-05,
+      "loss": 0.3672,
+      "step": 62300
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1662326065927527e-05,
+      "loss": 0.3605,
+      "step": 62310
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1659349653992114e-05,
+      "loss": 0.357,
+      "step": 62320
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.1656373242056705e-05,
+      "loss": 0.3518,
+      "step": 62330
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 2.165339683012129e-05,
+      "loss": 0.3524,
+      "step": 62340
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1650420418185882e-05,
+      "loss": 0.3618,
+      "step": 62350
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.164744400625047e-05,
+      "loss": 0.3452,
+      "step": 62360
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1644467594315052e-05,
+      "loss": 0.3706,
+      "step": 62370
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 2.1641491182379643e-05,
+      "loss": 0.3715,
+      "step": 62380
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.163851477044423e-05,
+      "loss": 0.3498,
+      "step": 62390
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.163553835850882e-05,
+      "loss": 0.3427,
+      "step": 62400
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.1632561946573407e-05,
+      "loss": 0.3553,
+      "step": 62410
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.1629585534637997e-05,
+      "loss": 0.3635,
+      "step": 62420
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 2.1626609122702584e-05,
+      "loss": 0.3546,
+      "step": 62430
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1623632710767174e-05,
+      "loss": 0.3621,
+      "step": 62440
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.162065629883176e-05,
+      "loss": 0.3564,
+      "step": 62450
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.161767988689635e-05,
+      "loss": 0.3449,
+      "step": 62460
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.1614703474960938e-05,
+      "loss": 0.3654,
+      "step": 62470
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.161172706302552e-05,
+      "loss": 0.3597,
+      "step": 62480
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1608750651090112e-05,
+      "loss": 0.3446,
+      "step": 62490
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.16057742391547e-05,
+      "loss": 0.3565,
+      "step": 62500
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.160279782721929e-05,
+      "loss": 0.3632,
+      "step": 62510
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 2.1599821415283876e-05,
+      "loss": 0.3677,
+      "step": 62520
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1596845003348466e-05,
+      "loss": 0.346,
+      "step": 62530
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1593868591413053e-05,
+      "loss": 0.3711,
+      "step": 62540
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.1590892179477644e-05,
+      "loss": 0.3829,
+      "step": 62550
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 2.158791576754223e-05,
+      "loss": 0.3607,
+      "step": 62560
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.158493935560682e-05,
+      "loss": 0.3558,
+      "step": 62570
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1581962943671404e-05,
+      "loss": 0.3456,
+      "step": 62580
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.157898653173599e-05,
+      "loss": 0.3678,
+      "step": 62590
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.157601011980058e-05,
+      "loss": 0.3649,
+      "step": 62600
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 2.1573033707865168e-05,
+      "loss": 0.3583,
+      "step": 62610
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.157005729592976e-05,
+      "loss": 0.3727,
+      "step": 62620
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1567080883994345e-05,
+      "loss": 0.3499,
+      "step": 62630
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1564104472058936e-05,
+      "loss": 0.3536,
+      "step": 62640
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 2.1561128060123523e-05,
+      "loss": 0.3485,
+      "step": 62650
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1558151648188113e-05,
+      "loss": 0.3626,
+      "step": 62660
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.15551752362527e-05,
+      "loss": 0.3599,
+      "step": 62670
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1552198824317287e-05,
+      "loss": 0.3532,
+      "step": 62680
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1549222412381874e-05,
+      "loss": 0.3796,
+      "step": 62690
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.1546246000446464e-05,
+      "loss": 0.3691,
+      "step": 62700
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.154326958851105e-05,
+      "loss": 0.3506,
+      "step": 62710
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.154029317657564e-05,
+      "loss": 0.3694,
+      "step": 62720
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1537316764640228e-05,
+      "loss": 0.3602,
+      "step": 62730
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 2.1534340352704818e-05,
+      "loss": 0.3546,
+      "step": 62740
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1531363940769405e-05,
+      "loss": 0.3706,
+      "step": 62750
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1528387528833995e-05,
+      "loss": 0.378,
+      "step": 62760
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1525411116898582e-05,
+      "loss": 0.3702,
+      "step": 62770
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1522434704963166e-05,
+      "loss": 0.3709,
+      "step": 62780
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 2.1519458293027756e-05,
+      "loss": 0.3706,
+      "step": 62790
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1516481881092343e-05,
+      "loss": 0.345,
+      "step": 62800
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.1513505469156933e-05,
+      "loss": 0.3565,
+      "step": 62810
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.151052905722152e-05,
+      "loss": 0.3501,
+      "step": 62820
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 2.150755264528611e-05,
+      "loss": 0.3437,
+      "step": 62830
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1504576233350697e-05,
+      "loss": 0.3643,
+      "step": 62840
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1501599821415288e-05,
+      "loss": 0.3467,
+      "step": 62850
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1498623409479875e-05,
+      "loss": 0.3612,
+      "step": 62860
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.1495646997544465e-05,
+      "loss": 0.3668,
+      "step": 62870
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 2.149267058560905e-05,
+      "loss": 0.3657,
+      "step": 62880
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1489694173673635e-05,
+      "loss": 0.3734,
+      "step": 62890
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1486717761738226e-05,
+      "loss": 0.3787,
+      "step": 62900
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1483741349802813e-05,
+      "loss": 0.3531,
+      "step": 62910
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.1480764937867403e-05,
+      "loss": 0.3649,
+      "step": 62920
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.147778852593199e-05,
+      "loss": 0.3723,
+      "step": 62930
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.147481211399658e-05,
+      "loss": 0.3614,
+      "step": 62940
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1471835702061167e-05,
+      "loss": 0.3574,
+      "step": 62950
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1468859290125757e-05,
+      "loss": 0.3423,
+      "step": 62960
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 2.1465882878190344e-05,
+      "loss": 0.3718,
+      "step": 62970
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1462906466254934e-05,
+      "loss": 0.3713,
+      "step": 62980
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1459930054319518e-05,
+      "loss": 0.3561,
+      "step": 62990
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1456953642384105e-05,
+      "loss": 0.3623,
+      "step": 63000
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 2.1453977230448695e-05,
+      "loss": 0.3772,
+      "step": 63010
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.1451000818513282e-05,
+      "loss": 0.3548,
+      "step": 63020
+    },
+    {
+      "epoch": 14.0,
+      "eval_cer": 4.549145299145299,
+      "eval_loss": 0.44927114248275757,
+      "eval_runtime": 1.2762,
+      "eval_samples_per_second": 7.836,
+      "eval_steps_per_second": 0.784,
+      "eval_wer": 0.6623376623376623,
+      "step": 63028
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.1448024406577872e-05,
+      "loss": 0.3511,
+      "step": 63030
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.144504799464246e-05,
+      "loss": 0.3583,
+      "step": 63040
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.144207158270705e-05,
+      "loss": 0.3522,
+      "step": 63050
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 2.1439095170771636e-05,
+      "loss": 0.3607,
+      "step": 63060
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1436118758836227e-05,
+      "loss": 0.3683,
+      "step": 63070
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1433142346900814e-05,
+      "loss": 0.3525,
+      "step": 63080
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.14301659349654e-05,
+      "loss": 0.3384,
+      "step": 63090
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 2.1427189523029987e-05,
+      "loss": 0.374,
+      "step": 63100
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1424213111094578e-05,
+      "loss": 0.3361,
+      "step": 63110
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1421236699159165e-05,
+      "loss": 0.3454,
+      "step": 63120
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1418260287223755e-05,
+      "loss": 0.368,
+      "step": 63130
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.141528387528834e-05,
+      "loss": 0.3602,
+      "step": 63140
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 2.1412307463352932e-05,
+      "loss": 0.3367,
+      "step": 63150
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.140933105141752e-05,
+      "loss": 0.3385,
+      "step": 63160
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.140635463948211e-05,
+      "loss": 0.3404,
+      "step": 63170
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.1403378227546696e-05,
+      "loss": 0.3433,
+      "step": 63180
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 2.140040181561128e-05,
+      "loss": 0.3418,
+      "step": 63190
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.139742540367587e-05,
+      "loss": 0.3392,
+      "step": 63200
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1394448991740457e-05,
+      "loss": 0.3573,
+      "step": 63210
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1391472579805047e-05,
+      "loss": 0.3459,
+      "step": 63220
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1388496167869634e-05,
+      "loss": 0.3624,
+      "step": 63230
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 2.1385519755934224e-05,
+      "loss": 0.3522,
+      "step": 63240
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.138254334399881e-05,
+      "loss": 0.352,
+      "step": 63250
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.13795669320634e-05,
+      "loss": 0.3585,
+      "step": 63260
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.137659052012799e-05,
+      "loss": 0.3591,
+      "step": 63270
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 2.137361410819258e-05,
+      "loss": 0.3568,
+      "step": 63280
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1370637696257162e-05,
+      "loss": 0.3551,
+      "step": 63290
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.136766128432175e-05,
+      "loss": 0.3513,
+      "step": 63300
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.136468487238634e-05,
+      "loss": 0.3499,
+      "step": 63310
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1361708460450926e-05,
+      "loss": 0.3445,
+      "step": 63320
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 2.1358732048515516e-05,
+      "loss": 0.3589,
+      "step": 63330
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1355755636580103e-05,
+      "loss": 0.3549,
+      "step": 63340
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.1352779224644694e-05,
+      "loss": 0.3482,
+      "step": 63350
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.134980281270928e-05,
+      "loss": 0.354,
+      "step": 63360
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 2.134682640077387e-05,
+      "loss": 0.3506,
+      "step": 63370
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1343849988838458e-05,
+      "loss": 0.3414,
+      "step": 63380
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.1340873576903048e-05,
+      "loss": 0.3523,
+      "step": 63390
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.133789716496763e-05,
+      "loss": 0.357,
+      "step": 63400
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.133492075303222e-05,
+      "loss": 0.3446,
+      "step": 63410
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 2.133194434109681e-05,
+      "loss": 0.3471,
+      "step": 63420
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1328967929161396e-05,
+      "loss": 0.3543,
+      "step": 63430
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1325991517225986e-05,
+      "loss": 0.3568,
+      "step": 63440
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1323015105290573e-05,
+      "loss": 0.3478,
+      "step": 63450
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 2.1320038693355163e-05,
+      "loss": 0.3509,
+      "step": 63460
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.131706228141975e-05,
+      "loss": 0.3522,
+      "step": 63470
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.131408586948434e-05,
+      "loss": 0.377,
+      "step": 63480
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1311109457548927e-05,
+      "loss": 0.3539,
+      "step": 63490
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.1308133045613514e-05,
+      "loss": 0.3562,
+      "step": 63500
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 2.13051566336781e-05,
+      "loss": 0.3527,
+      "step": 63510
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.130218022174269e-05,
+      "loss": 0.3457,
+      "step": 63520
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.1299203809807278e-05,
+      "loss": 0.3588,
+      "step": 63530
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.129622739787187e-05,
+      "loss": 0.3568,
+      "step": 63540
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 2.1293250985936455e-05,
+      "loss": 0.3463,
+      "step": 63550
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1290274574001046e-05,
+      "loss": 0.3536,
+      "step": 63560
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1287298162065633e-05,
+      "loss": 0.3584,
+      "step": 63570
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1284321750130223e-05,
+      "loss": 0.3512,
+      "step": 63580
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.128134533819481e-05,
+      "loss": 0.3453,
+      "step": 63590
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 2.1278368926259393e-05,
+      "loss": 0.3604,
+      "step": 63600
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.1275392514323984e-05,
+      "loss": 0.338,
+      "step": 63610
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.127241610238857e-05,
+      "loss": 0.3603,
+      "step": 63620
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.126943969045316e-05,
+      "loss": 0.3553,
+      "step": 63630
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 2.1266463278517748e-05,
+      "loss": 0.3584,
+      "step": 63640
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1263486866582338e-05,
+      "loss": 0.3479,
+      "step": 63650
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1260510454646925e-05,
+      "loss": 0.3606,
+      "step": 63660
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1257534042711515e-05,
+      "loss": 0.3428,
+      "step": 63670
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1254557630776102e-05,
+      "loss": 0.3456,
+      "step": 63680
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 2.1251581218840692e-05,
+      "loss": 0.3344,
+      "step": 63690
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1248604806905276e-05,
+      "loss": 0.3551,
+      "step": 63700
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1245628394969863e-05,
+      "loss": 0.3581,
+      "step": 63710
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.1242651983034453e-05,
+      "loss": 0.3648,
+      "step": 63720
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 2.123967557109904e-05,
+      "loss": 0.3492,
+      "step": 63730
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.123669915916363e-05,
+      "loss": 0.3472,
+      "step": 63740
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1233722747228217e-05,
+      "loss": 0.3555,
+      "step": 63750
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1230746335292807e-05,
+      "loss": 0.3456,
+      "step": 63760
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1227769923357394e-05,
+      "loss": 0.3551,
+      "step": 63770
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 2.1224793511421985e-05,
+      "loss": 0.3504,
+      "step": 63780
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.122181709948657e-05,
+      "loss": 0.359,
+      "step": 63790
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1218840687551162e-05,
+      "loss": 0.364,
+      "step": 63800
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1215864275615745e-05,
+      "loss": 0.3501,
+      "step": 63810
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1212887863680332e-05,
+      "loss": 0.3528,
+      "step": 63820
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 2.1209911451744922e-05,
+      "loss": 0.349,
+      "step": 63830
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.120693503980951e-05,
+      "loss": 0.35,
+      "step": 63840
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.12039586278741e-05,
+      "loss": 0.3618,
+      "step": 63850
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1200982215938686e-05,
+      "loss": 0.3568,
+      "step": 63860
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 2.1198005804003277e-05,
+      "loss": 0.3568,
+      "step": 63870
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1195029392067864e-05,
+      "loss": 0.3536,
+      "step": 63880
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1192052980132454e-05,
+      "loss": 0.3496,
+      "step": 63890
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.118907656819704e-05,
+      "loss": 0.352,
+      "step": 63900
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1186100156261628e-05,
+      "loss": 0.3321,
+      "step": 63910
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 2.1183123744326215e-05,
+      "loss": 0.3554,
+      "step": 63920
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1180147332390805e-05,
+      "loss": 0.3492,
+      "step": 63930
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1177170920455392e-05,
+      "loss": 0.3444,
+      "step": 63940
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.1174194508519982e-05,
+      "loss": 0.3419,
+      "step": 63950
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 2.117121809658457e-05,
+      "loss": 0.3688,
+      "step": 63960
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.116824168464916e-05,
+      "loss": 0.3742,
+      "step": 63970
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1165265272713746e-05,
+      "loss": 0.3498,
+      "step": 63980
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1162288860778337e-05,
+      "loss": 0.374,
+      "step": 63990
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1159312448842923e-05,
+      "loss": 0.3662,
+      "step": 64000
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 2.1156336036907507e-05,
+      "loss": 0.3547,
+      "step": 64010
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1153359624972097e-05,
+      "loss": 0.3441,
+      "step": 64020
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1150383213036684e-05,
+      "loss": 0.3533,
+      "step": 64030
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.1147406801101274e-05,
+      "loss": 0.3336,
+      "step": 64040
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 2.114443038916586e-05,
+      "loss": 0.3633,
+      "step": 64050
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.114145397723045e-05,
+      "loss": 0.3585,
+      "step": 64060
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.113847756529504e-05,
+      "loss": 0.3627,
+      "step": 64070
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.113550115335963e-05,
+      "loss": 0.3521,
+      "step": 64080
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1132524741424216e-05,
+      "loss": 0.3399,
+      "step": 64090
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 2.1129548329488806e-05,
+      "loss": 0.3462,
+      "step": 64100
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.112657191755339e-05,
+      "loss": 0.3619,
+      "step": 64110
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1123595505617976e-05,
+      "loss": 0.3582,
+      "step": 64120
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1120619093682567e-05,
+      "loss": 0.3376,
+      "step": 64130
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 2.1117642681747154e-05,
+      "loss": 0.346,
+      "step": 64140
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1114666269811744e-05,
+      "loss": 0.3414,
+      "step": 64150
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.111168985787633e-05,
+      "loss": 0.3497,
+      "step": 64160
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.110871344594092e-05,
+      "loss": 0.3567,
+      "step": 64170
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1105737034005508e-05,
+      "loss": 0.3659,
+      "step": 64180
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 2.1102760622070098e-05,
+      "loss": 0.3498,
+      "step": 64190
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1099784210134685e-05,
+      "loss": 0.3482,
+      "step": 64200
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1096807798199275e-05,
+      "loss": 0.3428,
+      "step": 64210
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.109383138626386e-05,
+      "loss": 0.3606,
+      "step": 64220
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 2.1090854974328446e-05,
+      "loss": 0.344,
+      "step": 64230
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1087878562393036e-05,
+      "loss": 0.3622,
+      "step": 64240
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1084902150457623e-05,
+      "loss": 0.3491,
+      "step": 64250
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.1081925738522213e-05,
+      "loss": 0.3466,
+      "step": 64260
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.10789493265868e-05,
+      "loss": 0.3497,
+      "step": 64270
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 2.107597291465139e-05,
+      "loss": 0.3466,
+      "step": 64280
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1072996502715977e-05,
+      "loss": 0.3572,
+      "step": 64290
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1070020090780568e-05,
+      "loss": 0.3542,
+      "step": 64300
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.1067043678845155e-05,
+      "loss": 0.3565,
+      "step": 64310
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 2.106406726690974e-05,
+      "loss": 0.3459,
+      "step": 64320
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.106109085497433e-05,
+      "loss": 0.3553,
+      "step": 64330
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.105811444303892e-05,
+      "loss": 0.365,
+      "step": 64340
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1055138031103506e-05,
+      "loss": 0.3422,
+      "step": 64350
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1052161619168096e-05,
+      "loss": 0.3419,
+      "step": 64360
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 2.1049185207232683e-05,
+      "loss": 0.3445,
+      "step": 64370
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.1046208795297273e-05,
+      "loss": 0.3701,
+      "step": 64380
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.104323238336186e-05,
+      "loss": 0.3365,
+      "step": 64390
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.104025597142645e-05,
+      "loss": 0.3603,
+      "step": 64400
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 2.1037279559491037e-05,
+      "loss": 0.3518,
+      "step": 64410
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.103430314755562e-05,
+      "loss": 0.3436,
+      "step": 64420
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.103132673562021e-05,
+      "loss": 0.3493,
+      "step": 64430
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1028350323684798e-05,
+      "loss": 0.345,
+      "step": 64440
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1025373911749388e-05,
+      "loss": 0.3457,
+      "step": 64450
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 2.1022397499813975e-05,
+      "loss": 0.3694,
+      "step": 64460
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1019421087878565e-05,
+      "loss": 0.3653,
+      "step": 64470
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1016444675943152e-05,
+      "loss": 0.3696,
+      "step": 64480
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.1013468264007742e-05,
+      "loss": 0.3521,
+      "step": 64490
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 2.101049185207233e-05,
+      "loss": 0.3391,
+      "step": 64500
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.100751544013692e-05,
+      "loss": 0.3588,
+      "step": 64510
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.1004539028201503e-05,
+      "loss": 0.3815,
+      "step": 64520
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.100156261626609e-05,
+      "loss": 0.3411,
+      "step": 64530
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.099858620433068e-05,
+      "loss": 0.3479,
+      "step": 64540
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 2.0995609792395267e-05,
+      "loss": 0.3389,
+      "step": 64550
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.0992633380459858e-05,
+      "loss": 0.36,
+      "step": 64560
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.0989656968524444e-05,
+      "loss": 0.3473,
+      "step": 64570
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.0986680556589035e-05,
+      "loss": 0.3439,
+      "step": 64580
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 2.098370414465362e-05,
+      "loss": 0.3564,
+      "step": 64590
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.0980727732718212e-05,
+      "loss": 0.3441,
+      "step": 64600
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.09777513207828e-05,
+      "loss": 0.338,
+      "step": 64610
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.0974774908847386e-05,
+      "loss": 0.3483,
+      "step": 64620
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.0971798496911973e-05,
+      "loss": 0.3624,
+      "step": 64630
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 2.096882208497656e-05,
+      "loss": 0.3601,
+      "step": 64640
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.096584567304115e-05,
+      "loss": 0.3341,
+      "step": 64650
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.0962869261105737e-05,
+      "loss": 0.3492,
+      "step": 64660
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.0959892849170327e-05,
+      "loss": 0.3463,
+      "step": 64670
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 2.0956916437234914e-05,
+      "loss": 0.3506,
+      "step": 64680
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.0953940025299504e-05,
+      "loss": 0.3471,
+      "step": 64690
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.095096361336409e-05,
+      "loss": 0.3515,
+      "step": 64700
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.094798720142868e-05,
+      "loss": 0.355,
+      "step": 64710
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.0945010789493268e-05,
+      "loss": 0.3463,
+      "step": 64720
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 2.0942034377557855e-05,
+      "loss": 0.3643,
+      "step": 64730
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.0939057965622442e-05,
+      "loss": 0.3658,
+      "step": 64740
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.0936081553687032e-05,
+      "loss": 0.3444,
+      "step": 64750
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.093310514175162e-05,
+      "loss": 0.3371,
+      "step": 64760
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 2.093012872981621e-05,
+      "loss": 0.3566,
+      "step": 64770
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.0927152317880796e-05,
+      "loss": 0.3595,
+      "step": 64780
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.0924175905945387e-05,
+      "loss": 0.344,
+      "step": 64790
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.0921199494009974e-05,
+      "loss": 0.3482,
+      "step": 64800
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.0918223082074564e-05,
+      "loss": 0.3672,
+      "step": 64810
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 2.091524667013915e-05,
+      "loss": 0.3566,
+      "step": 64820
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.0912270258203734e-05,
+      "loss": 0.3551,
+      "step": 64830
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.0909293846268325e-05,
+      "loss": 0.3475,
+      "step": 64840
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.090631743433291e-05,
+      "loss": 0.351,
+      "step": 64850
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 2.0903341022397502e-05,
+      "loss": 0.3423,
+      "step": 64860
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.090036461046209e-05,
+      "loss": 0.3548,
+      "step": 64870
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.089738819852668e-05,
+      "loss": 0.3479,
+      "step": 64880
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.0894411786591266e-05,
+      "loss": 0.3542,
+      "step": 64890
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.0891435374655856e-05,
+      "loss": 0.3567,
+      "step": 64900
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 2.0888458962720443e-05,
+      "loss": 0.3466,
+      "step": 64910
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.0885482550785033e-05,
+      "loss": 0.3475,
+      "step": 64920
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.0882506138849617e-05,
+      "loss": 0.3644,
+      "step": 64930
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.0879529726914204e-05,
+      "loss": 0.3545,
+      "step": 64940
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 2.0876553314978794e-05,
+      "loss": 0.3564,
+      "step": 64950
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.087357690304338e-05,
+      "loss": 0.3567,
+      "step": 64960
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.087060049110797e-05,
+      "loss": 0.3609,
+      "step": 64970
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.0867624079172558e-05,
+      "loss": 0.3491,
+      "step": 64980
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.086464766723715e-05,
+      "loss": 0.3607,
+      "step": 64990
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 2.0861671255301735e-05,
+      "loss": 0.3466,
+      "step": 65000
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.0858694843366326e-05,
+      "loss": 0.3481,
+      "step": 65010
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.0855718431430912e-05,
+      "loss": 0.3536,
+      "step": 65020
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.08527420194955e-05,
+      "loss": 0.3537,
+      "step": 65030
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 2.0849765607560086e-05,
+      "loss": 0.3512,
+      "step": 65040
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.0846789195624673e-05,
+      "loss": 0.3561,
+      "step": 65050
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.0843812783689263e-05,
+      "loss": 0.3545,
+      "step": 65060
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.084083637175385e-05,
+      "loss": 0.3531,
+      "step": 65070
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.083785995981844e-05,
+      "loss": 0.3586,
+      "step": 65080
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 2.0834883547883028e-05,
+      "loss": 0.3662,
+      "step": 65090
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.0831907135947618e-05,
+      "loss": 0.3534,
+      "step": 65100
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.0828930724012205e-05,
+      "loss": 0.368,
+      "step": 65110
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.0825954312076795e-05,
+      "loss": 0.3609,
+      "step": 65120
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 2.0822977900141382e-05,
+      "loss": 0.3494,
+      "step": 65130
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.082000148820597e-05,
+      "loss": 0.3472,
+      "step": 65140
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.0817025076270556e-05,
+      "loss": 0.3644,
+      "step": 65150
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.0814048664335146e-05,
+      "loss": 0.3508,
+      "step": 65160
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.0811072252399733e-05,
+      "loss": 0.3564,
+      "step": 65170
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 2.0808095840464323e-05,
+      "loss": 0.3594,
+      "step": 65180
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.080511942852891e-05,
+      "loss": 0.3541,
+      "step": 65190
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.08021430165935e-05,
+      "loss": 0.3561,
+      "step": 65200
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.0799166604658087e-05,
+      "loss": 0.3636,
+      "step": 65210
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 2.0796190192722678e-05,
+      "loss": 0.3817,
+      "step": 65220
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.0793213780787264e-05,
+      "loss": 0.3584,
+      "step": 65230
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.0790237368851848e-05,
+      "loss": 0.3715,
+      "step": 65240
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.0787260956916438e-05,
+      "loss": 0.3536,
+      "step": 65250
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.0784284544981025e-05,
+      "loss": 0.3475,
+      "step": 65260
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 2.0781308133045615e-05,
+      "loss": 0.3597,
+      "step": 65270
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.0778331721110202e-05,
+      "loss": 0.3476,
+      "step": 65280
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.0775355309174793e-05,
+      "loss": 0.3571,
+      "step": 65290
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.077237889723938e-05,
+      "loss": 0.3392,
+      "step": 65300
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 2.076940248530397e-05,
+      "loss": 0.3552,
+      "step": 65310
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.0766426073368557e-05,
+      "loss": 0.3511,
+      "step": 65320
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.0763449661433147e-05,
+      "loss": 0.3475,
+      "step": 65330
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.076047324949773e-05,
+      "loss": 0.351,
+      "step": 65340
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.0757496837562317e-05,
+      "loss": 0.3595,
+      "step": 65350
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 2.0754520425626908e-05,
+      "loss": 0.3666,
+      "step": 65360
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.0751544013691495e-05,
+      "loss": 0.3722,
+      "step": 65370
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.0748567601756085e-05,
+      "loss": 0.3479,
+      "step": 65380
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.0745591189820672e-05,
+      "loss": 0.3381,
+      "step": 65390
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 2.0742614777885262e-05,
+      "loss": 0.3402,
+      "step": 65400
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.073963836594985e-05,
+      "loss": 0.341,
+      "step": 65410
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.073666195401444e-05,
+      "loss": 0.3595,
+      "step": 65420
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.0733685542079026e-05,
+      "loss": 0.3618,
+      "step": 65430
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.0730709130143613e-05,
+      "loss": 0.3561,
+      "step": 65440
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 2.07277327182082e-05,
+      "loss": 0.3441,
+      "step": 65450
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.072475630627279e-05,
+      "loss": 0.3554,
+      "step": 65460
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.0721779894337377e-05,
+      "loss": 0.3433,
+      "step": 65470
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.0718803482401964e-05,
+      "loss": 0.3417,
+      "step": 65480
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 2.0715827070466554e-05,
+      "loss": 0.3652,
+      "step": 65490
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.071285065853114e-05,
+      "loss": 0.3497,
+      "step": 65500
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.070987424659573e-05,
+      "loss": 0.3391,
+      "step": 65510
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.070689783466032e-05,
+      "loss": 0.3674,
+      "step": 65520
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.070392142272491e-05,
+      "loss": 0.3587,
+      "step": 65530
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 2.0700945010789496e-05,
+      "loss": 0.3593,
+      "step": 65540
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.0697968598854082e-05,
+      "loss": 0.3647,
+      "step": 65550
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.069499218691867e-05,
+      "loss": 0.3514,
+      "step": 65560
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.069201577498326e-05,
+      "loss": 0.3374,
+      "step": 65570
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 2.0689039363047847e-05,
+      "loss": 0.3497,
+      "step": 65580
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.0686062951112437e-05,
+      "loss": 0.3524,
+      "step": 65590
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.0683086539177024e-05,
+      "loss": 0.3365,
+      "step": 65600
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.0680110127241614e-05,
+      "loss": 0.3433,
+      "step": 65610
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.06771337153062e-05,
+      "loss": 0.3561,
+      "step": 65620
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 2.067415730337079e-05,
+      "loss": 0.3577,
+      "step": 65630
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.0671180891435378e-05,
+      "loss": 0.358,
+      "step": 65640
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.066820447949996e-05,
+      "loss": 0.3626,
+      "step": 65650
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.0665228067564552e-05,
+      "loss": 0.3515,
+      "step": 65660
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 2.066225165562914e-05,
+      "loss": 0.3332,
+      "step": 65670
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.065927524369373e-05,
+      "loss": 0.3592,
+      "step": 65680
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.0656298831758316e-05,
+      "loss": 0.3434,
+      "step": 65690
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.0653322419822906e-05,
+      "loss": 0.3426,
+      "step": 65700
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.0650346007887493e-05,
+      "loss": 0.34,
+      "step": 65710
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 2.0647369595952083e-05,
+      "loss": 0.3447,
+      "step": 65720
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.064439318401667e-05,
+      "loss": 0.3314,
+      "step": 65730
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.064141677208126e-05,
+      "loss": 0.3594,
+      "step": 65740
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.0638440360145844e-05,
+      "loss": 0.3404,
+      "step": 65750
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 2.063546394821043e-05,
+      "loss": 0.3568,
+      "step": 65760
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.063248753627502e-05,
+      "loss": 0.3545,
+      "step": 65770
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.0629511124339608e-05,
+      "loss": 0.3554,
+      "step": 65780
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.06265347124042e-05,
+      "loss": 0.3569,
+      "step": 65790
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.0623558300468785e-05,
+      "loss": 0.359,
+      "step": 65800
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 2.0620581888533376e-05,
+      "loss": 0.351,
+      "step": 65810
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.0617605476597963e-05,
+      "loss": 0.3408,
+      "step": 65820
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.0614629064662553e-05,
+      "loss": 0.349,
+      "step": 65830
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.061165265272714e-05,
+      "loss": 0.3673,
+      "step": 65840
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 2.0608676240791727e-05,
+      "loss": 0.3562,
+      "step": 65850
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.0605699828856314e-05,
+      "loss": 0.3423,
+      "step": 65860
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.0602723416920904e-05,
+      "loss": 0.344,
+      "step": 65870
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.059974700498549e-05,
+      "loss": 0.369,
+      "step": 65880
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.059677059305008e-05,
+      "loss": 0.3535,
+      "step": 65890
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 2.0593794181114668e-05,
+      "loss": 0.3538,
+      "step": 65900
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.0590817769179255e-05,
+      "loss": 0.3465,
+      "step": 65910
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.0587841357243845e-05,
+      "loss": 0.3615,
+      "step": 65920
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.0584864945308432e-05,
+      "loss": 0.3732,
+      "step": 65930
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 2.0581888533373022e-05,
+      "loss": 0.3693,
+      "step": 65940
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.057891212143761e-05,
+      "loss": 0.3587,
+      "step": 65950
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.0575935709502196e-05,
+      "loss": 0.3753,
+      "step": 65960
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.0572959297566783e-05,
+      "loss": 0.3503,
+      "step": 65970
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.0569982885631373e-05,
+      "loss": 0.3434,
+      "step": 65980
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 2.056700647369596e-05,
+      "loss": 0.3586,
+      "step": 65990
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.056403006176055e-05,
+      "loss": 0.3563,
+      "step": 66000
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.0561053649825137e-05,
+      "loss": 0.3402,
+      "step": 66010
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.0558077237889728e-05,
+      "loss": 0.354,
+      "step": 66020
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 2.0555100825954315e-05,
+      "loss": 0.3532,
+      "step": 66030
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.0552124414018905e-05,
+      "loss": 0.3532,
+      "step": 66040
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.0549148002083492e-05,
+      "loss": 0.3387,
+      "step": 66050
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.0546171590148075e-05,
+      "loss": 0.3598,
+      "step": 66060
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.0543195178212666e-05,
+      "loss": 0.3469,
+      "step": 66070
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 2.0540218766277252e-05,
+      "loss": 0.3507,
+      "step": 66080
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.0537242354341843e-05,
+      "loss": 0.3547,
+      "step": 66090
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.053426594240643e-05,
+      "loss": 0.3592,
+      "step": 66100
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.053128953047102e-05,
+      "loss": 0.3416,
+      "step": 66110
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 2.0528313118535607e-05,
+      "loss": 0.3478,
+      "step": 66120
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.0525336706600197e-05,
+      "loss": 0.3534,
+      "step": 66130
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.0522360294664784e-05,
+      "loss": 0.3468,
+      "step": 66140
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.0519383882729374e-05,
+      "loss": 0.3406,
+      "step": 66150
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.0516407470793958e-05,
+      "loss": 0.3442,
+      "step": 66160
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 2.0513431058858545e-05,
+      "loss": 0.3515,
+      "step": 66170
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.0510454646923135e-05,
+      "loss": 0.3334,
+      "step": 66180
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.0507478234987722e-05,
+      "loss": 0.3551,
+      "step": 66190
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.0504501823052312e-05,
+      "loss": 0.3406,
+      "step": 66200
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 2.05015254111169e-05,
+      "loss": 0.3544,
+      "step": 66210
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.049854899918149e-05,
+      "loss": 0.3456,
+      "step": 66220
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.0495572587246076e-05,
+      "loss": 0.3534,
+      "step": 66230
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.0492596175310667e-05,
+      "loss": 0.3558,
+      "step": 66240
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.0489619763375253e-05,
+      "loss": 0.332,
+      "step": 66250
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 2.048664335143984e-05,
+      "loss": 0.3476,
+      "step": 66260
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.0483666939504427e-05,
+      "loss": 0.3528,
+      "step": 66270
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.0480690527569018e-05,
+      "loss": 0.3542,
+      "step": 66280
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.0477714115633604e-05,
+      "loss": 0.3519,
+      "step": 66290
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 2.0474737703698195e-05,
+      "loss": 0.3516,
+      "step": 66300
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.047176129176278e-05,
+      "loss": 0.3563,
+      "step": 66310
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.0468784879827372e-05,
+      "loss": 0.359,
+      "step": 66320
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.046580846789196e-05,
+      "loss": 0.3546,
+      "step": 66330
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.0462832055956546e-05,
+      "loss": 0.3444,
+      "step": 66340
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 2.0459855644021136e-05,
+      "loss": 0.3391,
+      "step": 66350
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.045687923208572e-05,
+      "loss": 0.3468,
+      "step": 66360
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.045390282015031e-05,
+      "loss": 0.3449,
+      "step": 66370
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.0450926408214897e-05,
+      "loss": 0.338,
+      "step": 66380
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 2.0447949996279487e-05,
+      "loss": 0.3379,
+      "step": 66390
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.0444973584344074e-05,
+      "loss": 0.3455,
+      "step": 66400
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.0441997172408664e-05,
+      "loss": 0.3519,
+      "step": 66410
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.043902076047325e-05,
+      "loss": 0.365,
+      "step": 66420
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.043604434853784e-05,
+      "loss": 0.3533,
+      "step": 66430
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 2.0433067936602428e-05,
+      "loss": 0.338,
+      "step": 66440
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.043009152466702e-05,
+      "loss": 0.3476,
+      "step": 66450
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.0427115112731605e-05,
+      "loss": 0.3456,
+      "step": 66460
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.042413870079619e-05,
+      "loss": 0.3381,
+      "step": 66470
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 2.042116228886078e-05,
+      "loss": 0.3417,
+      "step": 66480
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.0418185876925366e-05,
+      "loss": 0.3521,
+      "step": 66490
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.0415209464989956e-05,
+      "loss": 0.3493,
+      "step": 66500
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.0412233053054543e-05,
+      "loss": 0.3528,
+      "step": 66510
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.0409256641119134e-05,
+      "loss": 0.3541,
+      "step": 66520
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 2.040628022918372e-05,
+      "loss": 0.3582,
+      "step": 66530
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.040330381724831e-05,
+      "loss": 0.3432,
+      "step": 66540
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.0400327405312898e-05,
+      "loss": 0.3518,
+      "step": 66550
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.0397350993377488e-05,
+      "loss": 0.3427,
+      "step": 66560
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 2.039437458144207e-05,
+      "loss": 0.3456,
+      "step": 66570
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.039139816950666e-05,
+      "loss": 0.3529,
+      "step": 66580
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.038842175757125e-05,
+      "loss": 0.3406,
+      "step": 66590
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.0385445345635836e-05,
+      "loss": 0.3539,
+      "step": 66600
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.0382468933700426e-05,
+      "loss": 0.334,
+      "step": 66610
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 2.0379492521765013e-05,
+      "loss": 0.3509,
+      "step": 66620
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.0376516109829603e-05,
+      "loss": 0.3495,
+      "step": 66630
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.037353969789419e-05,
+      "loss": 0.3505,
+      "step": 66640
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.037056328595878e-05,
+      "loss": 0.3407,
+      "step": 66650
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 2.0367586874023367e-05,
+      "loss": 0.3573,
+      "step": 66660
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.0364610462087954e-05,
+      "loss": 0.3338,
+      "step": 66670
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.036163405015254e-05,
+      "loss": 0.3488,
+      "step": 66680
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.035865763821713e-05,
+      "loss": 0.3566,
+      "step": 66690
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.0355681226281718e-05,
+      "loss": 0.3561,
+      "step": 66700
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 2.035270481434631e-05,
+      "loss": 0.3556,
+      "step": 66710
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.0349728402410895e-05,
+      "loss": 0.3395,
+      "step": 66720
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.0346751990475486e-05,
+      "loss": 0.3393,
+      "step": 66730
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.0343775578540073e-05,
+      "loss": 0.3548,
+      "step": 66740
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 2.0340799166604663e-05,
+      "loss": 0.3526,
+      "step": 66750
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.033782275466925e-05,
+      "loss": 0.3598,
+      "step": 66760
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.0334846342733833e-05,
+      "loss": 0.3582,
+      "step": 66770
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.0331869930798423e-05,
+      "loss": 0.3474,
+      "step": 66780
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.032889351886301e-05,
+      "loss": 0.3488,
+      "step": 66790
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 2.03259171069276e-05,
+      "loss": 0.3522,
+      "step": 66800
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.0322940694992188e-05,
+      "loss": 0.3618,
+      "step": 66810
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.0319964283056778e-05,
+      "loss": 0.3472,
+      "step": 66820
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.0316987871121365e-05,
+      "loss": 0.3379,
+      "step": 66830
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 2.0314011459185955e-05,
+      "loss": 0.3601,
+      "step": 66840
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.0311035047250542e-05,
+      "loss": 0.3529,
+      "step": 66850
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.0308058635315132e-05,
+      "loss": 0.3421,
+      "step": 66860
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.030508222337972e-05,
+      "loss": 0.3508,
+      "step": 66870
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.0302105811444303e-05,
+      "loss": 0.3427,
+      "step": 66880
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 2.0299129399508893e-05,
+      "loss": 0.3438,
+      "step": 66890
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.029615298757348e-05,
+      "loss": 0.3553,
+      "step": 66900
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.029317657563807e-05,
+      "loss": 0.3442,
+      "step": 66910
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.0290200163702657e-05,
+      "loss": 0.3491,
+      "step": 66920
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 2.0287223751767247e-05,
+      "loss": 0.337,
+      "step": 66930
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.0284247339831834e-05,
+      "loss": 0.3611,
+      "step": 66940
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.0281270927896424e-05,
+      "loss": 0.3434,
+      "step": 66950
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.027829451596101e-05,
+      "loss": 0.3456,
+      "step": 66960
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.02753181040256e-05,
+      "loss": 0.3556,
+      "step": 66970
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 2.0272341692090185e-05,
+      "loss": 0.3439,
+      "step": 66980
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.0269365280154772e-05,
+      "loss": 0.3577,
+      "step": 66990
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.0266388868219362e-05,
+      "loss": 0.3484,
+      "step": 67000
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.026341245628395e-05,
+      "loss": 0.351,
+      "step": 67010
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.026043604434854e-05,
+      "loss": 0.3438,
+      "step": 67020
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.0257459632413126e-05,
+      "loss": 0.3671,
+      "step": 67030
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.0254483220477717e-05,
+      "loss": 0.3348,
+      "step": 67040
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.0251506808542304e-05,
+      "loss": 0.3539,
+      "step": 67050
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.0248530396606894e-05,
+      "loss": 0.3415,
+      "step": 67060
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 2.024555398467148e-05,
+      "loss": 0.358,
+      "step": 67070
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.0242577572736068e-05,
+      "loss": 0.3389,
+      "step": 67080
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.0239601160800655e-05,
+      "loss": 0.3592,
+      "step": 67090
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.0236624748865245e-05,
+      "loss": 0.355,
+      "step": 67100
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 2.0233648336929832e-05,
+      "loss": 0.3473,
+      "step": 67110
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.0230671924994422e-05,
+      "loss": 0.3449,
+      "step": 67120
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.022769551305901e-05,
+      "loss": 0.3547,
+      "step": 67130
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.02247191011236e-05,
+      "loss": 0.3496,
+      "step": 67140
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.0221742689188186e-05,
+      "loss": 0.3533,
+      "step": 67150
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 2.0218766277252776e-05,
+      "loss": 0.3552,
+      "step": 67160
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.0215789865317363e-05,
+      "loss": 0.3501,
+      "step": 67170
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.0212813453381947e-05,
+      "loss": 0.3659,
+      "step": 67180
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.0209837041446537e-05,
+      "loss": 0.3422,
+      "step": 67190
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 2.0206860629511124e-05,
+      "loss": 0.3504,
+      "step": 67200
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.0203884217575714e-05,
+      "loss": 0.3589,
+      "step": 67210
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.02009078056403e-05,
+      "loss": 0.3513,
+      "step": 67220
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.019793139370489e-05,
+      "loss": 0.3453,
+      "step": 67230
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.019495498176948e-05,
+      "loss": 0.3614,
+      "step": 67240
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 2.019197856983407e-05,
+      "loss": 0.34,
+      "step": 67250
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.0189002157898656e-05,
+      "loss": 0.3427,
+      "step": 67260
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.0186025745963246e-05,
+      "loss": 0.3447,
+      "step": 67270
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.0183049334027833e-05,
+      "loss": 0.344,
+      "step": 67280
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 2.0180072922092416e-05,
+      "loss": 0.3487,
+      "step": 67290
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.0177096510157007e-05,
+      "loss": 0.3569,
+      "step": 67300
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.0174120098221594e-05,
+      "loss": 0.359,
+      "step": 67310
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.0171143686286184e-05,
+      "loss": 0.3524,
+      "step": 67320
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.016816727435077e-05,
+      "loss": 0.3484,
+      "step": 67330
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 2.016519086241536e-05,
+      "loss": 0.3446,
+      "step": 67340
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.0162214450479948e-05,
+      "loss": 0.3599,
+      "step": 67350
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.0159238038544538e-05,
+      "loss": 0.3414,
+      "step": 67360
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.0156261626609125e-05,
+      "loss": 0.3554,
+      "step": 67370
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 2.0153285214673715e-05,
+      "loss": 0.355,
+      "step": 67380
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.01503088027383e-05,
+      "loss": 0.3521,
+      "step": 67390
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.0147332390802886e-05,
+      "loss": 0.3437,
+      "step": 67400
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.0144355978867476e-05,
+      "loss": 0.3644,
+      "step": 67410
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.0141379566932063e-05,
+      "loss": 0.3519,
+      "step": 67420
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 2.0138403154996653e-05,
+      "loss": 0.3429,
+      "step": 67430
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.013542674306124e-05,
+      "loss": 0.3742,
+      "step": 67440
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.013245033112583e-05,
+      "loss": 0.3556,
+      "step": 67450
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.0129473919190417e-05,
+      "loss": 0.3458,
+      "step": 67460
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 2.0126497507255008e-05,
+      "loss": 0.3335,
+      "step": 67470
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.0123521095319595e-05,
+      "loss": 0.3544,
+      "step": 67480
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.012054468338418e-05,
+      "loss": 0.3399,
+      "step": 67490
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.011756827144877e-05,
+      "loss": 0.3358,
+      "step": 67500
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.011459185951336e-05,
+      "loss": 0.3498,
+      "step": 67510
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 2.0111615447577945e-05,
+      "loss": 0.3438,
+      "step": 67520
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.0108639035642536e-05,
+      "loss": 0.3444,
+      "step": 67530
+    },
+    {
+      "epoch": 15.0,
+      "eval_cer": 4.532051282051282,
+      "eval_loss": 0.4469204545021057,
+      "eval_runtime": 1.2304,
+      "eval_samples_per_second": 8.127,
+      "eval_steps_per_second": 0.813,
+      "eval_wer": 0.6103896103896104,
+      "step": 67530
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.0105662623707123e-05,
+      "loss": 0.3671,
+      "step": 67540
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.0102686211771713e-05,
+      "loss": 0.3516,
+      "step": 67550
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 2.00997097998363e-05,
+      "loss": 0.3438,
+      "step": 67560
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.009673338790089e-05,
+      "loss": 0.3317,
+      "step": 67570
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.0093756975965477e-05,
+      "loss": 0.3395,
+      "step": 67580
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.009078056403006e-05,
+      "loss": 0.3338,
+      "step": 67590
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.008780415209465e-05,
+      "loss": 0.3412,
+      "step": 67600
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 2.0084827740159238e-05,
+      "loss": 0.3323,
+      "step": 67610
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.0081851328223828e-05,
+      "loss": 0.3558,
+      "step": 67620
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.0078874916288415e-05,
+      "loss": 0.3255,
+      "step": 67630
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.0075898504353005e-05,
+      "loss": 0.3515,
+      "step": 67640
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 2.0072922092417592e-05,
+      "loss": 0.3286,
+      "step": 67650
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.0069945680482182e-05,
+      "loss": 0.3361,
+      "step": 67660
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.006696926854677e-05,
+      "loss": 0.3383,
+      "step": 67670
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.006399285661136e-05,
+      "loss": 0.3359,
+      "step": 67680
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.0061016444675946e-05,
+      "loss": 0.3369,
+      "step": 67690
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 2.005804003274053e-05,
+      "loss": 0.3424,
+      "step": 67700
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.005506362080512e-05,
+      "loss": 0.3359,
+      "step": 67710
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.0052087208869707e-05,
+      "loss": 0.3392,
+      "step": 67720
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.0049110796934297e-05,
+      "loss": 0.3324,
+      "step": 67730
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 2.0046134384998884e-05,
+      "loss": 0.3442,
+      "step": 67740
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.0043157973063475e-05,
+      "loss": 0.3334,
+      "step": 67750
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.004018156112806e-05,
+      "loss": 0.339,
+      "step": 67760
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.0037205149192652e-05,
+      "loss": 0.3496,
+      "step": 67770
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.003422873725724e-05,
+      "loss": 0.3285,
+      "step": 67780
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 2.003125232532183e-05,
+      "loss": 0.3366,
+      "step": 67790
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.0028275913386413e-05,
+      "loss": 0.3456,
+      "step": 67800
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.0025299501451e-05,
+      "loss": 0.3308,
+      "step": 67810
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.002232308951559e-05,
+      "loss": 0.3395,
+      "step": 67820
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 2.0019346677580177e-05,
+      "loss": 0.3434,
+      "step": 67830
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0016370265644767e-05,
+      "loss": 0.3338,
+      "step": 67840
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0013393853709354e-05,
+      "loss": 0.342,
+      "step": 67850
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.0010417441773944e-05,
+      "loss": 0.3481,
+      "step": 67860
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.000744102983853e-05,
+      "loss": 0.3382,
+      "step": 67870
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 2.000446461790312e-05,
+      "loss": 0.3274,
+      "step": 67880
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 2.0001488205967708e-05,
+      "loss": 0.3346,
+      "step": 67890
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.9998511794032295e-05,
+      "loss": 0.3333,
+      "step": 67900
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.9995535382096885e-05,
+      "loss": 0.3439,
+      "step": 67910
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.9992558970161472e-05,
+      "loss": 0.3331,
+      "step": 67920
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.998958255822606e-05,
+      "loss": 0.3454,
+      "step": 67930
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.998660614629065e-05,
+      "loss": 0.3367,
+      "step": 67940
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.9983629734355236e-05,
+      "loss": 0.3295,
+      "step": 67950
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.9980653322419827e-05,
+      "loss": 0.3366,
+      "step": 67960
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.9977676910484414e-05,
+      "loss": 0.3446,
+      "step": 67970
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.9974700498549e-05,
+      "loss": 0.3465,
+      "step": 67980
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.9971724086613587e-05,
+      "loss": 0.3381,
+      "step": 67990
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.9968747674678178e-05,
+      "loss": 0.3322,
+      "step": 68000
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.9965771262742765e-05,
+      "loss": 0.3313,
+      "step": 68010
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.9962794850807355e-05,
+      "loss": 0.3454,
+      "step": 68020
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.9959818438871942e-05,
+      "loss": 0.3233,
+      "step": 68030
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.995684202693653e-05,
+      "loss": 0.3353,
+      "step": 68040
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.995386561500112e-05,
+      "loss": 0.351,
+      "step": 68050
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.9950889203065706e-05,
+      "loss": 0.3473,
+      "step": 68060
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.9947912791130296e-05,
+      "loss": 0.3511,
+      "step": 68070
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.9944936379194883e-05,
+      "loss": 0.3427,
+      "step": 68080
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.994195996725947e-05,
+      "loss": 0.3322,
+      "step": 68090
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.993898355532406e-05,
+      "loss": 0.3503,
+      "step": 68100
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.9936007143388647e-05,
+      "loss": 0.3378,
+      "step": 68110
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.9933030731453234e-05,
+      "loss": 0.3336,
+      "step": 68120
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.993005431951782e-05,
+      "loss": 0.3399,
+      "step": 68130
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.992707790758241e-05,
+      "loss": 0.3536,
+      "step": 68140
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.9924101495646998e-05,
+      "loss": 0.3412,
+      "step": 68150
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.992112508371159e-05,
+      "loss": 0.3515,
+      "step": 68160
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.9918148671776175e-05,
+      "loss": 0.3406,
+      "step": 68170
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.9915172259840762e-05,
+      "loss": 0.3289,
+      "step": 68180
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.9912195847905352e-05,
+      "loss": 0.3473,
+      "step": 68190
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.990921943596994e-05,
+      "loss": 0.345,
+      "step": 68200
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.990624302403453e-05,
+      "loss": 0.3464,
+      "step": 68210
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.9903266612099117e-05,
+      "loss": 0.3427,
+      "step": 68220
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.9900290200163703e-05,
+      "loss": 0.3473,
+      "step": 68230
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.989731378822829e-05,
+      "loss": 0.3408,
+      "step": 68240
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.989433737629288e-05,
+      "loss": 0.3558,
+      "step": 68250
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.9891360964357467e-05,
+      "loss": 0.3479,
+      "step": 68260
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.9888384552422058e-05,
+      "loss": 0.3485,
+      "step": 68270
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.9885408140486645e-05,
+      "loss": 0.3278,
+      "step": 68280
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.988243172855123e-05,
+      "loss": 0.3449,
+      "step": 68290
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.9879455316615822e-05,
+      "loss": 0.3338,
+      "step": 68300
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.987647890468041e-05,
+      "loss": 0.3201,
+      "step": 68310
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.9873502492745e-05,
+      "loss": 0.3481,
+      "step": 68320
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.9870526080809586e-05,
+      "loss": 0.336,
+      "step": 68330
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.9867549668874173e-05,
+      "loss": 0.3265,
+      "step": 68340
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.9864573256938763e-05,
+      "loss": 0.3434,
+      "step": 68350
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.986159684500335e-05,
+      "loss": 0.347,
+      "step": 68360
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.985862043306794e-05,
+      "loss": 0.3493,
+      "step": 68370
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.9855644021132527e-05,
+      "loss": 0.3328,
+      "step": 68380
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.9852667609197114e-05,
+      "loss": 0.333,
+      "step": 68390
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.98496911972617e-05,
+      "loss": 0.3435,
+      "step": 68400
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.984671478532629e-05,
+      "loss": 0.3371,
+      "step": 68410
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.9843738373390878e-05,
+      "loss": 0.3514,
+      "step": 68420
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.984076196145547e-05,
+      "loss": 0.3425,
+      "step": 68430
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.9837785549520055e-05,
+      "loss": 0.3276,
+      "step": 68440
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.9834809137584642e-05,
+      "loss": 0.3234,
+      "step": 68450
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.9831832725649233e-05,
+      "loss": 0.3331,
+      "step": 68460
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.982885631371382e-05,
+      "loss": 0.3346,
+      "step": 68470
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.982587990177841e-05,
+      "loss": 0.3431,
+      "step": 68480
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.9822903489842997e-05,
+      "loss": 0.345,
+      "step": 68490
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.9819927077907584e-05,
+      "loss": 0.3428,
+      "step": 68500
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.9816950665972174e-05,
+      "loss": 0.3434,
+      "step": 68510
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.981397425403676e-05,
+      "loss": 0.3495,
+      "step": 68520
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.981099784210135e-05,
+      "loss": 0.3321,
+      "step": 68530
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.9808021430165935e-05,
+      "loss": 0.3351,
+      "step": 68540
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.9805045018230525e-05,
+      "loss": 0.3429,
+      "step": 68550
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.9802068606295112e-05,
+      "loss": 0.3414,
+      "step": 68560
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.9799092194359702e-05,
+      "loss": 0.3461,
+      "step": 68570
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.979611578242429e-05,
+      "loss": 0.3343,
+      "step": 68580
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.9793139370488876e-05,
+      "loss": 0.3559,
+      "step": 68590
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.9790162958553466e-05,
+      "loss": 0.3341,
+      "step": 68600
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.9787186546618053e-05,
+      "loss": 0.3447,
+      "step": 68610
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.9784210134682643e-05,
+      "loss": 0.3537,
+      "step": 68620
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.978123372274723e-05,
+      "loss": 0.3539,
+      "step": 68630
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.9778257310811817e-05,
+      "loss": 0.3478,
+      "step": 68640
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.9775280898876404e-05,
+      "loss": 0.3359,
+      "step": 68650
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.9772304486940994e-05,
+      "loss": 0.3483,
+      "step": 68660
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.976932807500558e-05,
+      "loss": 0.366,
+      "step": 68670
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.976635166307017e-05,
+      "loss": 0.3516,
+      "step": 68680
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.976337525113476e-05,
+      "loss": 0.3433,
+      "step": 68690
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.9760398839199345e-05,
+      "loss": 0.3513,
+      "step": 68700
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.9757422427263936e-05,
+      "loss": 0.3453,
+      "step": 68710
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.9754446015328522e-05,
+      "loss": 0.3356,
+      "step": 68720
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.9751469603393113e-05,
+      "loss": 0.3295,
+      "step": 68730
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.97484931914577e-05,
+      "loss": 0.3294,
+      "step": 68740
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.9745516779522287e-05,
+      "loss": 0.3421,
+      "step": 68750
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.9742540367586877e-05,
+      "loss": 0.3427,
+      "step": 68760
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.9739563955651464e-05,
+      "loss": 0.3444,
+      "step": 68770
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.9736587543716054e-05,
+      "loss": 0.3537,
+      "step": 68780
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.973361113178064e-05,
+      "loss": 0.354,
+      "step": 68790
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.9730634719845228e-05,
+      "loss": 0.3441,
+      "step": 68800
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.9727658307909815e-05,
+      "loss": 0.3489,
+      "step": 68810
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.9724681895974405e-05,
+      "loss": 0.3386,
+      "step": 68820
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.9721705484038992e-05,
+      "loss": 0.331,
+      "step": 68830
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.9718729072103582e-05,
+      "loss": 0.3484,
+      "step": 68840
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.971575266016817e-05,
+      "loss": 0.353,
+      "step": 68850
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.9712776248232756e-05,
+      "loss": 0.3594,
+      "step": 68860
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.9709799836297346e-05,
+      "loss": 0.3459,
+      "step": 68870
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.9706823424361933e-05,
+      "loss": 0.3372,
+      "step": 68880
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.9703847012426523e-05,
+      "loss": 0.3419,
+      "step": 68890
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.970087060049111e-05,
+      "loss": 0.3396,
+      "step": 68900
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.9697894188555697e-05,
+      "loss": 0.3448,
+      "step": 68910
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.9694917776620288e-05,
+      "loss": 0.3342,
+      "step": 68920
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.9691941364684874e-05,
+      "loss": 0.3452,
+      "step": 68930
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.9688964952749465e-05,
+      "loss": 0.3466,
+      "step": 68940
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.9685988540814048e-05,
+      "loss": 0.3418,
+      "step": 68950
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.968301212887864e-05,
+      "loss": 0.3439,
+      "step": 68960
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.9680035716943225e-05,
+      "loss": 0.3359,
+      "step": 68970
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.9677059305007816e-05,
+      "loss": 0.3583,
+      "step": 68980
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.9674082893072403e-05,
+      "loss": 0.3504,
+      "step": 68990
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.967110648113699e-05,
+      "loss": 0.3356,
+      "step": 69000
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.966813006920158e-05,
+      "loss": 0.3361,
+      "step": 69010
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.9665153657266167e-05,
+      "loss": 0.3263,
+      "step": 69020
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.9662177245330757e-05,
+      "loss": 0.3469,
+      "step": 69030
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.9659200833395344e-05,
+      "loss": 0.3529,
+      "step": 69040
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.965622442145993e-05,
+      "loss": 0.3505,
+      "step": 69050
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.9653248009524518e-05,
+      "loss": 0.3488,
+      "step": 69060
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.9650271597589108e-05,
+      "loss": 0.3465,
+      "step": 69070
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.9647295185653695e-05,
+      "loss": 0.3452,
+      "step": 69080
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.9644318773718285e-05,
+      "loss": 0.3394,
+      "step": 69090
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.9641342361782872e-05,
+      "loss": 0.3309,
+      "step": 69100
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.963836594984746e-05,
+      "loss": 0.3484,
+      "step": 69110
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.963538953791205e-05,
+      "loss": 0.3366,
+      "step": 69120
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.9632413125976636e-05,
+      "loss": 0.3405,
+      "step": 69130
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.9629436714041226e-05,
+      "loss": 0.3455,
+      "step": 69140
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.9626460302105813e-05,
+      "loss": 0.3449,
+      "step": 69150
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.96234838901704e-05,
+      "loss": 0.3483,
+      "step": 69160
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.962050747823499e-05,
+      "loss": 0.3411,
+      "step": 69170
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.9617531066299577e-05,
+      "loss": 0.338,
+      "step": 69180
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.9614554654364168e-05,
+      "loss": 0.3546,
+      "step": 69190
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.9611578242428755e-05,
+      "loss": 0.3429,
+      "step": 69200
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.960860183049334e-05,
+      "loss": 0.3426,
+      "step": 69210
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.960562541855793e-05,
+      "loss": 0.3455,
+      "step": 69220
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.960264900662252e-05,
+      "loss": 0.3507,
+      "step": 69230
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.9599672594687106e-05,
+      "loss": 0.3367,
+      "step": 69240
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.9596696182751696e-05,
+      "loss": 0.3232,
+      "step": 69250
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.9593719770816283e-05,
+      "loss": 0.3406,
+      "step": 69260
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.959074335888087e-05,
+      "loss": 0.3538,
+      "step": 69270
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.958776694694546e-05,
+      "loss": 0.3344,
+      "step": 69280
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.9584790535010047e-05,
+      "loss": 0.3431,
+      "step": 69290
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.9581814123074637e-05,
+      "loss": 0.3262,
+      "step": 69300
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.9578837711139224e-05,
+      "loss": 0.3386,
+      "step": 69310
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.957586129920381e-05,
+      "loss": 0.3418,
+      "step": 69320
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.95728848872684e-05,
+      "loss": 0.338,
+      "step": 69330
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.9569908475332988e-05,
+      "loss": 0.3258,
+      "step": 69340
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.956693206339758e-05,
+      "loss": 0.3398,
+      "step": 69350
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.9563955651462162e-05,
+      "loss": 0.3391,
+      "step": 69360
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.9560979239526752e-05,
+      "loss": 0.3532,
+      "step": 69370
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.955800282759134e-05,
+      "loss": 0.34,
+      "step": 69380
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.955502641565593e-05,
+      "loss": 0.3362,
+      "step": 69390
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.9552050003720516e-05,
+      "loss": 0.3369,
+      "step": 69400
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.9549073591785103e-05,
+      "loss": 0.3413,
+      "step": 69410
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.9546097179849693e-05,
+      "loss": 0.358,
+      "step": 69420
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.954312076791428e-05,
+      "loss": 0.3312,
+      "step": 69430
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.954014435597887e-05,
+      "loss": 0.3393,
+      "step": 69440
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.9537167944043458e-05,
+      "loss": 0.3355,
+      "step": 69450
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.9534191532108044e-05,
+      "loss": 0.3591,
+      "step": 69460
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.953121512017263e-05,
+      "loss": 0.35,
+      "step": 69470
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.952823870823722e-05,
+      "loss": 0.3472,
+      "step": 69480
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.952526229630181e-05,
+      "loss": 0.335,
+      "step": 69490
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.95222858843664e-05,
+      "loss": 0.3503,
+      "step": 69500
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.9519309472430986e-05,
+      "loss": 0.3362,
+      "step": 69510
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.9516333060495573e-05,
+      "loss": 0.3295,
+      "step": 69520
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.9513356648560163e-05,
+      "loss": 0.3359,
+      "step": 69530
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.951038023662475e-05,
+      "loss": 0.3408,
+      "step": 69540
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.950740382468934e-05,
+      "loss": 0.3441,
+      "step": 69550
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.9504427412753927e-05,
+      "loss": 0.3248,
+      "step": 69560
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.9501451000818514e-05,
+      "loss": 0.3417,
+      "step": 69570
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.9498474588883104e-05,
+      "loss": 0.3285,
+      "step": 69580
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.949549817694769e-05,
+      "loss": 0.3326,
+      "step": 69590
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.949252176501228e-05,
+      "loss": 0.333,
+      "step": 69600
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.9489545353076865e-05,
+      "loss": 0.3431,
+      "step": 69610
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.9486568941141455e-05,
+      "loss": 0.3342,
+      "step": 69620
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.9483592529206042e-05,
+      "loss": 0.3425,
+      "step": 69630
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.9480616117270632e-05,
+      "loss": 0.3404,
+      "step": 69640
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.947763970533522e-05,
+      "loss": 0.339,
+      "step": 69650
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.947466329339981e-05,
+      "loss": 0.343,
+      "step": 69660
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.9471686881464396e-05,
+      "loss": 0.3338,
+      "step": 69670
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.9468710469528983e-05,
+      "loss": 0.3337,
+      "step": 69680
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.9465734057593574e-05,
+      "loss": 0.3404,
+      "step": 69690
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.946275764565816e-05,
+      "loss": 0.3337,
+      "step": 69700
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.945978123372275e-05,
+      "loss": 0.3391,
+      "step": 69710
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.9456804821787338e-05,
+      "loss": 0.3465,
+      "step": 69720
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.9453828409851925e-05,
+      "loss": 0.3399,
+      "step": 69730
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.9450851997916515e-05,
+      "loss": 0.3438,
+      "step": 69740
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.9447875585981102e-05,
+      "loss": 0.3354,
+      "step": 69750
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.9444899174045692e-05,
+      "loss": 0.3339,
+      "step": 69760
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.9441922762110276e-05,
+      "loss": 0.3476,
+      "step": 69770
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.9438946350174866e-05,
+      "loss": 0.332,
+      "step": 69780
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.9435969938239453e-05,
+      "loss": 0.3306,
+      "step": 69790
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.9432993526304043e-05,
+      "loss": 0.3435,
+      "step": 69800
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.943001711436863e-05,
+      "loss": 0.3406,
+      "step": 69810
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.9427040702433217e-05,
+      "loss": 0.3637,
+      "step": 69820
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.9424064290497807e-05,
+      "loss": 0.3408,
+      "step": 69830
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.9421087878562394e-05,
+      "loss": 0.3394,
+      "step": 69840
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.9418111466626984e-05,
+      "loss": 0.3431,
+      "step": 69850
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.941513505469157e-05,
+      "loss": 0.3429,
+      "step": 69860
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.9412158642756158e-05,
+      "loss": 0.3599,
+      "step": 69870
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.940918223082075e-05,
+      "loss": 0.3295,
+      "step": 69880
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.9406205818885335e-05,
+      "loss": 0.3379,
+      "step": 69890
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.9403229406949922e-05,
+      "loss": 0.3274,
+      "step": 69900
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.9400252995014512e-05,
+      "loss": 0.3466,
+      "step": 69910
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.93972765830791e-05,
+      "loss": 0.3405,
+      "step": 69920
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.9394300171143686e-05,
+      "loss": 0.3292,
+      "step": 69930
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.9391323759208277e-05,
+      "loss": 0.3577,
+      "step": 69940
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.9388347347272863e-05,
+      "loss": 0.3381,
+      "step": 69950
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.9385370935337454e-05,
+      "loss": 0.3432,
+      "step": 69960
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.938239452340204e-05,
+      "loss": 0.3458,
+      "step": 69970
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.9379418111466628e-05,
+      "loss": 0.3422,
+      "step": 69980
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.9376441699531218e-05,
+      "loss": 0.3351,
+      "step": 69990
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.9373465287595805e-05,
+      "loss": 0.3285,
+      "step": 70000
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.9370488875660395e-05,
+      "loss": 0.3309,
+      "step": 70010
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.936751246372498e-05,
+      "loss": 0.3411,
+      "step": 70020
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.936453605178957e-05,
+      "loss": 0.3302,
+      "step": 70030
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.9361559639854156e-05,
+      "loss": 0.3508,
+      "step": 70040
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.9358583227918746e-05,
+      "loss": 0.3394,
+      "step": 70050
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.9355606815983333e-05,
+      "loss": 0.3356,
+      "step": 70060
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.9352630404047923e-05,
+      "loss": 0.3473,
+      "step": 70070
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.934965399211251e-05,
+      "loss": 0.3343,
+      "step": 70080
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.9346677580177097e-05,
+      "loss": 0.3395,
+      "step": 70090
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.9343701168241687e-05,
+      "loss": 0.3294,
+      "step": 70100
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.9340724756306274e-05,
+      "loss": 0.3348,
+      "step": 70110
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.9337748344370864e-05,
+      "loss": 0.341,
+      "step": 70120
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.933477193243545e-05,
+      "loss": 0.3395,
+      "step": 70130
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.9331795520500038e-05,
+      "loss": 0.332,
+      "step": 70140
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.932881910856463e-05,
+      "loss": 0.3381,
+      "step": 70150
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.9325842696629215e-05,
+      "loss": 0.3328,
+      "step": 70160
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.9322866284693806e-05,
+      "loss": 0.3393,
+      "step": 70170
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.931988987275839e-05,
+      "loss": 0.3397,
+      "step": 70180
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.931691346082298e-05,
+      "loss": 0.3551,
+      "step": 70190
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.9313937048887566e-05,
+      "loss": 0.3389,
+      "step": 70200
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.9310960636952157e-05,
+      "loss": 0.3477,
+      "step": 70210
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.9307984225016744e-05,
+      "loss": 0.3427,
+      "step": 70220
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.930500781308133e-05,
+      "loss": 0.3484,
+      "step": 70230
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.930203140114592e-05,
+      "loss": 0.3553,
+      "step": 70240
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.9299054989210508e-05,
+      "loss": 0.3454,
+      "step": 70250
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.9296078577275098e-05,
+      "loss": 0.332,
+      "step": 70260
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.9293102165339685e-05,
+      "loss": 0.3272,
+      "step": 70270
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.9290125753404272e-05,
+      "loss": 0.3217,
+      "step": 70280
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.9287149341468862e-05,
+      "loss": 0.3414,
+      "step": 70290
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.928417292953345e-05,
+      "loss": 0.343,
+      "step": 70300
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.928119651759804e-05,
+      "loss": 0.3333,
+      "step": 70310
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.9278220105662626e-05,
+      "loss": 0.3414,
+      "step": 70320
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.9275243693727213e-05,
+      "loss": 0.3263,
+      "step": 70330
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.92722672817918e-05,
+      "loss": 0.3502,
+      "step": 70340
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.926929086985639e-05,
+      "loss": 0.3569,
+      "step": 70350
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.9266314457920977e-05,
+      "loss": 0.3364,
+      "step": 70360
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.9263338045985567e-05,
+      "loss": 0.3557,
+      "step": 70370
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.9260361634050154e-05,
+      "loss": 0.3265,
+      "step": 70380
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.925738522211474e-05,
+      "loss": 0.3475,
+      "step": 70390
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.925440881017933e-05,
+      "loss": 0.3382,
+      "step": 70400
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.925143239824392e-05,
+      "loss": 0.3404,
+      "step": 70410
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.924845598630851e-05,
+      "loss": 0.3353,
+      "step": 70420
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.9245479574373092e-05,
+      "loss": 0.3489,
+      "step": 70430
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.9242503162437682e-05,
+      "loss": 0.3373,
+      "step": 70440
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.923952675050227e-05,
+      "loss": 0.3358,
+      "step": 70450
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.923655033856686e-05,
+      "loss": 0.3555,
+      "step": 70460
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.9233573926631447e-05,
+      "loss": 0.3426,
+      "step": 70470
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.9230597514696033e-05,
+      "loss": 0.3351,
+      "step": 70480
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.9227621102760624e-05,
+      "loss": 0.3537,
+      "step": 70490
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.922464469082521e-05,
+      "loss": 0.3332,
+      "step": 70500
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.92216682788898e-05,
+      "loss": 0.3449,
+      "step": 70510
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.9218691866954388e-05,
+      "loss": 0.3462,
+      "step": 70520
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.9215715455018978e-05,
+      "loss": 0.341,
+      "step": 70530
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.9212739043083565e-05,
+      "loss": 0.3346,
+      "step": 70540
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.9209762631148152e-05,
+      "loss": 0.3554,
+      "step": 70550
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.9206786219212742e-05,
+      "loss": 0.3438,
+      "step": 70560
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.920380980727733e-05,
+      "loss": 0.3421,
+      "step": 70570
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.920083339534192e-05,
+      "loss": 0.332,
+      "step": 70580
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.9197856983406503e-05,
+      "loss": 0.343,
+      "step": 70590
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.9194880571471093e-05,
+      "loss": 0.3465,
+      "step": 70600
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.919190415953568e-05,
+      "loss": 0.3325,
+      "step": 70610
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.918892774760027e-05,
+      "loss": 0.3289,
+      "step": 70620
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.9185951335664857e-05,
+      "loss": 0.3361,
+      "step": 70630
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.9182974923729444e-05,
+      "loss": 0.3427,
+      "step": 70640
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.9179998511794034e-05,
+      "loss": 0.3544,
+      "step": 70650
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.917702209985862e-05,
+      "loss": 0.3521,
+      "step": 70660
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.917404568792321e-05,
+      "loss": 0.3364,
+      "step": 70670
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.91710692759878e-05,
+      "loss": 0.3381,
+      "step": 70680
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.9168092864052385e-05,
+      "loss": 0.3511,
+      "step": 70690
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.9165116452116976e-05,
+      "loss": 0.3389,
+      "step": 70700
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.9162140040181563e-05,
+      "loss": 0.3493,
+      "step": 70710
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.9159163628246153e-05,
+      "loss": 0.3565,
+      "step": 70720
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.915618721631074e-05,
+      "loss": 0.3375,
+      "step": 70730
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.9153210804375327e-05,
+      "loss": 0.3278,
+      "step": 70740
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.9150234392439914e-05,
+      "loss": 0.3457,
+      "step": 70750
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.9147257980504504e-05,
+      "loss": 0.3409,
+      "step": 70760
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.914428156856909e-05,
+      "loss": 0.3563,
+      "step": 70770
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.914130515663368e-05,
+      "loss": 0.3319,
+      "step": 70780
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.9138328744698268e-05,
+      "loss": 0.3448,
+      "step": 70790
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.9135352332762855e-05,
+      "loss": 0.3506,
+      "step": 70800
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.9132375920827445e-05,
+      "loss": 0.3488,
+      "step": 70810
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.9129399508892032e-05,
+      "loss": 0.325,
+      "step": 70820
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.9126423096956622e-05,
+      "loss": 0.349,
+      "step": 70830
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.9123446685021206e-05,
+      "loss": 0.3318,
+      "step": 70840
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.9120470273085796e-05,
+      "loss": 0.3305,
+      "step": 70850
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.9117493861150383e-05,
+      "loss": 0.3518,
+      "step": 70860
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.9114517449214973e-05,
+      "loss": 0.3304,
+      "step": 70870
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.911154103727956e-05,
+      "loss": 0.3537,
+      "step": 70880
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.9108564625344147e-05,
+      "loss": 0.3511,
+      "step": 70890
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.9105588213408737e-05,
+      "loss": 0.342,
+      "step": 70900
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.9102611801473324e-05,
+      "loss": 0.3319,
+      "step": 70910
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.9099635389537915e-05,
+      "loss": 0.3346,
+      "step": 70920
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.90966589776025e-05,
+      "loss": 0.3275,
+      "step": 70930
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.9093682565667092e-05,
+      "loss": 0.3349,
+      "step": 70940
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.909070615373168e-05,
+      "loss": 0.3324,
+      "step": 70950
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.9087729741796266e-05,
+      "loss": 0.3378,
+      "step": 70960
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.9084753329860856e-05,
+      "loss": 0.3475,
+      "step": 70970
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.9081776917925443e-05,
+      "loss": 0.3404,
+      "step": 70980
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.9078800505990033e-05,
+      "loss": 0.3366,
+      "step": 70990
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.9075824094054617e-05,
+      "loss": 0.333,
+      "step": 71000
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.9072847682119207e-05,
+      "loss": 0.3299,
+      "step": 71010
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.9069871270183794e-05,
+      "loss": 0.3435,
+      "step": 71020
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.9066894858248384e-05,
+      "loss": 0.3338,
+      "step": 71030
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.906391844631297e-05,
+      "loss": 0.3439,
+      "step": 71040
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.9060942034377558e-05,
+      "loss": 0.3289,
+      "step": 71050
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.9057965622442148e-05,
+      "loss": 0.3408,
+      "step": 71060
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.9054989210506735e-05,
+      "loss": 0.3568,
+      "step": 71070
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.9052012798571325e-05,
+      "loss": 0.3285,
+      "step": 71080
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.9049036386635912e-05,
+      "loss": 0.3465,
+      "step": 71090
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.90460599747005e-05,
+      "loss": 0.3172,
+      "step": 71100
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.904308356276509e-05,
+      "loss": 0.3533,
+      "step": 71110
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.9040107150829676e-05,
+      "loss": 0.344,
+      "step": 71120
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.9037130738894267e-05,
+      "loss": 0.3377,
+      "step": 71130
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.9034154326958854e-05,
+      "loss": 0.338,
+      "step": 71140
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.903117791502344e-05,
+      "loss": 0.3302,
+      "step": 71150
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.9028201503088027e-05,
+      "loss": 0.3484,
+      "step": 71160
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.9025225091152618e-05,
+      "loss": 0.3455,
+      "step": 71170
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.9022248679217204e-05,
+      "loss": 0.3418,
+      "step": 71180
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.9019272267281795e-05,
+      "loss": 0.3628,
+      "step": 71190
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.901629585534638e-05,
+      "loss": 0.3403,
+      "step": 71200
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.901331944341097e-05,
+      "loss": 0.3339,
+      "step": 71210
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.901034303147556e-05,
+      "loss": 0.3423,
+      "step": 71220
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.9007366619540146e-05,
+      "loss": 0.3558,
+      "step": 71230
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.9004390207604736e-05,
+      "loss": 0.3522,
+      "step": 71240
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.900141379566932e-05,
+      "loss": 0.337,
+      "step": 71250
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.899843738373391e-05,
+      "loss": 0.3393,
+      "step": 71260
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.8995460971798497e-05,
+      "loss": 0.3245,
+      "step": 71270
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.8992484559863087e-05,
+      "loss": 0.3549,
+      "step": 71280
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.8989508147927674e-05,
+      "loss": 0.3579,
+      "step": 71290
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.898653173599226e-05,
+      "loss": 0.3512,
+      "step": 71300
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.898355532405685e-05,
+      "loss": 0.3474,
+      "step": 71310
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.8980578912121438e-05,
+      "loss": 0.3423,
+      "step": 71320
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.8977602500186028e-05,
+      "loss": 0.3468,
+      "step": 71330
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.8974626088250615e-05,
+      "loss": 0.3334,
+      "step": 71340
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.8971649676315202e-05,
+      "loss": 0.3445,
+      "step": 71350
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.8968673264379792e-05,
+      "loss": 0.3416,
+      "step": 71360
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.896569685244438e-05,
+      "loss": 0.3448,
+      "step": 71370
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.896272044050897e-05,
+      "loss": 0.3432,
+      "step": 71380
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.8959744028573556e-05,
+      "loss": 0.3397,
+      "step": 71390
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.8956767616638147e-05,
+      "loss": 0.3394,
+      "step": 71400
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.895379120470273e-05,
+      "loss": 0.3432,
+      "step": 71410
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.895081479276732e-05,
+      "loss": 0.3313,
+      "step": 71420
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.8947838380831907e-05,
+      "loss": 0.3457,
+      "step": 71430
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.8944861968896498e-05,
+      "loss": 0.3337,
+      "step": 71440
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.8941885556961085e-05,
+      "loss": 0.3346,
+      "step": 71450
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.893890914502567e-05,
+      "loss": 0.3333,
+      "step": 71460
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.8935932733090262e-05,
+      "loss": 0.3388,
+      "step": 71470
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.893295632115485e-05,
+      "loss": 0.3634,
+      "step": 71480
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.892997990921944e-05,
+      "loss": 0.3585,
+      "step": 71490
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.8927003497284026e-05,
+      "loss": 0.3524,
+      "step": 71500
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.8924027085348613e-05,
+      "loss": 0.3542,
+      "step": 71510
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.8921050673413203e-05,
+      "loss": 0.3434,
+      "step": 71520
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.891807426147779e-05,
+      "loss": 0.3592,
+      "step": 71530
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.891509784954238e-05,
+      "loss": 0.3368,
+      "step": 71540
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.8912121437606967e-05,
+      "loss": 0.3478,
+      "step": 71550
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.8909145025671554e-05,
+      "loss": 0.3387,
+      "step": 71560
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.890616861373614e-05,
+      "loss": 0.3342,
+      "step": 71570
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.890319220180073e-05,
+      "loss": 0.3439,
+      "step": 71580
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.8900215789865318e-05,
+      "loss": 0.3489,
+      "step": 71590
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.889723937792991e-05,
+      "loss": 0.3381,
+      "step": 71600
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.8894262965994495e-05,
+      "loss": 0.3277,
+      "step": 71610
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.8891286554059082e-05,
+      "loss": 0.3339,
+      "step": 71620
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.8888310142123673e-05,
+      "loss": 0.3414,
+      "step": 71630
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.888533373018826e-05,
+      "loss": 0.3232,
+      "step": 71640
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.888235731825285e-05,
+      "loss": 0.3314,
+      "step": 71650
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.8879380906317433e-05,
+      "loss": 0.3593,
+      "step": 71660
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.8876404494382024e-05,
+      "loss": 0.3554,
+      "step": 71670
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.887342808244661e-05,
+      "loss": 0.3455,
+      "step": 71680
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.88704516705112e-05,
+      "loss": 0.3359,
+      "step": 71690
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.8867475258575788e-05,
+      "loss": 0.3342,
+      "step": 71700
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.8864498846640375e-05,
+      "loss": 0.35,
+      "step": 71710
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.8861522434704965e-05,
+      "loss": 0.3515,
+      "step": 71720
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.885854602276955e-05,
+      "loss": 0.3434,
+      "step": 71730
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.8855569610834142e-05,
+      "loss": 0.3323,
+      "step": 71740
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.885259319889873e-05,
+      "loss": 0.3499,
+      "step": 71750
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.8849616786963316e-05,
+      "loss": 0.3451,
+      "step": 71760
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.8846640375027906e-05,
+      "loss": 0.337,
+      "step": 71770
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.8843663963092493e-05,
+      "loss": 0.3372,
+      "step": 71780
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.8840687551157083e-05,
+      "loss": 0.3481,
+      "step": 71790
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.883771113922167e-05,
+      "loss": 0.3307,
+      "step": 71800
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.883473472728626e-05,
+      "loss": 0.349,
+      "step": 71810
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.8831758315350844e-05,
+      "loss": 0.346,
+      "step": 71820
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.8828781903415434e-05,
+      "loss": 0.3457,
+      "step": 71830
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.882580549148002e-05,
+      "loss": 0.3344,
+      "step": 71840
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.882282907954461e-05,
+      "loss": 0.3386,
+      "step": 71850
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.88198526676092e-05,
+      "loss": 0.3207,
+      "step": 71860
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.8816876255673785e-05,
+      "loss": 0.3384,
+      "step": 71870
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.8813899843738376e-05,
+      "loss": 0.3319,
+      "step": 71880
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.8810923431802962e-05,
+      "loss": 0.3391,
+      "step": 71890
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.8807947019867553e-05,
+      "loss": 0.3456,
+      "step": 71900
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.880497060793214e-05,
+      "loss": 0.3575,
+      "step": 71910
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.8801994195996726e-05,
+      "loss": 0.3541,
+      "step": 71920
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.8799017784061317e-05,
+      "loss": 0.3415,
+      "step": 71930
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.8796041372125904e-05,
+      "loss": 0.3344,
+      "step": 71940
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.8793064960190494e-05,
+      "loss": 0.3399,
+      "step": 71950
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.879008854825508e-05,
+      "loss": 0.34,
+      "step": 71960
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.8787112136319668e-05,
+      "loss": 0.3448,
+      "step": 71970
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.8784135724384255e-05,
+      "loss": 0.3393,
+      "step": 71980
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.8781159312448845e-05,
+      "loss": 0.3514,
+      "step": 71990
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.8778182900513432e-05,
+      "loss": 0.3465,
+      "step": 72000
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.8775206488578022e-05,
+      "loss": 0.3494,
+      "step": 72010
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.877223007664261e-05,
+      "loss": 0.3398,
+      "step": 72020
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.8769253664707196e-05,
+      "loss": 0.3422,
+      "step": 72030
+    },
+    {
+      "epoch": 16.0,
+      "eval_cer": 4.551282051282051,
+      "eval_loss": 0.4860053062438965,
+      "eval_runtime": 1.3046,
+      "eval_samples_per_second": 7.665,
+      "eval_steps_per_second": 0.767,
+      "eval_wer": 0.6363636363636364,
+      "step": 72032
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.8766277252771786e-05,
+      "loss": 0.3422,
+      "step": 72040
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.8763300840836373e-05,
+      "loss": 0.34,
+      "step": 72050
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.8760324428900963e-05,
+      "loss": 0.3389,
+      "step": 72060
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.875734801696555e-05,
+      "loss": 0.3199,
+      "step": 72070
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.8754371605030137e-05,
+      "loss": 0.322,
+      "step": 72080
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.8751395193094724e-05,
+      "loss": 0.3402,
+      "step": 72090
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.8748418781159314e-05,
+      "loss": 0.3332,
+      "step": 72100
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.87454423692239e-05,
+      "loss": 0.342,
+      "step": 72110
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.8742465957288488e-05,
+      "loss": 0.3176,
+      "step": 72120
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.873948954535308e-05,
+      "loss": 0.3325,
+      "step": 72130
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.8736513133417665e-05,
+      "loss": 0.3346,
+      "step": 72140
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.8733536721482256e-05,
+      "loss": 0.3183,
+      "step": 72150
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 1.8730560309546843e-05,
+      "loss": 0.3304,
+      "step": 72160
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 1.872758389761143e-05,
+      "loss": 0.3332,
+      "step": 72170
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 1.872460748567602e-05,
+      "loss": 0.327,
+      "step": 72180
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 1.8721631073740607e-05,
+      "loss": 0.3183,
+      "step": 72190
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 1.8718654661805197e-05,
+      "loss": 0.3168,
+      "step": 72200
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 1.8715678249869784e-05,
+      "loss": 0.3284,
+      "step": 72210
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 1.871270183793437e-05,
+      "loss": 0.3234,
+      "step": 72220
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 1.8709725425998958e-05,
+      "loss": 0.3287,
+      "step": 72230
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 1.8706749014063548e-05,
+      "loss": 0.3378,
+      "step": 72240
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 1.8703772602128135e-05,
+      "loss": 0.3288,
+      "step": 72250
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 1.8700796190192725e-05,
+      "loss": 0.3323,
+      "step": 72260
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 1.8697819778257312e-05,
+      "loss": 0.3145,
+      "step": 72270
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 1.86948433663219e-05,
+      "loss": 0.3346,
+      "step": 72280
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 1.869186695438649e-05,
+      "loss": 0.3325,
+      "step": 72290
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 1.8688890542451076e-05,
+      "loss": 0.3263,
+      "step": 72300
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 1.8685914130515666e-05,
+      "loss": 0.3254,
+      "step": 72310
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 1.8682937718580253e-05,
+      "loss": 0.3412,
+      "step": 72320
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 1.867996130664484e-05,
+      "loss": 0.3253,
+      "step": 72330
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 1.867698489470943e-05,
+      "loss": 0.3285,
+      "step": 72340
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 1.8674008482774017e-05,
+      "loss": 0.3444,
+      "step": 72350
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 1.8671032070838608e-05,
+      "loss": 0.3268,
+      "step": 72360
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 1.8668055658903195e-05,
+      "loss": 0.3147,
+      "step": 72370
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 1.866507924696778e-05,
+      "loss": 0.3215,
+      "step": 72380
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 1.866210283503237e-05,
+      "loss": 0.3209,
+      "step": 72390
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 1.865912642309696e-05,
+      "loss": 0.3384,
+      "step": 72400
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 1.8656150011161546e-05,
+      "loss": 0.3436,
+      "step": 72410
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 1.8653173599226136e-05,
+      "loss": 0.3392,
+      "step": 72420
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 1.8650197187290723e-05,
+      "loss": 0.3215,
+      "step": 72430
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 1.864722077535531e-05,
+      "loss": 0.3318,
+      "step": 72440
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 1.86442443634199e-05,
+      "loss": 0.3252,
+      "step": 72450
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 1.8641267951484487e-05,
+      "loss": 0.3347,
+      "step": 72460
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 1.8638291539549077e-05,
+      "loss": 0.3281,
+      "step": 72470
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.8635315127613664e-05,
+      "loss": 0.3318,
+      "step": 72480
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.863233871567825e-05,
+      "loss": 0.3273,
+      "step": 72490
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.862936230374284e-05,
+      "loss": 0.3314,
+      "step": 72500
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.8626385891807428e-05,
+      "loss": 0.331,
+      "step": 72510
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 1.8623409479872015e-05,
+      "loss": 0.3313,
+      "step": 72520
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 1.8620433067936602e-05,
+      "loss": 0.3207,
+      "step": 72530
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 1.8617456656001192e-05,
+      "loss": 0.3389,
+      "step": 72540
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 1.861448024406578e-05,
+      "loss": 0.3193,
+      "step": 72550
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 1.861150383213037e-05,
+      "loss": 0.3393,
+      "step": 72560
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.8608527420194956e-05,
+      "loss": 0.3221,
+      "step": 72570
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.8605551008259543e-05,
+      "loss": 0.319,
+      "step": 72580
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.8602574596324133e-05,
+      "loss": 0.3443,
+      "step": 72590
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.859959818438872e-05,
+      "loss": 0.341,
+      "step": 72600
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 1.859662177245331e-05,
+      "loss": 0.3357,
+      "step": 72610
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 1.8593645360517897e-05,
+      "loss": 0.341,
+      "step": 72620
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 1.8590668948582484e-05,
+      "loss": 0.3369,
+      "step": 72630
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 1.858769253664707e-05,
+      "loss": 0.3394,
+      "step": 72640
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 1.858471612471166e-05,
+      "loss": 0.3384,
+      "step": 72650
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.858173971277625e-05,
+      "loss": 0.3225,
+      "step": 72660
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.857876330084084e-05,
+      "loss": 0.3177,
+      "step": 72670
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.8575786888905426e-05,
+      "loss": 0.3356,
+      "step": 72680
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.8572810476970013e-05,
+      "loss": 0.3453,
+      "step": 72690
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 1.8569834065034603e-05,
+      "loss": 0.3343,
+      "step": 72700
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 1.856685765309919e-05,
+      "loss": 0.3173,
+      "step": 72710
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 1.856388124116378e-05,
+      "loss": 0.3236,
+      "step": 72720
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 1.8560904829228367e-05,
+      "loss": 0.3563,
+      "step": 72730
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 1.8557928417292954e-05,
+      "loss": 0.3334,
+      "step": 72740
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 1.8554952005357544e-05,
+      "loss": 0.3284,
+      "step": 72750
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 1.855197559342213e-05,
+      "loss": 0.3218,
+      "step": 72760
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 1.854899918148672e-05,
+      "loss": 0.3362,
+      "step": 72770
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 1.8546022769551308e-05,
+      "loss": 0.3364,
+      "step": 72780
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 1.8543046357615895e-05,
+      "loss": 0.3392,
+      "step": 72790
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 1.8540069945680482e-05,
+      "loss": 0.3384,
+      "step": 72800
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 1.8537093533745072e-05,
+      "loss": 0.3353,
+      "step": 72810
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 1.853411712180966e-05,
+      "loss": 0.3301,
+      "step": 72820
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 1.853114070987425e-05,
+      "loss": 0.3415,
+      "step": 72830
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 1.8528164297938836e-05,
+      "loss": 0.3262,
+      "step": 72840
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 1.8525187886003423e-05,
+      "loss": 0.3455,
+      "step": 72850
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 1.8522211474068014e-05,
+      "loss": 0.331,
+      "step": 72860
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 1.85192350621326e-05,
+      "loss": 0.3213,
+      "step": 72870
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 1.851625865019719e-05,
+      "loss": 0.3194,
+      "step": 72880
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 1.8513282238261778e-05,
+      "loss": 0.3248,
+      "step": 72890
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 1.8510305826326365e-05,
+      "loss": 0.3386,
+      "step": 72900
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 1.8507329414390955e-05,
+      "loss": 0.3386,
+      "step": 72910
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 1.8504353002455542e-05,
+      "loss": 0.3363,
+      "step": 72920
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 1.8501376590520132e-05,
+      "loss": 0.3336,
+      "step": 72930
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 1.8498400178584716e-05,
+      "loss": 0.3249,
+      "step": 72940
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 1.8495423766649306e-05,
+      "loss": 0.3279,
+      "step": 72950
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 1.8492447354713893e-05,
+      "loss": 0.3463,
+      "step": 72960
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 1.8489470942778483e-05,
+      "loss": 0.3253,
+      "step": 72970
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 1.848649453084307e-05,
+      "loss": 0.3318,
+      "step": 72980
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 1.8483518118907657e-05,
+      "loss": 0.3284,
+      "step": 72990
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 1.8480541706972247e-05,
+      "loss": 0.3387,
+      "step": 73000
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 1.8477565295036834e-05,
+      "loss": 0.3432,
+      "step": 73010
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.8474588883101424e-05,
+      "loss": 0.3286,
+      "step": 73020
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.847161247116601e-05,
+      "loss": 0.3347,
+      "step": 73030
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.8468636059230598e-05,
+      "loss": 0.3379,
+      "step": 73040
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.8465659647295185e-05,
+      "loss": 0.338,
+      "step": 73050
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 1.8462683235359775e-05,
+      "loss": 0.3448,
+      "step": 73060
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 1.8459706823424362e-05,
+      "loss": 0.3434,
+      "step": 73070
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 1.8456730411488952e-05,
+      "loss": 0.3427,
+      "step": 73080
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 1.845375399955354e-05,
+      "loss": 0.324,
+      "step": 73090
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 1.8450777587618126e-05,
+      "loss": 0.3374,
+      "step": 73100
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.8447801175682717e-05,
+      "loss": 0.3249,
+      "step": 73110
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.8444824763747303e-05,
+      "loss": 0.3307,
+      "step": 73120
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.8441848351811894e-05,
+      "loss": 0.329,
+      "step": 73130
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.843887193987648e-05,
+      "loss": 0.3202,
+      "step": 73140
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 1.8435895527941068e-05,
+      "loss": 0.3307,
+      "step": 73150
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 1.8432919116005658e-05,
+      "loss": 0.3265,
+      "step": 73160
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 1.8429942704070245e-05,
+      "loss": 0.3252,
+      "step": 73170
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 1.8426966292134835e-05,
+      "loss": 0.3276,
+      "step": 73180
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 1.8423989880199422e-05,
+      "loss": 0.3408,
+      "step": 73190
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.842101346826401e-05,
+      "loss": 0.3312,
+      "step": 73200
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.8418037056328596e-05,
+      "loss": 0.3246,
+      "step": 73210
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.8415060644393186e-05,
+      "loss": 0.3408,
+      "step": 73220
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.8412084232457773e-05,
+      "loss": 0.3305,
+      "step": 73230
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 1.8409107820522363e-05,
+      "loss": 0.3362,
+      "step": 73240
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 1.840613140858695e-05,
+      "loss": 0.3331,
+      "step": 73250
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 1.8403154996651537e-05,
+      "loss": 0.3455,
+      "step": 73260
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 1.8400178584716127e-05,
+      "loss": 0.3305,
+      "step": 73270
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 1.8397202172780714e-05,
+      "loss": 0.344,
+      "step": 73280
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 1.8394225760845304e-05,
+      "loss": 0.3212,
+      "step": 73290
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 1.839124934890989e-05,
+      "loss": 0.3347,
+      "step": 73300
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 1.8388272936974478e-05,
+      "loss": 0.3274,
+      "step": 73310
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 1.838529652503907e-05,
+      "loss": 0.3304,
+      "step": 73320
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 1.8382320113103655e-05,
+      "loss": 0.3254,
+      "step": 73330
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 1.8379343701168246e-05,
+      "loss": 0.3216,
+      "step": 73340
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 1.837636728923283e-05,
+      "loss": 0.3401,
+      "step": 73350
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 1.837339087729742e-05,
+      "loss": 0.3424,
+      "step": 73360
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 1.8370414465362006e-05,
+      "loss": 0.3151,
+      "step": 73370
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 1.8367438053426597e-05,
+      "loss": 0.3291,
+      "step": 73380
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 1.8364461641491184e-05,
+      "loss": 0.3293,
+      "step": 73390
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 1.836148522955577e-05,
+      "loss": 0.333,
+      "step": 73400
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 1.835850881762036e-05,
+      "loss": 0.3297,
+      "step": 73410
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 1.8355532405684948e-05,
+      "loss": 0.3278,
+      "step": 73420
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 1.8352555993749538e-05,
+      "loss": 0.3452,
+      "step": 73430
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 1.8349579581814125e-05,
+      "loss": 0.33,
+      "step": 73440
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 1.8346603169878712e-05,
+      "loss": 0.3272,
+      "step": 73450
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 1.83436267579433e-05,
+      "loss": 0.3414,
+      "step": 73460
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 1.834065034600789e-05,
+      "loss": 0.3361,
+      "step": 73470
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 1.8337673934072476e-05,
+      "loss": 0.3154,
+      "step": 73480
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 1.8334697522137066e-05,
+      "loss": 0.3276,
+      "step": 73490
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 1.8331721110201653e-05,
+      "loss": 0.3335,
+      "step": 73500
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 1.832874469826624e-05,
+      "loss": 0.3337,
+      "step": 73510
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 1.832576828633083e-05,
+      "loss": 0.3272,
+      "step": 73520
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 1.8322791874395417e-05,
+      "loss": 0.3556,
+      "step": 73530
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 1.8319815462460007e-05,
+      "loss": 0.3322,
+      "step": 73540
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 1.8316839050524594e-05,
+      "loss": 0.3391,
+      "step": 73550
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.831386263858918e-05,
+      "loss": 0.3154,
+      "step": 73560
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.831088622665377e-05,
+      "loss": 0.3318,
+      "step": 73570
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.830790981471836e-05,
+      "loss": 0.3399,
+      "step": 73580
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.830493340278295e-05,
+      "loss": 0.3399,
+      "step": 73590
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 1.8301956990847536e-05,
+      "loss": 0.35,
+      "step": 73600
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 1.8298980578912122e-05,
+      "loss": 0.3316,
+      "step": 73610
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 1.829600416697671e-05,
+      "loss": 0.3268,
+      "step": 73620
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 1.82930277550413e-05,
+      "loss": 0.3307,
+      "step": 73630
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 1.8290051343105887e-05,
+      "loss": 0.3382,
+      "step": 73640
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.8287074931170477e-05,
+      "loss": 0.3341,
+      "step": 73650
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.8284098519235064e-05,
+      "loss": 0.3254,
+      "step": 73660
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.828112210729965e-05,
+      "loss": 0.3255,
+      "step": 73670
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.827814569536424e-05,
+      "loss": 0.3361,
+      "step": 73680
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 1.8275169283428828e-05,
+      "loss": 0.3287,
+      "step": 73690
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 1.8272192871493418e-05,
+      "loss": 0.3328,
+      "step": 73700
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 1.8269216459558005e-05,
+      "loss": 0.3251,
+      "step": 73710
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 1.8266240047622592e-05,
+      "loss": 0.3409,
+      "step": 73720
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 1.8263263635687182e-05,
+      "loss": 0.3249,
+      "step": 73730
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.826028722375177e-05,
+      "loss": 0.3284,
+      "step": 73740
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.825731081181636e-05,
+      "loss": 0.3326,
+      "step": 73750
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.8254334399880943e-05,
+      "loss": 0.3208,
+      "step": 73760
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.8251357987945533e-05,
+      "loss": 0.3308,
+      "step": 73770
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 1.824838157601012e-05,
+      "loss": 0.3402,
+      "step": 73780
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 1.824540516407471e-05,
+      "loss": 0.3308,
+      "step": 73790
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 1.8242428752139297e-05,
+      "loss": 0.3373,
+      "step": 73800
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 1.8239452340203884e-05,
+      "loss": 0.3409,
+      "step": 73810
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 1.8236475928268474e-05,
+      "loss": 0.3256,
+      "step": 73820
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 1.823349951633306e-05,
+      "loss": 0.3297,
+      "step": 73830
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 1.823052310439765e-05,
+      "loss": 0.3324,
+      "step": 73840
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 1.822754669246224e-05,
+      "loss": 0.3361,
+      "step": 73850
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 1.8224570280526825e-05,
+      "loss": 0.3466,
+      "step": 73860
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 1.8221593868591412e-05,
+      "loss": 0.3366,
+      "step": 73870
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 1.8218617456656003e-05,
+      "loss": 0.3302,
+      "step": 73880
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 1.821564104472059e-05,
+      "loss": 0.3348,
+      "step": 73890
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 1.821266463278518e-05,
+      "loss": 0.3249,
+      "step": 73900
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 1.8209688220849767e-05,
+      "loss": 0.3267,
+      "step": 73910
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 1.8206711808914354e-05,
+      "loss": 0.3356,
+      "step": 73920
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 1.8203735396978944e-05,
+      "loss": 0.3293,
+      "step": 73930
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 1.820075898504353e-05,
+      "loss": 0.3168,
+      "step": 73940
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 1.819778257310812e-05,
+      "loss": 0.3515,
+      "step": 73950
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 1.8194806161172708e-05,
+      "loss": 0.3214,
+      "step": 73960
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 1.8191829749237295e-05,
+      "loss": 0.3304,
+      "step": 73970
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 1.8188853337301885e-05,
+      "loss": 0.3465,
+      "step": 73980
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 1.8185876925366472e-05,
+      "loss": 0.3377,
+      "step": 73990
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 1.8182900513431062e-05,
+      "loss": 0.3243,
+      "step": 74000
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 1.817992410149565e-05,
+      "loss": 0.3408,
+      "step": 74010
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 1.8176947689560236e-05,
+      "loss": 0.3357,
+      "step": 74020
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 1.8173971277624823e-05,
+      "loss": 0.3248,
+      "step": 74030
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 1.8170994865689413e-05,
+      "loss": 0.3471,
+      "step": 74040
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 1.8168018453754e-05,
+      "loss": 0.3452,
+      "step": 74050
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 1.816504204181859e-05,
+      "loss": 0.332,
+      "step": 74060
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 1.8162065629883177e-05,
+      "loss": 0.3327,
+      "step": 74070
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 1.8159089217947764e-05,
+      "loss": 0.3383,
+      "step": 74080
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 1.8156112806012355e-05,
+      "loss": 0.3328,
+      "step": 74090
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.815313639407694e-05,
+      "loss": 0.3414,
+      "step": 74100
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.8150159982141532e-05,
+      "loss": 0.346,
+      "step": 74110
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.814718357020612e-05,
+      "loss": 0.3208,
+      "step": 74120
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.8144207158270706e-05,
+      "loss": 0.3444,
+      "step": 74130
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 1.8141230746335296e-05,
+      "loss": 0.3211,
+      "step": 74140
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 1.8138254334399883e-05,
+      "loss": 0.331,
+      "step": 74150
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 1.8135277922464473e-05,
+      "loss": 0.3456,
+      "step": 74160
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 1.8132301510529057e-05,
+      "loss": 0.3308,
+      "step": 74170
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 1.8129325098593647e-05,
+      "loss": 0.3333,
+      "step": 74180
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.8126348686658234e-05,
+      "loss": 0.3251,
+      "step": 74190
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.8123372274722824e-05,
+      "loss": 0.3311,
+      "step": 74200
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.812039586278741e-05,
+      "loss": 0.3421,
+      "step": 74210
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.8117419450851998e-05,
+      "loss": 0.3565,
+      "step": 74220
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 1.8114443038916588e-05,
+      "loss": 0.3349,
+      "step": 74230
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 1.8111466626981175e-05,
+      "loss": 0.3275,
+      "step": 74240
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 1.8108490215045765e-05,
+      "loss": 0.329,
+      "step": 74250
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 1.8105513803110352e-05,
+      "loss": 0.3385,
+      "step": 74260
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 1.810253739117494e-05,
+      "loss": 0.3288,
+      "step": 74270
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.809956097923953e-05,
+      "loss": 0.3238,
+      "step": 74280
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.8096584567304116e-05,
+      "loss": 0.3204,
+      "step": 74290
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.8093608155368703e-05,
+      "loss": 0.3437,
+      "step": 74300
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.8090631743433293e-05,
+      "loss": 0.3305,
+      "step": 74310
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 1.808765533149788e-05,
+      "loss": 0.316,
+      "step": 74320
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 1.8084678919562467e-05,
+      "loss": 0.3328,
+      "step": 74330
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 1.8081702507627058e-05,
+      "loss": 0.3252,
+      "step": 74340
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 1.8078726095691644e-05,
+      "loss": 0.3346,
+      "step": 74350
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 1.8075749683756235e-05,
+      "loss": 0.3426,
+      "step": 74360
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 1.807277327182082e-05,
+      "loss": 0.3255,
+      "step": 74370
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 1.806979685988541e-05,
+      "loss": 0.3438,
+      "step": 74380
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 1.806682044795e-05,
+      "loss": 0.3281,
+      "step": 74390
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 1.8063844036014586e-05,
+      "loss": 0.3291,
+      "step": 74400
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 1.8060867624079176e-05,
+      "loss": 0.3252,
+      "step": 74410
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.8057891212143763e-05,
+      "loss": 0.3302,
+      "step": 74420
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.805491480020835e-05,
+      "loss": 0.3386,
+      "step": 74430
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.8051938388272937e-05,
+      "loss": 0.318,
+      "step": 74440
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 1.8048961976337527e-05,
+      "loss": 0.337,
+      "step": 74450
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 1.8045985564402114e-05,
+      "loss": 0.3363,
+      "step": 74460
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 1.8043009152466704e-05,
+      "loss": 0.3392,
+      "step": 74470
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 1.804003274053129e-05,
+      "loss": 0.3397,
+      "step": 74480
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 1.8037056328595878e-05,
+      "loss": 0.3284,
+      "step": 74490
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 1.8034079916660468e-05,
+      "loss": 0.3391,
+      "step": 74500
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 1.8031103504725055e-05,
+      "loss": 0.3169,
+      "step": 74510
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 1.8028127092789645e-05,
+      "loss": 0.335,
+      "step": 74520
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 1.8025150680854232e-05,
+      "loss": 0.3588,
+      "step": 74530
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 1.802217426891882e-05,
+      "loss": 0.322,
+      "step": 74540
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 1.801919785698341e-05,
+      "loss": 0.3223,
+      "step": 74550
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 1.8016221445047996e-05,
+      "loss": 0.3339,
+      "step": 74560
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 1.8013245033112587e-05,
+      "loss": 0.3373,
+      "step": 74570
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 1.801026862117717e-05,
+      "loss": 0.3435,
+      "step": 74580
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 1.800729220924176e-05,
+      "loss": 0.341,
+      "step": 74590
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 1.8004315797306347e-05,
+      "loss": 0.3377,
+      "step": 74600
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 1.8001339385370938e-05,
+      "loss": 0.3342,
+      "step": 74610
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 1.7998362973435525e-05,
+      "loss": 0.3426,
+      "step": 74620
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 1.799538656150011e-05,
+      "loss": 0.3309,
+      "step": 74630
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.7992410149564702e-05,
+      "loss": 0.331,
+      "step": 74640
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.798943373762929e-05,
+      "loss": 0.3276,
+      "step": 74650
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.798645732569388e-05,
+      "loss": 0.3318,
+      "step": 74660
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.7983480913758466e-05,
+      "loss": 0.3215,
+      "step": 74670
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 1.7980504501823053e-05,
+      "loss": 0.3272,
+      "step": 74680
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 1.7977528089887643e-05,
+      "loss": 0.3432,
+      "step": 74690
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 1.797455167795223e-05,
+      "loss": 0.3224,
+      "step": 74700
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 1.797157526601682e-05,
+      "loss": 0.3263,
+      "step": 74710
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 1.7968598854081407e-05,
+      "loss": 0.3261,
+      "step": 74720
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.7965622442145994e-05,
+      "loss": 0.3284,
+      "step": 74730
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.796264603021058e-05,
+      "loss": 0.3291,
+      "step": 74740
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.795966961827517e-05,
+      "loss": 0.3374,
+      "step": 74750
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.7956693206339758e-05,
+      "loss": 0.3293,
+      "step": 74760
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 1.795371679440435e-05,
+      "loss": 0.3253,
+      "step": 74770
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 1.7950740382468935e-05,
+      "loss": 0.3258,
+      "step": 74780
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 1.7947763970533522e-05,
+      "loss": 0.3257,
+      "step": 74790
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 1.7944787558598113e-05,
+      "loss": 0.3212,
+      "step": 74800
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 1.79418111466627e-05,
+      "loss": 0.3363,
+      "step": 74810
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.793883473472729e-05,
+      "loss": 0.3239,
+      "step": 74820
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.7935858322791873e-05,
+      "loss": 0.3383,
+      "step": 74830
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.7932881910856463e-05,
+      "loss": 0.3416,
+      "step": 74840
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.792990549892105e-05,
+      "loss": 0.3444,
+      "step": 74850
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 1.792692908698564e-05,
+      "loss": 0.3231,
+      "step": 74860
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 1.7923952675050228e-05,
+      "loss": 0.3272,
+      "step": 74870
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 1.7920976263114818e-05,
+      "loss": 0.3309,
+      "step": 74880
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 1.7917999851179405e-05,
+      "loss": 0.3306,
+      "step": 74890
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 1.791502343924399e-05,
+      "loss": 0.3291,
+      "step": 74900
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 1.7912047027308582e-05,
+      "loss": 0.3472,
+      "step": 74910
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 1.790907061537317e-05,
+      "loss": 0.3364,
+      "step": 74920
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 1.790609420343776e-05,
+      "loss": 0.3193,
+      "step": 74930
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 1.7903117791502346e-05,
+      "loss": 0.3222,
+      "step": 74940
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 1.7900141379566933e-05,
+      "loss": 0.3299,
+      "step": 74950
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 1.7897164967631523e-05,
+      "loss": 0.3347,
+      "step": 74960
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 1.789418855569611e-05,
+      "loss": 0.3346,
+      "step": 74970
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 1.78912121437607e-05,
+      "loss": 0.32,
+      "step": 74980
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 1.7888235731825284e-05,
+      "loss": 0.3201,
+      "step": 74990
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 1.7885259319889874e-05,
+      "loss": 0.3209,
+      "step": 75000
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 1.788228290795446e-05,
+      "loss": 0.3386,
+      "step": 75010
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 1.787930649601905e-05,
+      "loss": 0.321,
+      "step": 75020
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 1.7876330084083638e-05,
+      "loss": 0.3268,
+      "step": 75030
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 1.7873353672148225e-05,
+      "loss": 0.3351,
+      "step": 75040
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 1.7870377260212815e-05,
+      "loss": 0.3265,
+      "step": 75050
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 1.7867400848277402e-05,
+      "loss": 0.3338,
+      "step": 75060
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 1.7864424436341993e-05,
+      "loss": 0.3305,
+      "step": 75070
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 1.786144802440658e-05,
+      "loss": 0.3269,
+      "step": 75080
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 1.7858471612471166e-05,
+      "loss": 0.3503,
+      "step": 75090
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 1.7855495200535757e-05,
+      "loss": 0.3152,
+      "step": 75100
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 1.7852518788600344e-05,
+      "loss": 0.3138,
+      "step": 75110
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 1.7849542376664934e-05,
+      "loss": 0.3309,
+      "step": 75120
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 1.784656596472952e-05,
+      "loss": 0.3393,
+      "step": 75130
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 1.7843589552794108e-05,
+      "loss": 0.3309,
+      "step": 75140
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 1.7840613140858695e-05,
+      "loss": 0.3445,
+      "step": 75150
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 1.7837636728923285e-05,
+      "loss": 0.3243,
+      "step": 75160
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 1.7834660316987872e-05,
+      "loss": 0.312,
+      "step": 75170
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.7831683905052462e-05,
+      "loss": 0.328,
+      "step": 75180
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.782870749311705e-05,
+      "loss": 0.331,
+      "step": 75190
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.7825731081181636e-05,
+      "loss": 0.3278,
+      "step": 75200
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.7822754669246226e-05,
+      "loss": 0.3357,
+      "step": 75210
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 1.7819778257310813e-05,
+      "loss": 0.3305,
+      "step": 75220
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 1.7816801845375403e-05,
+      "loss": 0.3263,
+      "step": 75230
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 1.7813825433439987e-05,
+      "loss": 0.3321,
+      "step": 75240
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 1.7810849021504577e-05,
+      "loss": 0.3271,
+      "step": 75250
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 1.7807872609569164e-05,
+      "loss": 0.3287,
+      "step": 75260
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.7804896197633754e-05,
+      "loss": 0.3421,
+      "step": 75270
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.780191978569834e-05,
+      "loss": 0.3273,
+      "step": 75280
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.779894337376293e-05,
+      "loss": 0.3384,
+      "step": 75290
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.779596696182752e-05,
+      "loss": 0.3293,
+      "step": 75300
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 1.7792990549892105e-05,
+      "loss": 0.3183,
+      "step": 75310
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 1.7790014137956696e-05,
+      "loss": 0.3356,
+      "step": 75320
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 1.7787037726021283e-05,
+      "loss": 0.3365,
+      "step": 75330
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 1.7784061314085873e-05,
+      "loss": 0.3209,
+      "step": 75340
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 1.778108490215046e-05,
+      "loss": 0.3233,
+      "step": 75350
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.7778108490215047e-05,
+      "loss": 0.3263,
+      "step": 75360
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.7775132078279637e-05,
+      "loss": 0.3466,
+      "step": 75370
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.7772155666344224e-05,
+      "loss": 0.3317,
+      "step": 75380
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.7769179254408814e-05,
+      "loss": 0.3358,
+      "step": 75390
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 1.7766202842473398e-05,
+      "loss": 0.3294,
+      "step": 75400
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.7763226430537988e-05,
+      "loss": 0.3218,
+      "step": 75410
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.7760250018602575e-05,
+      "loss": 0.3239,
+      "step": 75420
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.7757273606667165e-05,
+      "loss": 0.3375,
+      "step": 75430
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 1.7754297194731752e-05,
+      "loss": 0.3268,
+      "step": 75440
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.775132078279634e-05,
+      "loss": 0.3147,
+      "step": 75450
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.774834437086093e-05,
+      "loss": 0.3454,
+      "step": 75460
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.7745367958925516e-05,
+      "loss": 0.3279,
+      "step": 75470
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.7742391546990106e-05,
+      "loss": 0.3236,
+      "step": 75480
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 1.7739415135054693e-05,
+      "loss": 0.3552,
+      "step": 75490
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.773643872311928e-05,
+      "loss": 0.3319,
+      "step": 75500
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.773346231118387e-05,
+      "loss": 0.3339,
+      "step": 75510
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.7730485899248457e-05,
+      "loss": 0.3385,
+      "step": 75520
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 1.7727509487313048e-05,
+      "loss": 0.3226,
+      "step": 75530
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.7724533075377634e-05,
+      "loss": 0.3332,
+      "step": 75540
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.772155666344222e-05,
+      "loss": 0.3271,
+      "step": 75550
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.7718580251506808e-05,
+      "loss": 0.3366,
+      "step": 75560
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.77156038395714e-05,
+      "loss": 0.3425,
+      "step": 75570
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 1.7712627427635985e-05,
+      "loss": 0.3215,
+      "step": 75580
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.7709651015700576e-05,
+      "loss": 0.3342,
+      "step": 75590
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.7706674603765163e-05,
+      "loss": 0.3358,
+      "step": 75600
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.770369819182975e-05,
+      "loss": 0.3316,
+      "step": 75610
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 1.770072177989434e-05,
+      "loss": 0.3375,
+      "step": 75620
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.7697745367958927e-05,
+      "loss": 0.333,
+      "step": 75630
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.7694768956023517e-05,
+      "loss": 0.3343,
+      "step": 75640
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.76917925440881e-05,
+      "loss": 0.3378,
+      "step": 75650
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.768881613215269e-05,
+      "loss": 0.315,
+      "step": 75660
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 1.7685839720217278e-05,
+      "loss": 0.3218,
+      "step": 75670
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.7682863308281868e-05,
+      "loss": 0.3247,
+      "step": 75680
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.7679886896346455e-05,
+      "loss": 0.3252,
+      "step": 75690
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.7676910484411042e-05,
+      "loss": 0.3445,
+      "step": 75700
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 1.7673934072475632e-05,
+      "loss": 0.3245,
+      "step": 75710
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.767095766054022e-05,
+      "loss": 0.3305,
+      "step": 75720
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.766798124860481e-05,
+      "loss": 0.3205,
+      "step": 75730
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.7665004836669396e-05,
+      "loss": 0.3371,
+      "step": 75740
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.7662028424733986e-05,
+      "loss": 0.3209,
+      "step": 75750
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 1.7659052012798573e-05,
+      "loss": 0.3214,
+      "step": 75760
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.765607560086316e-05,
+      "loss": 0.3301,
+      "step": 75770
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.765309918892775e-05,
+      "loss": 0.3282,
+      "step": 75780
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.7650122776992337e-05,
+      "loss": 0.3158,
+      "step": 75790
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 1.7647146365056928e-05,
+      "loss": 0.3349,
+      "step": 75800
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.764416995312151e-05,
+      "loss": 0.3213,
+      "step": 75810
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.76411935411861e-05,
+      "loss": 0.3306,
+      "step": 75820
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.763821712925069e-05,
+      "loss": 0.3388,
+      "step": 75830
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.763524071731528e-05,
+      "loss": 0.3322,
+      "step": 75840
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 1.7632264305379866e-05,
+      "loss": 0.3303,
+      "step": 75850
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.7629287893444453e-05,
+      "loss": 0.3392,
+      "step": 75860
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.7626311481509043e-05,
+      "loss": 0.3436,
+      "step": 75870
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.762333506957363e-05,
+      "loss": 0.3283,
+      "step": 75880
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 1.762035865763822e-05,
+      "loss": 0.3263,
+      "step": 75890
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.7617382245702807e-05,
+      "loss": 0.3332,
+      "step": 75900
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.7614405833767394e-05,
+      "loss": 0.3255,
+      "step": 75910
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.7611429421831984e-05,
+      "loss": 0.3273,
+      "step": 75920
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.760845300989657e-05,
+      "loss": 0.3356,
+      "step": 75930
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 1.760547659796116e-05,
+      "loss": 0.3366,
+      "step": 75940
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.7602500186025748e-05,
+      "loss": 0.3273,
+      "step": 75950
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.7599523774090335e-05,
+      "loss": 0.3238,
+      "step": 75960
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.7596547362154922e-05,
+      "loss": 0.335,
+      "step": 75970
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 1.7593570950219512e-05,
+      "loss": 0.3347,
+      "step": 75980
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.75905945382841e-05,
+      "loss": 0.33,
+      "step": 75990
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.758761812634869e-05,
+      "loss": 0.3273,
+      "step": 76000
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.7584641714413276e-05,
+      "loss": 0.3418,
+      "step": 76010
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.7581665302477863e-05,
+      "loss": 0.3351,
+      "step": 76020
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 1.7578688890542454e-05,
+      "loss": 0.3511,
+      "step": 76030
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.757571247860704e-05,
+      "loss": 0.3274,
+      "step": 76040
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.757273606667163e-05,
+      "loss": 0.3403,
+      "step": 76050
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.7569759654736218e-05,
+      "loss": 0.3283,
+      "step": 76060
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 1.7566783242800805e-05,
+      "loss": 0.3209,
+      "step": 76070
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.756380683086539e-05,
+      "loss": 0.3376,
+      "step": 76080
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.7560830418929982e-05,
+      "loss": 0.3399,
+      "step": 76090
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.755785400699457e-05,
+      "loss": 0.332,
+      "step": 76100
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.7554877595059155e-05,
+      "loss": 0.3278,
+      "step": 76110
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 1.7551901183123746e-05,
+      "loss": 0.3248,
+      "step": 76120
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.7548924771188333e-05,
+      "loss": 0.3386,
+      "step": 76130
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.7545948359252923e-05,
+      "loss": 0.3118,
+      "step": 76140
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.754297194731751e-05,
+      "loss": 0.3281,
+      "step": 76150
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 1.75399955353821e-05,
+      "loss": 0.3411,
+      "step": 76160
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.7537019123446687e-05,
+      "loss": 0.3394,
+      "step": 76170
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.7534042711511274e-05,
+      "loss": 0.3263,
+      "step": 76180
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.7531066299575864e-05,
+      "loss": 0.328,
+      "step": 76190
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.752808988764045e-05,
+      "loss": 0.3294,
+      "step": 76200
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 1.752511347570504e-05,
+      "loss": 0.3328,
+      "step": 76210
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.7522137063769625e-05,
+      "loss": 0.3441,
+      "step": 76220
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.7519160651834215e-05,
+      "loss": 0.3304,
+      "step": 76230
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.7516184239898802e-05,
+      "loss": 0.3168,
+      "step": 76240
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 1.7513207827963392e-05,
+      "loss": 0.3386,
+      "step": 76250
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.751023141602798e-05,
+      "loss": 0.3379,
+      "step": 76260
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.7507255004092566e-05,
+      "loss": 0.3403,
+      "step": 76270
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.7504278592157156e-05,
+      "loss": 0.3295,
+      "step": 76280
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.7501302180221743e-05,
+      "loss": 0.334,
+      "step": 76290
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 1.7498325768286334e-05,
+      "loss": 0.3345,
+      "step": 76300
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.749534935635092e-05,
+      "loss": 0.3452,
+      "step": 76310
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.7492372944415507e-05,
+      "loss": 0.3181,
+      "step": 76320
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.7489396532480098e-05,
+      "loss": 0.3303,
+      "step": 76330
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 1.7486420120544685e-05,
+      "loss": 0.3321,
+      "step": 76340
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.7483443708609275e-05,
+      "loss": 0.3303,
+      "step": 76350
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.7480467296673862e-05,
+      "loss": 0.333,
+      "step": 76360
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.747749088473845e-05,
+      "loss": 0.3181,
+      "step": 76370
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.7474514472803036e-05,
+      "loss": 0.3178,
+      "step": 76380
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 1.7471538060867626e-05,
+      "loss": 0.3245,
+      "step": 76390
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.7468561648932213e-05,
+      "loss": 0.3304,
+      "step": 76400
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.7465585236996803e-05,
+      "loss": 0.3246,
+      "step": 76410
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.746260882506139e-05,
+      "loss": 0.3198,
+      "step": 76420
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 1.7459632413125977e-05,
+      "loss": 0.3189,
+      "step": 76430
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.7456656001190567e-05,
+      "loss": 0.3261,
+      "step": 76440
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.7453679589255154e-05,
+      "loss": 0.3291,
+      "step": 76450
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.7450703177319744e-05,
+      "loss": 0.3316,
+      "step": 76460
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.744772676538433e-05,
+      "loss": 0.3393,
+      "step": 76470
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 1.7444750353448918e-05,
+      "loss": 0.3319,
+      "step": 76480
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.7441773941513505e-05,
+      "loss": 0.3238,
+      "step": 76490
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.7438797529578095e-05,
+      "loss": 0.3317,
+      "step": 76500
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.7435821117642682e-05,
+      "loss": 0.3503,
+      "step": 76510
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 1.743284470570727e-05,
+      "loss": 0.3313,
+      "step": 76520
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.742986829377186e-05,
+      "loss": 0.3332,
+      "step": 76530
+    },
+    {
+      "epoch": 17.0,
+      "eval_cer": 4.547008547008547,
+      "eval_loss": 0.43273496627807617,
+      "eval_runtime": 1.2705,
+      "eval_samples_per_second": 7.871,
+      "eval_steps_per_second": 0.787,
+      "eval_wer": 0.6493506493506493,
+      "step": 76534
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.7426891881836446e-05,
+      "loss": 0.331,
+      "step": 76540
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.7423915469901037e-05,
+      "loss": 0.3349,
+      "step": 76550
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.7420939057965624e-05,
+      "loss": 0.3103,
+      "step": 76560
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 1.741796264603021e-05,
+      "loss": 0.338,
+      "step": 76570
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.74149862340948e-05,
+      "loss": 0.3329,
+      "step": 76580
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.7412009822159388e-05,
+      "loss": 0.3315,
+      "step": 76590
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.7409033410223978e-05,
+      "loss": 0.3238,
+      "step": 76600
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 1.7406056998288565e-05,
+      "loss": 0.3187,
+      "step": 76610
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.7403080586353155e-05,
+      "loss": 0.3181,
+      "step": 76620
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.740010417441774e-05,
+      "loss": 0.3251,
+      "step": 76630
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.739712776248233e-05,
+      "loss": 0.3319,
+      "step": 76640
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.7394151350546916e-05,
+      "loss": 0.3282,
+      "step": 76650
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 1.7391174938611506e-05,
+      "loss": 0.3249,
+      "step": 76660
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.7388198526676093e-05,
+      "loss": 0.3211,
+      "step": 76670
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.738522211474068e-05,
+      "loss": 0.3141,
+      "step": 76680
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.738224570280527e-05,
+      "loss": 0.3222,
+      "step": 76690
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 1.7379269290869857e-05,
+      "loss": 0.3218,
+      "step": 76700
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.7376292878934447e-05,
+      "loss": 0.3153,
+      "step": 76710
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.7373316466999034e-05,
+      "loss": 0.3154,
+      "step": 76720
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.737034005506362e-05,
+      "loss": 0.3272,
+      "step": 76730
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.736736364312821e-05,
+      "loss": 0.3219,
+      "step": 76740
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 1.73643872311928e-05,
+      "loss": 0.3254,
+      "step": 76750
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.736141081925739e-05,
+      "loss": 0.3125,
+      "step": 76760
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.7358434407321976e-05,
+      "loss": 0.333,
+      "step": 76770
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.7355457995386562e-05,
+      "loss": 0.3238,
+      "step": 76780
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 1.735248158345115e-05,
+      "loss": 0.3326,
+      "step": 76790
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.734950517151574e-05,
+      "loss": 0.3287,
+      "step": 76800
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.7346528759580327e-05,
+      "loss": 0.3243,
+      "step": 76810
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.7343552347644917e-05,
+      "loss": 0.3159,
+      "step": 76820
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.7340575935709504e-05,
+      "loss": 0.3216,
+      "step": 76830
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 1.733759952377409e-05,
+      "loss": 0.3303,
+      "step": 76840
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.733462311183868e-05,
+      "loss": 0.31,
+      "step": 76850
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.7331646699903268e-05,
+      "loss": 0.3397,
+      "step": 76860
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.7328670287967858e-05,
+      "loss": 0.3308,
+      "step": 76870
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 1.7325693876032445e-05,
+      "loss": 0.3147,
+      "step": 76880
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.7322717464097032e-05,
+      "loss": 0.3292,
+      "step": 76890
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.7319741052161622e-05,
+      "loss": 0.3182,
+      "step": 76900
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.731676464022621e-05,
+      "loss": 0.3202,
+      "step": 76910
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.7313788228290796e-05,
+      "loss": 0.3211,
+      "step": 76920
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 1.7310811816355383e-05,
+      "loss": 0.3227,
+      "step": 76930
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.7307835404419973e-05,
+      "loss": 0.3085,
+      "step": 76940
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.730485899248456e-05,
+      "loss": 0.3249,
+      "step": 76950
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.730188258054915e-05,
+      "loss": 0.3302,
+      "step": 76960
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 1.7298906168613737e-05,
+      "loss": 0.3271,
+      "step": 76970
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.7295929756678324e-05,
+      "loss": 0.3205,
+      "step": 76980
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.7292953344742914e-05,
+      "loss": 0.3232,
+      "step": 76990
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.72899769328075e-05,
+      "loss": 0.3175,
+      "step": 77000
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.728700052087209e-05,
+      "loss": 0.3223,
+      "step": 77010
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 1.728402410893668e-05,
+      "loss": 0.3374,
+      "step": 77020
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.728104769700127e-05,
+      "loss": 0.3255,
+      "step": 77030
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.7278071285065852e-05,
+      "loss": 0.3369,
+      "step": 77040
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.7275094873130443e-05,
+      "loss": 0.3299,
+      "step": 77050
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 1.727211846119503e-05,
+      "loss": 0.3328,
+      "step": 77060
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.726914204925962e-05,
+      "loss": 0.3291,
+      "step": 77070
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.7266165637324207e-05,
+      "loss": 0.3281,
+      "step": 77080
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.7263189225388794e-05,
+      "loss": 0.3193,
+      "step": 77090
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.7260212813453384e-05,
+      "loss": 0.3301,
+      "step": 77100
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 1.725723640151797e-05,
+      "loss": 0.3165,
+      "step": 77110
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.725425998958256e-05,
+      "loss": 0.3175,
+      "step": 77120
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.7251283577647148e-05,
+      "loss": 0.323,
+      "step": 77130
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.7248307165711735e-05,
+      "loss": 0.3232,
+      "step": 77140
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 1.7245330753776325e-05,
+      "loss": 0.334,
+      "step": 77150
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.7242354341840912e-05,
+      "loss": 0.3272,
+      "step": 77160
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.7239377929905502e-05,
+      "loss": 0.3207,
+      "step": 77170
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.723640151797009e-05,
+      "loss": 0.3261,
+      "step": 77180
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.7233425106034676e-05,
+      "loss": 0.3192,
+      "step": 77190
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 1.7230448694099263e-05,
+      "loss": 0.3386,
+      "step": 77200
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.7227472282163853e-05,
+      "loss": 0.3238,
+      "step": 77210
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.722449587022844e-05,
+      "loss": 0.3159,
+      "step": 77220
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.722151945829303e-05,
+      "loss": 0.3217,
+      "step": 77230
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 1.7218543046357617e-05,
+      "loss": 0.3361,
+      "step": 77240
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.7215566634422204e-05,
+      "loss": 0.3333,
+      "step": 77250
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.7212590222486795e-05,
+      "loss": 0.3277,
+      "step": 77260
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.720961381055138e-05,
+      "loss": 0.3211,
+      "step": 77270
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.7206637398615972e-05,
+      "loss": 0.32,
+      "step": 77280
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 1.720366098668056e-05,
+      "loss": 0.3302,
+      "step": 77290
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.7200684574745146e-05,
+      "loss": 0.3235,
+      "step": 77300
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.7197708162809736e-05,
+      "loss": 0.325,
+      "step": 77310
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.7194731750874323e-05,
+      "loss": 0.3272,
+      "step": 77320
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 1.7191755338938913e-05,
+      "loss": 0.3285,
+      "step": 77330
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.7188778927003497e-05,
+      "loss": 0.3153,
+      "step": 77340
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.7185802515068087e-05,
+      "loss": 0.3183,
+      "step": 77350
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.7182826103132674e-05,
+      "loss": 0.3108,
+      "step": 77360
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.7179849691197264e-05,
+      "loss": 0.3311,
+      "step": 77370
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 1.717687327926185e-05,
+      "loss": 0.3368,
+      "step": 77380
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.7173896867326438e-05,
+      "loss": 0.3135,
+      "step": 77390
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.7170920455391028e-05,
+      "loss": 0.3256,
+      "step": 77400
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.7167944043455615e-05,
+      "loss": 0.3221,
+      "step": 77410
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 1.7164967631520205e-05,
+      "loss": 0.324,
+      "step": 77420
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.7161991219584792e-05,
+      "loss": 0.3296,
+      "step": 77430
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.715901480764938e-05,
+      "loss": 0.3275,
+      "step": 77440
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.7156038395713966e-05,
+      "loss": 0.3226,
+      "step": 77450
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.7153061983778556e-05,
+      "loss": 0.3261,
+      "step": 77460
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 1.7150085571843143e-05,
+      "loss": 0.3282,
+      "step": 77470
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.7147109159907733e-05,
+      "loss": 0.3242,
+      "step": 77480
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.714413274797232e-05,
+      "loss": 0.3105,
+      "step": 77490
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.7141156336036907e-05,
+      "loss": 0.3275,
+      "step": 77500
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 1.7138179924101498e-05,
+      "loss": 0.3194,
+      "step": 77510
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.7135203512166084e-05,
+      "loss": 0.305,
+      "step": 77520
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.7132227100230675e-05,
+      "loss": 0.3104,
+      "step": 77530
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.712925068829526e-05,
+      "loss": 0.3187,
+      "step": 77540
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.712627427635985e-05,
+      "loss": 0.332,
+      "step": 77550
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 1.712329786442444e-05,
+      "loss": 0.3271,
+      "step": 77560
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.7120321452489026e-05,
+      "loss": 0.322,
+      "step": 77570
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.7117345040553616e-05,
+      "loss": 0.3216,
+      "step": 77580
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.7114368628618203e-05,
+      "loss": 0.3079,
+      "step": 77590
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 1.711139221668279e-05,
+      "loss": 0.3416,
+      "step": 77600
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.7108415804747377e-05,
+      "loss": 0.3425,
+      "step": 77610
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.7105439392811967e-05,
+      "loss": 0.3192,
+      "step": 77620
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.7102462980876554e-05,
+      "loss": 0.3231,
+      "step": 77630
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.7099486568941144e-05,
+      "loss": 0.3285,
+      "step": 77640
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 1.709651015700573e-05,
+      "loss": 0.3248,
+      "step": 77650
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.7093533745070318e-05,
+      "loss": 0.3284,
+      "step": 77660
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.7090557333134908e-05,
+      "loss": 0.3292,
+      "step": 77670
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.7087580921199495e-05,
+      "loss": 0.3253,
+      "step": 77680
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 1.7084604509264085e-05,
+      "loss": 0.3134,
+      "step": 77690
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.7081628097328672e-05,
+      "loss": 0.3208,
+      "step": 77700
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.707865168539326e-05,
+      "loss": 0.3319,
+      "step": 77710
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.707567527345785e-05,
+      "loss": 0.3144,
+      "step": 77720
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.7072698861522436e-05,
+      "loss": 0.3082,
+      "step": 77730
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 1.7069722449587027e-05,
+      "loss": 0.3276,
+      "step": 77740
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.706674603765161e-05,
+      "loss": 0.3337,
+      "step": 77750
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.70637696257162e-05,
+      "loss": 0.3134,
+      "step": 77760
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.7060793213780787e-05,
+      "loss": 0.3241,
+      "step": 77770
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 1.7057816801845378e-05,
+      "loss": 0.3203,
+      "step": 77780
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.7054840389909965e-05,
+      "loss": 0.3128,
+      "step": 77790
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.705186397797455e-05,
+      "loss": 0.3242,
+      "step": 77800
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.7048887566039142e-05,
+      "loss": 0.3218,
+      "step": 77810
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.704591115410373e-05,
+      "loss": 0.3245,
+      "step": 77820
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 1.704293474216832e-05,
+      "loss": 0.3126,
+      "step": 77830
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.7039958330232906e-05,
+      "loss": 0.3169,
+      "step": 77840
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.7036981918297493e-05,
+      "loss": 0.3176,
+      "step": 77850
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.703400550636208e-05,
+      "loss": 0.3121,
+      "step": 77860
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 1.703102909442667e-05,
+      "loss": 0.3182,
+      "step": 77870
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.7028052682491257e-05,
+      "loss": 0.3229,
+      "step": 77880
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.7025076270555847e-05,
+      "loss": 0.3209,
+      "step": 77890
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.7022099858620434e-05,
+      "loss": 0.3112,
+      "step": 77900
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.701912344668502e-05,
+      "loss": 0.316,
+      "step": 77910
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 1.701614703474961e-05,
+      "loss": 0.332,
+      "step": 77920
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.7013170622814198e-05,
+      "loss": 0.3272,
+      "step": 77930
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.701019421087879e-05,
+      "loss": 0.3406,
+      "step": 77940
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.7007217798943375e-05,
+      "loss": 0.339,
+      "step": 77950
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 1.7004241387007962e-05,
+      "loss": 0.3155,
+      "step": 77960
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.7001264975072552e-05,
+      "loss": 0.3196,
+      "step": 77970
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.699828856313714e-05,
+      "loss": 0.3296,
+      "step": 77980
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.699531215120173e-05,
+      "loss": 0.3147,
+      "step": 77990
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.6992335739266317e-05,
+      "loss": 0.3274,
+      "step": 78000
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 1.6989359327330903e-05,
+      "loss": 0.3108,
+      "step": 78010
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.698638291539549e-05,
+      "loss": 0.325,
+      "step": 78020
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.698340650346008e-05,
+      "loss": 0.3165,
+      "step": 78030
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.6980430091524668e-05,
+      "loss": 0.3126,
+      "step": 78040
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 1.6977453679589258e-05,
+      "loss": 0.3118,
+      "step": 78050
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.6974477267653845e-05,
+      "loss": 0.3154,
+      "step": 78060
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.697150085571843e-05,
+      "loss": 0.3308,
+      "step": 78070
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.6968524443783022e-05,
+      "loss": 0.3291,
+      "step": 78080
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.696554803184761e-05,
+      "loss": 0.3114,
+      "step": 78090
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 1.69625716199122e-05,
+      "loss": 0.3207,
+      "step": 78100
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.6959595207976786e-05,
+      "loss": 0.3211,
+      "step": 78110
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.6956618796041373e-05,
+      "loss": 0.3233,
+      "step": 78120
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.6953642384105963e-05,
+      "loss": 0.3447,
+      "step": 78130
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 1.695066597217055e-05,
+      "loss": 0.3369,
+      "step": 78140
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.694768956023514e-05,
+      "loss": 0.3173,
+      "step": 78150
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.6944713148299724e-05,
+      "loss": 0.3384,
+      "step": 78160
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.6941736736364314e-05,
+      "loss": 0.3197,
+      "step": 78170
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.69387603244289e-05,
+      "loss": 0.3203,
+      "step": 78180
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 1.693578391249349e-05,
+      "loss": 0.3297,
+      "step": 78190
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.6932807500558078e-05,
+      "loss": 0.3124,
+      "step": 78200
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.6929831088622665e-05,
+      "loss": 0.3304,
+      "step": 78210
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.6926854676687255e-05,
+      "loss": 0.3343,
+      "step": 78220
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 1.6923878264751842e-05,
+      "loss": 0.3299,
+      "step": 78230
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.6920901852816433e-05,
+      "loss": 0.3239,
+      "step": 78240
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.691792544088102e-05,
+      "loss": 0.3356,
+      "step": 78250
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.6914949028945606e-05,
+      "loss": 0.3255,
+      "step": 78260
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.6911972617010193e-05,
+      "loss": 0.3163,
+      "step": 78270
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 1.6908996205074784e-05,
+      "loss": 0.3304,
+      "step": 78280
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.690601979313937e-05,
+      "loss": 0.3151,
+      "step": 78290
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.690304338120396e-05,
+      "loss": 0.3215,
+      "step": 78300
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.6900066969268548e-05,
+      "loss": 0.3256,
+      "step": 78310
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 1.6897090557333135e-05,
+      "loss": 0.3395,
+      "step": 78320
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.6894114145397725e-05,
+      "loss": 0.3345,
+      "step": 78330
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.6891137733462312e-05,
+      "loss": 0.3203,
+      "step": 78340
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.6888161321526902e-05,
+      "loss": 0.33,
+      "step": 78350
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.688518490959149e-05,
+      "loss": 0.3201,
+      "step": 78360
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 1.6882208497656076e-05,
+      "loss": 0.3334,
+      "step": 78370
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.6879232085720666e-05,
+      "loss": 0.324,
+      "step": 78380
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.6876255673785253e-05,
+      "loss": 0.3195,
+      "step": 78390
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.6873279261849843e-05,
+      "loss": 0.3282,
+      "step": 78400
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 1.687030284991443e-05,
+      "loss": 0.3053,
+      "step": 78410
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.6867326437979017e-05,
+      "loss": 0.3177,
+      "step": 78420
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.6864350026043604e-05,
+      "loss": 0.3285,
+      "step": 78430
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.6861373614108194e-05,
+      "loss": 0.329,
+      "step": 78440
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.685839720217278e-05,
+      "loss": 0.3122,
+      "step": 78450
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 1.685542079023737e-05,
+      "loss": 0.3242,
+      "step": 78460
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.685244437830196e-05,
+      "loss": 0.3263,
+      "step": 78470
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.6849467966366545e-05,
+      "loss": 0.314,
+      "step": 78480
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.6846491554431136e-05,
+      "loss": 0.3388,
+      "step": 78490
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 1.6843515142495722e-05,
+      "loss": 0.3194,
+      "step": 78500
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.6840538730560313e-05,
+      "loss": 0.3376,
+      "step": 78510
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.68375623186249e-05,
+      "loss": 0.3245,
+      "step": 78520
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.6834585906689487e-05,
+      "loss": 0.3162,
+      "step": 78530
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.6831609494754077e-05,
+      "loss": 0.313,
+      "step": 78540
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 1.6828633082818664e-05,
+      "loss": 0.319,
+      "step": 78550
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.6825656670883254e-05,
+      "loss": 0.3146,
+      "step": 78560
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.6822680258947838e-05,
+      "loss": 0.3134,
+      "step": 78570
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.6819703847012428e-05,
+      "loss": 0.3103,
+      "step": 78580
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 1.6816727435077015e-05,
+      "loss": 0.3191,
+      "step": 78590
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.6813751023141605e-05,
+      "loss": 0.3083,
+      "step": 78600
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.6810774611206192e-05,
+      "loss": 0.3205,
+      "step": 78610
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.680779819927078e-05,
+      "loss": 0.3029,
+      "step": 78620
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.680482178733537e-05,
+      "loss": 0.32,
+      "step": 78630
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 1.6801845375399956e-05,
+      "loss": 0.3249,
+      "step": 78640
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.6798868963464546e-05,
+      "loss": 0.3139,
+      "step": 78650
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.6795892551529133e-05,
+      "loss": 0.3322,
+      "step": 78660
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.679291613959372e-05,
+      "loss": 0.3234,
+      "step": 78670
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 1.678993972765831e-05,
+      "loss": 0.323,
+      "step": 78680
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.6786963315722897e-05,
+      "loss": 0.3157,
+      "step": 78690
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.6783986903787484e-05,
+      "loss": 0.3316,
+      "step": 78700
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.6781010491852074e-05,
+      "loss": 0.3299,
+      "step": 78710
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.677803407991666e-05,
+      "loss": 0.3142,
+      "step": 78720
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 1.6775057667981248e-05,
+      "loss": 0.3128,
+      "step": 78730
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.677208125604584e-05,
+      "loss": 0.3305,
+      "step": 78740
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.6769104844110425e-05,
+      "loss": 0.3397,
+      "step": 78750
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.6766128432175016e-05,
+      "loss": 0.3272,
+      "step": 78760
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 1.6763152020239603e-05,
+      "loss": 0.327,
+      "step": 78770
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.676017560830419e-05,
+      "loss": 0.3262,
+      "step": 78780
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.675719919636878e-05,
+      "loss": 0.3265,
+      "step": 78790
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.6754222784433367e-05,
+      "loss": 0.3184,
+      "step": 78800
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.6751246372497957e-05,
+      "loss": 0.3194,
+      "step": 78810
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 1.6748269960562544e-05,
+      "loss": 0.3288,
+      "step": 78820
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.674529354862713e-05,
+      "loss": 0.3156,
+      "step": 78830
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.6742317136691718e-05,
+      "loss": 0.3235,
+      "step": 78840
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.6739340724756308e-05,
+      "loss": 0.3048,
+      "step": 78850
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.6736364312820895e-05,
+      "loss": 0.3082,
+      "step": 78860
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 1.6733387900885485e-05,
+      "loss": 0.3365,
+      "step": 78870
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.6730411488950072e-05,
+      "loss": 0.3314,
+      "step": 78880
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.672743507701466e-05,
+      "loss": 0.3248,
+      "step": 78890
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.672445866507925e-05,
+      "loss": 0.3252,
+      "step": 78900
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 1.6721482253143836e-05,
+      "loss": 0.3227,
+      "step": 78910
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.6718505841208426e-05,
+      "loss": 0.3272,
+      "step": 78920
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.6715529429273013e-05,
+      "loss": 0.3115,
+      "step": 78930
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.67125530173376e-05,
+      "loss": 0.3195,
+      "step": 78940
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.670957660540219e-05,
+      "loss": 0.3328,
+      "step": 78950
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 1.6706600193466777e-05,
+      "loss": 0.3339,
+      "step": 78960
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.6703623781531368e-05,
+      "loss": 0.323,
+      "step": 78970
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.670064736959595e-05,
+      "loss": 0.3309,
+      "step": 78980
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.669767095766054e-05,
+      "loss": 0.3162,
+      "step": 78990
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 1.669469454572513e-05,
+      "loss": 0.3168,
+      "step": 79000
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.669171813378972e-05,
+      "loss": 0.3212,
+      "step": 79010
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.6688741721854306e-05,
+      "loss": 0.3278,
+      "step": 79020
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.6685765309918892e-05,
+      "loss": 0.311,
+      "step": 79030
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.6682788897983483e-05,
+      "loss": 0.3304,
+      "step": 79040
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 1.667981248604807e-05,
+      "loss": 0.3287,
+      "step": 79050
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.667683607411266e-05,
+      "loss": 0.3136,
+      "step": 79060
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.6673859662177247e-05,
+      "loss": 0.3161,
+      "step": 79070
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.6670883250241834e-05,
+      "loss": 0.321,
+      "step": 79080
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 1.6667906838306424e-05,
+      "loss": 0.3105,
+      "step": 79090
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.666493042637101e-05,
+      "loss": 0.3267,
+      "step": 79100
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.66619540144356e-05,
+      "loss": 0.3189,
+      "step": 79110
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.6658977602500188e-05,
+      "loss": 0.3235,
+      "step": 79120
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.6656001190564775e-05,
+      "loss": 0.3281,
+      "step": 79130
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 1.6653024778629362e-05,
+      "loss": 0.3228,
+      "step": 79140
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.6650048366693952e-05,
+      "loss": 0.3224,
+      "step": 79150
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.664707195475854e-05,
+      "loss": 0.3283,
+      "step": 79160
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.664409554282313e-05,
+      "loss": 0.3258,
+      "step": 79170
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 1.6641119130887716e-05,
+      "loss": 0.318,
+      "step": 79180
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.6638142718952303e-05,
+      "loss": 0.3204,
+      "step": 79190
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.6635166307016893e-05,
+      "loss": 0.3226,
+      "step": 79200
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.663218989508148e-05,
+      "loss": 0.3218,
+      "step": 79210
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.662921348314607e-05,
+      "loss": 0.3185,
+      "step": 79220
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 1.6626237071210658e-05,
+      "loss": 0.317,
+      "step": 79230
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.6623260659275244e-05,
+      "loss": 0.3262,
+      "step": 79240
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.662028424733983e-05,
+      "loss": 0.3154,
+      "step": 79250
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.661730783540442e-05,
+      "loss": 0.3207,
+      "step": 79260
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 1.661433142346901e-05,
+      "loss": 0.3258,
+      "step": 79270
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.66113550115336e-05,
+      "loss": 0.3496,
+      "step": 79280
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.6608378599598186e-05,
+      "loss": 0.3203,
+      "step": 79290
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.6605402187662773e-05,
+      "loss": 0.3271,
+      "step": 79300
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.6602425775727363e-05,
+      "loss": 0.3236,
+      "step": 79310
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 1.659944936379195e-05,
+      "loss": 0.3115,
+      "step": 79320
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.659647295185654e-05,
+      "loss": 0.3292,
+      "step": 79330
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.6593496539921127e-05,
+      "loss": 0.3234,
+      "step": 79340
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.6590520127985714e-05,
+      "loss": 0.3223,
+      "step": 79350
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 1.6587543716050304e-05,
+      "loss": 0.3214,
+      "step": 79360
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.658456730411489e-05,
+      "loss": 0.3292,
+      "step": 79370
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.658159089217948e-05,
+      "loss": 0.3291,
+      "step": 79380
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.6578614480244065e-05,
+      "loss": 0.3192,
+      "step": 79390
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.6575638068308655e-05,
+      "loss": 0.3313,
+      "step": 79400
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 1.6572661656373242e-05,
+      "loss": 0.3072,
+      "step": 79410
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.6569685244437832e-05,
+      "loss": 0.3153,
+      "step": 79420
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.656670883250242e-05,
+      "loss": 0.3278,
+      "step": 79430
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.6563732420567006e-05,
+      "loss": 0.3317,
+      "step": 79440
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 1.6560756008631596e-05,
+      "loss": 0.319,
+      "step": 79450
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.6557779596696183e-05,
+      "loss": 0.3203,
+      "step": 79460
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.6554803184760774e-05,
+      "loss": 0.3185,
+      "step": 79470
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.655182677282536e-05,
+      "loss": 0.3206,
+      "step": 79480
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.6548850360889947e-05,
+      "loss": 0.3292,
+      "step": 79490
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 1.6545873948954538e-05,
+      "loss": 0.3097,
+      "step": 79500
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.6542897537019125e-05,
+      "loss": 0.3332,
+      "step": 79510
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.6539921125083715e-05,
+      "loss": 0.3316,
+      "step": 79520
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.6536944713148302e-05,
+      "loss": 0.3204,
+      "step": 79530
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 1.653396830121289e-05,
+      "loss": 0.3355,
+      "step": 79540
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.6530991889277476e-05,
+      "loss": 0.299,
+      "step": 79550
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.6528015477342066e-05,
+      "loss": 0.3229,
+      "step": 79560
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.6525039065406653e-05,
+      "loss": 0.3303,
+      "step": 79570
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.6522062653471243e-05,
+      "loss": 0.3232,
+      "step": 79580
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 1.651908624153583e-05,
+      "loss": 0.319,
+      "step": 79590
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.6516109829600417e-05,
+      "loss": 0.3306,
+      "step": 79600
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.6513133417665007e-05,
+      "loss": 0.3287,
+      "step": 79610
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.6510157005729594e-05,
+      "loss": 0.3268,
+      "step": 79620
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 1.6507180593794184e-05,
+      "loss": 0.3181,
+      "step": 79630
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.650420418185877e-05,
+      "loss": 0.3252,
+      "step": 79640
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.6501227769923358e-05,
+      "loss": 0.3352,
+      "step": 79650
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.6498251357987945e-05,
+      "loss": 0.3193,
+      "step": 79660
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.6495274946052535e-05,
+      "loss": 0.3204,
+      "step": 79670
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 1.6492298534117122e-05,
+      "loss": 0.322,
+      "step": 79680
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.6489322122181713e-05,
+      "loss": 0.3126,
+      "step": 79690
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.64863457102463e-05,
+      "loss": 0.317,
+      "step": 79700
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.6483369298310886e-05,
+      "loss": 0.3306,
+      "step": 79710
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 1.6480392886375477e-05,
+      "loss": 0.307,
+      "step": 79720
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.6477416474440064e-05,
+      "loss": 0.3138,
+      "step": 79730
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.6474440062504654e-05,
+      "loss": 0.3328,
+      "step": 79740
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.647146365056924e-05,
+      "loss": 0.3321,
+      "step": 79750
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.6468487238633828e-05,
+      "loss": 0.3172,
+      "step": 79760
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 1.6465510826698418e-05,
+      "loss": 0.3271,
+      "step": 79770
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.6462534414763005e-05,
+      "loss": 0.3181,
+      "step": 79780
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.6459558002827595e-05,
+      "loss": 0.3454,
+      "step": 79790
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.645658159089218e-05,
+      "loss": 0.3199,
+      "step": 79800
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 1.645360517895677e-05,
+      "loss": 0.3183,
+      "step": 79810
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.6450628767021356e-05,
+      "loss": 0.328,
+      "step": 79820
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.6447652355085946e-05,
+      "loss": 0.3174,
+      "step": 79830
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.6444675943150533e-05,
+      "loss": 0.31,
+      "step": 79840
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.644169953121512e-05,
+      "loss": 0.3216,
+      "step": 79850
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 1.643872311927971e-05,
+      "loss": 0.317,
+      "step": 79860
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.6435746707344297e-05,
+      "loss": 0.3305,
+      "step": 79870
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.6432770295408887e-05,
+      "loss": 0.3428,
+      "step": 79880
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.6429793883473474e-05,
+      "loss": 0.3235,
+      "step": 79890
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 1.642681747153806e-05,
+      "loss": 0.3218,
+      "step": 79900
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.642384105960265e-05,
+      "loss": 0.313,
+      "step": 79910
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.6420864647667238e-05,
+      "loss": 0.314,
+      "step": 79920
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.641788823573183e-05,
+      "loss": 0.318,
+      "step": 79930
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.6414911823796415e-05,
+      "loss": 0.3239,
+      "step": 79940
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 1.6411935411861002e-05,
+      "loss": 0.3082,
+      "step": 79950
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.640895899992559e-05,
+      "loss": 0.3225,
+      "step": 79960
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.640598258799018e-05,
+      "loss": 0.3221,
+      "step": 79970
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.6403006176054766e-05,
+      "loss": 0.3266,
+      "step": 79980
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 1.6400029764119357e-05,
+      "loss": 0.3257,
+      "step": 79990
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.6397053352183944e-05,
+      "loss": 0.3245,
+      "step": 80000
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.639407694024853e-05,
+      "loss": 0.3193,
+      "step": 80010
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.639110052831312e-05,
+      "loss": 0.3374,
+      "step": 80020
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.6388124116377708e-05,
+      "loss": 0.3195,
+      "step": 80030
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 1.6385147704442298e-05,
+      "loss": 0.3184,
+      "step": 80040
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.638217129250688e-05,
+      "loss": 0.3336,
+      "step": 80050
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.6379194880571472e-05,
+      "loss": 0.3234,
+      "step": 80060
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.637621846863606e-05,
+      "loss": 0.2995,
+      "step": 80070
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 1.637324205670065e-05,
+      "loss": 0.3219,
+      "step": 80080
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.6370265644765236e-05,
+      "loss": 0.3222,
+      "step": 80090
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.6367289232829826e-05,
+      "loss": 0.3203,
+      "step": 80100
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.6364312820894413e-05,
+      "loss": 0.3299,
+      "step": 80110
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.6361336408959e-05,
+      "loss": 0.3245,
+      "step": 80120
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 1.635835999702359e-05,
+      "loss": 0.3338,
+      "step": 80130
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.6355383585088177e-05,
+      "loss": 0.3242,
+      "step": 80140
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.6352407173152767e-05,
+      "loss": 0.3125,
+      "step": 80150
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.6349430761217354e-05,
+      "loss": 0.3295,
+      "step": 80160
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 1.634645434928194e-05,
+      "loss": 0.3297,
+      "step": 80170
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.634347793734653e-05,
+      "loss": 0.3222,
+      "step": 80180
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.634050152541112e-05,
+      "loss": 0.323,
+      "step": 80190
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.633752511347571e-05,
+      "loss": 0.3381,
+      "step": 80200
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.6334548701540292e-05,
+      "loss": 0.3199,
+      "step": 80210
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 1.6331572289604883e-05,
+      "loss": 0.3198,
+      "step": 80220
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.632859587766947e-05,
+      "loss": 0.3285,
+      "step": 80230
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.632561946573406e-05,
+      "loss": 0.322,
+      "step": 80240
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.6322643053798647e-05,
+      "loss": 0.3229,
+      "step": 80250
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 1.6319666641863234e-05,
+      "loss": 0.3198,
+      "step": 80260
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.6316690229927824e-05,
+      "loss": 0.3145,
+      "step": 80270
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.631371381799241e-05,
+      "loss": 0.331,
+      "step": 80280
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.6310737406057e-05,
+      "loss": 0.3097,
+      "step": 80290
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.6307760994121588e-05,
+      "loss": 0.3087,
+      "step": 80300
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 1.6304784582186175e-05,
+      "loss": 0.3293,
+      "step": 80310
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.6301808170250765e-05,
+      "loss": 0.3166,
+      "step": 80320
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.6298831758315352e-05,
+      "loss": 0.3249,
+      "step": 80330
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.6295855346379942e-05,
+      "loss": 0.3153,
+      "step": 80340
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 1.629287893444453e-05,
+      "loss": 0.3088,
+      "step": 80350
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.6289902522509116e-05,
+      "loss": 0.3384,
+      "step": 80360
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.6286926110573703e-05,
+      "loss": 0.3149,
+      "step": 80370
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.6283949698638293e-05,
+      "loss": 0.3192,
+      "step": 80380
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.628097328670288e-05,
+      "loss": 0.3216,
+      "step": 80390
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 1.627799687476747e-05,
+      "loss": 0.3181,
+      "step": 80400
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.6275020462832057e-05,
+      "loss": 0.3188,
+      "step": 80410
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.6272044050896644e-05,
+      "loss": 0.3297,
+      "step": 80420
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.6269067638961235e-05,
+      "loss": 0.3243,
+      "step": 80430
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 1.626609122702582e-05,
+      "loss": 0.3262,
+      "step": 80440
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.6263114815090412e-05,
+      "loss": 0.3216,
+      "step": 80450
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.6260138403155e-05,
+      "loss": 0.3221,
+      "step": 80460
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.6257161991219586e-05,
+      "loss": 0.322,
+      "step": 80470
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.6254185579284172e-05,
+      "loss": 0.3309,
+      "step": 80480
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 1.6251209167348763e-05,
+      "loss": 0.3079,
+      "step": 80490
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.624823275541335e-05,
+      "loss": 0.3287,
+      "step": 80500
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.624525634347794e-05,
+      "loss": 0.3169,
+      "step": 80510
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.6242279931542527e-05,
+      "loss": 0.323,
+      "step": 80520
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 1.6239303519607114e-05,
+      "loss": 0.312,
+      "step": 80530
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.6236327107671704e-05,
+      "loss": 0.3149,
+      "step": 80540
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.623335069573629e-05,
+      "loss": 0.3236,
+      "step": 80550
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.623037428380088e-05,
+      "loss": 0.3244,
+      "step": 80560
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.6227397871865468e-05,
+      "loss": 0.3073,
+      "step": 80570
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 1.6224421459930055e-05,
+      "loss": 0.323,
+      "step": 80580
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.6221445047994645e-05,
+      "loss": 0.3243,
+      "step": 80590
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.6218468636059232e-05,
+      "loss": 0.3192,
+      "step": 80600
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.6215492224123822e-05,
+      "loss": 0.3219,
+      "step": 80610
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 1.6212515812188406e-05,
+      "loss": 0.3392,
+      "step": 80620
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.6209539400252996e-05,
+      "loss": 0.3314,
+      "step": 80630
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.6206562988317583e-05,
+      "loss": 0.3279,
+      "step": 80640
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.6203586576382173e-05,
+      "loss": 0.3325,
+      "step": 80650
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.620061016444676e-05,
+      "loss": 0.3126,
+      "step": 80660
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 1.6197633752511347e-05,
+      "loss": 0.3122,
+      "step": 80670
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.6194657340575937e-05,
+      "loss": 0.322,
+      "step": 80680
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.6191680928640524e-05,
+      "loss": 0.3293,
+      "step": 80690
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.6188704516705115e-05,
+      "loss": 0.3328,
+      "step": 80700
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 1.61857281047697e-05,
+      "loss": 0.312,
+      "step": 80710
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.618275169283429e-05,
+      "loss": 0.3091,
+      "step": 80720
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.617977528089888e-05,
+      "loss": 0.3192,
+      "step": 80730
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.6176798868963466e-05,
+      "loss": 0.3215,
+      "step": 80740
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.6173822457028056e-05,
+      "loss": 0.3239,
+      "step": 80750
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 1.6170846045092643e-05,
+      "loss": 0.3231,
+      "step": 80760
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.616786963315723e-05,
+      "loss": 0.3296,
+      "step": 80770
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.6164893221221817e-05,
+      "loss": 0.3365,
+      "step": 80780
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.6161916809286407e-05,
+      "loss": 0.3267,
+      "step": 80790
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 1.6158940397350994e-05,
+      "loss": 0.3233,
+      "step": 80800
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.6155963985415584e-05,
+      "loss": 0.3114,
+      "step": 80810
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.615298757348017e-05,
+      "loss": 0.3321,
+      "step": 80820
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.6150011161544758e-05,
+      "loss": 0.3226,
+      "step": 80830
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.6147034749609348e-05,
+      "loss": 0.3255,
+      "step": 80840
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.6144058337673935e-05,
+      "loss": 0.32,
+      "step": 80850
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.6141081925738525e-05,
+      "loss": 0.3179,
+      "step": 80860
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.6138105513803112e-05,
+      "loss": 0.3142,
+      "step": 80870
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.61351291018677e-05,
+      "loss": 0.3139,
+      "step": 80880
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.613215268993229e-05,
+      "loss": 0.3276,
+      "step": 80890
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.6129176277996876e-05,
+      "loss": 0.3401,
+      "step": 80900
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.6126199866061463e-05,
+      "loss": 0.3282,
+      "step": 80910
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.612322345412605e-05,
+      "loss": 0.3223,
+      "step": 80920
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.612024704219064e-05,
+      "loss": 0.3246,
+      "step": 80930
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.6117270630255227e-05,
+      "loss": 0.318,
+      "step": 80940
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.6114294218319818e-05,
+      "loss": 0.3182,
+      "step": 80950
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.6111317806384405e-05,
+      "loss": 0.3183,
+      "step": 80960
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.6108341394448995e-05,
+      "loss": 0.3378,
+      "step": 80970
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.6105364982513582e-05,
+      "loss": 0.3146,
+      "step": 80980
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.610238857057817e-05,
+      "loss": 0.3161,
+      "step": 80990
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.609941215864276e-05,
+      "loss": 0.3336,
+      "step": 81000
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.6096435746707346e-05,
+      "loss": 0.3172,
+      "step": 81010
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.6093459334771936e-05,
+      "loss": 0.325,
+      "step": 81020
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.609048292283652e-05,
+      "loss": 0.3087,
+      "step": 81030
+    },
+    {
+      "epoch": 18.0,
+      "eval_cer": 4.551282051282051,
+      "eval_loss": 0.4698427617549896,
+      "eval_runtime": 1.2573,
+      "eval_samples_per_second": 7.953,
+      "eval_steps_per_second": 0.795,
+      "eval_wer": 0.6493506493506493,
+      "step": 81036
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.608750651090111e-05,
+      "loss": 0.3172,
+      "step": 81040
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.6084530098965697e-05,
+      "loss": 0.308,
+      "step": 81050
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.6081553687030287e-05,
+      "loss": 0.3417,
+      "step": 81060
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.6078577275094874e-05,
+      "loss": 0.3213,
+      "step": 81070
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.607560086315946e-05,
+      "loss": 0.3191,
+      "step": 81080
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.607262445122405e-05,
+      "loss": 0.3138,
+      "step": 81090
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.6069648039288638e-05,
+      "loss": 0.3136,
+      "step": 81100
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.606667162735323e-05,
+      "loss": 0.31,
+      "step": 81110
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.6063695215417815e-05,
+      "loss": 0.3238,
+      "step": 81120
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.6060718803482402e-05,
+      "loss": 0.331,
+      "step": 81130
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.6057742391546992e-05,
+      "loss": 0.3045,
+      "step": 81140
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.605476597961158e-05,
+      "loss": 0.3159,
+      "step": 81150
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.605178956767617e-05,
+      "loss": 0.309,
+      "step": 81160
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.6048813155740757e-05,
+      "loss": 0.3139,
+      "step": 81170
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.6045836743805343e-05,
+      "loss": 0.3063,
+      "step": 81180
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.604286033186993e-05,
+      "loss": 0.3162,
+      "step": 81190
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.603988391993452e-05,
+      "loss": 0.3122,
+      "step": 81200
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.6036907507999107e-05,
+      "loss": 0.3015,
+      "step": 81210
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.6033931096063698e-05,
+      "loss": 0.3283,
+      "step": 81220
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.6030954684128285e-05,
+      "loss": 0.32,
+      "step": 81230
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.602797827219287e-05,
+      "loss": 0.3176,
+      "step": 81240
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.6025001860257462e-05,
+      "loss": 0.3155,
+      "step": 81250
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.602202544832205e-05,
+      "loss": 0.3161,
+      "step": 81260
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.601904903638664e-05,
+      "loss": 0.3331,
+      "step": 81270
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.6016072624451226e-05,
+      "loss": 0.319,
+      "step": 81280
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.6013096212515813e-05,
+      "loss": 0.3146,
+      "step": 81290
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.6010119800580403e-05,
+      "loss": 0.2892,
+      "step": 81300
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.600714338864499e-05,
+      "loss": 0.3178,
+      "step": 81310
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.600416697670958e-05,
+      "loss": 0.3211,
+      "step": 81320
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.6001190564774164e-05,
+      "loss": 0.323,
+      "step": 81330
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.5998214152838754e-05,
+      "loss": 0.3188,
+      "step": 81340
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.599523774090334e-05,
+      "loss": 0.3115,
+      "step": 81350
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.599226132896793e-05,
+      "loss": 0.3197,
+      "step": 81360
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.5989284917032518e-05,
+      "loss": 0.3127,
+      "step": 81370
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.598630850509711e-05,
+      "loss": 0.3155,
+      "step": 81380
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.5983332093161695e-05,
+      "loss": 0.3198,
+      "step": 81390
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.5980355681226282e-05,
+      "loss": 0.3009,
+      "step": 81400
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.5977379269290873e-05,
+      "loss": 0.3211,
+      "step": 81410
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.597440285735546e-05,
+      "loss": 0.3304,
+      "step": 81420
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.597142644542005e-05,
+      "loss": 0.3231,
+      "step": 81430
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.5968450033484633e-05,
+      "loss": 0.3407,
+      "step": 81440
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.5965473621549224e-05,
+      "loss": 0.3296,
+      "step": 81450
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.596249720961381e-05,
+      "loss": 0.3072,
+      "step": 81460
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.59595207976784e-05,
+      "loss": 0.3119,
+      "step": 81470
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.5956544385742988e-05,
+      "loss": 0.3152,
+      "step": 81480
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.5953567973807575e-05,
+      "loss": 0.3065,
+      "step": 81490
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.5950591561872165e-05,
+      "loss": 0.3067,
+      "step": 81500
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.5947615149936752e-05,
+      "loss": 0.3173,
+      "step": 81510
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.5944638738001342e-05,
+      "loss": 0.321,
+      "step": 81520
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.594166232606593e-05,
+      "loss": 0.3192,
+      "step": 81530
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.5938685914130516e-05,
+      "loss": 0.3175,
+      "step": 81540
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.5935709502195106e-05,
+      "loss": 0.3058,
+      "step": 81550
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.5932733090259693e-05,
+      "loss": 0.311,
+      "step": 81560
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.5929756678324283e-05,
+      "loss": 0.3232,
+      "step": 81570
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.592678026638887e-05,
+      "loss": 0.3092,
+      "step": 81580
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.5923803854453457e-05,
+      "loss": 0.3123,
+      "step": 81590
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.5920827442518044e-05,
+      "loss": 0.3094,
+      "step": 81600
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.5917851030582634e-05,
+      "loss": 0.3116,
+      "step": 81610
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.591487461864722e-05,
+      "loss": 0.314,
+      "step": 81620
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.591189820671181e-05,
+      "loss": 0.3148,
+      "step": 81630
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.59089217947764e-05,
+      "loss": 0.307,
+      "step": 81640
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.5905945382840985e-05,
+      "loss": 0.3029,
+      "step": 81650
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.5902968970905576e-05,
+      "loss": 0.3088,
+      "step": 81660
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.5899992558970162e-05,
+      "loss": 0.315,
+      "step": 81670
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.5897016147034753e-05,
+      "loss": 0.3181,
+      "step": 81680
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.589403973509934e-05,
+      "loss": 0.3098,
+      "step": 81690
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.5891063323163927e-05,
+      "loss": 0.3268,
+      "step": 81700
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.5888086911228517e-05,
+      "loss": 0.3377,
+      "step": 81710
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.5885110499293104e-05,
+      "loss": 0.3077,
+      "step": 81720
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.5882134087357694e-05,
+      "loss": 0.3131,
+      "step": 81730
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.5879157675422278e-05,
+      "loss": 0.3275,
+      "step": 81740
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.5876181263486868e-05,
+      "loss": 0.3072,
+      "step": 81750
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.5873204851551455e-05,
+      "loss": 0.3175,
+      "step": 81760
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.5870228439616045e-05,
+      "loss": 0.3025,
+      "step": 81770
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.5867252027680632e-05,
+      "loss": 0.3219,
+      "step": 81780
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.586427561574522e-05,
+      "loss": 0.3144,
+      "step": 81790
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.586129920380981e-05,
+      "loss": 0.3297,
+      "step": 81800
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.5858322791874396e-05,
+      "loss": 0.3266,
+      "step": 81810
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.5855346379938986e-05,
+      "loss": 0.3139,
+      "step": 81820
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.5852369968003573e-05,
+      "loss": 0.3147,
+      "step": 81830
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.5849393556068163e-05,
+      "loss": 0.3185,
+      "step": 81840
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5846417144132747e-05,
+      "loss": 0.3075,
+      "step": 81850
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5843440732197337e-05,
+      "loss": 0.3245,
+      "step": 81860
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5840464320261924e-05,
+      "loss": 0.3141,
+      "step": 81870
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5837487908326514e-05,
+      "loss": 0.309,
+      "step": 81880
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.58345114963911e-05,
+      "loss": 0.3063,
+      "step": 81890
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.5831535084455688e-05,
+      "loss": 0.3005,
+      "step": 81900
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.582855867252028e-05,
+      "loss": 0.3104,
+      "step": 81910
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.5825582260584865e-05,
+      "loss": 0.3311,
+      "step": 81920
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.5822605848649456e-05,
+      "loss": 0.3267,
+      "step": 81930
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.5819629436714043e-05,
+      "loss": 0.3324,
+      "step": 81940
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.581665302477863e-05,
+      "loss": 0.3127,
+      "step": 81950
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.581367661284322e-05,
+      "loss": 0.3172,
+      "step": 81960
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.5810700200907807e-05,
+      "loss": 0.2998,
+      "step": 81970
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5807723788972397e-05,
+      "loss": 0.31,
+      "step": 81980
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5804747377036984e-05,
+      "loss": 0.3174,
+      "step": 81990
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.580177096510157e-05,
+      "loss": 0.3085,
+      "step": 82000
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5798794553166158e-05,
+      "loss": 0.3223,
+      "step": 82010
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5795818141230748e-05,
+      "loss": 0.3147,
+      "step": 82020
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.5792841729295335e-05,
+      "loss": 0.3179,
+      "step": 82030
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.5789865317359925e-05,
+      "loss": 0.3197,
+      "step": 82040
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.5786888905424512e-05,
+      "loss": 0.3237,
+      "step": 82050
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.57839124934891e-05,
+      "loss": 0.3145,
+      "step": 82060
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.578093608155369e-05,
+      "loss": 0.3326,
+      "step": 82070
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.5777959669618276e-05,
+      "loss": 0.3097,
+      "step": 82080
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.5774983257682866e-05,
+      "loss": 0.3081,
+      "step": 82090
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.5772006845747453e-05,
+      "loss": 0.3184,
+      "step": 82100
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.576903043381204e-05,
+      "loss": 0.3174,
+      "step": 82110
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.576605402187663e-05,
+      "loss": 0.3166,
+      "step": 82120
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.5763077609941217e-05,
+      "loss": 0.3088,
+      "step": 82130
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.5760101198005808e-05,
+      "loss": 0.3105,
+      "step": 82140
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.575712478607039e-05,
+      "loss": 0.3074,
+      "step": 82150
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.575414837413498e-05,
+      "loss": 0.3168,
+      "step": 82160
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.575117196219957e-05,
+      "loss": 0.3069,
+      "step": 82170
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.574819555026416e-05,
+      "loss": 0.3183,
+      "step": 82180
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.5745219138328746e-05,
+      "loss": 0.316,
+      "step": 82190
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.5742242726393332e-05,
+      "loss": 0.3305,
+      "step": 82200
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.5739266314457923e-05,
+      "loss": 0.3192,
+      "step": 82210
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.573628990252251e-05,
+      "loss": 0.3204,
+      "step": 82220
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.57333134905871e-05,
+      "loss": 0.3108,
+      "step": 82230
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.5730337078651687e-05,
+      "loss": 0.3194,
+      "step": 82240
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.5727360666716277e-05,
+      "loss": 0.3144,
+      "step": 82250
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.572438425478086e-05,
+      "loss": 0.3212,
+      "step": 82260
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.572140784284545e-05,
+      "loss": 0.3192,
+      "step": 82270
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.5718431430910038e-05,
+      "loss": 0.3036,
+      "step": 82280
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.5715455018974628e-05,
+      "loss": 0.3267,
+      "step": 82290
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.5712478607039215e-05,
+      "loss": 0.3194,
+      "step": 82300
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.5709502195103802e-05,
+      "loss": 0.3,
+      "step": 82310
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.5706525783168392e-05,
+      "loss": 0.3299,
+      "step": 82320
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.570354937123298e-05,
+      "loss": 0.3091,
+      "step": 82330
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.570057295929757e-05,
+      "loss": 0.3239,
+      "step": 82340
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.5697596547362156e-05,
+      "loss": 0.3097,
+      "step": 82350
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.5694620135426743e-05,
+      "loss": 0.3049,
+      "step": 82360
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.5691643723491333e-05,
+      "loss": 0.3181,
+      "step": 82370
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.568866731155592e-05,
+      "loss": 0.3166,
+      "step": 82380
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.568569089962051e-05,
+      "loss": 0.3007,
+      "step": 82390
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.5682714487685098e-05,
+      "loss": 0.3104,
+      "step": 82400
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.5679738075749684e-05,
+      "loss": 0.3056,
+      "step": 82410
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.567676166381427e-05,
+      "loss": 0.3306,
+      "step": 82420
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.567378525187886e-05,
+      "loss": 0.319,
+      "step": 82430
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.567080883994345e-05,
+      "loss": 0.3203,
+      "step": 82440
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.566783242800804e-05,
+      "loss": 0.3267,
+      "step": 82450
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.5664856016072626e-05,
+      "loss": 0.3279,
+      "step": 82460
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.5661879604137213e-05,
+      "loss": 0.3241,
+      "step": 82470
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.5658903192201803e-05,
+      "loss": 0.3216,
+      "step": 82480
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.565592678026639e-05,
+      "loss": 0.3084,
+      "step": 82490
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.565295036833098e-05,
+      "loss": 0.3,
+      "step": 82500
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.5649973956395567e-05,
+      "loss": 0.3071,
+      "step": 82510
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.5646997544460154e-05,
+      "loss": 0.2902,
+      "step": 82520
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.5644021132524744e-05,
+      "loss": 0.3268,
+      "step": 82530
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.564104472058933e-05,
+      "loss": 0.3078,
+      "step": 82540
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.563806830865392e-05,
+      "loss": 0.308,
+      "step": 82550
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.5635091896718505e-05,
+      "loss": 0.2997,
+      "step": 82560
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.5632115484783095e-05,
+      "loss": 0.3221,
+      "step": 82570
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.5629139072847682e-05,
+      "loss": 0.307,
+      "step": 82580
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.5626162660912272e-05,
+      "loss": 0.3042,
+      "step": 82590
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.562318624897686e-05,
+      "loss": 0.3076,
+      "step": 82600
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.5620209837041446e-05,
+      "loss": 0.3254,
+      "step": 82610
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.5617233425106036e-05,
+      "loss": 0.3178,
+      "step": 82620
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.5614257013170623e-05,
+      "loss": 0.3056,
+      "step": 82630
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.5611280601235214e-05,
+      "loss": 0.3055,
+      "step": 82640
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.56083041892998e-05,
+      "loss": 0.3022,
+      "step": 82650
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.5605327777364387e-05,
+      "loss": 0.3173,
+      "step": 82660
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.5602351365428974e-05,
+      "loss": 0.3223,
+      "step": 82670
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.5599374953493565e-05,
+      "loss": 0.3214,
+      "step": 82680
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.559639854155815e-05,
+      "loss": 0.3062,
+      "step": 82690
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.5593422129622742e-05,
+      "loss": 0.3322,
+      "step": 82700
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.559044571768733e-05,
+      "loss": 0.3202,
+      "step": 82710
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.5587469305751916e-05,
+      "loss": 0.3098,
+      "step": 82720
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.5584492893816506e-05,
+      "loss": 0.3071,
+      "step": 82730
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.5581516481881093e-05,
+      "loss": 0.3154,
+      "step": 82740
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.5578540069945683e-05,
+      "loss": 0.3144,
+      "step": 82750
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.557556365801027e-05,
+      "loss": 0.3205,
+      "step": 82760
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.5572587246074857e-05,
+      "loss": 0.3144,
+      "step": 82770
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.5569610834139447e-05,
+      "loss": 0.3283,
+      "step": 82780
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.5566634422204034e-05,
+      "loss": 0.3198,
+      "step": 82790
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.5563658010268624e-05,
+      "loss": 0.3096,
+      "step": 82800
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.556068159833321e-05,
+      "loss": 0.3144,
+      "step": 82810
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.5557705186397798e-05,
+      "loss": 0.3121,
+      "step": 82820
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.5554728774462385e-05,
+      "loss": 0.3056,
+      "step": 82830
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.5551752362526975e-05,
+      "loss": 0.2886,
+      "step": 82840
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.5548775950591562e-05,
+      "loss": 0.3222,
+      "step": 82850
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.5545799538656152e-05,
+      "loss": 0.3176,
+      "step": 82860
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.554282312672074e-05,
+      "loss": 0.3234,
+      "step": 82870
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.5539846714785326e-05,
+      "loss": 0.3022,
+      "step": 82880
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.5536870302849917e-05,
+      "loss": 0.3186,
+      "step": 82890
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.5533893890914503e-05,
+      "loss": 0.3276,
+      "step": 82900
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.5530917478979094e-05,
+      "loss": 0.3093,
+      "step": 82910
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.552794106704368e-05,
+      "loss": 0.3118,
+      "step": 82920
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.5524964655108268e-05,
+      "loss": 0.3148,
+      "step": 82930
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.5521988243172858e-05,
+      "loss": 0.3122,
+      "step": 82940
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.5519011831237445e-05,
+      "loss": 0.3168,
+      "step": 82950
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.5516035419302035e-05,
+      "loss": 0.3195,
+      "step": 82960
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.551305900736662e-05,
+      "loss": 0.3163,
+      "step": 82970
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.551008259543121e-05,
+      "loss": 0.2995,
+      "step": 82980
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.5507106183495796e-05,
+      "loss": 0.3138,
+      "step": 82990
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.5504129771560386e-05,
+      "loss": 0.3096,
+      "step": 83000
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.5501153359624973e-05,
+      "loss": 0.313,
+      "step": 83010
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.549817694768956e-05,
+      "loss": 0.3072,
+      "step": 83020
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.549520053575415e-05,
+      "loss": 0.3198,
+      "step": 83030
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.5492224123818737e-05,
+      "loss": 0.3119,
+      "step": 83040
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.5489247711883327e-05,
+      "loss": 0.3127,
+      "step": 83050
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.5486271299947914e-05,
+      "loss": 0.3259,
+      "step": 83060
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.54832948880125e-05,
+      "loss": 0.3074,
+      "step": 83070
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.548031847607709e-05,
+      "loss": 0.2944,
+      "step": 83080
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.5477342064141678e-05,
+      "loss": 0.338,
+      "step": 83090
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.5474365652206265e-05,
+      "loss": 0.3219,
+      "step": 83100
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.5471389240270855e-05,
+      "loss": 0.3561,
+      "step": 83110
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.5468412828335442e-05,
+      "loss": 0.3181,
+      "step": 83120
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.546543641640003e-05,
+      "loss": 0.316,
+      "step": 83130
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.546246000446462e-05,
+      "loss": 0.311,
+      "step": 83140
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.5459483592529206e-05,
+      "loss": 0.3264,
+      "step": 83150
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.5456507180593797e-05,
+      "loss": 0.3179,
+      "step": 83160
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.5453530768658384e-05,
+      "loss": 0.2932,
+      "step": 83170
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.545055435672297e-05,
+      "loss": 0.3024,
+      "step": 83180
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.544757794478756e-05,
+      "loss": 0.3192,
+      "step": 83190
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.5444601532852148e-05,
+      "loss": 0.3305,
+      "step": 83200
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.5441625120916738e-05,
+      "loss": 0.3039,
+      "step": 83210
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.5438648708981325e-05,
+      "loss": 0.2955,
+      "step": 83220
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.5435672297045912e-05,
+      "loss": 0.3163,
+      "step": 83230
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.54326958851105e-05,
+      "loss": 0.3275,
+      "step": 83240
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.542971947317509e-05,
+      "loss": 0.3089,
+      "step": 83250
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.5426743061239676e-05,
+      "loss": 0.3098,
+      "step": 83260
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.5423766649304266e-05,
+      "loss": 0.3164,
+      "step": 83270
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.5420790237368853e-05,
+      "loss": 0.3045,
+      "step": 83280
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.541781382543344e-05,
+      "loss": 0.3092,
+      "step": 83290
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.541483741349803e-05,
+      "loss": 0.3257,
+      "step": 83300
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.5411861001562617e-05,
+      "loss": 0.308,
+      "step": 83310
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.5408884589627207e-05,
+      "loss": 0.3151,
+      "step": 83320
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.5405908177691794e-05,
+      "loss": 0.3193,
+      "step": 83330
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.540293176575638e-05,
+      "loss": 0.3031,
+      "step": 83340
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.539995535382097e-05,
+      "loss": 0.3118,
+      "step": 83350
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.539697894188556e-05,
+      "loss": 0.3085,
+      "step": 83360
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.539400252995015e-05,
+      "loss": 0.3037,
+      "step": 83370
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.5391026118014732e-05,
+      "loss": 0.3117,
+      "step": 83380
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.5388049706079323e-05,
+      "loss": 0.3312,
+      "step": 83390
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.538507329414391e-05,
+      "loss": 0.3043,
+      "step": 83400
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.53820968822085e-05,
+      "loss": 0.3204,
+      "step": 83410
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.5379120470273087e-05,
+      "loss": 0.3219,
+      "step": 83420
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.5376144058337673e-05,
+      "loss": 0.3177,
+      "step": 83430
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.5373167646402264e-05,
+      "loss": 0.3218,
+      "step": 83440
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.537019123446685e-05,
+      "loss": 0.3157,
+      "step": 83450
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.536721482253144e-05,
+      "loss": 0.3013,
+      "step": 83460
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.5364238410596028e-05,
+      "loss": 0.3242,
+      "step": 83470
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.5361261998660615e-05,
+      "loss": 0.331,
+      "step": 83480
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.5358285586725205e-05,
+      "loss": 0.3166,
+      "step": 83490
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.5355309174789792e-05,
+      "loss": 0.3044,
+      "step": 83500
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.5352332762854382e-05,
+      "loss": 0.3305,
+      "step": 83510
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.534935635091897e-05,
+      "loss": 0.3264,
+      "step": 83520
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.5346379938983556e-05,
+      "loss": 0.3206,
+      "step": 83530
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.5343403527048143e-05,
+      "loss": 0.3102,
+      "step": 83540
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.5340427115112733e-05,
+      "loss": 0.3034,
+      "step": 83550
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.533745070317732e-05,
+      "loss": 0.318,
+      "step": 83560
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.533447429124191e-05,
+      "loss": 0.3208,
+      "step": 83570
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.5331497879306497e-05,
+      "loss": 0.3168,
+      "step": 83580
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 1.5328521467371084e-05,
+      "loss": 0.3076,
+      "step": 83590
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.5325545055435674e-05,
+      "loss": 0.2983,
+      "step": 83600
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.532256864350026e-05,
+      "loss": 0.3047,
+      "step": 83610
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.531959223156485e-05,
+      "loss": 0.3249,
+      "step": 83620
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.531661581962944e-05,
+      "loss": 0.3109,
+      "step": 83630
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 1.5313639407694025e-05,
+      "loss": 0.3155,
+      "step": 83640
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.5310662995758612e-05,
+      "loss": 0.311,
+      "step": 83650
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.5307686583823203e-05,
+      "loss": 0.3135,
+      "step": 83660
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.530471017188779e-05,
+      "loss": 0.3249,
+      "step": 83670
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 1.530173375995238e-05,
+      "loss": 0.3167,
+      "step": 83680
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.5298757348016967e-05,
+      "loss": 0.3172,
+      "step": 83690
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.5295780936081554e-05,
+      "loss": 0.3152,
+      "step": 83700
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.5292804524146144e-05,
+      "loss": 0.3212,
+      "step": 83710
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.528982811221073e-05,
+      "loss": 0.3088,
+      "step": 83720
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 1.528685170027532e-05,
+      "loss": 0.3132,
+      "step": 83730
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.5283875288339908e-05,
+      "loss": 0.3291,
+      "step": 83740
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.5280898876404495e-05,
+      "loss": 0.3069,
+      "step": 83750
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.5277922464469085e-05,
+      "loss": 0.3134,
+      "step": 83760
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 1.5274946052533672e-05,
+      "loss": 0.3225,
+      "step": 83770
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.5271969640598262e-05,
+      "loss": 0.306,
+      "step": 83780
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.5268993228662846e-05,
+      "loss": 0.3069,
+      "step": 83790
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.5266016816727436e-05,
+      "loss": 0.3223,
+      "step": 83800
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.5263040404792023e-05,
+      "loss": 0.3054,
+      "step": 83810
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 1.5260063992856613e-05,
+      "loss": 0.3207,
+      "step": 83820
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.5257087580921202e-05,
+      "loss": 0.3177,
+      "step": 83830
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.5254111168985787e-05,
+      "loss": 0.3218,
+      "step": 83840
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.5251134757050376e-05,
+      "loss": 0.3224,
+      "step": 83850
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 1.5248158345114964e-05,
+      "loss": 0.3292,
+      "step": 83860
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.5245181933179553e-05,
+      "loss": 0.3189,
+      "step": 83870
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.5242205521244142e-05,
+      "loss": 0.3139,
+      "step": 83880
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.5239229109308728e-05,
+      "loss": 0.3099,
+      "step": 83890
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.5236252697373317e-05,
+      "loss": 0.3073,
+      "step": 83900
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 1.5233276285437906e-05,
+      "loss": 0.3175,
+      "step": 83910
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.5230299873502494e-05,
+      "loss": 0.3119,
+      "step": 83920
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.5227323461567083e-05,
+      "loss": 0.3198,
+      "step": 83930
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.522434704963167e-05,
+      "loss": 0.3022,
+      "step": 83940
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 1.5221370637696258e-05,
+      "loss": 0.2942,
+      "step": 83950
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.5218394225760847e-05,
+      "loss": 0.3251,
+      "step": 83960
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.5215417813825435e-05,
+      "loss": 0.3165,
+      "step": 83970
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.5212441401890024e-05,
+      "loss": 0.314,
+      "step": 83980
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.5209464989954613e-05,
+      "loss": 0.3253,
+      "step": 83990
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 1.5206488578019198e-05,
+      "loss": 0.3007,
+      "step": 84000
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.5203512166083786e-05,
+      "loss": 0.3169,
+      "step": 84010
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.5200535754148375e-05,
+      "loss": 0.3309,
+      "step": 84020
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.5197559342212964e-05,
+      "loss": 0.324,
+      "step": 84030
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 1.5194582930277552e-05,
+      "loss": 0.309,
+      "step": 84040
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.5191606518342139e-05,
+      "loss": 0.3182,
+      "step": 84050
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.5188630106406728e-05,
+      "loss": 0.3082,
+      "step": 84060
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.5185653694471316e-05,
+      "loss": 0.313,
+      "step": 84070
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.5182677282535905e-05,
+      "loss": 0.3189,
+      "step": 84080
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 1.5179700870600494e-05,
+      "loss": 0.325,
+      "step": 84090
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.517672445866508e-05,
+      "loss": 0.3157,
+      "step": 84100
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.5173748046729669e-05,
+      "loss": 0.2981,
+      "step": 84110
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.5170771634794258e-05,
+      "loss": 0.3145,
+      "step": 84120
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 1.5167795222858846e-05,
+      "loss": 0.3188,
+      "step": 84130
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.5164818810923435e-05,
+      "loss": 0.3148,
+      "step": 84140
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.516184239898802e-05,
+      "loss": 0.3131,
+      "step": 84150
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.5158865987052609e-05,
+      "loss": 0.3051,
+      "step": 84160
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.5155889575117197e-05,
+      "loss": 0.3199,
+      "step": 84170
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 1.5152913163181786e-05,
+      "loss": 0.305,
+      "step": 84180
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.5149936751246374e-05,
+      "loss": 0.3073,
+      "step": 84190
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.5146960339310961e-05,
+      "loss": 0.3246,
+      "step": 84200
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.514398392737555e-05,
+      "loss": 0.3186,
+      "step": 84210
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 1.5141007515440138e-05,
+      "loss": 0.3106,
+      "step": 84220
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.5138031103504727e-05,
+      "loss": 0.3349,
+      "step": 84230
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.5135054691569316e-05,
+      "loss": 0.3276,
+      "step": 84240
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.51320782796339e-05,
+      "loss": 0.2998,
+      "step": 84250
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.512910186769849e-05,
+      "loss": 0.3148,
+      "step": 84260
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.5126125455763078e-05,
+      "loss": 0.313,
+      "step": 84270
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.5123149043827667e-05,
+      "loss": 0.3091,
+      "step": 84280
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.5120172631892255e-05,
+      "loss": 0.2989,
+      "step": 84290
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.5117196219956842e-05,
+      "loss": 0.3263,
+      "step": 84300
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.511421980802143e-05,
+      "loss": 0.3095,
+      "step": 84310
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.511124339608602e-05,
+      "loss": 0.3103,
+      "step": 84320
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.5108266984150608e-05,
+      "loss": 0.3066,
+      "step": 84330
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.5105290572215196e-05,
+      "loss": 0.3088,
+      "step": 84340
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.5102314160279783e-05,
+      "loss": 0.3134,
+      "step": 84350
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 1.5099337748344372e-05,
+      "loss": 0.3066,
+      "step": 84360
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.509636133640896e-05,
+      "loss": 0.3239,
+      "step": 84370
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.509338492447355e-05,
+      "loss": 0.3329,
+      "step": 84380
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.5090408512538138e-05,
+      "loss": 0.3068,
+      "step": 84390
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.5087432100602723e-05,
+      "loss": 0.314,
+      "step": 84400
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.5084455688667312e-05,
+      "loss": 0.3143,
+      "step": 84410
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.50814792767319e-05,
+      "loss": 0.3151,
+      "step": 84420
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.5078502864796489e-05,
+      "loss": 0.3214,
+      "step": 84430
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.5075526452861077e-05,
+      "loss": 0.3232,
+      "step": 84440
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.5072550040925666e-05,
+      "loss": 0.3133,
+      "step": 84450
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.5069573628990253e-05,
+      "loss": 0.327,
+      "step": 84460
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.5066597217054841e-05,
+      "loss": 0.3048,
+      "step": 84470
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.506362080511943e-05,
+      "loss": 0.3306,
+      "step": 84480
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.5060644393184019e-05,
+      "loss": 0.3057,
+      "step": 84490
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.5057667981248607e-05,
+      "loss": 0.3195,
+      "step": 84500
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.5054691569313194e-05,
+      "loss": 0.3073,
+      "step": 84510
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.5051715157377783e-05,
+      "loss": 0.3086,
+      "step": 84520
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.5048738745442371e-05,
+      "loss": 0.2964,
+      "step": 84530
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.504576233350696e-05,
+      "loss": 0.3234,
+      "step": 84540
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.5042785921571548e-05,
+      "loss": 0.308,
+      "step": 84550
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.5039809509636134e-05,
+      "loss": 0.3243,
+      "step": 84560
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.5036833097700722e-05,
+      "loss": 0.3222,
+      "step": 84570
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 1.5033856685765311e-05,
+      "loss": 0.3194,
+      "step": 84580
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.50308802738299e-05,
+      "loss": 0.3209,
+      "step": 84590
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.5027903861894488e-05,
+      "loss": 0.3212,
+      "step": 84600
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.5024927449959075e-05,
+      "loss": 0.307,
+      "step": 84610
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.5021951038023664e-05,
+      "loss": 0.3155,
+      "step": 84620
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.5018974626088252e-05,
+      "loss": 0.3198,
+      "step": 84630
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.501599821415284e-05,
+      "loss": 0.3064,
+      "step": 84640
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.501302180221743e-05,
+      "loss": 0.3217,
+      "step": 84650
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.5010045390282015e-05,
+      "loss": 0.3168,
+      "step": 84660
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.5007068978346603e-05,
+      "loss": 0.3147,
+      "step": 84670
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.5004092566411192e-05,
+      "loss": 0.3127,
+      "step": 84680
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.500111615447578e-05,
+      "loss": 0.3143,
+      "step": 84690
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.4998139742540369e-05,
+      "loss": 0.3128,
+      "step": 84700
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.4995163330604956e-05,
+      "loss": 0.3288,
+      "step": 84710
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.4992186918669544e-05,
+      "loss": 0.3143,
+      "step": 84720
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.4989210506734133e-05,
+      "loss": 0.3224,
+      "step": 84730
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.4986234094798722e-05,
+      "loss": 0.3269,
+      "step": 84740
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.498325768286331e-05,
+      "loss": 0.3191,
+      "step": 84750
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 1.4980281270927897e-05,
+      "loss": 0.3195,
+      "step": 84760
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.4977304858992486e-05,
+      "loss": 0.3138,
+      "step": 84770
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.4974328447057074e-05,
+      "loss": 0.3136,
+      "step": 84780
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.4971352035121663e-05,
+      "loss": 0.3204,
+      "step": 84790
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.4968375623186251e-05,
+      "loss": 0.3001,
+      "step": 84800
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.4965399211250837e-05,
+      "loss": 0.3242,
+      "step": 84810
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.4962422799315425e-05,
+      "loss": 0.3085,
+      "step": 84820
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.4959446387380014e-05,
+      "loss": 0.2929,
+      "step": 84830
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.4956469975444602e-05,
+      "loss": 0.3028,
+      "step": 84840
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.4953493563509191e-05,
+      "loss": 0.3184,
+      "step": 84850
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.495051715157378e-05,
+      "loss": 0.3076,
+      "step": 84860
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.4947540739638366e-05,
+      "loss": 0.3062,
+      "step": 84870
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.4944564327702955e-05,
+      "loss": 0.3186,
+      "step": 84880
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.4941587915767544e-05,
+      "loss": 0.3129,
+      "step": 84890
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.4938611503832132e-05,
+      "loss": 0.3109,
+      "step": 84900
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.4935635091896721e-05,
+      "loss": 0.315,
+      "step": 84910
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.4932658679961308e-05,
+      "loss": 0.3049,
+      "step": 84920
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.4929682268025896e-05,
+      "loss": 0.3255,
+      "step": 84930
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.4926705856090485e-05,
+      "loss": 0.3056,
+      "step": 84940
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.4923729444155074e-05,
+      "loss": 0.3122,
+      "step": 84950
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.4920753032219662e-05,
+      "loss": 0.3202,
+      "step": 84960
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.4917776620284247e-05,
+      "loss": 0.3211,
+      "step": 84970
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.4914800208348836e-05,
+      "loss": 0.3172,
+      "step": 84980
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 1.4911823796413425e-05,
+      "loss": 0.3124,
+      "step": 84990
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.4908847384478013e-05,
+      "loss": 0.3178,
+      "step": 85000
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.4905870972542602e-05,
+      "loss": 0.3176,
+      "step": 85010
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.4902894560607189e-05,
+      "loss": 0.3159,
+      "step": 85020
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.4899918148671777e-05,
+      "loss": 0.3165,
+      "step": 85030
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.4896941736736366e-05,
+      "loss": 0.3117,
+      "step": 85040
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.4893965324800954e-05,
+      "loss": 0.3082,
+      "step": 85050
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.4890988912865543e-05,
+      "loss": 0.3102,
+      "step": 85060
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.488801250093013e-05,
+      "loss": 0.3085,
+      "step": 85070
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.4885036088994718e-05,
+      "loss": 0.3275,
+      "step": 85080
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.4882059677059305e-05,
+      "loss": 0.3119,
+      "step": 85090
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.4879083265123894e-05,
+      "loss": 0.3308,
+      "step": 85100
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.4876106853188483e-05,
+      "loss": 0.3126,
+      "step": 85110
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.487313044125307e-05,
+      "loss": 0.3223,
+      "step": 85120
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.4870154029317658e-05,
+      "loss": 0.318,
+      "step": 85130
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.4867177617382247e-05,
+      "loss": 0.3119,
+      "step": 85140
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.4864201205446835e-05,
+      "loss": 0.32,
+      "step": 85150
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.4861224793511424e-05,
+      "loss": 0.3074,
+      "step": 85160
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 1.485824838157601e-05,
+      "loss": 0.3155,
+      "step": 85170
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.48552719696406e-05,
+      "loss": 0.3135,
+      "step": 85180
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.4852295557705188e-05,
+      "loss": 0.3184,
+      "step": 85190
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.4849319145769777e-05,
+      "loss": 0.3234,
+      "step": 85200
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.4846342733834365e-05,
+      "loss": 0.3311,
+      "step": 85210
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.484336632189895e-05,
+      "loss": 0.2978,
+      "step": 85220
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.4840389909963539e-05,
+      "loss": 0.317,
+      "step": 85230
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.4837413498028127e-05,
+      "loss": 0.3344,
+      "step": 85240
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.4834437086092716e-05,
+      "loss": 0.3039,
+      "step": 85250
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.4831460674157305e-05,
+      "loss": 0.2969,
+      "step": 85260
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.4828484262221892e-05,
+      "loss": 0.3214,
+      "step": 85270
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.482550785028648e-05,
+      "loss": 0.3124,
+      "step": 85280
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.4822531438351069e-05,
+      "loss": 0.3221,
+      "step": 85290
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.4819555026415657e-05,
+      "loss": 0.3157,
+      "step": 85300
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.4816578614480246e-05,
+      "loss": 0.3205,
+      "step": 85310
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.4813602202544835e-05,
+      "loss": 0.3197,
+      "step": 85320
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.4810625790609421e-05,
+      "loss": 0.3127,
+      "step": 85330
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.480764937867401e-05,
+      "loss": 0.2975,
+      "step": 85340
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.4804672966738599e-05,
+      "loss": 0.3283,
+      "step": 85350
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.4801696554803187e-05,
+      "loss": 0.3028,
+      "step": 85360
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.4798720142867776e-05,
+      "loss": 0.3247,
+      "step": 85370
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.4795743730932361e-05,
+      "loss": 0.301,
+      "step": 85380
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 1.479276731899695e-05,
+      "loss": 0.3149,
+      "step": 85390
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.4789790907061538e-05,
+      "loss": 0.3012,
+      "step": 85400
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.4786814495126127e-05,
+      "loss": 0.3059,
+      "step": 85410
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.4783838083190715e-05,
+      "loss": 0.3241,
+      "step": 85420
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.4780861671255302e-05,
+      "loss": 0.3138,
+      "step": 85430
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.4777885259319891e-05,
+      "loss": 0.3093,
+      "step": 85440
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.477490884738448e-05,
+      "loss": 0.3161,
+      "step": 85450
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.4771932435449068e-05,
+      "loss": 0.3173,
+      "step": 85460
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.4768956023513657e-05,
+      "loss": 0.33,
+      "step": 85470
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.4765979611578244e-05,
+      "loss": 0.3218,
+      "step": 85480
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.4763003199642832e-05,
+      "loss": 0.3176,
+      "step": 85490
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.476002678770742e-05,
+      "loss": 0.3187,
+      "step": 85500
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.475705037577201e-05,
+      "loss": 0.3006,
+      "step": 85510
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.4754073963836596e-05,
+      "loss": 0.3281,
+      "step": 85520
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.4751097551901183e-05,
+      "loss": 0.3216,
+      "step": 85530
+    },
+    {
+      "epoch": 19.0,
+      "eval_cer": 4.549145299145299,
+      "eval_loss": 0.4755484461784363,
+      "eval_runtime": 1.2921,
+      "eval_samples_per_second": 7.74,
+      "eval_steps_per_second": 0.774,
+      "eval_wer": 0.6493506493506493,
+      "step": 85538
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.4748121139965772e-05,
+      "loss": 0.3056,
+      "step": 85540
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.474514472803036e-05,
+      "loss": 0.3192,
+      "step": 85550
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.4742168316094949e-05,
+      "loss": 0.311,
+      "step": 85560
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 1.4739191904159538e-05,
+      "loss": 0.304,
+      "step": 85570
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.4736215492224124e-05,
+      "loss": 0.3115,
+      "step": 85580
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.4733239080288713e-05,
+      "loss": 0.3006,
+      "step": 85590
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.4730262668353302e-05,
+      "loss": 0.313,
+      "step": 85600
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.472728625641789e-05,
+      "loss": 0.3144,
+      "step": 85610
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.4724309844482479e-05,
+      "loss": 0.3053,
+      "step": 85620
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.4721333432547064e-05,
+      "loss": 0.2966,
+      "step": 85630
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.4718357020611653e-05,
+      "loss": 0.3148,
+      "step": 85640
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.4715380608676241e-05,
+      "loss": 0.3004,
+      "step": 85650
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.471240419674083e-05,
+      "loss": 0.3136,
+      "step": 85660
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.4709427784805418e-05,
+      "loss": 0.3149,
+      "step": 85670
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.4706451372870005e-05,
+      "loss": 0.3152,
+      "step": 85680
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.4703474960934594e-05,
+      "loss": 0.3165,
+      "step": 85690
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.4700498548999182e-05,
+      "loss": 0.3188,
+      "step": 85700
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.4697522137063771e-05,
+      "loss": 0.3032,
+      "step": 85710
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.469454572512836e-05,
+      "loss": 0.3,
+      "step": 85720
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.4691569313192948e-05,
+      "loss": 0.295,
+      "step": 85730
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.4688592901257535e-05,
+      "loss": 0.3124,
+      "step": 85740
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.4685616489322124e-05,
+      "loss": 0.306,
+      "step": 85750
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.4682640077386712e-05,
+      "loss": 0.3033,
+      "step": 85760
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.4679663665451301e-05,
+      "loss": 0.2994,
+      "step": 85770
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.467668725351589e-05,
+      "loss": 0.3046,
+      "step": 85780
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.4673710841580475e-05,
+      "loss": 0.3017,
+      "step": 85790
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 1.4670734429645063e-05,
+      "loss": 0.3173,
+      "step": 85800
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4667758017709652e-05,
+      "loss": 0.3191,
+      "step": 85810
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.466478160577424e-05,
+      "loss": 0.2992,
+      "step": 85820
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4661805193838829e-05,
+      "loss": 0.3021,
+      "step": 85830
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4658828781903416e-05,
+      "loss": 0.2852,
+      "step": 85840
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4655852369968005e-05,
+      "loss": 0.3062,
+      "step": 85850
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4652875958032593e-05,
+      "loss": 0.2995,
+      "step": 85860
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4649899546097182e-05,
+      "loss": 0.3088,
+      "step": 85870
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.464692313416177e-05,
+      "loss": 0.3152,
+      "step": 85880
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4643946722226357e-05,
+      "loss": 0.2966,
+      "step": 85890
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4640970310290946e-05,
+      "loss": 0.3072,
+      "step": 85900
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4637993898355534e-05,
+      "loss": 0.3166,
+      "step": 85910
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4635017486420123e-05,
+      "loss": 0.304,
+      "step": 85920
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4632041074484712e-05,
+      "loss": 0.3094,
+      "step": 85930
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4629064662549297e-05,
+      "loss": 0.3155,
+      "step": 85940
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4626088250613885e-05,
+      "loss": 0.3214,
+      "step": 85950
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4623111838678474e-05,
+      "loss": 0.3098,
+      "step": 85960
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4620135426743063e-05,
+      "loss": 0.3203,
+      "step": 85970
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4617159014807651e-05,
+      "loss": 0.307,
+      "step": 85980
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4614182602872238e-05,
+      "loss": 0.3103,
+      "step": 85990
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4611206190936827e-05,
+      "loss": 0.3263,
+      "step": 86000
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4608229779001415e-05,
+      "loss": 0.3054,
+      "step": 86010
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 1.4605253367066004e-05,
+      "loss": 0.3259,
+      "step": 86020
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4602276955130592e-05,
+      "loss": 0.2951,
+      "step": 86030
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4599300543195178e-05,
+      "loss": 0.3146,
+      "step": 86040
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4596324131259766e-05,
+      "loss": 0.3102,
+      "step": 86050
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4593347719324355e-05,
+      "loss": 0.3066,
+      "step": 86060
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.4590371307388943e-05,
+      "loss": 0.3093,
+      "step": 86070
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4587394895453532e-05,
+      "loss": 0.2991,
+      "step": 86080
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4584418483518119e-05,
+      "loss": 0.3082,
+      "step": 86090
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4581442071582708e-05,
+      "loss": 0.3194,
+      "step": 86100
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.4578465659647296e-05,
+      "loss": 0.312,
+      "step": 86110
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.4575489247711885e-05,
+      "loss": 0.3029,
+      "step": 86120
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.4572512835776473e-05,
+      "loss": 0.3045,
+      "step": 86130
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.456953642384106e-05,
+      "loss": 0.3105,
+      "step": 86140
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.4566560011905649e-05,
+      "loss": 0.3129,
+      "step": 86150
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.4563583599970237e-05,
+      "loss": 0.3212,
+      "step": 86160
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.4560607188034826e-05,
+      "loss": 0.3014,
+      "step": 86170
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.4557630776099415e-05,
+      "loss": 0.3039,
+      "step": 86180
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.4554654364164003e-05,
+      "loss": 0.3022,
+      "step": 86190
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 1.4551677952228588e-05,
+      "loss": 0.2929,
+      "step": 86200
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.4548701540293177e-05,
+      "loss": 0.3081,
+      "step": 86210
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.4545725128357766e-05,
+      "loss": 0.3158,
+      "step": 86220
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.4542748716422354e-05,
+      "loss": 0.2991,
+      "step": 86230
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.4539772304486943e-05,
+      "loss": 0.3127,
+      "step": 86240
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.453679589255153e-05,
+      "loss": 0.3041,
+      "step": 86250
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.4533819480616118e-05,
+      "loss": 0.2965,
+      "step": 86260
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.4530843068680707e-05,
+      "loss": 0.3247,
+      "step": 86270
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.4527866656745295e-05,
+      "loss": 0.3007,
+      "step": 86280
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.4524890244809884e-05,
+      "loss": 0.3032,
+      "step": 86290
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.4521913832874471e-05,
+      "loss": 0.2948,
+      "step": 86300
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.451893742093906e-05,
+      "loss": 0.3083,
+      "step": 86310
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.4515961009003648e-05,
+      "loss": 0.2957,
+      "step": 86320
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.4512984597068237e-05,
+      "loss": 0.3052,
+      "step": 86330
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.4510008185132825e-05,
+      "loss": 0.3173,
+      "step": 86340
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.450703177319741e-05,
+      "loss": 0.3106,
+      "step": 86350
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.4504055361261999e-05,
+      "loss": 0.3139,
+      "step": 86360
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.4501078949326588e-05,
+      "loss": 0.3104,
+      "step": 86370
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.4498102537391176e-05,
+      "loss": 0.3008,
+      "step": 86380
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.4495126125455765e-05,
+      "loss": 0.3099,
+      "step": 86390
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.4492149713520352e-05,
+      "loss": 0.3271,
+      "step": 86400
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.448917330158494e-05,
+      "loss": 0.3057,
+      "step": 86410
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.4486196889649529e-05,
+      "loss": 0.3179,
+      "step": 86420
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 1.4483220477714118e-05,
+      "loss": 0.3095,
+      "step": 86430
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.4480244065778706e-05,
+      "loss": 0.3047,
+      "step": 86440
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.4477267653843291e-05,
+      "loss": 0.3123,
+      "step": 86450
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.447429124190788e-05,
+      "loss": 0.2953,
+      "step": 86460
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.4471314829972469e-05,
+      "loss": 0.3167,
+      "step": 86470
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.4468338418037057e-05,
+      "loss": 0.303,
+      "step": 86480
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.4465362006101646e-05,
+      "loss": 0.2945,
+      "step": 86490
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.4462385594166233e-05,
+      "loss": 0.2998,
+      "step": 86500
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.4459409182230821e-05,
+      "loss": 0.3118,
+      "step": 86510
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.445643277029541e-05,
+      "loss": 0.2855,
+      "step": 86520
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.4453456358359998e-05,
+      "loss": 0.3195,
+      "step": 86530
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.4450479946424587e-05,
+      "loss": 0.2986,
+      "step": 86540
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.4447503534489174e-05,
+      "loss": 0.3079,
+      "step": 86550
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.4444527122553762e-05,
+      "loss": 0.3168,
+      "step": 86560
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.4441550710618351e-05,
+      "loss": 0.3215,
+      "step": 86570
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.443857429868294e-05,
+      "loss": 0.3041,
+      "step": 86580
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.4435597886747528e-05,
+      "loss": 0.3035,
+      "step": 86590
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.4432621474812117e-05,
+      "loss": 0.3081,
+      "step": 86600
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.4429645062876702e-05,
+      "loss": 0.3031,
+      "step": 86610
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.442666865094129e-05,
+      "loss": 0.2981,
+      "step": 86620
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.442369223900588e-05,
+      "loss": 0.3123,
+      "step": 86630
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.4420715827070468e-05,
+      "loss": 0.3124,
+      "step": 86640
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.4417739415135056e-05,
+      "loss": 0.3124,
+      "step": 86650
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.4414763003199643e-05,
+      "loss": 0.3235,
+      "step": 86660
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.4411786591264232e-05,
+      "loss": 0.3171,
+      "step": 86670
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.440881017932882e-05,
+      "loss": 0.3159,
+      "step": 86680
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.4405833767393409e-05,
+      "loss": 0.3129,
+      "step": 86690
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.4402857355457998e-05,
+      "loss": 0.3042,
+      "step": 86700
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.4399880943522585e-05,
+      "loss": 0.298,
+      "step": 86710
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.4396904531587173e-05,
+      "loss": 0.3161,
+      "step": 86720
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.4393928119651762e-05,
+      "loss": 0.3117,
+      "step": 86730
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.439095170771635e-05,
+      "loss": 0.3114,
+      "step": 86740
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.4387975295780939e-05,
+      "loss": 0.3131,
+      "step": 86750
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.4384998883845524e-05,
+      "loss": 0.2929,
+      "step": 86760
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.4382022471910113e-05,
+      "loss": 0.3106,
+      "step": 86770
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.4379046059974701e-05,
+      "loss": 0.2969,
+      "step": 86780
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.437606964803929e-05,
+      "loss": 0.3084,
+      "step": 86790
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.4373093236103879e-05,
+      "loss": 0.3086,
+      "step": 86800
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.4370116824168465e-05,
+      "loss": 0.3204,
+      "step": 86810
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.4367140412233054e-05,
+      "loss": 0.3158,
+      "step": 86820
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.4364164000297643e-05,
+      "loss": 0.3057,
+      "step": 86830
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.4361187588362231e-05,
+      "loss": 0.3009,
+      "step": 86840
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.435821117642682e-05,
+      "loss": 0.3049,
+      "step": 86850
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.4355234764491405e-05,
+      "loss": 0.3002,
+      "step": 86860
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.4352258352555994e-05,
+      "loss": 0.3054,
+      "step": 86870
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.4349281940620582e-05,
+      "loss": 0.3059,
+      "step": 86880
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.434630552868517e-05,
+      "loss": 0.3017,
+      "step": 86890
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.434332911674976e-05,
+      "loss": 0.2989,
+      "step": 86900
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.4340352704814346e-05,
+      "loss": 0.3169,
+      "step": 86910
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.4337376292878935e-05,
+      "loss": 0.2993,
+      "step": 86920
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.4334399880943523e-05,
+      "loss": 0.2946,
+      "step": 86930
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.4331423469008112e-05,
+      "loss": 0.3149,
+      "step": 86940
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.43284470570727e-05,
+      "loss": 0.3086,
+      "step": 86950
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.4325470645137288e-05,
+      "loss": 0.307,
+      "step": 86960
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.4322494233201876e-05,
+      "loss": 0.3039,
+      "step": 86970
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.4319517821266465e-05,
+      "loss": 0.3275,
+      "step": 86980
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.4316541409331053e-05,
+      "loss": 0.3186,
+      "step": 86990
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.4313564997395642e-05,
+      "loss": 0.3172,
+      "step": 87000
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.4310588585460227e-05,
+      "loss": 0.3101,
+      "step": 87010
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.4307612173524816e-05,
+      "loss": 0.3094,
+      "step": 87020
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.4304635761589404e-05,
+      "loss": 0.3047,
+      "step": 87030
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.4301659349653993e-05,
+      "loss": 0.296,
+      "step": 87040
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.4298682937718581e-05,
+      "loss": 0.312,
+      "step": 87050
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.429570652578317e-05,
+      "loss": 0.2997,
+      "step": 87060
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.4292730113847757e-05,
+      "loss": 0.3193,
+      "step": 87070
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.4289753701912346e-05,
+      "loss": 0.3119,
+      "step": 87080
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.4286777289976934e-05,
+      "loss": 0.2993,
+      "step": 87090
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.4283800878041523e-05,
+      "loss": 0.2983,
+      "step": 87100
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.4280824466106111e-05,
+      "loss": 0.31,
+      "step": 87110
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.4277848054170698e-05,
+      "loss": 0.307,
+      "step": 87120
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.4274871642235287e-05,
+      "loss": 0.3176,
+      "step": 87130
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.4271895230299875e-05,
+      "loss": 0.3157,
+      "step": 87140
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.4268918818364464e-05,
+      "loss": 0.3164,
+      "step": 87150
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.4265942406429053e-05,
+      "loss": 0.3001,
+      "step": 87160
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.4262965994493638e-05,
+      "loss": 0.3296,
+      "step": 87170
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.4259989582558226e-05,
+      "loss": 0.3162,
+      "step": 87180
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.4257013170622815e-05,
+      "loss": 0.3139,
+      "step": 87190
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.4254036758687404e-05,
+      "loss": 0.3028,
+      "step": 87200
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.4251060346751992e-05,
+      "loss": 0.3012,
+      "step": 87210
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.4248083934816579e-05,
+      "loss": 0.2996,
+      "step": 87220
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.4245107522881168e-05,
+      "loss": 0.3191,
+      "step": 87230
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.4242131110945756e-05,
+      "loss": 0.2991,
+      "step": 87240
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.4239154699010345e-05,
+      "loss": 0.3217,
+      "step": 87250
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.4236178287074933e-05,
+      "loss": 0.3197,
+      "step": 87260
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.423320187513952e-05,
+      "loss": 0.3154,
+      "step": 87270
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.4230225463204109e-05,
+      "loss": 0.3033,
+      "step": 87280
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.4227249051268696e-05,
+      "loss": 0.3206,
+      "step": 87290
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.4224272639333284e-05,
+      "loss": 0.3259,
+      "step": 87300
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.4221296227397873e-05,
+      "loss": 0.3016,
+      "step": 87310
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.421831981546246e-05,
+      "loss": 0.3003,
+      "step": 87320
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.4215343403527049e-05,
+      "loss": 0.3104,
+      "step": 87330
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4212366991591637e-05,
+      "loss": 0.2947,
+      "step": 87340
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4209390579656226e-05,
+      "loss": 0.2983,
+      "step": 87350
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4206414167720814e-05,
+      "loss": 0.3105,
+      "step": 87360
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4203437755785401e-05,
+      "loss": 0.2973,
+      "step": 87370
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.420046134384999e-05,
+      "loss": 0.2926,
+      "step": 87380
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4197484931914578e-05,
+      "loss": 0.306,
+      "step": 87390
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4194508519979167e-05,
+      "loss": 0.2984,
+      "step": 87400
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4191532108043756e-05,
+      "loss": 0.3044,
+      "step": 87410
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.418855569610834e-05,
+      "loss": 0.2999,
+      "step": 87420
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.418557928417293e-05,
+      "loss": 0.3121,
+      "step": 87430
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4182602872237518e-05,
+      "loss": 0.3103,
+      "step": 87440
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4179626460302107e-05,
+      "loss": 0.294,
+      "step": 87450
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4176650048366695e-05,
+      "loss": 0.2891,
+      "step": 87460
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4173673636431284e-05,
+      "loss": 0.2905,
+      "step": 87470
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.417069722449587e-05,
+      "loss": 0.2949,
+      "step": 87480
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.416772081256046e-05,
+      "loss": 0.3037,
+      "step": 87490
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4164744400625048e-05,
+      "loss": 0.3062,
+      "step": 87500
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4161767988689636e-05,
+      "loss": 0.2995,
+      "step": 87510
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4158791576754225e-05,
+      "loss": 0.3046,
+      "step": 87520
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4155815164818812e-05,
+      "loss": 0.3199,
+      "step": 87530
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.41528387528834e-05,
+      "loss": 0.302,
+      "step": 87540
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4149862340947989e-05,
+      "loss": 0.3135,
+      "step": 87550
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4146885929012578e-05,
+      "loss": 0.2923,
+      "step": 87560
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.4143909517077166e-05,
+      "loss": 0.3022,
+      "step": 87570
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.4140933105141752e-05,
+      "loss": 0.317,
+      "step": 87580
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.413795669320634e-05,
+      "loss": 0.3066,
+      "step": 87590
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.4134980281270929e-05,
+      "loss": 0.3099,
+      "step": 87600
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.4132003869335517e-05,
+      "loss": 0.3229,
+      "step": 87610
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.4129027457400106e-05,
+      "loss": 0.3175,
+      "step": 87620
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.4126051045464693e-05,
+      "loss": 0.3105,
+      "step": 87630
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.4123074633529281e-05,
+      "loss": 0.3112,
+      "step": 87640
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.412009822159387e-05,
+      "loss": 0.3072,
+      "step": 87650
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.4117121809658459e-05,
+      "loss": 0.3207,
+      "step": 87660
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.4114145397723047e-05,
+      "loss": 0.2927,
+      "step": 87670
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.4111168985787634e-05,
+      "loss": 0.2983,
+      "step": 87680
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.4108192573852223e-05,
+      "loss": 0.3169,
+      "step": 87690
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.4105216161916811e-05,
+      "loss": 0.3011,
+      "step": 87700
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.41022397499814e-05,
+      "loss": 0.3119,
+      "step": 87710
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.4099263338045987e-05,
+      "loss": 0.3134,
+      "step": 87720
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.4096286926110574e-05,
+      "loss": 0.2961,
+      "step": 87730
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.4093310514175162e-05,
+      "loss": 0.3026,
+      "step": 87740
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.409033410223975e-05,
+      "loss": 0.2964,
+      "step": 87750
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.408735769030434e-05,
+      "loss": 0.3121,
+      "step": 87760
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.4084381278368928e-05,
+      "loss": 0.3075,
+      "step": 87770
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.4081404866433515e-05,
+      "loss": 0.3069,
+      "step": 87780
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.4078428454498103e-05,
+      "loss": 0.3028,
+      "step": 87790
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.4075452042562692e-05,
+      "loss": 0.3042,
+      "step": 87800
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.407247563062728e-05,
+      "loss": 0.3013,
+      "step": 87810
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.406949921869187e-05,
+      "loss": 0.31,
+      "step": 87820
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.4066522806756454e-05,
+      "loss": 0.2998,
+      "step": 87830
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.4063546394821043e-05,
+      "loss": 0.3078,
+      "step": 87840
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.4060569982885632e-05,
+      "loss": 0.3145,
+      "step": 87850
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.405759357095022e-05,
+      "loss": 0.3041,
+      "step": 87860
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.4054617159014809e-05,
+      "loss": 0.3095,
+      "step": 87870
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.4051640747079396e-05,
+      "loss": 0.3057,
+      "step": 87880
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.4048664335143984e-05,
+      "loss": 0.3078,
+      "step": 87890
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.4045687923208573e-05,
+      "loss": 0.3077,
+      "step": 87900
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.4042711511273162e-05,
+      "loss": 0.3057,
+      "step": 87910
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.403973509933775e-05,
+      "loss": 0.3084,
+      "step": 87920
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.4036758687402339e-05,
+      "loss": 0.3077,
+      "step": 87930
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.4033782275466926e-05,
+      "loss": 0.3101,
+      "step": 87940
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.4030805863531514e-05,
+      "loss": 0.3166,
+      "step": 87950
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.4027829451596103e-05,
+      "loss": 0.3025,
+      "step": 87960
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.4024853039660691e-05,
+      "loss": 0.3203,
+      "step": 87970
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.402187662772528e-05,
+      "loss": 0.3042,
+      "step": 87980
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.4018900215789865e-05,
+      "loss": 0.3156,
+      "step": 87990
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.4015923803854454e-05,
+      "loss": 0.3261,
+      "step": 88000
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.4012947391919042e-05,
+      "loss": 0.3168,
+      "step": 88010
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.4009970979983631e-05,
+      "loss": 0.2989,
+      "step": 88020
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.400699456804822e-05,
+      "loss": 0.3039,
+      "step": 88030
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.4004018156112806e-05,
+      "loss": 0.298,
+      "step": 88040
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.4001041744177395e-05,
+      "loss": 0.3049,
+      "step": 88050
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.3998065332241984e-05,
+      "loss": 0.3178,
+      "step": 88060
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.3995088920306572e-05,
+      "loss": 0.3071,
+      "step": 88070
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.399211250837116e-05,
+      "loss": 0.3145,
+      "step": 88080
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.3989136096435748e-05,
+      "loss": 0.3046,
+      "step": 88090
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.3986159684500336e-05,
+      "loss": 0.3078,
+      "step": 88100
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.3983183272564925e-05,
+      "loss": 0.3075,
+      "step": 88110
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.3980206860629514e-05,
+      "loss": 0.3119,
+      "step": 88120
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.3977230448694102e-05,
+      "loss": 0.3145,
+      "step": 88130
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.3974254036758687e-05,
+      "loss": 0.3118,
+      "step": 88140
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.3971277624823276e-05,
+      "loss": 0.3082,
+      "step": 88150
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.3968301212887864e-05,
+      "loss": 0.3212,
+      "step": 88160
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.3965324800952453e-05,
+      "loss": 0.3009,
+      "step": 88170
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.3962348389017042e-05,
+      "loss": 0.301,
+      "step": 88180
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.3959371977081629e-05,
+      "loss": 0.3083,
+      "step": 88190
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.3956395565146217e-05,
+      "loss": 0.3107,
+      "step": 88200
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.3953419153210806e-05,
+      "loss": 0.3204,
+      "step": 88210
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.3950442741275394e-05,
+      "loss": 0.3148,
+      "step": 88220
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.3947466329339983e-05,
+      "loss": 0.3091,
+      "step": 88230
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.3944489917404568e-05,
+      "loss": 0.3066,
+      "step": 88240
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.3941513505469157e-05,
+      "loss": 0.2919,
+      "step": 88250
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.3938537093533745e-05,
+      "loss": 0.3054,
+      "step": 88260
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.3935560681598334e-05,
+      "loss": 0.3047,
+      "step": 88270
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.3932584269662923e-05,
+      "loss": 0.3085,
+      "step": 88280
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.392960785772751e-05,
+      "loss": 0.3087,
+      "step": 88290
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.3926631445792098e-05,
+      "loss": 0.3008,
+      "step": 88300
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.3923655033856687e-05,
+      "loss": 0.287,
+      "step": 88310
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 1.3920678621921275e-05,
+      "loss": 0.3166,
+      "step": 88320
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.3917702209985864e-05,
+      "loss": 0.3036,
+      "step": 88330
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.3914725798050452e-05,
+      "loss": 0.3111,
+      "step": 88340
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.391174938611504e-05,
+      "loss": 0.3021,
+      "step": 88350
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.3908772974179628e-05,
+      "loss": 0.2965,
+      "step": 88360
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 1.3905796562244216e-05,
+      "loss": 0.3162,
+      "step": 88370
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.3902820150308805e-05,
+      "loss": 0.3035,
+      "step": 88380
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.3899843738373394e-05,
+      "loss": 0.3055,
+      "step": 88390
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.3896867326437979e-05,
+      "loss": 0.3046,
+      "step": 88400
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 1.3893890914502567e-05,
+      "loss": 0.3233,
+      "step": 88410
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.3890914502567156e-05,
+      "loss": 0.3098,
+      "step": 88420
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.3887938090631745e-05,
+      "loss": 0.3156,
+      "step": 88430
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.3884961678696333e-05,
+      "loss": 0.3011,
+      "step": 88440
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.388198526676092e-05,
+      "loss": 0.3116,
+      "step": 88450
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 1.3879008854825509e-05,
+      "loss": 0.3061,
+      "step": 88460
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.3876032442890097e-05,
+      "loss": 0.3084,
+      "step": 88470
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.3873056030954686e-05,
+      "loss": 0.3115,
+      "step": 88480
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.3870079619019275e-05,
+      "loss": 0.3091,
+      "step": 88490
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 1.3867103207083861e-05,
+      "loss": 0.303,
+      "step": 88500
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.386412679514845e-05,
+      "loss": 0.2923,
+      "step": 88510
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.3861150383213039e-05,
+      "loss": 0.3187,
+      "step": 88520
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.3858173971277627e-05,
+      "loss": 0.2972,
+      "step": 88530
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.3855197559342216e-05,
+      "loss": 0.3027,
+      "step": 88540
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 1.3852221147406801e-05,
+      "loss": 0.3224,
+      "step": 88550
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.384924473547139e-05,
+      "loss": 0.3022,
+      "step": 88560
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.3846268323535978e-05,
+      "loss": 0.3188,
+      "step": 88570
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.3843291911600567e-05,
+      "loss": 0.3071,
+      "step": 88580
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 1.3840315499665155e-05,
+      "loss": 0.3118,
+      "step": 88590
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.3837339087729742e-05,
+      "loss": 0.3082,
+      "step": 88600
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.383436267579433e-05,
+      "loss": 0.3277,
+      "step": 88610
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.383138626385892e-05,
+      "loss": 0.2943,
+      "step": 88620
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.3828409851923508e-05,
+      "loss": 0.2962,
+      "step": 88630
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 1.3825433439988097e-05,
+      "loss": 0.2991,
+      "step": 88640
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.3822457028052682e-05,
+      "loss": 0.313,
+      "step": 88650
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.381948061611727e-05,
+      "loss": 0.3074,
+      "step": 88660
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.3816504204181859e-05,
+      "loss": 0.3079,
+      "step": 88670
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 1.3813527792246448e-05,
+      "loss": 0.2981,
+      "step": 88680
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.3810551380311036e-05,
+      "loss": 0.3179,
+      "step": 88690
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.3807574968375623e-05,
+      "loss": 0.3144,
+      "step": 88700
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.3804598556440212e-05,
+      "loss": 0.3131,
+      "step": 88710
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.38016221445048e-05,
+      "loss": 0.3159,
+      "step": 88720
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 1.3798645732569389e-05,
+      "loss": 0.3057,
+      "step": 88730
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.3795669320633977e-05,
+      "loss": 0.3029,
+      "step": 88740
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.3792692908698564e-05,
+      "loss": 0.3095,
+      "step": 88750
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.3789716496763153e-05,
+      "loss": 0.3117,
+      "step": 88760
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 1.3786740084827742e-05,
+      "loss": 0.3194,
+      "step": 88770
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.378376367289233e-05,
+      "loss": 0.3206,
+      "step": 88780
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.3780787260956919e-05,
+      "loss": 0.2961,
+      "step": 88790
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.3777810849021507e-05,
+      "loss": 0.3018,
+      "step": 88800
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.3774834437086093e-05,
+      "loss": 0.3197,
+      "step": 88810
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 1.3771858025150681e-05,
+      "loss": 0.2909,
+      "step": 88820
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.376888161321527e-05,
+      "loss": 0.3368,
+      "step": 88830
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.3765905201279858e-05,
+      "loss": 0.3103,
+      "step": 88840
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.3762928789344447e-05,
+      "loss": 0.3147,
+      "step": 88850
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 1.3759952377409034e-05,
+      "loss": 0.312,
+      "step": 88860
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.3756975965473622e-05,
+      "loss": 0.3098,
+      "step": 88870
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.3753999553538211e-05,
+      "loss": 0.3094,
+      "step": 88880
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.37510231416028e-05,
+      "loss": 0.3095,
+      "step": 88890
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.3748046729667388e-05,
+      "loss": 0.3086,
+      "step": 88900
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 1.3745070317731975e-05,
+      "loss": 0.3098,
+      "step": 88910
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.3742093905796564e-05,
+      "loss": 0.2945,
+      "step": 88920
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.3739117493861152e-05,
+      "loss": 0.3143,
+      "step": 88930
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.3736141081925741e-05,
+      "loss": 0.2968,
+      "step": 88940
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 1.373316466999033e-05,
+      "loss": 0.3124,
+      "step": 88950
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.3730188258054915e-05,
+      "loss": 0.3058,
+      "step": 88960
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.3727211846119503e-05,
+      "loss": 0.3178,
+      "step": 88970
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.3724235434184092e-05,
+      "loss": 0.3055,
+      "step": 88980
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.372125902224868e-05,
+      "loss": 0.3223,
+      "step": 88990
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 1.3718282610313269e-05,
+      "loss": 0.3023,
+      "step": 89000
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.3715306198377856e-05,
+      "loss": 0.312,
+      "step": 89010
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.3712329786442445e-05,
+      "loss": 0.2995,
+      "step": 89020
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.3709353374507033e-05,
+      "loss": 0.2998,
+      "step": 89030
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 1.3706376962571622e-05,
+      "loss": 0.3126,
+      "step": 89040
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.370340055063621e-05,
+      "loss": 0.2995,
+      "step": 89050
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.3700424138700796e-05,
+      "loss": 0.3219,
+      "step": 89060
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.3697447726765384e-05,
+      "loss": 0.3119,
+      "step": 89070
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.3694471314829973e-05,
+      "loss": 0.3061,
+      "step": 89080
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 1.3691494902894561e-05,
+      "loss": 0.3001,
+      "step": 89090
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.368851849095915e-05,
+      "loss": 0.3214,
+      "step": 89100
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.3685542079023737e-05,
+      "loss": 0.2924,
+      "step": 89110
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.3682565667088325e-05,
+      "loss": 0.3167,
+      "step": 89120
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 1.3679589255152914e-05,
+      "loss": 0.3019,
+      "step": 89130
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.3676612843217503e-05,
+      "loss": 0.3159,
+      "step": 89140
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.3673636431282091e-05,
+      "loss": 0.3058,
+      "step": 89150
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.3670660019346678e-05,
+      "loss": 0.3068,
+      "step": 89160
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.3667683607411267e-05,
+      "loss": 0.3012,
+      "step": 89170
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.3664707195475855e-05,
+      "loss": 0.2993,
+      "step": 89180
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.3661730783540444e-05,
+      "loss": 0.319,
+      "step": 89190
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.3658754371605032e-05,
+      "loss": 0.2861,
+      "step": 89200
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.3655777959669621e-05,
+      "loss": 0.3031,
+      "step": 89210
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.3652801547734206e-05,
+      "loss": 0.2957,
+      "step": 89220
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.3649825135798795e-05,
+      "loss": 0.3223,
+      "step": 89230
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.3646848723863383e-05,
+      "loss": 0.3006,
+      "step": 89240
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.3643872311927972e-05,
+      "loss": 0.3077,
+      "step": 89250
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.364089589999256e-05,
+      "loss": 0.3109,
+      "step": 89260
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.3637919488057147e-05,
+      "loss": 0.3059,
+      "step": 89270
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.3634943076121736e-05,
+      "loss": 0.3169,
+      "step": 89280
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.3631966664186325e-05,
+      "loss": 0.3119,
+      "step": 89290
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.3628990252250913e-05,
+      "loss": 0.3031,
+      "step": 89300
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.3626013840315502e-05,
+      "loss": 0.2973,
+      "step": 89310
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.3623037428380089e-05,
+      "loss": 0.3106,
+      "step": 89320
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.3620061016444677e-05,
+      "loss": 0.3197,
+      "step": 89330
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.3617084604509266e-05,
+      "loss": 0.2969,
+      "step": 89340
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.3614108192573855e-05,
+      "loss": 0.3042,
+      "step": 89350
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.3611131780638443e-05,
+      "loss": 0.3028,
+      "step": 89360
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.3608155368703028e-05,
+      "loss": 0.3123,
+      "step": 89370
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.3605178956767617e-05,
+      "loss": 0.3207,
+      "step": 89380
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.3602202544832206e-05,
+      "loss": 0.3139,
+      "step": 89390
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.3599226132896794e-05,
+      "loss": 0.3161,
+      "step": 89400
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.3596249720961383e-05,
+      "loss": 0.3184,
+      "step": 89410
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.359327330902597e-05,
+      "loss": 0.3016,
+      "step": 89420
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.3590296897090558e-05,
+      "loss": 0.3093,
+      "step": 89430
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.3587320485155147e-05,
+      "loss": 0.3051,
+      "step": 89440
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.3584344073219735e-05,
+      "loss": 0.298,
+      "step": 89450
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.3581367661284324e-05,
+      "loss": 0.3081,
+      "step": 89460
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.3578391249348911e-05,
+      "loss": 0.3003,
+      "step": 89470
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.35754148374135e-05,
+      "loss": 0.3124,
+      "step": 89480
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 1.3572438425478086e-05,
+      "loss": 0.3047,
+      "step": 89490
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.3569462013542675e-05,
+      "loss": 0.2928,
+      "step": 89500
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.3566485601607264e-05,
+      "loss": 0.2978,
+      "step": 89510
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.356350918967185e-05,
+      "loss": 0.2997,
+      "step": 89520
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.3560532777736439e-05,
+      "loss": 0.3121,
+      "step": 89530
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.3557556365801028e-05,
+      "loss": 0.3069,
+      "step": 89540
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.3554579953865616e-05,
+      "loss": 0.3177,
+      "step": 89550
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.3551603541930205e-05,
+      "loss": 0.3042,
+      "step": 89560
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.3548627129994792e-05,
+      "loss": 0.3093,
+      "step": 89570
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 1.354565071805938e-05,
+      "loss": 0.304,
+      "step": 89580
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.3542674306123969e-05,
+      "loss": 0.2984,
+      "step": 89590
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.3539697894188558e-05,
+      "loss": 0.3064,
+      "step": 89600
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.3536721482253146e-05,
+      "loss": 0.315,
+      "step": 89610
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.3533745070317731e-05,
+      "loss": 0.3137,
+      "step": 89620
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.353076865838232e-05,
+      "loss": 0.2963,
+      "step": 89630
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.3527792246446908e-05,
+      "loss": 0.3037,
+      "step": 89640
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.3524815834511497e-05,
+      "loss": 0.3185,
+      "step": 89650
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.3521839422576086e-05,
+      "loss": 0.3076,
+      "step": 89660
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.3518863010640674e-05,
+      "loss": 0.3179,
+      "step": 89670
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3515886598705261e-05,
+      "loss": 0.3018,
+      "step": 89680
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.351291018676985e-05,
+      "loss": 0.2911,
+      "step": 89690
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3509933774834438e-05,
+      "loss": 0.3042,
+      "step": 89700
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3506957362899027e-05,
+      "loss": 0.3081,
+      "step": 89710
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3503980950963616e-05,
+      "loss": 0.304,
+      "step": 89720
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3501004539028202e-05,
+      "loss": 0.3213,
+      "step": 89730
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3498028127092791e-05,
+      "loss": 0.3121,
+      "step": 89740
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.349505171515738e-05,
+      "loss": 0.3127,
+      "step": 89750
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.3492075303221968e-05,
+      "loss": 0.3066,
+      "step": 89760
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.3489098891286557e-05,
+      "loss": 0.3169,
+      "step": 89770
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.3486122479351142e-05,
+      "loss": 0.2959,
+      "step": 89780
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.348314606741573e-05,
+      "loss": 0.3123,
+      "step": 89790
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.348016965548032e-05,
+      "loss": 0.305,
+      "step": 89800
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.3477193243544908e-05,
+      "loss": 0.303,
+      "step": 89810
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3474216831609496e-05,
+      "loss": 0.3145,
+      "step": 89820
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3471240419674083e-05,
+      "loss": 0.2892,
+      "step": 89830
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3468264007738672e-05,
+      "loss": 0.2931,
+      "step": 89840
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.346528759580326e-05,
+      "loss": 0.3165,
+      "step": 89850
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.3462311183867849e-05,
+      "loss": 0.3069,
+      "step": 89860
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.3459334771932438e-05,
+      "loss": 0.3025,
+      "step": 89870
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.3456358359997025e-05,
+      "loss": 0.3174,
+      "step": 89880
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.3453381948061613e-05,
+      "loss": 0.3059,
+      "step": 89890
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.3450405536126202e-05,
+      "loss": 0.2977,
+      "step": 89900
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.344742912419079e-05,
+      "loss": 0.3148,
+      "step": 89910
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.3444452712255377e-05,
+      "loss": 0.3016,
+      "step": 89920
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.3441476300319964e-05,
+      "loss": 0.2984,
+      "step": 89930
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 1.3438499888384553e-05,
+      "loss": 0.2962,
+      "step": 89940
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.3435523476449141e-05,
+      "loss": 0.3114,
+      "step": 89950
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.343254706451373e-05,
+      "loss": 0.2962,
+      "step": 89960
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.3429570652578318e-05,
+      "loss": 0.3172,
+      "step": 89970
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.3426594240642905e-05,
+      "loss": 0.3072,
+      "step": 89980
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 1.3423617828707494e-05,
+      "loss": 0.3106,
+      "step": 89990
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.3420641416772083e-05,
+      "loss": 0.3199,
+      "step": 90000
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.3417665004836671e-05,
+      "loss": 0.3005,
+      "step": 90010
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.341468859290126e-05,
+      "loss": 0.3182,
+      "step": 90020
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.3411712180965845e-05,
+      "loss": 0.2932,
+      "step": 90030
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.3408735769030434e-05,
+      "loss": 0.3104,
+      "step": 90040
+    },
+    {
+      "epoch": 20.0,
+      "eval_cer": 4.555555555555555,
+      "eval_loss": 0.538056492805481,
+      "eval_runtime": 1.2412,
+      "eval_samples_per_second": 8.057,
+      "eval_steps_per_second": 0.806,
+      "eval_wer": 0.6493506493506493,
+      "step": 90040
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.3405759357095022e-05,
+      "loss": 0.3177,
+      "step": 90050
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.340278294515961e-05,
+      "loss": 0.3099,
+      "step": 90060
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.33998065332242e-05,
+      "loss": 0.2884,
+      "step": 90070
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.3396830121288788e-05,
+      "loss": 0.2941,
+      "step": 90080
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.3393853709353375e-05,
+      "loss": 0.3095,
+      "step": 90090
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.3390877297417963e-05,
+      "loss": 0.3004,
+      "step": 90100
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.3387900885482552e-05,
+      "loss": 0.309,
+      "step": 90110
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 1.338492447354714e-05,
+      "loss": 0.3007,
+      "step": 90120
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.338194806161173e-05,
+      "loss": 0.2902,
+      "step": 90130
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.3378971649676316e-05,
+      "loss": 0.3057,
+      "step": 90140
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.3375995237740905e-05,
+      "loss": 0.2889,
+      "step": 90150
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.3373018825805493e-05,
+      "loss": 0.3179,
+      "step": 90160
+    },
+    {
+      "epoch": 20.02,
+      "learning_rate": 1.3370042413870082e-05,
+      "loss": 0.2994,
+      "step": 90170
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.336706600193467e-05,
+      "loss": 0.2928,
+      "step": 90180
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.3364089589999256e-05,
+      "loss": 0.2963,
+      "step": 90190
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.3361113178063844e-05,
+      "loss": 0.3013,
+      "step": 90200
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 1.3358136766128433e-05,
+      "loss": 0.2997,
+      "step": 90210
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.3355160354193021e-05,
+      "loss": 0.3129,
+      "step": 90220
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.335218394225761e-05,
+      "loss": 0.3002,
+      "step": 90230
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.3349207530322197e-05,
+      "loss": 0.2998,
+      "step": 90240
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.3346231118386786e-05,
+      "loss": 0.3004,
+      "step": 90250
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 1.3343254706451374e-05,
+      "loss": 0.2999,
+      "step": 90260
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.3340278294515963e-05,
+      "loss": 0.3025,
+      "step": 90270
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.3337301882580551e-05,
+      "loss": 0.3013,
+      "step": 90280
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.3334325470645138e-05,
+      "loss": 0.2923,
+      "step": 90290
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 1.3331349058709727e-05,
+      "loss": 0.3064,
+      "step": 90300
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.3328372646774315e-05,
+      "loss": 0.3013,
+      "step": 90310
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.3325396234838904e-05,
+      "loss": 0.2988,
+      "step": 90320
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.3322419822903493e-05,
+      "loss": 0.3045,
+      "step": 90330
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.3319443410968078e-05,
+      "loss": 0.3078,
+      "step": 90340
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 1.3316466999032666e-05,
+      "loss": 0.3063,
+      "step": 90350
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.3313490587097255e-05,
+      "loss": 0.3046,
+      "step": 90360
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.3310514175161844e-05,
+      "loss": 0.3029,
+      "step": 90370
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.3307537763226432e-05,
+      "loss": 0.3013,
+      "step": 90380
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 1.3304561351291019e-05,
+      "loss": 0.305,
+      "step": 90390
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.3301584939355608e-05,
+      "loss": 0.2971,
+      "step": 90400
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.3298608527420196e-05,
+      "loss": 0.3021,
+      "step": 90410
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.3295632115484785e-05,
+      "loss": 0.2948,
+      "step": 90420
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.3292655703549373e-05,
+      "loss": 0.3009,
+      "step": 90430
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 1.3289679291613959e-05,
+      "loss": 0.2944,
+      "step": 90440
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.3286702879678547e-05,
+      "loss": 0.3092,
+      "step": 90450
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.3283726467743136e-05,
+      "loss": 0.2909,
+      "step": 90460
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.3280750055807724e-05,
+      "loss": 0.3005,
+      "step": 90470
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 1.3277773643872313e-05,
+      "loss": 0.3067,
+      "step": 90480
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.32747972319369e-05,
+      "loss": 0.2985,
+      "step": 90490
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.3271820820001489e-05,
+      "loss": 0.3047,
+      "step": 90500
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.3268844408066077e-05,
+      "loss": 0.2989,
+      "step": 90510
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.3265867996130666e-05,
+      "loss": 0.2893,
+      "step": 90520
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 1.3262891584195254e-05,
+      "loss": 0.3089,
+      "step": 90530
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.3259915172259843e-05,
+      "loss": 0.3223,
+      "step": 90540
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.325693876032443e-05,
+      "loss": 0.2889,
+      "step": 90550
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.3253962348389018e-05,
+      "loss": 0.2889,
+      "step": 90560
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 1.3250985936453607e-05,
+      "loss": 0.3067,
+      "step": 90570
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.3248009524518196e-05,
+      "loss": 0.2915,
+      "step": 90580
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.3245033112582784e-05,
+      "loss": 0.2984,
+      "step": 90590
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.324205670064737e-05,
+      "loss": 0.3094,
+      "step": 90600
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.3239080288711958e-05,
+      "loss": 0.3071,
+      "step": 90610
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 1.3236103876776547e-05,
+      "loss": 0.3037,
+      "step": 90620
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.3233127464841135e-05,
+      "loss": 0.3119,
+      "step": 90630
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.3230151052905724e-05,
+      "loss": 0.3046,
+      "step": 90640
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.322717464097031e-05,
+      "loss": 0.3082,
+      "step": 90650
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 1.32241982290349e-05,
+      "loss": 0.3086,
+      "step": 90660
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.3221221817099488e-05,
+      "loss": 0.293,
+      "step": 90670
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.3218245405164076e-05,
+      "loss": 0.2942,
+      "step": 90680
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.3215268993228665e-05,
+      "loss": 0.3033,
+      "step": 90690
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.3212292581293252e-05,
+      "loss": 0.3013,
+      "step": 90700
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 1.320931616935784e-05,
+      "loss": 0.308,
+      "step": 90710
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.3206339757422429e-05,
+      "loss": 0.2854,
+      "step": 90720
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.3203363345487018e-05,
+      "loss": 0.3034,
+      "step": 90730
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.3200386933551606e-05,
+      "loss": 0.3083,
+      "step": 90740
+    },
+    {
+      "epoch": 20.15,
+      "learning_rate": 1.3197410521616191e-05,
+      "loss": 0.291,
+      "step": 90750
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.319443410968078e-05,
+      "loss": 0.2947,
+      "step": 90760
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.3191457697745369e-05,
+      "loss": 0.3028,
+      "step": 90770
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.3188481285809957e-05,
+      "loss": 0.3029,
+      "step": 90780
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.3185504873874546e-05,
+      "loss": 0.2982,
+      "step": 90790
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 1.3182528461939133e-05,
+      "loss": 0.3136,
+      "step": 90800
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.3179552050003721e-05,
+      "loss": 0.3045,
+      "step": 90810
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.317657563806831e-05,
+      "loss": 0.31,
+      "step": 90820
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.3173599226132899e-05,
+      "loss": 0.2998,
+      "step": 90830
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 1.3170622814197487e-05,
+      "loss": 0.2966,
+      "step": 90840
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.3167646402262072e-05,
+      "loss": 0.3077,
+      "step": 90850
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.3164669990326661e-05,
+      "loss": 0.3049,
+      "step": 90860
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.316169357839125e-05,
+      "loss": 0.311,
+      "step": 90870
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.3158717166455838e-05,
+      "loss": 0.3101,
+      "step": 90880
+    },
+    {
+      "epoch": 20.18,
+      "learning_rate": 1.3155740754520427e-05,
+      "loss": 0.2943,
+      "step": 90890
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.3152764342585014e-05,
+      "loss": 0.2931,
+      "step": 90900
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.3149787930649602e-05,
+      "loss": 0.2922,
+      "step": 90910
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.314681151871419e-05,
+      "loss": 0.2999,
+      "step": 90920
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 1.314383510677878e-05,
+      "loss": 0.309,
+      "step": 90930
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.3140858694843368e-05,
+      "loss": 0.3136,
+      "step": 90940
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.3137882282907957e-05,
+      "loss": 0.2973,
+      "step": 90950
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.3134905870972543e-05,
+      "loss": 0.3128,
+      "step": 90960
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.3131929459037132e-05,
+      "loss": 0.3151,
+      "step": 90970
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 1.312895304710172e-05,
+      "loss": 0.3054,
+      "step": 90980
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.312597663516631e-05,
+      "loss": 0.3178,
+      "step": 90990
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.3123000223230898e-05,
+      "loss": 0.2897,
+      "step": 91000
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.3120023811295483e-05,
+      "loss": 0.2959,
+      "step": 91010
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 1.3117047399360072e-05,
+      "loss": 0.3018,
+      "step": 91020
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.311407098742466e-05,
+      "loss": 0.296,
+      "step": 91030
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.3111094575489249e-05,
+      "loss": 0.3093,
+      "step": 91040
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.3108118163553837e-05,
+      "loss": 0.3039,
+      "step": 91050
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.3105141751618424e-05,
+      "loss": 0.3052,
+      "step": 91060
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 1.3102165339683013e-05,
+      "loss": 0.3013,
+      "step": 91070
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.3099188927747601e-05,
+      "loss": 0.2893,
+      "step": 91080
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.309621251581219e-05,
+      "loss": 0.3045,
+      "step": 91090
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.3093236103876779e-05,
+      "loss": 0.2913,
+      "step": 91100
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 1.3090259691941366e-05,
+      "loss": 0.3002,
+      "step": 91110
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.3087283280005954e-05,
+      "loss": 0.3009,
+      "step": 91120
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.3084306868070543e-05,
+      "loss": 0.2895,
+      "step": 91130
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.3081330456135131e-05,
+      "loss": 0.2908,
+      "step": 91140
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.307835404419972e-05,
+      "loss": 0.3016,
+      "step": 91150
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 1.3075377632264305e-05,
+      "loss": 0.3014,
+      "step": 91160
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.3072401220328894e-05,
+      "loss": 0.3036,
+      "step": 91170
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.3069424808393482e-05,
+      "loss": 0.2997,
+      "step": 91180
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.3066448396458071e-05,
+      "loss": 0.2965,
+      "step": 91190
+    },
+    {
+      "epoch": 20.25,
+      "learning_rate": 1.306347198452266e-05,
+      "loss": 0.2981,
+      "step": 91200
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.3060495572587246e-05,
+      "loss": 0.2915,
+      "step": 91210
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.3057519160651835e-05,
+      "loss": 0.3028,
+      "step": 91220
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.3054542748716424e-05,
+      "loss": 0.2941,
+      "step": 91230
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.3051566336781012e-05,
+      "loss": 0.3028,
+      "step": 91240
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 1.30485899248456e-05,
+      "loss": 0.3115,
+      "step": 91250
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.3045613512910188e-05,
+      "loss": 0.3027,
+      "step": 91260
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.3042637100974775e-05,
+      "loss": 0.3085,
+      "step": 91270
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.3039660689039363e-05,
+      "loss": 0.3031,
+      "step": 91280
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 1.3036684277103952e-05,
+      "loss": 0.3058,
+      "step": 91290
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.303370786516854e-05,
+      "loss": 0.3022,
+      "step": 91300
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.3030731453233127e-05,
+      "loss": 0.2967,
+      "step": 91310
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.3027755041297716e-05,
+      "loss": 0.3206,
+      "step": 91320
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.3024778629362304e-05,
+      "loss": 0.3054,
+      "step": 91330
+    },
+    {
+      "epoch": 20.28,
+      "learning_rate": 1.3021802217426893e-05,
+      "loss": 0.298,
+      "step": 91340
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.3018825805491482e-05,
+      "loss": 0.2989,
+      "step": 91350
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.3015849393556069e-05,
+      "loss": 0.3071,
+      "step": 91360
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.3012872981620657e-05,
+      "loss": 0.2976,
+      "step": 91370
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 1.3009896569685246e-05,
+      "loss": 0.2978,
+      "step": 91380
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.3006920157749834e-05,
+      "loss": 0.2924,
+      "step": 91390
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.3003943745814423e-05,
+      "loss": 0.2944,
+      "step": 91400
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.3000967333879012e-05,
+      "loss": 0.3024,
+      "step": 91410
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.2997990921943597e-05,
+      "loss": 0.2927,
+      "step": 91420
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 1.2995014510008185e-05,
+      "loss": 0.3066,
+      "step": 91430
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.2992038098072774e-05,
+      "loss": 0.3007,
+      "step": 91440
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.2989061686137362e-05,
+      "loss": 0.2852,
+      "step": 91450
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.2986085274201951e-05,
+      "loss": 0.3009,
+      "step": 91460
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 1.2983108862266538e-05,
+      "loss": 0.302,
+      "step": 91470
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.2980132450331127e-05,
+      "loss": 0.2866,
+      "step": 91480
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.2977156038395715e-05,
+      "loss": 0.2885,
+      "step": 91490
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.2974179626460304e-05,
+      "loss": 0.3162,
+      "step": 91500
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.2971203214524892e-05,
+      "loss": 0.2997,
+      "step": 91510
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 1.296822680258948e-05,
+      "loss": 0.3034,
+      "step": 91520
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.2965250390654068e-05,
+      "loss": 0.2895,
+      "step": 91530
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.2962273978718656e-05,
+      "loss": 0.3049,
+      "step": 91540
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.2959297566783245e-05,
+      "loss": 0.293,
+      "step": 91550
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 1.2956321154847834e-05,
+      "loss": 0.3062,
+      "step": 91560
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.2953344742912419e-05,
+      "loss": 0.295,
+      "step": 91570
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.2950368330977007e-05,
+      "loss": 0.3063,
+      "step": 91580
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.2947391919041596e-05,
+      "loss": 0.3054,
+      "step": 91590
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.2944415507106185e-05,
+      "loss": 0.2869,
+      "step": 91600
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 1.2941439095170773e-05,
+      "loss": 0.311,
+      "step": 91610
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.293846268323536e-05,
+      "loss": 0.3054,
+      "step": 91620
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.2935486271299949e-05,
+      "loss": 0.3008,
+      "step": 91630
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.2932509859364537e-05,
+      "loss": 0.2979,
+      "step": 91640
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 1.2929533447429126e-05,
+      "loss": 0.2952,
+      "step": 91650
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.2926557035493714e-05,
+      "loss": 0.2897,
+      "step": 91660
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.2923580623558301e-05,
+      "loss": 0.3057,
+      "step": 91670
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.292060421162289e-05,
+      "loss": 0.2868,
+      "step": 91680
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.2917627799687477e-05,
+      "loss": 0.3025,
+      "step": 91690
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 1.2914651387752065e-05,
+      "loss": 0.3011,
+      "step": 91700
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.2911674975816654e-05,
+      "loss": 0.2937,
+      "step": 91710
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.2908698563881241e-05,
+      "loss": 0.3035,
+      "step": 91720
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.290572215194583e-05,
+      "loss": 0.3029,
+      "step": 91730
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 1.2902745740010418e-05,
+      "loss": 0.2981,
+      "step": 91740
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.2899769328075007e-05,
+      "loss": 0.2958,
+      "step": 91750
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.2896792916139595e-05,
+      "loss": 0.2973,
+      "step": 91760
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.2893816504204182e-05,
+      "loss": 0.3011,
+      "step": 91770
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.289084009226877e-05,
+      "loss": 0.2946,
+      "step": 91780
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 1.288786368033336e-05,
+      "loss": 0.2997,
+      "step": 91790
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.2884887268397948e-05,
+      "loss": 0.2989,
+      "step": 91800
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.2881910856462537e-05,
+      "loss": 0.3051,
+      "step": 91810
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.2878934444527125e-05,
+      "loss": 0.3114,
+      "step": 91820
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 1.287595803259171e-05,
+      "loss": 0.3125,
+      "step": 91830
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.2872981620656299e-05,
+      "loss": 0.2965,
+      "step": 91840
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.2870005208720888e-05,
+      "loss": 0.2866,
+      "step": 91850
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.2867028796785476e-05,
+      "loss": 0.2986,
+      "step": 91860
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.2864052384850065e-05,
+      "loss": 0.2921,
+      "step": 91870
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 1.2861075972914652e-05,
+      "loss": 0.3039,
+      "step": 91880
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.285809956097924e-05,
+      "loss": 0.2991,
+      "step": 91890
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.2855123149043829e-05,
+      "loss": 0.3042,
+      "step": 91900
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.2852146737108417e-05,
+      "loss": 0.2954,
+      "step": 91910
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 1.2849170325173006e-05,
+      "loss": 0.3019,
+      "step": 91920
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.2846193913237593e-05,
+      "loss": 0.2837,
+      "step": 91930
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.2843217501302182e-05,
+      "loss": 0.2998,
+      "step": 91940
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.284024108936677e-05,
+      "loss": 0.3067,
+      "step": 91950
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.2837264677431359e-05,
+      "loss": 0.299,
+      "step": 91960
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 1.2834288265495947e-05,
+      "loss": 0.3045,
+      "step": 91970
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.2831311853560533e-05,
+      "loss": 0.3079,
+      "step": 91980
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.2828335441625121e-05,
+      "loss": 0.2936,
+      "step": 91990
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.282535902968971e-05,
+      "loss": 0.3054,
+      "step": 92000
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 1.2822382617754298e-05,
+      "loss": 0.2995,
+      "step": 92010
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.2819406205818887e-05,
+      "loss": 0.2951,
+      "step": 92020
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.2816429793883474e-05,
+      "loss": 0.2953,
+      "step": 92030
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.2813453381948062e-05,
+      "loss": 0.2953,
+      "step": 92040
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.2810476970012651e-05,
+      "loss": 0.3041,
+      "step": 92050
+    },
+    {
+      "epoch": 20.44,
+      "learning_rate": 1.280750055807724e-05,
+      "loss": 0.3143,
+      "step": 92060
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.2804524146141828e-05,
+      "loss": 0.3002,
+      "step": 92070
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.2801547734206415e-05,
+      "loss": 0.3159,
+      "step": 92080
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.2798571322271004e-05,
+      "loss": 0.3105,
+      "step": 92090
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 1.2795594910335592e-05,
+      "loss": 0.309,
+      "step": 92100
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.279261849840018e-05,
+      "loss": 0.3086,
+      "step": 92110
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.2789642086464768e-05,
+      "loss": 0.3038,
+      "step": 92120
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.2786665674529355e-05,
+      "loss": 0.2953,
+      "step": 92130
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.2783689262593943e-05,
+      "loss": 0.2903,
+      "step": 92140
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 1.2780712850658532e-05,
+      "loss": 0.3005,
+      "step": 92150
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.277773643872312e-05,
+      "loss": 0.3089,
+      "step": 92160
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.2774760026787709e-05,
+      "loss": 0.2967,
+      "step": 92170
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.2771783614852296e-05,
+      "loss": 0.3145,
+      "step": 92180
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 1.2768807202916884e-05,
+      "loss": 0.301,
+      "step": 92190
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.2765830790981473e-05,
+      "loss": 0.3042,
+      "step": 92200
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.2762854379046062e-05,
+      "loss": 0.2985,
+      "step": 92210
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.275987796711065e-05,
+      "loss": 0.2996,
+      "step": 92220
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.2756901555175235e-05,
+      "loss": 0.2937,
+      "step": 92230
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 1.2753925143239824e-05,
+      "loss": 0.3048,
+      "step": 92240
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.2750948731304413e-05,
+      "loss": 0.2999,
+      "step": 92250
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.2747972319369001e-05,
+      "loss": 0.2956,
+      "step": 92260
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.274499590743359e-05,
+      "loss": 0.311,
+      "step": 92270
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 1.2742019495498178e-05,
+      "loss": 0.3016,
+      "step": 92280
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.2739043083562765e-05,
+      "loss": 0.297,
+      "step": 92290
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.2736066671627354e-05,
+      "loss": 0.2977,
+      "step": 92300
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.2733090259691943e-05,
+      "loss": 0.3145,
+      "step": 92310
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.2730113847756531e-05,
+      "loss": 0.3058,
+      "step": 92320
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 1.272713743582112e-05,
+      "loss": 0.2968,
+      "step": 92330
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.2724161023885707e-05,
+      "loss": 0.2974,
+      "step": 92340
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.2721184611950295e-05,
+      "loss": 0.3198,
+      "step": 92350
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.2718208200014884e-05,
+      "loss": 0.2953,
+      "step": 92360
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 1.2715231788079472e-05,
+      "loss": 0.3105,
+      "step": 92370
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.2712255376144061e-05,
+      "loss": 0.3047,
+      "step": 92380
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.2709278964208646e-05,
+      "loss": 0.2921,
+      "step": 92390
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.2706302552273235e-05,
+      "loss": 0.2928,
+      "step": 92400
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.2703326140337823e-05,
+      "loss": 0.3126,
+      "step": 92410
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 1.2700349728402412e-05,
+      "loss": 0.2976,
+      "step": 92420
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.2697373316467e-05,
+      "loss": 0.2981,
+      "step": 92430
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.2694396904531587e-05,
+      "loss": 0.2993,
+      "step": 92440
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.2691420492596176e-05,
+      "loss": 0.2956,
+      "step": 92450
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 1.2688444080660765e-05,
+      "loss": 0.3052,
+      "step": 92460
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.2685467668725353e-05,
+      "loss": 0.2998,
+      "step": 92470
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.2682491256789942e-05,
+      "loss": 0.3004,
+      "step": 92480
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.2679514844854529e-05,
+      "loss": 0.3009,
+      "step": 92490
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.2676538432919117e-05,
+      "loss": 0.3076,
+      "step": 92500
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 1.2673562020983706e-05,
+      "loss": 0.2986,
+      "step": 92510
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.2670585609048295e-05,
+      "loss": 0.2913,
+      "step": 92520
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.2667609197112883e-05,
+      "loss": 0.3021,
+      "step": 92530
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.2664632785177468e-05,
+      "loss": 0.2954,
+      "step": 92540
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 1.2661656373242057e-05,
+      "loss": 0.3186,
+      "step": 92550
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.2658679961306645e-05,
+      "loss": 0.2847,
+      "step": 92560
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.2655703549371234e-05,
+      "loss": 0.3115,
+      "step": 92570
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.2652727137435823e-05,
+      "loss": 0.303,
+      "step": 92580
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.264975072550041e-05,
+      "loss": 0.2984,
+      "step": 92590
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 1.2646774313564998e-05,
+      "loss": 0.2972,
+      "step": 92600
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.2643797901629587e-05,
+      "loss": 0.2914,
+      "step": 92610
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.2640821489694175e-05,
+      "loss": 0.3004,
+      "step": 92620
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.2637845077758764e-05,
+      "loss": 0.3089,
+      "step": 92630
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 1.2634868665823349e-05,
+      "loss": 0.3068,
+      "step": 92640
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.2631892253887938e-05,
+      "loss": 0.2924,
+      "step": 92650
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.2628915841952526e-05,
+      "loss": 0.2917,
+      "step": 92660
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.2625939430017115e-05,
+      "loss": 0.2943,
+      "step": 92670
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.2622963018081704e-05,
+      "loss": 0.3012,
+      "step": 92680
+    },
+    {
+      "epoch": 20.58,
+      "learning_rate": 1.2619986606146292e-05,
+      "loss": 0.2891,
+      "step": 92690
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.2617010194210879e-05,
+      "loss": 0.2987,
+      "step": 92700
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.2614033782275468e-05,
+      "loss": 0.2946,
+      "step": 92710
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.2611057370340056e-05,
+      "loss": 0.299,
+      "step": 92720
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 1.2608080958404645e-05,
+      "loss": 0.3197,
+      "step": 92730
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.2605104546469233e-05,
+      "loss": 0.2956,
+      "step": 92740
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.260212813453382e-05,
+      "loss": 0.2973,
+      "step": 92750
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.2599151722598409e-05,
+      "loss": 0.2934,
+      "step": 92760
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.2596175310662997e-05,
+      "loss": 0.3074,
+      "step": 92770
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 1.2593198898727586e-05,
+      "loss": 0.2967,
+      "step": 92780
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.2590222486792175e-05,
+      "loss": 0.2918,
+      "step": 92790
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.258724607485676e-05,
+      "loss": 0.3059,
+      "step": 92800
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.2584269662921348e-05,
+      "loss": 0.2978,
+      "step": 92810
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 1.2581293250985937e-05,
+      "loss": 0.2916,
+      "step": 92820
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.2578316839050526e-05,
+      "loss": 0.3117,
+      "step": 92830
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.2575340427115114e-05,
+      "loss": 0.3018,
+      "step": 92840
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.2572364015179701e-05,
+      "loss": 0.2997,
+      "step": 92850
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.256938760324429e-05,
+      "loss": 0.3057,
+      "step": 92860
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 1.2566411191308878e-05,
+      "loss": 0.3014,
+      "step": 92870
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.2563434779373467e-05,
+      "loss": 0.3115,
+      "step": 92880
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.2560458367438055e-05,
+      "loss": 0.3013,
+      "step": 92890
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.2557481955502642e-05,
+      "loss": 0.2914,
+      "step": 92900
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 1.2554505543567231e-05,
+      "loss": 0.3029,
+      "step": 92910
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.255152913163182e-05,
+      "loss": 0.2944,
+      "step": 92920
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.2548552719696408e-05,
+      "loss": 0.3104,
+      "step": 92930
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.2545576307760997e-05,
+      "loss": 0.3118,
+      "step": 92940
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.2542599895825582e-05,
+      "loss": 0.2855,
+      "step": 92950
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 1.253962348389017e-05,
+      "loss": 0.3038,
+      "step": 92960
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.253664707195476e-05,
+      "loss": 0.3099,
+      "step": 92970
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.2533670660019348e-05,
+      "loss": 0.3102,
+      "step": 92980
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.2530694248083936e-05,
+      "loss": 0.2898,
+      "step": 92990
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 1.2527717836148523e-05,
+      "loss": 0.3096,
+      "step": 93000
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.2524741424213112e-05,
+      "loss": 0.3043,
+      "step": 93010
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.25217650122777e-05,
+      "loss": 0.314,
+      "step": 93020
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.2518788600342289e-05,
+      "loss": 0.3018,
+      "step": 93030
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.2515812188406878e-05,
+      "loss": 0.3076,
+      "step": 93040
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 1.2512835776471463e-05,
+      "loss": 0.2916,
+      "step": 93050
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.2509859364536051e-05,
+      "loss": 0.301,
+      "step": 93060
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.250688295260064e-05,
+      "loss": 0.2956,
+      "step": 93070
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.2503906540665229e-05,
+      "loss": 0.2895,
+      "step": 93080
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 1.2500930128729817e-05,
+      "loss": 0.3129,
+      "step": 93090
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.2497953716794404e-05,
+      "loss": 0.3003,
+      "step": 93100
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.2494977304858993e-05,
+      "loss": 0.2933,
+      "step": 93110
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.2492000892923581e-05,
+      "loss": 0.3025,
+      "step": 93120
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.248902448098817e-05,
+      "loss": 0.2996,
+      "step": 93130
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 1.2486048069052758e-05,
+      "loss": 0.3085,
+      "step": 93140
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.2483071657117347e-05,
+      "loss": 0.3097,
+      "step": 93150
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.2480095245181934e-05,
+      "loss": 0.2961,
+      "step": 93160
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.2477118833246523e-05,
+      "loss": 0.307,
+      "step": 93170
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 1.2474142421311111e-05,
+      "loss": 0.2961,
+      "step": 93180
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.24711660093757e-05,
+      "loss": 0.3113,
+      "step": 93190
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.2468189597440288e-05,
+      "loss": 0.3068,
+      "step": 93200
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.2465213185504874e-05,
+      "loss": 0.3004,
+      "step": 93210
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.2462236773569462e-05,
+      "loss": 0.2969,
+      "step": 93220
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 1.245926036163405e-05,
+      "loss": 0.2947,
+      "step": 93230
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.245628394969864e-05,
+      "loss": 0.3069,
+      "step": 93240
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.2453307537763228e-05,
+      "loss": 0.2961,
+      "step": 93250
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.2450331125827815e-05,
+      "loss": 0.2912,
+      "step": 93260
+    },
+    {
+      "epoch": 20.71,
+      "learning_rate": 1.2447354713892403e-05,
+      "loss": 0.2879,
+      "step": 93270
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.2444378301956992e-05,
+      "loss": 0.3091,
+      "step": 93280
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.244140189002158e-05,
+      "loss": 0.3048,
+      "step": 93290
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.243842547808617e-05,
+      "loss": 0.296,
+      "step": 93300
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.2435449066150756e-05,
+      "loss": 0.3041,
+      "step": 93310
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 1.2432472654215345e-05,
+      "loss": 0.3005,
+      "step": 93320
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.2429496242279933e-05,
+      "loss": 0.3004,
+      "step": 93330
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.2426519830344522e-05,
+      "loss": 0.3032,
+      "step": 93340
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.242354341840911e-05,
+      "loss": 0.2875,
+      "step": 93350
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 1.2420567006473696e-05,
+      "loss": 0.2946,
+      "step": 93360
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.2417590594538284e-05,
+      "loss": 0.3072,
+      "step": 93370
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.2414614182602873e-05,
+      "loss": 0.3117,
+      "step": 93380
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.2411637770667461e-05,
+      "loss": 0.3016,
+      "step": 93390
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.240866135873205e-05,
+      "loss": 0.3011,
+      "step": 93400
+    },
+    {
+      "epoch": 20.74,
+      "learning_rate": 1.2405684946796637e-05,
+      "loss": 0.2961,
+      "step": 93410
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.2402708534861226e-05,
+      "loss": 0.3039,
+      "step": 93420
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.2399732122925814e-05,
+      "loss": 0.3142,
+      "step": 93430
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.2396755710990403e-05,
+      "loss": 0.2967,
+      "step": 93440
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 1.2393779299054991e-05,
+      "loss": 0.3042,
+      "step": 93450
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.2390802887119578e-05,
+      "loss": 0.2935,
+      "step": 93460
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.2387826475184165e-05,
+      "loss": 0.296,
+      "step": 93470
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.2384850063248754e-05,
+      "loss": 0.3074,
+      "step": 93480
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.2381873651313342e-05,
+      "loss": 0.3169,
+      "step": 93490
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 1.2378897239377931e-05,
+      "loss": 0.3129,
+      "step": 93500
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.2375920827442518e-05,
+      "loss": 0.3099,
+      "step": 93510
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.2372944415507106e-05,
+      "loss": 0.3053,
+      "step": 93520
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.2369968003571695e-05,
+      "loss": 0.2924,
+      "step": 93530
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 1.2366991591636284e-05,
+      "loss": 0.3009,
+      "step": 93540
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.2364015179700872e-05,
+      "loss": 0.3028,
+      "step": 93550
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.236103876776546e-05,
+      "loss": 0.2968,
+      "step": 93560
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.2358062355830048e-05,
+      "loss": 0.303,
+      "step": 93570
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.2355085943894636e-05,
+      "loss": 0.2997,
+      "step": 93580
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 1.2352109531959225e-05,
+      "loss": 0.2869,
+      "step": 93590
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.2349133120023813e-05,
+      "loss": 0.293,
+      "step": 93600
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.2346156708088402e-05,
+      "loss": 0.3017,
+      "step": 93610
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.2343180296152987e-05,
+      "loss": 0.2993,
+      "step": 93620
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 1.2340203884217576e-05,
+      "loss": 0.2971,
+      "step": 93630
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.2337227472282164e-05,
+      "loss": 0.3061,
+      "step": 93640
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.2334251060346753e-05,
+      "loss": 0.3173,
+      "step": 93650
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.2331274648411342e-05,
+      "loss": 0.3136,
+      "step": 93660
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.2328298236475928e-05,
+      "loss": 0.3099,
+      "step": 93670
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 1.2325321824540517e-05,
+      "loss": 0.3033,
+      "step": 93680
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.2322345412605106e-05,
+      "loss": 0.3024,
+      "step": 93690
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.2319369000669694e-05,
+      "loss": 0.3112,
+      "step": 93700
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.2316392588734283e-05,
+      "loss": 0.3064,
+      "step": 93710
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 1.231341617679887e-05,
+      "loss": 0.2961,
+      "step": 93720
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.2310439764863458e-05,
+      "loss": 0.3095,
+      "step": 93730
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.2307463352928047e-05,
+      "loss": 0.3101,
+      "step": 93740
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.2304486940992636e-05,
+      "loss": 0.303,
+      "step": 93750
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.2301510529057224e-05,
+      "loss": 0.3028,
+      "step": 93760
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 1.229853411712181e-05,
+      "loss": 0.3067,
+      "step": 93770
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.2295557705186398e-05,
+      "loss": 0.2906,
+      "step": 93780
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.2292581293250987e-05,
+      "loss": 0.294,
+      "step": 93790
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.2289604881315575e-05,
+      "loss": 0.3014,
+      "step": 93800
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.2286628469380164e-05,
+      "loss": 0.3009,
+      "step": 93810
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 1.228365205744475e-05,
+      "loss": 0.2964,
+      "step": 93820
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.228067564550934e-05,
+      "loss": 0.3116,
+      "step": 93830
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.2277699233573928e-05,
+      "loss": 0.2883,
+      "step": 93840
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.2274722821638516e-05,
+      "loss": 0.3009,
+      "step": 93850
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 1.2271746409703105e-05,
+      "loss": 0.3152,
+      "step": 93860
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.2268769997767692e-05,
+      "loss": 0.3026,
+      "step": 93870
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.226579358583228e-05,
+      "loss": 0.3007,
+      "step": 93880
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.2262817173896869e-05,
+      "loss": 0.3037,
+      "step": 93890
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.2259840761961456e-05,
+      "loss": 0.2999,
+      "step": 93900
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 1.2256864350026045e-05,
+      "loss": 0.295,
+      "step": 93910
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.2253887938090631e-05,
+      "loss": 0.3068,
+      "step": 93920
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.225091152615522e-05,
+      "loss": 0.305,
+      "step": 93930
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.2247935114219809e-05,
+      "loss": 0.2911,
+      "step": 93940
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 1.2244958702284397e-05,
+      "loss": 0.3016,
+      "step": 93950
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.2241982290348986e-05,
+      "loss": 0.3116,
+      "step": 93960
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.2239005878413573e-05,
+      "loss": 0.2899,
+      "step": 93970
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.2236029466478161e-05,
+      "loss": 0.2989,
+      "step": 93980
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.223305305454275e-05,
+      "loss": 0.3054,
+      "step": 93990
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 1.2230076642607338e-05,
+      "loss": 0.3123,
+      "step": 94000
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.2227100230671927e-05,
+      "loss": 0.2909,
+      "step": 94010
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.2224123818736516e-05,
+      "loss": 0.2965,
+      "step": 94020
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.2221147406801101e-05,
+      "loss": 0.3155,
+      "step": 94030
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 1.221817099486569e-05,
+      "loss": 0.3008,
+      "step": 94040
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.2215194582930278e-05,
+      "loss": 0.2977,
+      "step": 94050
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.2212218170994867e-05,
+      "loss": 0.2994,
+      "step": 94060
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.2209241759059455e-05,
+      "loss": 0.3211,
+      "step": 94070
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.2206265347124042e-05,
+      "loss": 0.3089,
+      "step": 94080
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 1.220328893518863e-05,
+      "loss": 0.3055,
+      "step": 94090
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.220031252325322e-05,
+      "loss": 0.3073,
+      "step": 94100
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.2197336111317808e-05,
+      "loss": 0.3116,
+      "step": 94110
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.2194359699382397e-05,
+      "loss": 0.2913,
+      "step": 94120
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 1.2191383287446983e-05,
+      "loss": 0.3077,
+      "step": 94130
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.2188406875511572e-05,
+      "loss": 0.2991,
+      "step": 94140
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.218543046357616e-05,
+      "loss": 0.3002,
+      "step": 94150
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.218245405164075e-05,
+      "loss": 0.2963,
+      "step": 94160
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.2179477639705338e-05,
+      "loss": 0.3114,
+      "step": 94170
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 1.2176501227769923e-05,
+      "loss": 0.3053,
+      "step": 94180
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.2173524815834512e-05,
+      "loss": 0.3192,
+      "step": 94190
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.21705484038991e-05,
+      "loss": 0.2929,
+      "step": 94200
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.2167571991963689e-05,
+      "loss": 0.2972,
+      "step": 94210
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 1.2164595580028277e-05,
+      "loss": 0.3095,
+      "step": 94220
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.2161619168092864e-05,
+      "loss": 0.2982,
+      "step": 94230
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.2158642756157453e-05,
+      "loss": 0.2945,
+      "step": 94240
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.2155666344222041e-05,
+      "loss": 0.3075,
+      "step": 94250
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.215268993228663e-05,
+      "loss": 0.3093,
+      "step": 94260
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 1.2149713520351219e-05,
+      "loss": 0.3063,
+      "step": 94270
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.2146737108415806e-05,
+      "loss": 0.2904,
+      "step": 94280
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.2143760696480394e-05,
+      "loss": 0.3029,
+      "step": 94290
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.2140784284544983e-05,
+      "loss": 0.3156,
+      "step": 94300
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 1.2137807872609571e-05,
+      "loss": 0.3016,
+      "step": 94310
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.213483146067416e-05,
+      "loss": 0.2942,
+      "step": 94320
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.2131855048738745e-05,
+      "loss": 0.3086,
+      "step": 94330
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.2128878636803334e-05,
+      "loss": 0.3125,
+      "step": 94340
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.2125902224867922e-05,
+      "loss": 0.2987,
+      "step": 94350
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 1.2122925812932511e-05,
+      "loss": 0.2901,
+      "step": 94360
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.21199494009971e-05,
+      "loss": 0.3046,
+      "step": 94370
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.2116972989061686e-05,
+      "loss": 0.3089,
+      "step": 94380
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.2113996577126275e-05,
+      "loss": 0.3062,
+      "step": 94390
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 1.2111020165190864e-05,
+      "loss": 0.2903,
+      "step": 94400
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.2108043753255452e-05,
+      "loss": 0.3154,
+      "step": 94410
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.210506734132004e-05,
+      "loss": 0.2941,
+      "step": 94420
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.210209092938463e-05,
+      "loss": 0.301,
+      "step": 94430
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.2099114517449215e-05,
+      "loss": 0.3038,
+      "step": 94440
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 1.2096138105513803e-05,
+      "loss": 0.3149,
+      "step": 94450
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.2093161693578392e-05,
+      "loss": 0.2992,
+      "step": 94460
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.209018528164298e-05,
+      "loss": 0.2968,
+      "step": 94470
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.2087208869707569e-05,
+      "loss": 0.2993,
+      "step": 94480
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 1.2084232457772156e-05,
+      "loss": 0.3062,
+      "step": 94490
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.2081256045836744e-05,
+      "loss": 0.3094,
+      "step": 94500
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.2078279633901333e-05,
+      "loss": 0.3045,
+      "step": 94510
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.2075303221965922e-05,
+      "loss": 0.3022,
+      "step": 94520
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.207232681003051e-05,
+      "loss": 0.2973,
+      "step": 94530
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 1.2069350398095097e-05,
+      "loss": 0.3009,
+      "step": 94540
+    },
+    {
+      "epoch": 21.0,
+      "eval_cer": 4.542735042735043,
+      "eval_loss": 0.5427362322807312,
+      "eval_runtime": 1.2558,
+      "eval_samples_per_second": 7.963,
+      "eval_steps_per_second": 0.796,
+      "eval_wer": 0.6233766233766234,
+      "step": 94542
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.2066373986159686e-05,
+      "loss": 0.2985,
+      "step": 94550
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.2063397574224274e-05,
+      "loss": 0.3059,
+      "step": 94560
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.2060421162288863e-05,
+      "loss": 0.3003,
+      "step": 94570
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 1.2057444750353451e-05,
+      "loss": 0.2934,
+      "step": 94580
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.2054468338418037e-05,
+      "loss": 0.2794,
+      "step": 94590
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.2051491926482625e-05,
+      "loss": 0.2899,
+      "step": 94600
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.2048515514547214e-05,
+      "loss": 0.2997,
+      "step": 94610
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.2045539102611802e-05,
+      "loss": 0.3013,
+      "step": 94620
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 1.2042562690676391e-05,
+      "loss": 0.2879,
+      "step": 94630
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.2039586278740978e-05,
+      "loss": 0.2989,
+      "step": 94640
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.2036609866805567e-05,
+      "loss": 0.3046,
+      "step": 94650
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.2033633454870155e-05,
+      "loss": 0.2848,
+      "step": 94660
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 1.2030657042934744e-05,
+      "loss": 0.3021,
+      "step": 94670
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.2027680630999332e-05,
+      "loss": 0.2939,
+      "step": 94680
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.202470421906392e-05,
+      "loss": 0.2959,
+      "step": 94690
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.2021727807128508e-05,
+      "loss": 0.3003,
+      "step": 94700
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.2018751395193096e-05,
+      "loss": 0.2923,
+      "step": 94710
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 1.2015774983257685e-05,
+      "loss": 0.3001,
+      "step": 94720
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.2012798571322274e-05,
+      "loss": 0.3013,
+      "step": 94730
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.2009822159386859e-05,
+      "loss": 0.2815,
+      "step": 94740
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.2006845747451447e-05,
+      "loss": 0.305,
+      "step": 94750
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 1.2003869335516036e-05,
+      "loss": 0.3074,
+      "step": 94760
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.2000892923580625e-05,
+      "loss": 0.2982,
+      "step": 94770
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.1997916511645213e-05,
+      "loss": 0.2813,
+      "step": 94780
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.19949400997098e-05,
+      "loss": 0.3147,
+      "step": 94790
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.1991963687774389e-05,
+      "loss": 0.2939,
+      "step": 94800
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 1.1988987275838977e-05,
+      "loss": 0.3055,
+      "step": 94810
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.1986010863903566e-05,
+      "loss": 0.2979,
+      "step": 94820
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.1983034451968154e-05,
+      "loss": 0.3045,
+      "step": 94830
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.198005804003274e-05,
+      "loss": 0.2885,
+      "step": 94840
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 1.1977081628097328e-05,
+      "loss": 0.3021,
+      "step": 94850
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.1974105216161917e-05,
+      "loss": 0.2928,
+      "step": 94860
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.1971128804226505e-05,
+      "loss": 0.2957,
+      "step": 94870
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.1968152392291094e-05,
+      "loss": 0.3082,
+      "step": 94880
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.1965175980355683e-05,
+      "loss": 0.2878,
+      "step": 94890
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 1.196219956842027e-05,
+      "loss": 0.2909,
+      "step": 94900
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.1959223156484858e-05,
+      "loss": 0.3103,
+      "step": 94910
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.1956246744549447e-05,
+      "loss": 0.2847,
+      "step": 94920
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.1953270332614035e-05,
+      "loss": 0.2934,
+      "step": 94930
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 1.1950293920678624e-05,
+      "loss": 0.2862,
+      "step": 94940
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.194731750874321e-05,
+      "loss": 0.2872,
+      "step": 94950
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.19443410968078e-05,
+      "loss": 0.2944,
+      "step": 94960
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.1941364684872388e-05,
+      "loss": 0.3068,
+      "step": 94970
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.1938388272936977e-05,
+      "loss": 0.3023,
+      "step": 94980
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 1.1935411861001565e-05,
+      "loss": 0.3038,
+      "step": 94990
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.193243544906615e-05,
+      "loss": 0.2879,
+      "step": 95000
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.1929459037130739e-05,
+      "loss": 0.2934,
+      "step": 95010
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.1926482625195328e-05,
+      "loss": 0.2904,
+      "step": 95020
+    },
+    {
+      "epoch": 21.1,
+      "learning_rate": 1.1923506213259916e-05,
+      "loss": 0.297,
+      "step": 95030
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.1920529801324505e-05,
+      "loss": 0.3102,
+      "step": 95040
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.1917553389389092e-05,
+      "loss": 0.2831,
+      "step": 95050
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.191457697745368e-05,
+      "loss": 0.2958,
+      "step": 95060
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.1911600565518269e-05,
+      "loss": 0.303,
+      "step": 95070
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 1.1908624153582857e-05,
+      "loss": 0.2924,
+      "step": 95080
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.1905647741647446e-05,
+      "loss": 0.3082,
+      "step": 95090
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.1902671329712033e-05,
+      "loss": 0.3022,
+      "step": 95100
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.1899694917776621e-05,
+      "loss": 0.2944,
+      "step": 95110
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 1.189671850584121e-05,
+      "loss": 0.291,
+      "step": 95120
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.1893742093905799e-05,
+      "loss": 0.2844,
+      "step": 95130
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.1890765681970387e-05,
+      "loss": 0.3083,
+      "step": 95140
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.1887789270034972e-05,
+      "loss": 0.3123,
+      "step": 95150
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.1884812858099561e-05,
+      "loss": 0.3041,
+      "step": 95160
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 1.188183644616415e-05,
+      "loss": 0.2928,
+      "step": 95170
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.1878860034228738e-05,
+      "loss": 0.2947,
+      "step": 95180
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.1875883622293327e-05,
+      "loss": 0.2867,
+      "step": 95190
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.1872907210357914e-05,
+      "loss": 0.2999,
+      "step": 95200
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 1.1869930798422502e-05,
+      "loss": 0.2939,
+      "step": 95210
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.1866954386487091e-05,
+      "loss": 0.2888,
+      "step": 95220
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.186397797455168e-05,
+      "loss": 0.2914,
+      "step": 95230
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.1861001562616268e-05,
+      "loss": 0.2916,
+      "step": 95240
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.1858025150680853e-05,
+      "loss": 0.293,
+      "step": 95250
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 1.1855048738745442e-05,
+      "loss": 0.2842,
+      "step": 95260
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.185207232681003e-05,
+      "loss": 0.2972,
+      "step": 95270
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.1849095914874619e-05,
+      "loss": 0.3047,
+      "step": 95280
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.1846119502939208e-05,
+      "loss": 0.2798,
+      "step": 95290
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 1.1843143091003796e-05,
+      "loss": 0.3076,
+      "step": 95300
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.1840166679068383e-05,
+      "loss": 0.3001,
+      "step": 95310
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.1837190267132972e-05,
+      "loss": 0.2901,
+      "step": 95320
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.183421385519756e-05,
+      "loss": 0.2871,
+      "step": 95330
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.1831237443262149e-05,
+      "loss": 0.3,
+      "step": 95340
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 1.1828261031326738e-05,
+      "loss": 0.3021,
+      "step": 95350
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.1825284619391324e-05,
+      "loss": 0.2966,
+      "step": 95360
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.1822308207455913e-05,
+      "loss": 0.2934,
+      "step": 95370
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.1819331795520502e-05,
+      "loss": 0.2973,
+      "step": 95380
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 1.181635538358509e-05,
+      "loss": 0.3023,
+      "step": 95390
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.1813378971649679e-05,
+      "loss": 0.3041,
+      "step": 95400
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.1810402559714264e-05,
+      "loss": 0.2869,
+      "step": 95410
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.1807426147778853e-05,
+      "loss": 0.3006,
+      "step": 95420
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.1804449735843441e-05,
+      "loss": 0.2906,
+      "step": 95430
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 1.180147332390803e-05,
+      "loss": 0.2964,
+      "step": 95440
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.1798496911972618e-05,
+      "loss": 0.2972,
+      "step": 95450
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.1795520500037205e-05,
+      "loss": 0.2883,
+      "step": 95460
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.1792544088101794e-05,
+      "loss": 0.3023,
+      "step": 95470
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 1.1789567676166382e-05,
+      "loss": 0.2906,
+      "step": 95480
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.1786591264230971e-05,
+      "loss": 0.2845,
+      "step": 95490
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.178361485229556e-05,
+      "loss": 0.299,
+      "step": 95500
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.1780638440360147e-05,
+      "loss": 0.2976,
+      "step": 95510
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.1777662028424735e-05,
+      "loss": 0.2908,
+      "step": 95520
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 1.1774685616489324e-05,
+      "loss": 0.2806,
+      "step": 95530
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.1771709204553912e-05,
+      "loss": 0.293,
+      "step": 95540
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.1768732792618501e-05,
+      "loss": 0.3065,
+      "step": 95550
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.1765756380683086e-05,
+      "loss": 0.3052,
+      "step": 95560
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 1.1762779968747675e-05,
+      "loss": 0.2958,
+      "step": 95570
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.1759803556812263e-05,
+      "loss": 0.2857,
+      "step": 95580
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.1756827144876852e-05,
+      "loss": 0.2988,
+      "step": 95590
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.175385073294144e-05,
+      "loss": 0.296,
+      "step": 95600
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.1750874321006027e-05,
+      "loss": 0.2791,
+      "step": 95610
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 1.1747897909070616e-05,
+      "loss": 0.3036,
+      "step": 95620
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.1744921497135205e-05,
+      "loss": 0.2798,
+      "step": 95630
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.1741945085199793e-05,
+      "loss": 0.2866,
+      "step": 95640
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.1738968673264382e-05,
+      "loss": 0.2925,
+      "step": 95650
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 1.1735992261328969e-05,
+      "loss": 0.2815,
+      "step": 95660
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.1733015849393556e-05,
+      "loss": 0.2965,
+      "step": 95670
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.1730039437458144e-05,
+      "loss": 0.2962,
+      "step": 95680
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.1727063025522733e-05,
+      "loss": 0.2938,
+      "step": 95690
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.1724086613587321e-05,
+      "loss": 0.2929,
+      "step": 95700
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 1.1721110201651908e-05,
+      "loss": 0.2859,
+      "step": 95710
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.1718133789716497e-05,
+      "loss": 0.3069,
+      "step": 95720
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.1715157377781085e-05,
+      "loss": 0.2922,
+      "step": 95730
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.1712180965845674e-05,
+      "loss": 0.2913,
+      "step": 95740
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 1.1709204553910263e-05,
+      "loss": 0.309,
+      "step": 95750
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.1706228141974851e-05,
+      "loss": 0.2924,
+      "step": 95760
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.1703251730039438e-05,
+      "loss": 0.2777,
+      "step": 95770
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.1700275318104027e-05,
+      "loss": 0.2862,
+      "step": 95780
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.1697298906168615e-05,
+      "loss": 0.3033,
+      "step": 95790
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 1.1694322494233204e-05,
+      "loss": 0.2958,
+      "step": 95800
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.1691346082297792e-05,
+      "loss": 0.2869,
+      "step": 95810
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.1688369670362378e-05,
+      "loss": 0.2812,
+      "step": 95820
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.1685393258426966e-05,
+      "loss": 0.2832,
+      "step": 95830
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 1.1682416846491555e-05,
+      "loss": 0.3005,
+      "step": 95840
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.1679440434556143e-05,
+      "loss": 0.3011,
+      "step": 95850
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.1676464022620732e-05,
+      "loss": 0.2921,
+      "step": 95860
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.1673487610685319e-05,
+      "loss": 0.3042,
+      "step": 95870
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.1670511198749908e-05,
+      "loss": 0.2908,
+      "step": 95880
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 1.1667534786814496e-05,
+      "loss": 0.2885,
+      "step": 95890
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.1664558374879085e-05,
+      "loss": 0.2812,
+      "step": 95900
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.1661581962943673e-05,
+      "loss": 0.2938,
+      "step": 95910
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.165860555100826e-05,
+      "loss": 0.2972,
+      "step": 95920
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 1.1655629139072849e-05,
+      "loss": 0.2938,
+      "step": 95930
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.1652652727137437e-05,
+      "loss": 0.2949,
+      "step": 95940
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.1649676315202026e-05,
+      "loss": 0.296,
+      "step": 95950
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.1646699903266615e-05,
+      "loss": 0.2903,
+      "step": 95960
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.16437234913312e-05,
+      "loss": 0.2809,
+      "step": 95970
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 1.1640747079395788e-05,
+      "loss": 0.2951,
+      "step": 95980
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.1637770667460377e-05,
+      "loss": 0.2908,
+      "step": 95990
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.1634794255524966e-05,
+      "loss": 0.2888,
+      "step": 96000
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.1631817843589554e-05,
+      "loss": 0.3004,
+      "step": 96010
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 1.1628841431654141e-05,
+      "loss": 0.306,
+      "step": 96020
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.162586501971873e-05,
+      "loss": 0.286,
+      "step": 96030
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.1622888607783318e-05,
+      "loss": 0.2876,
+      "step": 96040
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.1619912195847907e-05,
+      "loss": 0.2936,
+      "step": 96050
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.1616935783912495e-05,
+      "loss": 0.2998,
+      "step": 96060
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 1.1613959371977082e-05,
+      "loss": 0.2912,
+      "step": 96070
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.1610982960041671e-05,
+      "loss": 0.2989,
+      "step": 96080
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.160800654810626e-05,
+      "loss": 0.3018,
+      "step": 96090
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.1605030136170846e-05,
+      "loss": 0.2868,
+      "step": 96100
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 1.1602053724235435e-05,
+      "loss": 0.3082,
+      "step": 96110
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.1599077312300022e-05,
+      "loss": 0.3035,
+      "step": 96120
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.159610090036461e-05,
+      "loss": 0.2895,
+      "step": 96130
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.1593124488429199e-05,
+      "loss": 0.2782,
+      "step": 96140
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.1590148076493788e-05,
+      "loss": 0.2946,
+      "step": 96150
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 1.1587171664558376e-05,
+      "loss": 0.2868,
+      "step": 96160
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.1584195252622965e-05,
+      "loss": 0.2903,
+      "step": 96170
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.1581218840687552e-05,
+      "loss": 0.2818,
+      "step": 96180
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.157824242875214e-05,
+      "loss": 0.2802,
+      "step": 96190
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 1.1575266016816729e-05,
+      "loss": 0.2943,
+      "step": 96200
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.1572289604881318e-05,
+      "loss": 0.3091,
+      "step": 96210
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.1569313192945906e-05,
+      "loss": 0.2992,
+      "step": 96220
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.1566336781010491e-05,
+      "loss": 0.2945,
+      "step": 96230
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.156336036907508e-05,
+      "loss": 0.2975,
+      "step": 96240
+    },
+    {
+      "epoch": 21.37,
+      "learning_rate": 1.1560383957139669e-05,
+      "loss": 0.2966,
+      "step": 96250
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.1557407545204257e-05,
+      "loss": 0.306,
+      "step": 96260
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.1554431133268846e-05,
+      "loss": 0.3087,
+      "step": 96270
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.1551454721333433e-05,
+      "loss": 0.2956,
+      "step": 96280
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 1.1548478309398021e-05,
+      "loss": 0.2915,
+      "step": 96290
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.154550189746261e-05,
+      "loss": 0.3109,
+      "step": 96300
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.1542525485527198e-05,
+      "loss": 0.3003,
+      "step": 96310
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.1539549073591787e-05,
+      "loss": 0.3007,
+      "step": 96320
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.1536572661656374e-05,
+      "loss": 0.2891,
+      "step": 96330
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 1.1533596249720963e-05,
+      "loss": 0.3021,
+      "step": 96340
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.1530619837785551e-05,
+      "loss": 0.2906,
+      "step": 96350
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.152764342585014e-05,
+      "loss": 0.2933,
+      "step": 96360
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.1524667013914728e-05,
+      "loss": 0.2797,
+      "step": 96370
+    },
+    {
+      "epoch": 21.4,
+      "learning_rate": 1.1521690601979313e-05,
+      "loss": 0.2912,
+      "step": 96380
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.1518714190043902e-05,
+      "loss": 0.2815,
+      "step": 96390
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.151573777810849e-05,
+      "loss": 0.2922,
+      "step": 96400
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.151276136617308e-05,
+      "loss": 0.2858,
+      "step": 96410
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.1509784954237668e-05,
+      "loss": 0.2925,
+      "step": 96420
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 1.1506808542302255e-05,
+      "loss": 0.2944,
+      "step": 96430
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.1503832130366843e-05,
+      "loss": 0.2953,
+      "step": 96440
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.1500855718431432e-05,
+      "loss": 0.2931,
+      "step": 96450
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.149787930649602e-05,
+      "loss": 0.2971,
+      "step": 96460
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 1.1494902894560609e-05,
+      "loss": 0.2872,
+      "step": 96470
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.1491926482625196e-05,
+      "loss": 0.2891,
+      "step": 96480
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.1488950070689785e-05,
+      "loss": 0.3014,
+      "step": 96490
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.1485973658754373e-05,
+      "loss": 0.2918,
+      "step": 96500
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.1482997246818962e-05,
+      "loss": 0.296,
+      "step": 96510
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 1.148002083488355e-05,
+      "loss": 0.3031,
+      "step": 96520
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.1477044422948136e-05,
+      "loss": 0.3062,
+      "step": 96530
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.1474068011012724e-05,
+      "loss": 0.2962,
+      "step": 96540
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.1471091599077313e-05,
+      "loss": 0.3002,
+      "step": 96550
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 1.1468115187141901e-05,
+      "loss": 0.2889,
+      "step": 96560
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.146513877520649e-05,
+      "loss": 0.2995,
+      "step": 96570
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.1462162363271077e-05,
+      "loss": 0.3024,
+      "step": 96580
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.1459185951335665e-05,
+      "loss": 0.2997,
+      "step": 96590
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.1456209539400254e-05,
+      "loss": 0.2994,
+      "step": 96600
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 1.1453233127464843e-05,
+      "loss": 0.2958,
+      "step": 96610
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.1450256715529431e-05,
+      "loss": 0.3113,
+      "step": 96620
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.144728030359402e-05,
+      "loss": 0.31,
+      "step": 96630
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.1444303891658605e-05,
+      "loss": 0.2999,
+      "step": 96640
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 1.1441327479723194e-05,
+      "loss": 0.3003,
+      "step": 96650
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.1438351067787782e-05,
+      "loss": 0.289,
+      "step": 96660
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.143537465585237e-05,
+      "loss": 0.2829,
+      "step": 96670
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.143239824391696e-05,
+      "loss": 0.2964,
+      "step": 96680
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.1429421831981546e-05,
+      "loss": 0.2994,
+      "step": 96690
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 1.1426445420046135e-05,
+      "loss": 0.3123,
+      "step": 96700
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.1423469008110724e-05,
+      "loss": 0.3115,
+      "step": 96710
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.1420492596175312e-05,
+      "loss": 0.2963,
+      "step": 96720
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.14175161842399e-05,
+      "loss": 0.2902,
+      "step": 96730
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 1.1414539772304488e-05,
+      "loss": 0.3039,
+      "step": 96740
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1411563360369076e-05,
+      "loss": 0.2961,
+      "step": 96750
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1408586948433665e-05,
+      "loss": 0.3038,
+      "step": 96760
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1405610536498253e-05,
+      "loss": 0.3031,
+      "step": 96770
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1402634124562842e-05,
+      "loss": 0.2982,
+      "step": 96780
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 1.1399657712627427e-05,
+      "loss": 0.2886,
+      "step": 96790
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.1396681300692016e-05,
+      "loss": 0.3186,
+      "step": 96800
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.1393704888756604e-05,
+      "loss": 0.301,
+      "step": 96810
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.1390728476821193e-05,
+      "loss": 0.2971,
+      "step": 96820
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 1.1387752064885782e-05,
+      "loss": 0.2923,
+      "step": 96830
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.1384775652950368e-05,
+      "loss": 0.2971,
+      "step": 96840
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.1381799241014957e-05,
+      "loss": 0.3064,
+      "step": 96850
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.1378822829079546e-05,
+      "loss": 0.2873,
+      "step": 96860
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.1375846417144134e-05,
+      "loss": 0.292,
+      "step": 96870
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 1.1372870005208723e-05,
+      "loss": 0.2885,
+      "step": 96880
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.136989359327331e-05,
+      "loss": 0.294,
+      "step": 96890
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.1366917181337898e-05,
+      "loss": 0.3091,
+      "step": 96900
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.1363940769402487e-05,
+      "loss": 0.3077,
+      "step": 96910
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 1.1360964357467075e-05,
+      "loss": 0.2876,
+      "step": 96920
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.1357987945531664e-05,
+      "loss": 0.2898,
+      "step": 96930
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.135501153359625e-05,
+      "loss": 0.2978,
+      "step": 96940
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.1352035121660838e-05,
+      "loss": 0.2946,
+      "step": 96950
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.1349058709725426e-05,
+      "loss": 0.3067,
+      "step": 96960
+    },
+    {
+      "epoch": 21.53,
+      "learning_rate": 1.1346082297790015e-05,
+      "loss": 0.2863,
+      "step": 96970
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.1343105885854604e-05,
+      "loss": 0.301,
+      "step": 96980
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.134012947391919e-05,
+      "loss": 0.3005,
+      "step": 96990
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.133715306198378e-05,
+      "loss": 0.3065,
+      "step": 97000
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 1.1334176650048368e-05,
+      "loss": 0.2927,
+      "step": 97010
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.1331200238112956e-05,
+      "loss": 0.2952,
+      "step": 97020
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.1328223826177545e-05,
+      "loss": 0.3013,
+      "step": 97030
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.1325247414242134e-05,
+      "loss": 0.3021,
+      "step": 97040
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.1322271002306719e-05,
+      "loss": 0.2987,
+      "step": 97050
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 1.1319294590371307e-05,
+      "loss": 0.3084,
+      "step": 97060
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.1316318178435896e-05,
+      "loss": 0.2995,
+      "step": 97070
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.1313341766500485e-05,
+      "loss": 0.2893,
+      "step": 97080
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.1310365354565073e-05,
+      "loss": 0.3021,
+      "step": 97090
+    },
+    {
+      "epoch": 21.56,
+      "learning_rate": 1.130738894262966e-05,
+      "loss": 0.2924,
+      "step": 97100
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1304412530694249e-05,
+      "loss": 0.2913,
+      "step": 97110
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1301436118758837e-05,
+      "loss": 0.2966,
+      "step": 97120
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1298459706823426e-05,
+      "loss": 0.2926,
+      "step": 97130
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1295483294888014e-05,
+      "loss": 0.3001,
+      "step": 97140
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 1.1292506882952601e-05,
+      "loss": 0.2971,
+      "step": 97150
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.128953047101719e-05,
+      "loss": 0.2895,
+      "step": 97160
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.1286554059081778e-05,
+      "loss": 0.287,
+      "step": 97170
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.1283577647146367e-05,
+      "loss": 0.3088,
+      "step": 97180
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 1.1280601235210956e-05,
+      "loss": 0.295,
+      "step": 97190
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.1277624823275541e-05,
+      "loss": 0.2974,
+      "step": 97200
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.127464841134013e-05,
+      "loss": 0.2929,
+      "step": 97210
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.1271671999404718e-05,
+      "loss": 0.2978,
+      "step": 97220
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.1268695587469307e-05,
+      "loss": 0.2988,
+      "step": 97230
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 1.1265719175533895e-05,
+      "loss": 0.284,
+      "step": 97240
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.1262742763598482e-05,
+      "loss": 0.2929,
+      "step": 97250
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.125976635166307e-05,
+      "loss": 0.2972,
+      "step": 97260
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.125678993972766e-05,
+      "loss": 0.2933,
+      "step": 97270
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 1.1253813527792248e-05,
+      "loss": 0.296,
+      "step": 97280
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.1250837115856836e-05,
+      "loss": 0.2798,
+      "step": 97290
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.1247860703921423e-05,
+      "loss": 0.2945,
+      "step": 97300
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.1244884291986012e-05,
+      "loss": 0.3153,
+      "step": 97310
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.12419078800506e-05,
+      "loss": 0.313,
+      "step": 97320
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 1.123893146811519e-05,
+      "loss": 0.3135,
+      "step": 97330
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.1235955056179778e-05,
+      "loss": 0.2879,
+      "step": 97340
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.1232978644244363e-05,
+      "loss": 0.2997,
+      "step": 97350
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.1230002232308952e-05,
+      "loss": 0.3012,
+      "step": 97360
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 1.122702582037354e-05,
+      "loss": 0.2771,
+      "step": 97370
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.1224049408438129e-05,
+      "loss": 0.2953,
+      "step": 97380
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.1221072996502717e-05,
+      "loss": 0.3136,
+      "step": 97390
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.1218096584567304e-05,
+      "loss": 0.2824,
+      "step": 97400
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.1215120172631893e-05,
+      "loss": 0.2899,
+      "step": 97410
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 1.1212143760696481e-05,
+      "loss": 0.2933,
+      "step": 97420
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.120916734876107e-05,
+      "loss": 0.2924,
+      "step": 97430
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.1206190936825659e-05,
+      "loss": 0.2967,
+      "step": 97440
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.1203214524890244e-05,
+      "loss": 0.2815,
+      "step": 97450
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 1.1200238112954832e-05,
+      "loss": 0.2975,
+      "step": 97460
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.1197261701019421e-05,
+      "loss": 0.2984,
+      "step": 97470
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.119428528908401e-05,
+      "loss": 0.2716,
+      "step": 97480
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.1191308877148598e-05,
+      "loss": 0.2916,
+      "step": 97490
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.1188332465213187e-05,
+      "loss": 0.3086,
+      "step": 97500
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 1.1185356053277774e-05,
+      "loss": 0.2748,
+      "step": 97510
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.1182379641342362e-05,
+      "loss": 0.2809,
+      "step": 97520
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.1179403229406951e-05,
+      "loss": 0.2957,
+      "step": 97530
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.117642681747154e-05,
+      "loss": 0.3002,
+      "step": 97540
+    },
+    {
+      "epoch": 21.66,
+      "learning_rate": 1.1173450405536128e-05,
+      "loss": 0.3031,
+      "step": 97550
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.1170473993600715e-05,
+      "loss": 0.2826,
+      "step": 97560
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.1167497581665304e-05,
+      "loss": 0.2885,
+      "step": 97570
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.1164521169729892e-05,
+      "loss": 0.2956,
+      "step": 97580
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.116154475779448e-05,
+      "loss": 0.279,
+      "step": 97590
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 1.115856834585907e-05,
+      "loss": 0.288,
+      "step": 97600
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.1155591933923655e-05,
+      "loss": 0.2918,
+      "step": 97610
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.1152615521988243e-05,
+      "loss": 0.305,
+      "step": 97620
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.1149639110052832e-05,
+      "loss": 0.3087,
+      "step": 97630
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 1.114666269811742e-05,
+      "loss": 0.307,
+      "step": 97640
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1143686286182009e-05,
+      "loss": 0.3015,
+      "step": 97650
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1140709874246596e-05,
+      "loss": 0.2897,
+      "step": 97660
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1137733462311184e-05,
+      "loss": 0.2943,
+      "step": 97670
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1134757050375773e-05,
+      "loss": 0.2959,
+      "step": 97680
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 1.1131780638440362e-05,
+      "loss": 0.2983,
+      "step": 97690
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.112880422650495e-05,
+      "loss": 0.306,
+      "step": 97700
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.1125827814569537e-05,
+      "loss": 0.2967,
+      "step": 97710
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.1122851402634126e-05,
+      "loss": 0.2975,
+      "step": 97720
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 1.1119874990698714e-05,
+      "loss": 0.2923,
+      "step": 97730
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1116898578763303e-05,
+      "loss": 0.2949,
+      "step": 97740
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1113922166827891e-05,
+      "loss": 0.3004,
+      "step": 97750
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1110945754892477e-05,
+      "loss": 0.2842,
+      "step": 97760
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1107969342957065e-05,
+      "loss": 0.2961,
+      "step": 97770
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 1.1104992931021654e-05,
+      "loss": 0.2932,
+      "step": 97780
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.1102016519086242e-05,
+      "loss": 0.2964,
+      "step": 97790
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.1099040107150831e-05,
+      "loss": 0.2881,
+      "step": 97800
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.1096063695215418e-05,
+      "loss": 0.2975,
+      "step": 97810
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 1.1093087283280007e-05,
+      "loss": 0.2973,
+      "step": 97820
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.1090110871344595e-05,
+      "loss": 0.2835,
+      "step": 97830
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.1087134459409184e-05,
+      "loss": 0.2998,
+      "step": 97840
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.1084158047473772e-05,
+      "loss": 0.3011,
+      "step": 97850
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.108118163553836e-05,
+      "loss": 0.2865,
+      "step": 97860
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 1.1078205223602946e-05,
+      "loss": 0.2962,
+      "step": 97870
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.1075228811667535e-05,
+      "loss": 0.2984,
+      "step": 97880
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.1072252399732123e-05,
+      "loss": 0.3043,
+      "step": 97890
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.1069275987796712e-05,
+      "loss": 0.2881,
+      "step": 97900
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 1.10662995758613e-05,
+      "loss": 0.3047,
+      "step": 97910
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.1063323163925887e-05,
+      "loss": 0.296,
+      "step": 97920
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.1060346751990476e-05,
+      "loss": 0.2736,
+      "step": 97930
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.1057370340055065e-05,
+      "loss": 0.2998,
+      "step": 97940
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.1054393928119653e-05,
+      "loss": 0.3036,
+      "step": 97950
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 1.1051417516184242e-05,
+      "loss": 0.2944,
+      "step": 97960
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.1048441104248829e-05,
+      "loss": 0.2917,
+      "step": 97970
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.1045464692313417e-05,
+      "loss": 0.2923,
+      "step": 97980
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.1042488280378006e-05,
+      "loss": 0.3057,
+      "step": 97990
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 1.1039511868442594e-05,
+      "loss": 0.3053,
+      "step": 98000
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.1036535456507183e-05,
+      "loss": 0.294,
+      "step": 98010
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.1033559044571768e-05,
+      "loss": 0.28,
+      "step": 98020
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.1030582632636357e-05,
+      "loss": 0.2988,
+      "step": 98030
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.1027606220700945e-05,
+      "loss": 0.2967,
+      "step": 98040
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 1.1024629808765534e-05,
+      "loss": 0.2995,
+      "step": 98050
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.1021653396830123e-05,
+      "loss": 0.2924,
+      "step": 98060
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.101867698489471e-05,
+      "loss": 0.292,
+      "step": 98070
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.1015700572959298e-05,
+      "loss": 0.2895,
+      "step": 98080
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 1.1012724161023887e-05,
+      "loss": 0.2867,
+      "step": 98090
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.1009747749088475e-05,
+      "loss": 0.301,
+      "step": 98100
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.1006771337153064e-05,
+      "loss": 0.2972,
+      "step": 98110
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.100379492521765e-05,
+      "loss": 0.3051,
+      "step": 98120
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.100081851328224e-05,
+      "loss": 0.2797,
+      "step": 98130
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 1.0997842101346828e-05,
+      "loss": 0.2925,
+      "step": 98140
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.0994865689411417e-05,
+      "loss": 0.2923,
+      "step": 98150
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.0991889277476005e-05,
+      "loss": 0.295,
+      "step": 98160
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.098891286554059e-05,
+      "loss": 0.2945,
+      "step": 98170
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 1.0985936453605179e-05,
+      "loss": 0.2957,
+      "step": 98180
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.0982960041669768e-05,
+      "loss": 0.285,
+      "step": 98190
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.0979983629734356e-05,
+      "loss": 0.2938,
+      "step": 98200
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.0977007217798945e-05,
+      "loss": 0.2884,
+      "step": 98210
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.0974030805863532e-05,
+      "loss": 0.2918,
+      "step": 98220
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 1.097105439392812e-05,
+      "loss": 0.2943,
+      "step": 98230
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.0968077981992709e-05,
+      "loss": 0.2879,
+      "step": 98240
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.0965101570057297e-05,
+      "loss": 0.29,
+      "step": 98250
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.0962125158121886e-05,
+      "loss": 0.3046,
+      "step": 98260
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 1.0959148746186473e-05,
+      "loss": 0.2912,
+      "step": 98270
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.0956172334251061e-05,
+      "loss": 0.2999,
+      "step": 98280
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.095319592231565e-05,
+      "loss": 0.2916,
+      "step": 98290
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.0950219510380237e-05,
+      "loss": 0.2969,
+      "step": 98300
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.0947243098444826e-05,
+      "loss": 0.3153,
+      "step": 98310
+    },
+    {
+      "epoch": 21.83,
+      "learning_rate": 1.0944266686509412e-05,
+      "loss": 0.2985,
+      "step": 98320
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.0941290274574001e-05,
+      "loss": 0.2977,
+      "step": 98330
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.093831386263859e-05,
+      "loss": 0.2861,
+      "step": 98340
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.0935337450703178e-05,
+      "loss": 0.2981,
+      "step": 98350
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 1.0932361038767767e-05,
+      "loss": 0.2975,
+      "step": 98360
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.0929384626832355e-05,
+      "loss": 0.2942,
+      "step": 98370
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.0926408214896942e-05,
+      "loss": 0.2921,
+      "step": 98380
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.0923431802961531e-05,
+      "loss": 0.3035,
+      "step": 98390
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.092045539102612e-05,
+      "loss": 0.2982,
+      "step": 98400
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 1.0917478979090708e-05,
+      "loss": 0.297,
+      "step": 98410
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.0914502567155297e-05,
+      "loss": 0.2951,
+      "step": 98420
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.0911526155219882e-05,
+      "loss": 0.2912,
+      "step": 98430
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.090854974328447e-05,
+      "loss": 0.2901,
+      "step": 98440
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 1.0905573331349059e-05,
+      "loss": 0.2981,
+      "step": 98450
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0902596919413648e-05,
+      "loss": 0.2999,
+      "step": 98460
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0899620507478236e-05,
+      "loss": 0.2896,
+      "step": 98470
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0896644095542823e-05,
+      "loss": 0.2905,
+      "step": 98480
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0893667683607412e-05,
+      "loss": 0.2978,
+      "step": 98490
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 1.0890691271672e-05,
+      "loss": 0.2976,
+      "step": 98500
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.0887714859736589e-05,
+      "loss": 0.3019,
+      "step": 98510
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.0884738447801178e-05,
+      "loss": 0.302,
+      "step": 98520
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.0881762035865764e-05,
+      "loss": 0.2934,
+      "step": 98530
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 1.0878785623930353e-05,
+      "loss": 0.2944,
+      "step": 98540
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.0875809211994942e-05,
+      "loss": 0.291,
+      "step": 98550
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.087283280005953e-05,
+      "loss": 0.3012,
+      "step": 98560
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.0869856388124119e-05,
+      "loss": 0.295,
+      "step": 98570
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.0866879976188704e-05,
+      "loss": 0.2928,
+      "step": 98580
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 1.0863903564253293e-05,
+      "loss": 0.3077,
+      "step": 98590
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.0860927152317881e-05,
+      "loss": 0.308,
+      "step": 98600
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.085795074038247e-05,
+      "loss": 0.2997,
+      "step": 98610
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.0854974328447058e-05,
+      "loss": 0.2782,
+      "step": 98620
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 1.0851997916511645e-05,
+      "loss": 0.2917,
+      "step": 98630
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.0849021504576234e-05,
+      "loss": 0.2835,
+      "step": 98640
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.0846045092640822e-05,
+      "loss": 0.3024,
+      "step": 98650
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.0843068680705411e-05,
+      "loss": 0.2974,
+      "step": 98660
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.084009226877e-05,
+      "loss": 0.2961,
+      "step": 98670
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 1.0837115856834587e-05,
+      "loss": 0.2824,
+      "step": 98680
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.0834139444899175e-05,
+      "loss": 0.2859,
+      "step": 98690
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.0831163032963764e-05,
+      "loss": 0.2999,
+      "step": 98700
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.0828186621028352e-05,
+      "loss": 0.2846,
+      "step": 98710
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 1.0825210209092941e-05,
+      "loss": 0.2949,
+      "step": 98720
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0822233797157526e-05,
+      "loss": 0.3045,
+      "step": 98730
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0819257385222115e-05,
+      "loss": 0.3076,
+      "step": 98740
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0816280973286703e-05,
+      "loss": 0.2966,
+      "step": 98750
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.0813304561351292e-05,
+      "loss": 0.3009,
+      "step": 98760
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 1.081032814941588e-05,
+      "loss": 0.3004,
+      "step": 98770
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.0807351737480469e-05,
+      "loss": 0.2965,
+      "step": 98780
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.0804375325545056e-05,
+      "loss": 0.2961,
+      "step": 98790
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.0801398913609645e-05,
+      "loss": 0.2945,
+      "step": 98800
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 1.0798422501674233e-05,
+      "loss": 0.297,
+      "step": 98810
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.0795446089738822e-05,
+      "loss": 0.2999,
+      "step": 98820
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.079246967780341e-05,
+      "loss": 0.3224,
+      "step": 98830
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.0789493265867996e-05,
+      "loss": 0.2943,
+      "step": 98840
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.0786516853932584e-05,
+      "loss": 0.291,
+      "step": 98850
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 1.0783540441997173e-05,
+      "loss": 0.2852,
+      "step": 98860
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.0780564030061761e-05,
+      "loss": 0.2983,
+      "step": 98870
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.077758761812635e-05,
+      "loss": 0.2915,
+      "step": 98880
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.0774611206190937e-05,
+      "loss": 0.2943,
+      "step": 98890
+    },
+    {
+      "epoch": 21.96,
+      "learning_rate": 1.0771634794255525e-05,
+      "loss": 0.3019,
+      "step": 98900
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.0768658382320114e-05,
+      "loss": 0.3013,
+      "step": 98910
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.0765681970384703e-05,
+      "loss": 0.2878,
+      "step": 98920
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.0762705558449291e-05,
+      "loss": 0.2886,
+      "step": 98930
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.0759729146513878e-05,
+      "loss": 0.2937,
+      "step": 98940
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 1.0756752734578467e-05,
+      "loss": 0.3068,
+      "step": 98950
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.0753776322643055e-05,
+      "loss": 0.2983,
+      "step": 98960
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.0750799910707644e-05,
+      "loss": 0.3156,
+      "step": 98970
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.0747823498772232e-05,
+      "loss": 0.2944,
+      "step": 98980
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 1.0744847086836818e-05,
+      "loss": 0.2821,
+      "step": 98990
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.0741870674901406e-05,
+      "loss": 0.2818,
+      "step": 99000
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.0738894262965995e-05,
+      "loss": 0.2909,
+      "step": 99010
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.0735917851030583e-05,
+      "loss": 0.2914,
+      "step": 99020
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.0732941439095172e-05,
+      "loss": 0.2961,
+      "step": 99030
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 1.0729965027159759e-05,
+      "loss": 0.2875,
+      "step": 99040
+    },
+    {
+      "epoch": 22.0,
+      "eval_cer": 4.553418803418803,
+      "eval_loss": 0.5452433824539185,
+      "eval_runtime": 1.2683,
+      "eval_samples_per_second": 7.885,
+      "eval_steps_per_second": 0.788,
+      "eval_wer": 0.6233766233766234,
+      "step": 99044
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.0726988615224348e-05,
+      "loss": 0.3042,
+      "step": 99050
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.0724012203288936e-05,
+      "loss": 0.2979,
+      "step": 99060
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.0721035791353525e-05,
+      "loss": 0.2876,
+      "step": 99070
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 1.0718059379418113e-05,
+      "loss": 0.2777,
+      "step": 99080
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.07150829674827e-05,
+      "loss": 0.2932,
+      "step": 99090
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.0712106555547289e-05,
+      "loss": 0.2828,
+      "step": 99100
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.0709130143611877e-05,
+      "loss": 0.2954,
+      "step": 99110
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.0706153731676466e-05,
+      "loss": 0.2801,
+      "step": 99120
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 1.0703177319741055e-05,
+      "loss": 0.2976,
+      "step": 99130
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.070020090780564e-05,
+      "loss": 0.2939,
+      "step": 99140
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.0697224495870228e-05,
+      "loss": 0.2905,
+      "step": 99150
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.0694248083934817e-05,
+      "loss": 0.2868,
+      "step": 99160
+    },
+    {
+      "epoch": 22.02,
+      "learning_rate": 1.0691271671999406e-05,
+      "loss": 0.2937,
+      "step": 99170
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.0688295260063994e-05,
+      "loss": 0.2852,
+      "step": 99180
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.0685318848128581e-05,
+      "loss": 0.3074,
+      "step": 99190
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.068234243619317e-05,
+      "loss": 0.2868,
+      "step": 99200
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.0679366024257758e-05,
+      "loss": 0.2915,
+      "step": 99210
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 1.0676389612322347e-05,
+      "loss": 0.3094,
+      "step": 99220
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.0673413200386935e-05,
+      "loss": 0.2942,
+      "step": 99230
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.0670436788451524e-05,
+      "loss": 0.3011,
+      "step": 99240
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.066746037651611e-05,
+      "loss": 0.2867,
+      "step": 99250
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 1.0664483964580698e-05,
+      "loss": 0.2896,
+      "step": 99260
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.0661507552645286e-05,
+      "loss": 0.2971,
+      "step": 99270
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.0658531140709875e-05,
+      "loss": 0.2827,
+      "step": 99280
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.0655554728774464e-05,
+      "loss": 0.2977,
+      "step": 99290
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.065257831683905e-05,
+      "loss": 0.2729,
+      "step": 99300
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 1.0649601904903639e-05,
+      "loss": 0.2869,
+      "step": 99310
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.0646625492968228e-05,
+      "loss": 0.3005,
+      "step": 99320
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.0643649081032816e-05,
+      "loss": 0.2953,
+      "step": 99330
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.0640672669097405e-05,
+      "loss": 0.2893,
+      "step": 99340
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 1.0637696257161992e-05,
+      "loss": 0.276,
+      "step": 99350
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.063471984522658e-05,
+      "loss": 0.296,
+      "step": 99360
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.0631743433291169e-05,
+      "loss": 0.3006,
+      "step": 99370
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.0628767021355758e-05,
+      "loss": 0.2995,
+      "step": 99380
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.0625790609420346e-05,
+      "loss": 0.2872,
+      "step": 99390
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 1.0622814197484931e-05,
+      "loss": 0.2958,
+      "step": 99400
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.061983778554952e-05,
+      "loss": 0.2788,
+      "step": 99410
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.0616861373614109e-05,
+      "loss": 0.2981,
+      "step": 99420
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.0613884961678697e-05,
+      "loss": 0.2939,
+      "step": 99430
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 1.0610908549743286e-05,
+      "loss": 0.2951,
+      "step": 99440
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.0607932137807873e-05,
+      "loss": 0.2878,
+      "step": 99450
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.0604955725872461e-05,
+      "loss": 0.2803,
+      "step": 99460
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.060197931393705e-05,
+      "loss": 0.2972,
+      "step": 99470
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.0599002902001638e-05,
+      "loss": 0.2928,
+      "step": 99480
+    },
+    {
+      "epoch": 22.09,
+      "learning_rate": 1.0596026490066227e-05,
+      "loss": 0.283,
+      "step": 99490
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.0593050078130814e-05,
+      "loss": 0.2985,
+      "step": 99500
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.0590073666195402e-05,
+      "loss": 0.2767,
+      "step": 99510
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.0587097254259991e-05,
+      "loss": 0.2945,
+      "step": 99520
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 1.058412084232458e-05,
+      "loss": 0.2911,
+      "step": 99530
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.0581144430389168e-05,
+      "loss": 0.288,
+      "step": 99540
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.0578168018453753e-05,
+      "loss": 0.285,
+      "step": 99550
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.0575191606518342e-05,
+      "loss": 0.2943,
+      "step": 99560
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.057221519458293e-05,
+      "loss": 0.2796,
+      "step": 99570
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 1.056923878264752e-05,
+      "loss": 0.2886,
+      "step": 99580
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.0566262370712108e-05,
+      "loss": 0.2882,
+      "step": 99590
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.0563285958776695e-05,
+      "loss": 0.2723,
+      "step": 99600
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.0560309546841283e-05,
+      "loss": 0.2819,
+      "step": 99610
+    },
+    {
+      "epoch": 22.12,
+      "learning_rate": 1.0557333134905872e-05,
+      "loss": 0.2866,
+      "step": 99620
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.055435672297046e-05,
+      "loss": 0.2869,
+      "step": 99630
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.0551380311035049e-05,
+      "loss": 0.2923,
+      "step": 99640
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.0548403899099638e-05,
+      "loss": 0.2783,
+      "step": 99650
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.0545427487164223e-05,
+      "loss": 0.2856,
+      "step": 99660
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 1.0542451075228811e-05,
+      "loss": 0.2931,
+      "step": 99670
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.05394746632934e-05,
+      "loss": 0.2935,
+      "step": 99680
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.0536498251357989e-05,
+      "loss": 0.2947,
+      "step": 99690
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.0533521839422577e-05,
+      "loss": 0.2801,
+      "step": 99700
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 1.0530545427487164e-05,
+      "loss": 0.2966,
+      "step": 99710
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.0527569015551753e-05,
+      "loss": 0.2986,
+      "step": 99720
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.0524592603616341e-05,
+      "loss": 0.2996,
+      "step": 99730
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.052161619168093e-05,
+      "loss": 0.2875,
+      "step": 99740
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.0518639779745519e-05,
+      "loss": 0.2889,
+      "step": 99750
+    },
+    {
+      "epoch": 22.15,
+      "learning_rate": 1.0515663367810105e-05,
+      "loss": 0.3052,
+      "step": 99760
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.0512686955874694e-05,
+      "loss": 0.2829,
+      "step": 99770
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.0509710543939283e-05,
+      "loss": 0.2952,
+      "step": 99780
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.0506734132003871e-05,
+      "loss": 0.2799,
+      "step": 99790
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 1.050375772006846e-05,
+      "loss": 0.2725,
+      "step": 99800
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.0500781308133045e-05,
+      "loss": 0.2925,
+      "step": 99810
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.0497804896197634e-05,
+      "loss": 0.2802,
+      "step": 99820
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.0494828484262222e-05,
+      "loss": 0.2837,
+      "step": 99830
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.049185207232681e-05,
+      "loss": 0.2951,
+      "step": 99840
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 1.04888756603914e-05,
+      "loss": 0.2969,
+      "step": 99850
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.0485899248455986e-05,
+      "loss": 0.2924,
+      "step": 99860
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.0482922836520575e-05,
+      "loss": 0.2989,
+      "step": 99870
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.0479946424585163e-05,
+      "loss": 0.289,
+      "step": 99880
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 1.0476970012649752e-05,
+      "loss": 0.2961,
+      "step": 99890
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.047399360071434e-05,
+      "loss": 0.2889,
+      "step": 99900
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.0471017188778928e-05,
+      "loss": 0.2922,
+      "step": 99910
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.0468040776843516e-05,
+      "loss": 0.3033,
+      "step": 99920
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.0465064364908105e-05,
+      "loss": 0.3015,
+      "step": 99930
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 1.0462087952972693e-05,
+      "loss": 0.2942,
+      "step": 99940
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.0459111541037282e-05,
+      "loss": 0.2956,
+      "step": 99950
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.0456135129101867e-05,
+      "loss": 0.2678,
+      "step": 99960
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.0453158717166456e-05,
+      "loss": 0.2841,
+      "step": 99970
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 1.0450182305231044e-05,
+      "loss": 0.2972,
+      "step": 99980
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0447205893295633e-05,
+      "loss": 0.2926,
+      "step": 99990
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0444229481360222e-05,
+      "loss": 0.2861,
+      "step": 100000
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0441253069424808e-05,
+      "loss": 0.276,
+      "step": 100010
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0438276657489397e-05,
+      "loss": 0.2849,
+      "step": 100020
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 1.0435300245553986e-05,
+      "loss": 0.2915,
+      "step": 100030
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.0432323833618574e-05,
+      "loss": 0.2877,
+      "step": 100040
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.0429347421683163e-05,
+      "loss": 0.2964,
+      "step": 100050
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.042637100974775e-05,
+      "loss": 0.2853,
+      "step": 100060
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 1.0423394597812337e-05,
+      "loss": 0.2907,
+      "step": 100070
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.0420418185876925e-05,
+      "loss": 0.2854,
+      "step": 100080
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.0417441773941514e-05,
+      "loss": 0.2871,
+      "step": 100090
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.0414465362006102e-05,
+      "loss": 0.2883,
+      "step": 100100
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.0411488950070691e-05,
+      "loss": 0.2893,
+      "step": 100110
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 1.0408512538135278e-05,
+      "loss": 0.2899,
+      "step": 100120
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.0405536126199866e-05,
+      "loss": 0.2947,
+      "step": 100130
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.0402559714264455e-05,
+      "loss": 0.2873,
+      "step": 100140
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.0399583302329044e-05,
+      "loss": 0.2937,
+      "step": 100150
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 1.0396606890393632e-05,
+      "loss": 0.2898,
+      "step": 100160
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.0393630478458219e-05,
+      "loss": 0.2948,
+      "step": 100170
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.0390654066522808e-05,
+      "loss": 0.2731,
+      "step": 100180
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.0387677654587396e-05,
+      "loss": 0.2783,
+      "step": 100190
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.0384701242651985e-05,
+      "loss": 0.2939,
+      "step": 100200
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 1.0381724830716573e-05,
+      "loss": 0.2947,
+      "step": 100210
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.0378748418781159e-05,
+      "loss": 0.2858,
+      "step": 100220
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.0375772006845747e-05,
+      "loss": 0.2931,
+      "step": 100230
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.0372795594910336e-05,
+      "loss": 0.3055,
+      "step": 100240
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 1.0369819182974924e-05,
+      "loss": 0.2995,
+      "step": 100250
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.0366842771039513e-05,
+      "loss": 0.2893,
+      "step": 100260
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.03638663591041e-05,
+      "loss": 0.2785,
+      "step": 100270
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.0360889947168689e-05,
+      "loss": 0.2819,
+      "step": 100280
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.0357913535233277e-05,
+      "loss": 0.2958,
+      "step": 100290
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 1.0354937123297866e-05,
+      "loss": 0.2944,
+      "step": 100300
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.0351960711362454e-05,
+      "loss": 0.2797,
+      "step": 100310
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.0348984299427041e-05,
+      "loss": 0.2891,
+      "step": 100320
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.034600788749163e-05,
+      "loss": 0.2838,
+      "step": 100330
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 1.0343031475556218e-05,
+      "loss": 0.2963,
+      "step": 100340
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.0340055063620807e-05,
+      "loss": 0.2939,
+      "step": 100350
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.0337078651685396e-05,
+      "loss": 0.3014,
+      "step": 100360
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.033410223974998e-05,
+      "loss": 0.2871,
+      "step": 100370
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.033112582781457e-05,
+      "loss": 0.2825,
+      "step": 100380
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 1.0328149415879158e-05,
+      "loss": 0.2837,
+      "step": 100390
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.0325173003943747e-05,
+      "loss": 0.2828,
+      "step": 100400
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.0322196592008335e-05,
+      "loss": 0.3048,
+      "step": 100410
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.0319220180072922e-05,
+      "loss": 0.2856,
+      "step": 100420
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 1.031624376813751e-05,
+      "loss": 0.2956,
+      "step": 100430
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.03132673562021e-05,
+      "loss": 0.2948,
+      "step": 100440
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.0310290944266688e-05,
+      "loss": 0.2964,
+      "step": 100450
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.0307314532331276e-05,
+      "loss": 0.2884,
+      "step": 100460
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.0304338120395863e-05,
+      "loss": 0.2758,
+      "step": 100470
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 1.0301361708460452e-05,
+      "loss": 0.2911,
+      "step": 100480
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.029838529652504e-05,
+      "loss": 0.2731,
+      "step": 100490
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.0295408884589627e-05,
+      "loss": 0.2788,
+      "step": 100500
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.0292432472654216e-05,
+      "loss": 0.2987,
+      "step": 100510
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 1.0289456060718805e-05,
+      "loss": 0.2919,
+      "step": 100520
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.0286479648783392e-05,
+      "loss": 0.2885,
+      "step": 100530
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.028350323684798e-05,
+      "loss": 0.297,
+      "step": 100540
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.0280526824912569e-05,
+      "loss": 0.2997,
+      "step": 100550
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.0277550412977157e-05,
+      "loss": 0.2869,
+      "step": 100560
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 1.0274574001041746e-05,
+      "loss": 0.2824,
+      "step": 100570
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0271597589106333e-05,
+      "loss": 0.3028,
+      "step": 100580
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0268621177170921e-05,
+      "loss": 0.284,
+      "step": 100590
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.026564476523551e-05,
+      "loss": 0.2935,
+      "step": 100600
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 1.0262668353300099e-05,
+      "loss": 0.2915,
+      "step": 100610
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.0259691941364687e-05,
+      "loss": 0.2728,
+      "step": 100620
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.0256715529429272e-05,
+      "loss": 0.3048,
+      "step": 100630
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.0253739117493861e-05,
+      "loss": 0.3021,
+      "step": 100640
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.025076270555845e-05,
+      "loss": 0.2909,
+      "step": 100650
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 1.0247786293623038e-05,
+      "loss": 0.2935,
+      "step": 100660
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0244809881687627e-05,
+      "loss": 0.2891,
+      "step": 100670
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0241833469752214e-05,
+      "loss": 0.2731,
+      "step": 100680
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.0238857057816802e-05,
+      "loss": 0.2991,
+      "step": 100690
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 1.023588064588139e-05,
+      "loss": 0.2912,
+      "step": 100700
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.023290423394598e-05,
+      "loss": 0.2937,
+      "step": 100710
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.0229927822010568e-05,
+      "loss": 0.2995,
+      "step": 100720
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.0226951410075155e-05,
+      "loss": 0.2893,
+      "step": 100730
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.0223974998139744e-05,
+      "loss": 0.2818,
+      "step": 100740
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 1.0220998586204332e-05,
+      "loss": 0.2824,
+      "step": 100750
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.021802217426892e-05,
+      "loss": 0.2903,
+      "step": 100760
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.021504576233351e-05,
+      "loss": 0.2843,
+      "step": 100770
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.0212069350398094e-05,
+      "loss": 0.2895,
+      "step": 100780
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 1.0209092938462683e-05,
+      "loss": 0.2887,
+      "step": 100790
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0206116526527272e-05,
+      "loss": 0.282,
+      "step": 100800
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.020314011459186e-05,
+      "loss": 0.2957,
+      "step": 100810
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0200163702656449e-05,
+      "loss": 0.3115,
+      "step": 100820
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0197187290721036e-05,
+      "loss": 0.2834,
+      "step": 100830
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 1.0194210878785624e-05,
+      "loss": 0.2873,
+      "step": 100840
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.0191234466850213e-05,
+      "loss": 0.2865,
+      "step": 100850
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.0188258054914802e-05,
+      "loss": 0.2994,
+      "step": 100860
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.018528164297939e-05,
+      "loss": 0.2957,
+      "step": 100870
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 1.0182305231043977e-05,
+      "loss": 0.2923,
+      "step": 100880
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0179328819108566e-05,
+      "loss": 0.2801,
+      "step": 100890
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0176352407173154e-05,
+      "loss": 0.2977,
+      "step": 100900
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0173375995237743e-05,
+      "loss": 0.3,
+      "step": 100910
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0170399583302331e-05,
+      "loss": 0.2908,
+      "step": 100920
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 1.0167423171366917e-05,
+      "loss": 0.2837,
+      "step": 100930
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.0164446759431505e-05,
+      "loss": 0.2861,
+      "step": 100940
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.0161470347496094e-05,
+      "loss": 0.2852,
+      "step": 100950
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.0158493935560682e-05,
+      "loss": 0.2808,
+      "step": 100960
+    },
+    {
+      "epoch": 22.42,
+      "learning_rate": 1.0155517523625271e-05,
+      "loss": 0.3117,
+      "step": 100970
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.015254111168986e-05,
+      "loss": 0.2908,
+      "step": 100980
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0149564699754446e-05,
+      "loss": 0.2888,
+      "step": 100990
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0146588287819035e-05,
+      "loss": 0.2779,
+      "step": 101000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0143611875883624e-05,
+      "loss": 0.2694,
+      "step": 101010
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 1.0140635463948212e-05,
+      "loss": 0.3044,
+      "step": 101020
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.01376590520128e-05,
+      "loss": 0.2868,
+      "step": 101030
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.0134682640077386e-05,
+      "loss": 0.3071,
+      "step": 101040
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.0131706228141975e-05,
+      "loss": 0.3003,
+      "step": 101050
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 1.0128729816206563e-05,
+      "loss": 0.2977,
+      "step": 101060
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0125753404271152e-05,
+      "loss": 0.3069,
+      "step": 101070
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.012277699233574e-05,
+      "loss": 0.2905,
+      "step": 101080
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0119800580400327e-05,
+      "loss": 0.2812,
+      "step": 101090
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0116824168464916e-05,
+      "loss": 0.28,
+      "step": 101100
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 1.0113847756529505e-05,
+      "loss": 0.3011,
+      "step": 101110
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0110871344594093e-05,
+      "loss": 0.2891,
+      "step": 101120
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0107894932658682e-05,
+      "loss": 0.2897,
+      "step": 101130
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0104918520723269e-05,
+      "loss": 0.2905,
+      "step": 101140
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 1.0101942108787857e-05,
+      "loss": 0.2801,
+      "step": 101150
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0098965696852446e-05,
+      "loss": 0.3017,
+      "step": 101160
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0095989284917034e-05,
+      "loss": 0.2837,
+      "step": 101170
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0093012872981623e-05,
+      "loss": 0.2847,
+      "step": 101180
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0090036461046208e-05,
+      "loss": 0.2812,
+      "step": 101190
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 1.0087060049110797e-05,
+      "loss": 0.3059,
+      "step": 101200
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0084083637175385e-05,
+      "loss": 0.2878,
+      "step": 101210
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0081107225239974e-05,
+      "loss": 0.2952,
+      "step": 101220
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.0078130813304563e-05,
+      "loss": 0.2868,
+      "step": 101230
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 1.007515440136915e-05,
+      "loss": 0.2969,
+      "step": 101240
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.0072177989433738e-05,
+      "loss": 0.2991,
+      "step": 101250
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.0069201577498327e-05,
+      "loss": 0.2997,
+      "step": 101260
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.0066225165562915e-05,
+      "loss": 0.2893,
+      "step": 101270
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.0063248753627504e-05,
+      "loss": 0.3006,
+      "step": 101280
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 1.006027234169209e-05,
+      "loss": 0.2788,
+      "step": 101290
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.005729592975668e-05,
+      "loss": 0.2985,
+      "step": 101300
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0054319517821268e-05,
+      "loss": 0.2865,
+      "step": 101310
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0051343105885856e-05,
+      "loss": 0.2838,
+      "step": 101320
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.0048366693950445e-05,
+      "loss": 0.2829,
+      "step": 101330
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 1.004539028201503e-05,
+      "loss": 0.3041,
+      "step": 101340
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.0042413870079619e-05,
+      "loss": 0.2853,
+      "step": 101350
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.0039437458144207e-05,
+      "loss": 0.2843,
+      "step": 101360
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.0036461046208796e-05,
+      "loss": 0.2816,
+      "step": 101370
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 1.0033484634273385e-05,
+      "loss": 0.2782,
+      "step": 101380
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.0030508222337973e-05,
+      "loss": 0.2924,
+      "step": 101390
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.002753181040256e-05,
+      "loss": 0.2913,
+      "step": 101400
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.0024555398467149e-05,
+      "loss": 0.2987,
+      "step": 101410
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.0021578986531737e-05,
+      "loss": 0.2866,
+      "step": 101420
+    },
+    {
+      "epoch": 22.52,
+      "learning_rate": 1.0018602574596326e-05,
+      "loss": 0.2811,
+      "step": 101430
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.0015626162660915e-05,
+      "loss": 0.2967,
+      "step": 101440
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.00126497507255e-05,
+      "loss": 0.2758,
+      "step": 101450
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.0009673338790088e-05,
+      "loss": 0.2854,
+      "step": 101460
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 1.0006696926854677e-05,
+      "loss": 0.2685,
+      "step": 101470
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.0003720514919266e-05,
+      "loss": 0.2831,
+      "step": 101480
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 1.0000744102983854e-05,
+      "loss": 0.2787,
+      "step": 101490
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 9.997767691048443e-06,
+      "loss": 0.3016,
+      "step": 101500
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 9.99479127911303e-06,
+      "loss": 0.3006,
+      "step": 101510
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 9.991814867177618e-06,
+      "loss": 0.2875,
+      "step": 101520
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.988838455242207e-06,
+      "loss": 0.2887,
+      "step": 101530
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.985862043306794e-06,
+      "loss": 0.2929,
+      "step": 101540
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.982885631371382e-06,
+      "loss": 0.2885,
+      "step": 101550
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 9.979909219435971e-06,
+      "loss": 0.2967,
+      "step": 101560
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.97693280750056e-06,
+      "loss": 0.2901,
+      "step": 101570
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.973956395565148e-06,
+      "loss": 0.3023,
+      "step": 101580
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.970979983629735e-06,
+      "loss": 0.2999,
+      "step": 101590
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.968003571694324e-06,
+      "loss": 0.2781,
+      "step": 101600
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.96502715975891e-06,
+      "loss": 0.2829,
+      "step": 101610
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.962050747823499e-06,
+      "loss": 0.2981,
+      "step": 101620
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.959074335888088e-06,
+      "loss": 0.2891,
+      "step": 101630
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.956097923952676e-06,
+      "loss": 0.2866,
+      "step": 101640
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 9.953121512017265e-06,
+      "loss": 0.2909,
+      "step": 101650
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.950145100081852e-06,
+      "loss": 0.2942,
+      "step": 101660
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.94716868814644e-06,
+      "loss": 0.3036,
+      "step": 101670
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.944192276211029e-06,
+      "loss": 0.2907,
+      "step": 101680
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.941215864275616e-06,
+      "loss": 0.2986,
+      "step": 101690
+    },
+    {
+      "epoch": 22.58,
+      "learning_rate": 9.938239452340204e-06,
+      "loss": 0.2949,
+      "step": 101700
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.935263040404793e-06,
+      "loss": 0.284,
+      "step": 101710
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.932286628469382e-06,
+      "loss": 0.2762,
+      "step": 101720
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.92931021653397e-06,
+      "loss": 0.2871,
+      "step": 101730
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.926333804598557e-06,
+      "loss": 0.2864,
+      "step": 101740
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.923357392663146e-06,
+      "loss": 0.2971,
+      "step": 101750
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.920380980727734e-06,
+      "loss": 0.2867,
+      "step": 101760
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.917404568792321e-06,
+      "loss": 0.2964,
+      "step": 101770
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.91442815685691e-06,
+      "loss": 0.2973,
+      "step": 101780
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 9.911451744921498e-06,
+      "loss": 0.2919,
+      "step": 101790
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.908475332986087e-06,
+      "loss": 0.2808,
+      "step": 101800
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.905498921050676e-06,
+      "loss": 0.2898,
+      "step": 101810
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.902522509115262e-06,
+      "loss": 0.2982,
+      "step": 101820
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.899546097179851e-06,
+      "loss": 0.3003,
+      "step": 101830
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.896569685244438e-06,
+      "loss": 0.3065,
+      "step": 101840
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.893593273309026e-06,
+      "loss": 0.2823,
+      "step": 101850
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.890616861373615e-06,
+      "loss": 0.2828,
+      "step": 101860
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.887640449438202e-06,
+      "loss": 0.3001,
+      "step": 101870
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 9.88466403750279e-06,
+      "loss": 0.2843,
+      "step": 101880
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 9.88168762556738e-06,
+      "loss": 0.2878,
+      "step": 101890
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 9.878711213631968e-06,
+      "loss": 0.2899,
+      "step": 101900
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 9.875734801696556e-06,
+      "loss": 0.3047,
+      "step": 101910
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 9.872758389761143e-06,
+      "loss": 0.2967,
+      "step": 101920
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.869781977825732e-06,
+      "loss": 0.2971,
+      "step": 101930
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.86680556589032e-06,
+      "loss": 0.2861,
+      "step": 101940
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.863829153954907e-06,
+      "loss": 0.2967,
+      "step": 101950
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.860852742019496e-06,
+      "loss": 0.2975,
+      "step": 101960
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.857876330084085e-06,
+      "loss": 0.2932,
+      "step": 101970
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 9.854899918148673e-06,
+      "loss": 0.2953,
+      "step": 101980
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 9.851923506213262e-06,
+      "loss": 0.2958,
+      "step": 101990
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 9.848947094277849e-06,
+      "loss": 0.2957,
+      "step": 102000
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 9.845970682342437e-06,
+      "loss": 0.2843,
+      "step": 102010
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.842994270407024e-06,
+      "loss": 0.2999,
+      "step": 102020
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.840017858471613e-06,
+      "loss": 0.2896,
+      "step": 102030
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.837041446536201e-06,
+      "loss": 0.2824,
+      "step": 102040
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.83406503460079e-06,
+      "loss": 0.2933,
+      "step": 102050
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 9.831088622665378e-06,
+      "loss": 0.2892,
+      "step": 102060
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.828112210729965e-06,
+      "loss": 0.2853,
+      "step": 102070
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.825135798794554e-06,
+      "loss": 0.2857,
+      "step": 102080
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.822159386859143e-06,
+      "loss": 0.2952,
+      "step": 102090
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.81918297492373e-06,
+      "loss": 0.2937,
+      "step": 102100
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.816206562988318e-06,
+      "loss": 0.3032,
+      "step": 102110
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.813230151052907e-06,
+      "loss": 0.2889,
+      "step": 102120
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.810253739117495e-06,
+      "loss": 0.2839,
+      "step": 102130
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.807277327182084e-06,
+      "loss": 0.2756,
+      "step": 102140
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 9.80430091524667e-06,
+      "loss": 0.2919,
+      "step": 102150
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.80132450331126e-06,
+      "loss": 0.278,
+      "step": 102160
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.798348091375848e-06,
+      "loss": 0.2954,
+      "step": 102170
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.795371679440435e-06,
+      "loss": 0.2724,
+      "step": 102180
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.792395267505023e-06,
+      "loss": 0.298,
+      "step": 102190
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.789418855569612e-06,
+      "loss": 0.3035,
+      "step": 102200
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.7864424436342e-06,
+      "loss": 0.2855,
+      "step": 102210
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.78346603169879e-06,
+      "loss": 0.2888,
+      "step": 102220
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.780489619763376e-06,
+      "loss": 0.2997,
+      "step": 102230
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 9.777513207827965e-06,
+      "loss": 0.2889,
+      "step": 102240
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.774536795892552e-06,
+      "loss": 0.2969,
+      "step": 102250
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.77156038395714e-06,
+      "loss": 0.2946,
+      "step": 102260
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.768583972021729e-06,
+      "loss": 0.2927,
+      "step": 102270
+    },
+    {
+      "epoch": 22.71,
+      "learning_rate": 9.765607560086316e-06,
+      "loss": 0.2749,
+      "step": 102280
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.762631148150904e-06,
+      "loss": 0.303,
+      "step": 102290
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.759654736215493e-06,
+      "loss": 0.2989,
+      "step": 102300
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.756678324280081e-06,
+      "loss": 0.2851,
+      "step": 102310
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.75370191234467e-06,
+      "loss": 0.2804,
+      "step": 102320
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.750725500409257e-06,
+      "loss": 0.2876,
+      "step": 102330
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.747749088473846e-06,
+      "loss": 0.2897,
+      "step": 102340
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.744772676538432e-06,
+      "loss": 0.29,
+      "step": 102350
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.741796264603021e-06,
+      "loss": 0.2829,
+      "step": 102360
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 9.73881985266761e-06,
+      "loss": 0.2894,
+      "step": 102370
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.735843440732198e-06,
+      "loss": 0.2922,
+      "step": 102380
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.732867028796787e-06,
+      "loss": 0.2951,
+      "step": 102390
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.729890616861375e-06,
+      "loss": 0.2887,
+      "step": 102400
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.726914204925962e-06,
+      "loss": 0.285,
+      "step": 102410
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 9.723937792990551e-06,
+      "loss": 0.3034,
+      "step": 102420
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.720961381055138e-06,
+      "loss": 0.2981,
+      "step": 102430
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.717984969119726e-06,
+      "loss": 0.3033,
+      "step": 102440
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.715008557184315e-06,
+      "loss": 0.2836,
+      "step": 102450
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.712032145248904e-06,
+      "loss": 0.2859,
+      "step": 102460
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.709055733313492e-06,
+      "loss": 0.2982,
+      "step": 102470
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.706079321378079e-06,
+      "loss": 0.2902,
+      "step": 102480
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.703102909442668e-06,
+      "loss": 0.2845,
+      "step": 102490
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.700126497507256e-06,
+      "loss": 0.2792,
+      "step": 102500
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 9.697150085571843e-06,
+      "loss": 0.2914,
+      "step": 102510
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.694173673636432e-06,
+      "loss": 0.2995,
+      "step": 102520
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.69119726170102e-06,
+      "loss": 0.2902,
+      "step": 102530
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.688220849765609e-06,
+      "loss": 0.2839,
+      "step": 102540
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.685244437830198e-06,
+      "loss": 0.2955,
+      "step": 102550
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 9.682268025894784e-06,
+      "loss": 0.2905,
+      "step": 102560
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 9.679291613959373e-06,
+      "loss": 0.3009,
+      "step": 102570
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 9.676315202023962e-06,
+      "loss": 0.2967,
+      "step": 102580
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 9.673338790088548e-06,
+      "loss": 0.295,
+      "step": 102590
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 9.670362378153137e-06,
+      "loss": 0.2817,
+      "step": 102600
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 9.667385966217726e-06,
+      "loss": 0.2845,
+      "step": 102610
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 9.664409554282314e-06,
+      "loss": 0.2962,
+      "step": 102620
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 9.661433142346903e-06,
+      "loss": 0.2978,
+      "step": 102630
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 9.65845673041149e-06,
+      "loss": 0.2938,
+      "step": 102640
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.655480318476078e-06,
+      "loss": 0.2779,
+      "step": 102650
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.652503906540665e-06,
+      "loss": 0.2777,
+      "step": 102660
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.649527494605254e-06,
+      "loss": 0.3048,
+      "step": 102670
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.646551082669842e-06,
+      "loss": 0.3045,
+      "step": 102680
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.643574670734431e-06,
+      "loss": 0.2779,
+      "step": 102690
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 9.64059825879902e-06,
+      "loss": 0.2856,
+      "step": 102700
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 9.637621846863607e-06,
+      "loss": 0.2844,
+      "step": 102710
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 9.634645434928195e-06,
+      "loss": 0.2943,
+      "step": 102720
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 9.631669022992784e-06,
+      "loss": 0.2937,
+      "step": 102730
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 9.62869261105737e-06,
+      "loss": 0.2786,
+      "step": 102740
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 9.62571619912196e-06,
+      "loss": 0.2773,
+      "step": 102750
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 9.622739787186546e-06,
+      "loss": 0.3075,
+      "step": 102760
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 9.619763375251135e-06,
+      "loss": 0.2841,
+      "step": 102770
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 9.616786963315723e-06,
+      "loss": 0.3,
+      "step": 102780
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.613810551380312e-06,
+      "loss": 0.294,
+      "step": 102790
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.6108341394449e-06,
+      "loss": 0.2807,
+      "step": 102800
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.607857727509489e-06,
+      "loss": 0.2853,
+      "step": 102810
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.604881315574076e-06,
+      "loss": 0.2766,
+      "step": 102820
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 9.601904903638665e-06,
+      "loss": 0.2985,
+      "step": 102830
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 9.598928491703251e-06,
+      "loss": 0.292,
+      "step": 102840
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 9.59595207976784e-06,
+      "loss": 0.2782,
+      "step": 102850
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 9.592975667832429e-06,
+      "loss": 0.2817,
+      "step": 102860
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 9.589999255897017e-06,
+      "loss": 0.2919,
+      "step": 102870
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 9.587022843961606e-06,
+      "loss": 0.2866,
+      "step": 102880
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 9.584046432026193e-06,
+      "loss": 0.2971,
+      "step": 102890
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 9.581070020090781e-06,
+      "loss": 0.2894,
+      "step": 102900
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 9.57809360815537e-06,
+      "loss": 0.2834,
+      "step": 102910
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 9.575117196219957e-06,
+      "loss": 0.2802,
+      "step": 102920
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 9.572140784284545e-06,
+      "loss": 0.2928,
+      "step": 102930
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 9.569164372349134e-06,
+      "loss": 0.2941,
+      "step": 102940
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 9.566187960413723e-06,
+      "loss": 0.2994,
+      "step": 102950
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 9.563211548478311e-06,
+      "loss": 0.2966,
+      "step": 102960
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 9.560235136542898e-06,
+      "loss": 0.2928,
+      "step": 102970
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 9.557258724607487e-06,
+      "loss": 0.2991,
+      "step": 102980
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 9.554282312672074e-06,
+      "loss": 0.2987,
+      "step": 102990
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 9.551305900736662e-06,
+      "loss": 0.3014,
+      "step": 103000
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.54832948880125e-06,
+      "loss": 0.2928,
+      "step": 103010
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.54535307686584e-06,
+      "loss": 0.2908,
+      "step": 103020
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.542376664930428e-06,
+      "loss": 0.2963,
+      "step": 103030
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.539400252995017e-06,
+      "loss": 0.2919,
+      "step": 103040
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.536423841059603e-06,
+      "loss": 0.2864,
+      "step": 103050
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 9.533447429124192e-06,
+      "loss": 0.3002,
+      "step": 103060
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 9.530471017188779e-06,
+      "loss": 0.2825,
+      "step": 103070
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 9.527494605253368e-06,
+      "loss": 0.2956,
+      "step": 103080
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 9.524518193317956e-06,
+      "loss": 0.2936,
+      "step": 103090
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 9.521541781382545e-06,
+      "loss": 0.3005,
+      "step": 103100
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 9.518565369447133e-06,
+      "loss": 0.2882,
+      "step": 103110
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 9.51558895751172e-06,
+      "loss": 0.291,
+      "step": 103120
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 9.512612545576309e-06,
+      "loss": 0.2801,
+      "step": 103130
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 9.509636133640897e-06,
+      "loss": 0.2858,
+      "step": 103140
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 9.506659721705484e-06,
+      "loss": 0.2915,
+      "step": 103150
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 9.503683309770073e-06,
+      "loss": 0.2805,
+      "step": 103160
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 9.50070689783466e-06,
+      "loss": 0.2918,
+      "step": 103170
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 9.497730485899248e-06,
+      "loss": 0.2945,
+      "step": 103180
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 9.494754073963837e-06,
+      "loss": 0.2897,
+      "step": 103190
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 9.491777662028426e-06,
+      "loss": 0.2853,
+      "step": 103200
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 9.488801250093014e-06,
+      "loss": 0.2744,
+      "step": 103210
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 9.485824838157601e-06,
+      "loss": 0.2968,
+      "step": 103220
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 9.48284842622219e-06,
+      "loss": 0.2885,
+      "step": 103230
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 9.479872014286778e-06,
+      "loss": 0.2855,
+      "step": 103240
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 9.476895602351365e-06,
+      "loss": 0.2808,
+      "step": 103250
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 9.473919190415954e-06,
+      "loss": 0.286,
+      "step": 103260
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 9.470942778480542e-06,
+      "loss": 0.2965,
+      "step": 103270
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 9.467966366545131e-06,
+      "loss": 0.2801,
+      "step": 103280
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 9.46498995460972e-06,
+      "loss": 0.2978,
+      "step": 103290
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 9.462013542674306e-06,
+      "loss": 0.2918,
+      "step": 103300
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 9.459037130738895e-06,
+      "loss": 0.3037,
+      "step": 103310
+    },
+    {
+      "epoch": 22.94,
+      "learning_rate": 9.456060718803484e-06,
+      "loss": 0.2826,
+      "step": 103320
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 9.45308430686807e-06,
+      "loss": 0.2851,
+      "step": 103330
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 9.450107894932659e-06,
+      "loss": 0.287,
+      "step": 103340
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 9.447131482997248e-06,
+      "loss": 0.2801,
+      "step": 103350
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 9.444155071061836e-06,
+      "loss": 0.2905,
+      "step": 103360
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.441178659126425e-06,
+      "loss": 0.3033,
+      "step": 103370
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.438202247191012e-06,
+      "loss": 0.288,
+      "step": 103380
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.4352258352556e-06,
+      "loss": 0.3016,
+      "step": 103390
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.432249423320187e-06,
+      "loss": 0.3037,
+      "step": 103400
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.429273011384776e-06,
+      "loss": 0.2847,
+      "step": 103410
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 9.426296599449364e-06,
+      "loss": 0.2853,
+      "step": 103420
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 9.423320187513953e-06,
+      "loss": 0.2958,
+      "step": 103430
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 9.420343775578542e-06,
+      "loss": 0.2839,
+      "step": 103440
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 9.41736736364313e-06,
+      "loss": 0.2735,
+      "step": 103450
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 9.414390951707717e-06,
+      "loss": 0.2847,
+      "step": 103460
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 9.411414539772306e-06,
+      "loss": 0.3046,
+      "step": 103470
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 9.408438127836893e-06,
+      "loss": 0.2758,
+      "step": 103480
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 9.405461715901481e-06,
+      "loss": 0.2902,
+      "step": 103490
+    },
+    {
+      "epoch": 22.98,
+      "learning_rate": 9.40248530396607e-06,
+      "loss": 0.2969,
+      "step": 103500
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 9.399508892030658e-06,
+      "loss": 0.2856,
+      "step": 103510
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 9.396532480095247e-06,
+      "loss": 0.2918,
+      "step": 103520
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 9.393556068159834e-06,
+      "loss": 0.2911,
+      "step": 103530
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 9.390579656224422e-06,
+      "loss": 0.2896,
+      "step": 103540
+    },
+    {
+      "epoch": 22.99,
+      "eval_cer": 4.547008547008547,
+      "eval_loss": 0.5644484758377075,
+      "eval_runtime": 1.2911,
+      "eval_samples_per_second": 7.745,
+      "eval_steps_per_second": 0.775,
+      "eval_wer": 0.6363636363636364,
+      "step": 103546
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 9.387603244289011e-06,
+      "loss": 0.2893,
+      "step": 103550
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 9.384626832353598e-06,
+      "loss": 0.2872,
+      "step": 103560
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 9.381650420418187e-06,
+      "loss": 0.3037,
+      "step": 103570
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 9.378674008482775e-06,
+      "loss": 0.2976,
+      "step": 103580
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 9.375697596547362e-06,
+      "loss": 0.2867,
+      "step": 103590
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.37272118461195e-06,
+      "loss": 0.2785,
+      "step": 103600
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.36974477267654e-06,
+      "loss": 0.2901,
+      "step": 103610
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.366768360741128e-06,
+      "loss": 0.281,
+      "step": 103620
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.363791948805715e-06,
+      "loss": 0.2853,
+      "step": 103630
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.360815536870303e-06,
+      "loss": 0.2859,
+      "step": 103640
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.357839124934892e-06,
+      "loss": 0.2763,
+      "step": 103650
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.354862712999479e-06,
+      "loss": 0.2854,
+      "step": 103660
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.351886301064067e-06,
+      "loss": 0.2688,
+      "step": 103670
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 9.348909889128656e-06,
+      "loss": 0.2871,
+      "step": 103680
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 9.345933477193245e-06,
+      "loss": 0.2942,
+      "step": 103690
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 9.342957065257833e-06,
+      "loss": 0.2843,
+      "step": 103700
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 9.33998065332242e-06,
+      "loss": 0.2729,
+      "step": 103710
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 9.337004241387009e-06,
+      "loss": 0.3009,
+      "step": 103720
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.334027829451597e-06,
+      "loss": 0.2727,
+      "step": 103730
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.331051417516184e-06,
+      "loss": 0.2885,
+      "step": 103740
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.328075005580773e-06,
+      "loss": 0.2846,
+      "step": 103750
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.325098593645361e-06,
+      "loss": 0.2957,
+      "step": 103760
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.32212218170995e-06,
+      "loss": 0.2903,
+      "step": 103770
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 9.319145769774539e-06,
+      "loss": 0.2749,
+      "step": 103780
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 9.316169357839125e-06,
+      "loss": 0.2879,
+      "step": 103790
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 9.313192945903714e-06,
+      "loss": 0.2894,
+      "step": 103800
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 9.310216533968301e-06,
+      "loss": 0.2889,
+      "step": 103810
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.30724012203289e-06,
+      "loss": 0.2891,
+      "step": 103820
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.304263710097478e-06,
+      "loss": 0.2898,
+      "step": 103830
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.301287298162067e-06,
+      "loss": 0.2905,
+      "step": 103840
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.298310886226655e-06,
+      "loss": 0.2891,
+      "step": 103850
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 9.295334474291242e-06,
+      "loss": 0.2698,
+      "step": 103860
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.29235806235583e-06,
+      "loss": 0.2944,
+      "step": 103870
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.28938165042042e-06,
+      "loss": 0.2996,
+      "step": 103880
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.286405238485006e-06,
+      "loss": 0.2707,
+      "step": 103890
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.283428826549595e-06,
+      "loss": 0.2892,
+      "step": 103900
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.280452414614183e-06,
+      "loss": 0.2884,
+      "step": 103910
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.277476002678772e-06,
+      "loss": 0.2819,
+      "step": 103920
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.27449959074336e-06,
+      "loss": 0.2841,
+      "step": 103930
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.271523178807948e-06,
+      "loss": 0.2783,
+      "step": 103940
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 9.268546766872536e-06,
+      "loss": 0.2937,
+      "step": 103950
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.265570354937125e-06,
+      "loss": 0.2933,
+      "step": 103960
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.262593943001712e-06,
+      "loss": 0.2769,
+      "step": 103970
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.2596175310663e-06,
+      "loss": 0.2824,
+      "step": 103980
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.256641119130889e-06,
+      "loss": 0.2754,
+      "step": 103990
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.253664707195477e-06,
+      "loss": 0.2775,
+      "step": 104000
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.250688295260066e-06,
+      "loss": 0.2844,
+      "step": 104010
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.247711883324653e-06,
+      "loss": 0.2914,
+      "step": 104020
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.244735471389242e-06,
+      "loss": 0.2767,
+      "step": 104030
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 9.241759059453828e-06,
+      "loss": 0.2814,
+      "step": 104040
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 9.238782647518417e-06,
+      "loss": 0.295,
+      "step": 104050
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 9.235806235583006e-06,
+      "loss": 0.279,
+      "step": 104060
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 9.232829823647592e-06,
+      "loss": 0.2942,
+      "step": 104070
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 9.229853411712181e-06,
+      "loss": 0.2668,
+      "step": 104080
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.22687699977677e-06,
+      "loss": 0.2766,
+      "step": 104090
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.223900587841358e-06,
+      "loss": 0.2831,
+      "step": 104100
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.220924175905947e-06,
+      "loss": 0.271,
+      "step": 104110
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.217947763970534e-06,
+      "loss": 0.2904,
+      "step": 104120
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.214971352035122e-06,
+      "loss": 0.2824,
+      "step": 104130
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 9.211994940099711e-06,
+      "loss": 0.2744,
+      "step": 104140
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 9.209018528164298e-06,
+      "loss": 0.2695,
+      "step": 104150
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 9.206042116228886e-06,
+      "loss": 0.2787,
+      "step": 104160
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 9.203065704293475e-06,
+      "loss": 0.2947,
+      "step": 104170
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.200089292358064e-06,
+      "loss": 0.2802,
+      "step": 104180
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.197112880422652e-06,
+      "loss": 0.2849,
+      "step": 104190
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.194136468487239e-06,
+      "loss": 0.28,
+      "step": 104200
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.191160056551828e-06,
+      "loss": 0.2882,
+      "step": 104210
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 9.188183644616415e-06,
+      "loss": 0.272,
+      "step": 104220
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.185207232681003e-06,
+      "loss": 0.2916,
+      "step": 104230
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.182230820745592e-06,
+      "loss": 0.2829,
+      "step": 104240
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.17925440881018e-06,
+      "loss": 0.2822,
+      "step": 104250
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.176277996874769e-06,
+      "loss": 0.2926,
+      "step": 104260
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.173301584939356e-06,
+      "loss": 0.2911,
+      "step": 104270
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.170325173003944e-06,
+      "loss": 0.2832,
+      "step": 104280
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.167348761068533e-06,
+      "loss": 0.2789,
+      "step": 104290
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.16437234913312e-06,
+      "loss": 0.2879,
+      "step": 104300
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 9.161395937197709e-06,
+      "loss": 0.2884,
+      "step": 104310
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.158419525262297e-06,
+      "loss": 0.28,
+      "step": 104320
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.155443113326886e-06,
+      "loss": 0.2618,
+      "step": 104330
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.152466701391474e-06,
+      "loss": 0.2782,
+      "step": 104340
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.149490289456061e-06,
+      "loss": 0.2865,
+      "step": 104350
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.14651387752065e-06,
+      "loss": 0.2987,
+      "step": 104360
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.143537465585238e-06,
+      "loss": 0.2907,
+      "step": 104370
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.140561053649825e-06,
+      "loss": 0.2868,
+      "step": 104380
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.137584641714414e-06,
+      "loss": 0.2703,
+      "step": 104390
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 9.134608229779003e-06,
+      "loss": 0.2731,
+      "step": 104400
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.131631817843591e-06,
+      "loss": 0.2935,
+      "step": 104410
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.12865540590818e-06,
+      "loss": 0.2975,
+      "step": 104420
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.125678993972767e-06,
+      "loss": 0.2833,
+      "step": 104430
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 9.122702582037355e-06,
+      "loss": 0.2809,
+      "step": 104440
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.119726170101942e-06,
+      "loss": 0.2941,
+      "step": 104450
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.11674975816653e-06,
+      "loss": 0.2769,
+      "step": 104460
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.11377334623112e-06,
+      "loss": 0.281,
+      "step": 104470
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.110796934295706e-06,
+      "loss": 0.2851,
+      "step": 104480
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.107820522360295e-06,
+      "loss": 0.2884,
+      "step": 104490
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 9.104844110424883e-06,
+      "loss": 0.2804,
+      "step": 104500
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 9.101867698489472e-06,
+      "loss": 0.2785,
+      "step": 104510
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 9.09889128655406e-06,
+      "loss": 0.294,
+      "step": 104520
+    },
+    {
+      "epoch": 23.21,
+      "learning_rate": 9.095914874618647e-06,
+      "loss": 0.2726,
+      "step": 104530
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.092938462683236e-06,
+      "loss": 0.2909,
+      "step": 104540
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.089962050747825e-06,
+      "loss": 0.2889,
+      "step": 104550
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.086985638812412e-06,
+      "loss": 0.2964,
+      "step": 104560
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.084009226877e-06,
+      "loss": 0.2811,
+      "step": 104570
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 9.081032814941589e-06,
+      "loss": 0.2897,
+      "step": 104580
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.078056403006177e-06,
+      "loss": 0.2742,
+      "step": 104590
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.075079991070766e-06,
+      "loss": 0.2753,
+      "step": 104600
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.072103579135353e-06,
+      "loss": 0.2867,
+      "step": 104610
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.069127167199941e-06,
+      "loss": 0.2975,
+      "step": 104620
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.066150755264528e-06,
+      "loss": 0.2847,
+      "step": 104630
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.063174343329117e-06,
+      "loss": 0.2831,
+      "step": 104640
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.060197931393705e-06,
+      "loss": 0.283,
+      "step": 104650
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.057221519458294e-06,
+      "loss": 0.2768,
+      "step": 104660
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 9.054245107522883e-06,
+      "loss": 0.2907,
+      "step": 104670
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.05126869558747e-06,
+      "loss": 0.282,
+      "step": 104680
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.048292283652058e-06,
+      "loss": 0.2909,
+      "step": 104690
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.045315871716647e-06,
+      "loss": 0.2758,
+      "step": 104700
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.042339459781234e-06,
+      "loss": 0.2884,
+      "step": 104710
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.039363047845822e-06,
+      "loss": 0.2932,
+      "step": 104720
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.03638663591041e-06,
+      "loss": 0.2783,
+      "step": 104730
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.033410223975e-06,
+      "loss": 0.2945,
+      "step": 104740
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.030433812039588e-06,
+      "loss": 0.2875,
+      "step": 104750
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 9.027457400104175e-06,
+      "loss": 0.2924,
+      "step": 104760
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.024480988168763e-06,
+      "loss": 0.2758,
+      "step": 104770
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.021504576233352e-06,
+      "loss": 0.2703,
+      "step": 104780
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.018528164297939e-06,
+      "loss": 0.2898,
+      "step": 104790
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 9.015551752362528e-06,
+      "loss": 0.2946,
+      "step": 104800
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.012575340427116e-06,
+      "loss": 0.2906,
+      "step": 104810
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.009598928491705e-06,
+      "loss": 0.2847,
+      "step": 104820
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.006622516556293e-06,
+      "loss": 0.2823,
+      "step": 104830
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.00364610462088e-06,
+      "loss": 0.2748,
+      "step": 104840
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.000669692685469e-06,
+      "loss": 0.2833,
+      "step": 104850
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 8.997693280750056e-06,
+      "loss": 0.275,
+      "step": 104860
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 8.994716868814644e-06,
+      "loss": 0.2809,
+      "step": 104870
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 8.991740456879233e-06,
+      "loss": 0.2764,
+      "step": 104880
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 8.988764044943822e-06,
+      "loss": 0.2832,
+      "step": 104890
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 8.98578763300841e-06,
+      "loss": 0.289,
+      "step": 104900
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 8.982811221072997e-06,
+      "loss": 0.2739,
+      "step": 104910
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 8.979834809137586e-06,
+      "loss": 0.2958,
+      "step": 104920
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 8.976858397202174e-06,
+      "loss": 0.2818,
+      "step": 104930
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 8.973881985266761e-06,
+      "loss": 0.2895,
+      "step": 104940
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 8.97090557333135e-06,
+      "loss": 0.2917,
+      "step": 104950
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 8.967929161395937e-06,
+      "loss": 0.2809,
+      "step": 104960
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 8.964952749460525e-06,
+      "loss": 0.2826,
+      "step": 104970
+    },
+    {
+      "epoch": 23.31,
+      "learning_rate": 8.961976337525114e-06,
+      "loss": 0.296,
+      "step": 104980
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 8.958999925589702e-06,
+      "loss": 0.2797,
+      "step": 104990
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 8.956023513654291e-06,
+      "loss": 0.3047,
+      "step": 105000
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 8.95304710171888e-06,
+      "loss": 0.286,
+      "step": 105010
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 8.950070689783466e-06,
+      "loss": 0.2782,
+      "step": 105020
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 8.947094277848055e-06,
+      "loss": 0.286,
+      "step": 105030
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 8.944117865912642e-06,
+      "loss": 0.2805,
+      "step": 105040
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 8.94114145397723e-06,
+      "loss": 0.2809,
+      "step": 105050
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 8.938165042041819e-06,
+      "loss": 0.2835,
+      "step": 105060
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 8.935188630106408e-06,
+      "loss": 0.2674,
+      "step": 105070
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 8.932212218170996e-06,
+      "loss": 0.2924,
+      "step": 105080
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 8.929235806235583e-06,
+      "loss": 0.2904,
+      "step": 105090
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 8.926259394300172e-06,
+      "loss": 0.2749,
+      "step": 105100
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 8.92328298236476e-06,
+      "loss": 0.2846,
+      "step": 105110
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 8.920306570429347e-06,
+      "loss": 0.2717,
+      "step": 105120
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 8.917330158493936e-06,
+      "loss": 0.294,
+      "step": 105130
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 8.914353746558524e-06,
+      "loss": 0.2808,
+      "step": 105140
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 8.911377334623113e-06,
+      "loss": 0.2801,
+      "step": 105150
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 8.908400922687702e-06,
+      "loss": 0.2944,
+      "step": 105160
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 8.905424510752289e-06,
+      "loss": 0.2791,
+      "step": 105170
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 8.902448098816877e-06,
+      "loss": 0.2731,
+      "step": 105180
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 8.899471686881466e-06,
+      "loss": 0.3083,
+      "step": 105190
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 8.896495274946053e-06,
+      "loss": 0.2865,
+      "step": 105200
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 8.893518863010641e-06,
+      "loss": 0.279,
+      "step": 105210
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 8.89054245107523e-06,
+      "loss": 0.2938,
+      "step": 105220
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 8.887566039139818e-06,
+      "loss": 0.2922,
+      "step": 105230
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 8.884589627204407e-06,
+      "loss": 0.296,
+      "step": 105240
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 8.881613215268994e-06,
+      "loss": 0.2879,
+      "step": 105250
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 8.878636803333583e-06,
+      "loss": 0.2865,
+      "step": 105260
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 8.87566039139817e-06,
+      "loss": 0.2792,
+      "step": 105270
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 8.872683979462758e-06,
+      "loss": 0.2837,
+      "step": 105280
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 8.869707567527347e-06,
+      "loss": 0.2785,
+      "step": 105290
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 8.866731155591935e-06,
+      "loss": 0.2964,
+      "step": 105300
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 8.863754743656524e-06,
+      "loss": 0.287,
+      "step": 105310
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 8.86077833172111e-06,
+      "loss": 0.2931,
+      "step": 105320
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 8.8578019197857e-06,
+      "loss": 0.2877,
+      "step": 105330
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 8.854825507850288e-06,
+      "loss": 0.2904,
+      "step": 105340
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 8.851849095914875e-06,
+      "loss": 0.2894,
+      "step": 105350
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 8.848872683979463e-06,
+      "loss": 0.275,
+      "step": 105360
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 8.84589627204405e-06,
+      "loss": 0.2887,
+      "step": 105370
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 8.842919860108639e-06,
+      "loss": 0.2926,
+      "step": 105380
+    },
+    {
+      "epoch": 23.4,
+      "learning_rate": 8.839943448173227e-06,
+      "loss": 0.2908,
+      "step": 105390
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 8.836967036237816e-06,
+      "loss": 0.2768,
+      "step": 105400
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 8.833990624302405e-06,
+      "loss": 0.2933,
+      "step": 105410
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 8.831014212366993e-06,
+      "loss": 0.2944,
+      "step": 105420
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 8.82803780043158e-06,
+      "loss": 0.2975,
+      "step": 105430
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 8.825061388496169e-06,
+      "loss": 0.2921,
+      "step": 105440
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 8.822084976560756e-06,
+      "loss": 0.2902,
+      "step": 105450
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 8.819108564625344e-06,
+      "loss": 0.29,
+      "step": 105460
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 8.816132152689933e-06,
+      "loss": 0.2983,
+      "step": 105470
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 8.813155740754521e-06,
+      "loss": 0.2862,
+      "step": 105480
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 8.81017932881911e-06,
+      "loss": 0.2788,
+      "step": 105490
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 8.807202916883697e-06,
+      "loss": 0.2936,
+      "step": 105500
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 8.804226504948285e-06,
+      "loss": 0.2742,
+      "step": 105510
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 8.801250093012874e-06,
+      "loss": 0.2948,
+      "step": 105520
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 8.798273681077461e-06,
+      "loss": 0.2852,
+      "step": 105530
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 8.79529726914205e-06,
+      "loss": 0.2879,
+      "step": 105540
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 8.792320857206638e-06,
+      "loss": 0.2863,
+      "step": 105550
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 8.789344445271227e-06,
+      "loss": 0.2775,
+      "step": 105560
+    },
+    {
+      "epoch": 23.44,
+      "learning_rate": 8.786368033335815e-06,
+      "loss": 0.2673,
+      "step": 105570
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 8.783391621400402e-06,
+      "loss": 0.2858,
+      "step": 105580
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 8.780415209464991e-06,
+      "loss": 0.2769,
+      "step": 105590
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 8.777438797529578e-06,
+      "loss": 0.3005,
+      "step": 105600
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 8.774462385594166e-06,
+      "loss": 0.2836,
+      "step": 105610
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 8.771485973658755e-06,
+      "loss": 0.2813,
+      "step": 105620
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 8.768509561723344e-06,
+      "loss": 0.2895,
+      "step": 105630
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 8.765533149787932e-06,
+      "loss": 0.2986,
+      "step": 105640
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 8.76255673785252e-06,
+      "loss": 0.3064,
+      "step": 105650
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 8.759580325917108e-06,
+      "loss": 0.2808,
+      "step": 105660
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 8.756603913981696e-06,
+      "loss": 0.2801,
+      "step": 105670
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 8.753627502046283e-06,
+      "loss": 0.2882,
+      "step": 105680
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 8.750651090110872e-06,
+      "loss": 0.2898,
+      "step": 105690
+    },
+    {
+      "epoch": 23.47,
+      "learning_rate": 8.74767467817546e-06,
+      "loss": 0.2712,
+      "step": 105700
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 8.744698266240049e-06,
+      "loss": 0.286,
+      "step": 105710
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 8.741721854304637e-06,
+      "loss": 0.2875,
+      "step": 105720
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 8.738745442369224e-06,
+      "loss": 0.2853,
+      "step": 105730
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 8.735769030433813e-06,
+      "loss": 0.2917,
+      "step": 105740
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 8.732792618498402e-06,
+      "loss": 0.2925,
+      "step": 105750
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 8.729816206562988e-06,
+      "loss": 0.2833,
+      "step": 105760
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 8.726839794627577e-06,
+      "loss": 0.2931,
+      "step": 105770
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 8.723863382692166e-06,
+      "loss": 0.2958,
+      "step": 105780
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 8.720886970756753e-06,
+      "loss": 0.2795,
+      "step": 105790
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 8.717910558821341e-06,
+      "loss": 0.2951,
+      "step": 105800
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 8.71493414688593e-06,
+      "loss": 0.2826,
+      "step": 105810
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 8.711957734950518e-06,
+      "loss": 0.283,
+      "step": 105820
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 8.708981323015105e-06,
+      "loss": 0.287,
+      "step": 105830
+    },
+    {
+      "epoch": 23.5,
+      "learning_rate": 8.706004911079694e-06,
+      "loss": 0.2932,
+      "step": 105840
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 8.703028499144282e-06,
+      "loss": 0.3109,
+      "step": 105850
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 8.70005208720887e-06,
+      "loss": 0.301,
+      "step": 105860
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 8.697075675273458e-06,
+      "loss": 0.2762,
+      "step": 105870
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 8.694099263338046e-06,
+      "loss": 0.2705,
+      "step": 105880
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 8.691122851402635e-06,
+      "loss": 0.2749,
+      "step": 105890
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 8.688146439467224e-06,
+      "loss": 0.2846,
+      "step": 105900
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 8.68517002753181e-06,
+      "loss": 0.2817,
+      "step": 105910
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 8.6821936155964e-06,
+      "loss": 0.2855,
+      "step": 105920
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 8.679217203660988e-06,
+      "loss": 0.2839,
+      "step": 105930
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 8.676240791725575e-06,
+      "loss": 0.282,
+      "step": 105940
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 8.673264379790163e-06,
+      "loss": 0.2904,
+      "step": 105950
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 8.670287967854752e-06,
+      "loss": 0.2819,
+      "step": 105960
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 8.66731155591934e-06,
+      "loss": 0.2967,
+      "step": 105970
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 8.664335143983929e-06,
+      "loss": 0.2815,
+      "step": 105980
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 8.661358732048516e-06,
+      "loss": 0.2881,
+      "step": 105990
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 8.658382320113105e-06,
+      "loss": 0.2798,
+      "step": 106000
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 8.655405908177691e-06,
+      "loss": 0.2727,
+      "step": 106010
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 8.65242949624228e-06,
+      "loss": 0.2875,
+      "step": 106020
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 8.649453084306869e-06,
+      "loss": 0.2815,
+      "step": 106030
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 8.646476672371457e-06,
+      "loss": 0.283,
+      "step": 106040
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 8.643500260436046e-06,
+      "loss": 0.2998,
+      "step": 106050
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 8.640523848500634e-06,
+      "loss": 0.2768,
+      "step": 106060
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 8.637547436565221e-06,
+      "loss": 0.2761,
+      "step": 106070
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 8.63457102462981e-06,
+      "loss": 0.2902,
+      "step": 106080
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 8.631594612694397e-06,
+      "loss": 0.2821,
+      "step": 106090
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 8.628618200758985e-06,
+      "loss": 0.3036,
+      "step": 106100
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 8.625641788823574e-06,
+      "loss": 0.2971,
+      "step": 106110
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 8.622665376888163e-06,
+      "loss": 0.2763,
+      "step": 106120
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 8.619688964952751e-06,
+      "loss": 0.2946,
+      "step": 106130
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 8.616712553017338e-06,
+      "loss": 0.2919,
+      "step": 106140
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 8.613736141081927e-06,
+      "loss": 0.2963,
+      "step": 106150
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 8.610759729146515e-06,
+      "loss": 0.2837,
+      "step": 106160
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 8.607783317211102e-06,
+      "loss": 0.285,
+      "step": 106170
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 8.60480690527569e-06,
+      "loss": 0.2876,
+      "step": 106180
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 8.60183049334028e-06,
+      "loss": 0.305,
+      "step": 106190
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 8.598854081404868e-06,
+      "loss": 0.2747,
+      "step": 106200
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 8.595877669469457e-06,
+      "loss": 0.2827,
+      "step": 106210
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 8.592901257534043e-06,
+      "loss": 0.2881,
+      "step": 106220
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 8.589924845598632e-06,
+      "loss": 0.2807,
+      "step": 106230
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 8.586948433663219e-06,
+      "loss": 0.2829,
+      "step": 106240
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 8.583972021727807e-06,
+      "loss": 0.2914,
+      "step": 106250
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 8.580995609792396e-06,
+      "loss": 0.2789,
+      "step": 106260
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 8.578019197856983e-06,
+      "loss": 0.2822,
+      "step": 106270
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 8.575042785921572e-06,
+      "loss": 0.2915,
+      "step": 106280
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 8.57206637398616e-06,
+      "loss": 0.2762,
+      "step": 106290
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 8.569089962050749e-06,
+      "loss": 0.2826,
+      "step": 106300
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 8.566113550115337e-06,
+      "loss": 0.2815,
+      "step": 106310
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 8.563137138179924e-06,
+      "loss": 0.2669,
+      "step": 106320
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 8.560160726244513e-06,
+      "loss": 0.2886,
+      "step": 106330
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 8.557184314309101e-06,
+      "loss": 0.2905,
+      "step": 106340
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 8.554207902373688e-06,
+      "loss": 0.2836,
+      "step": 106350
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 8.551231490438277e-06,
+      "loss": 0.2849,
+      "step": 106360
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 8.548255078502866e-06,
+      "loss": 0.2732,
+      "step": 106370
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 8.545278666567454e-06,
+      "loss": 0.2929,
+      "step": 106380
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 8.542302254632043e-06,
+      "loss": 0.289,
+      "step": 106390
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 8.53932584269663e-06,
+      "loss": 0.2912,
+      "step": 106400
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 8.536349430761218e-06,
+      "loss": 0.2774,
+      "step": 106410
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 8.533373018825805e-06,
+      "loss": 0.278,
+      "step": 106420
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 8.530396606890394e-06,
+      "loss": 0.2924,
+      "step": 106430
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 8.527420194954982e-06,
+      "loss": 0.2812,
+      "step": 106440
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 8.524443783019571e-06,
+      "loss": 0.2754,
+      "step": 106450
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 8.52146737108416e-06,
+      "loss": 0.2863,
+      "step": 106460
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 8.518490959148746e-06,
+      "loss": 0.2744,
+      "step": 106470
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 8.515514547213335e-06,
+      "loss": 0.2918,
+      "step": 106480
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 8.512538135277924e-06,
+      "loss": 0.2757,
+      "step": 106490
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 8.50956172334251e-06,
+      "loss": 0.2836,
+      "step": 106500
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 8.506585311407099e-06,
+      "loss": 0.2957,
+      "step": 106510
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 8.503608899471688e-06,
+      "loss": 0.2857,
+      "step": 106520
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 8.500632487536276e-06,
+      "loss": 0.3051,
+      "step": 106530
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 8.497656075600865e-06,
+      "loss": 0.2871,
+      "step": 106540
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 8.494679663665452e-06,
+      "loss": 0.2872,
+      "step": 106550
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 8.49170325173004e-06,
+      "loss": 0.2752,
+      "step": 106560
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 8.488726839794629e-06,
+      "loss": 0.2812,
+      "step": 106570
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 8.485750427859216e-06,
+      "loss": 0.2854,
+      "step": 106580
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 8.482774015923804e-06,
+      "loss": 0.2822,
+      "step": 106590
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 8.479797603988393e-06,
+      "loss": 0.2819,
+      "step": 106600
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 8.476821192052982e-06,
+      "loss": 0.2901,
+      "step": 106610
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 8.47384478011757e-06,
+      "loss": 0.2883,
+      "step": 106620
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 8.470868368182157e-06,
+      "loss": 0.2813,
+      "step": 106630
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 8.467891956246746e-06,
+      "loss": 0.2819,
+      "step": 106640
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 8.464915544311333e-06,
+      "loss": 0.2953,
+      "step": 106650
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 8.461939132375921e-06,
+      "loss": 0.2827,
+      "step": 106660
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 8.45896272044051e-06,
+      "loss": 0.3049,
+      "step": 106670
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 8.455986308505097e-06,
+      "loss": 0.2852,
+      "step": 106680
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 8.453009896569685e-06,
+      "loss": 0.2779,
+      "step": 106690
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 8.450033484634274e-06,
+      "loss": 0.2975,
+      "step": 106700
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 8.447057072698862e-06,
+      "loss": 0.2796,
+      "step": 106710
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 8.444080660763451e-06,
+      "loss": 0.2675,
+      "step": 106720
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 8.441104248828038e-06,
+      "loss": 0.2798,
+      "step": 106730
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 8.438127836892627e-06,
+      "loss": 0.2765,
+      "step": 106740
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 8.435151424957215e-06,
+      "loss": 0.2803,
+      "step": 106750
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 8.432175013021802e-06,
+      "loss": 0.2838,
+      "step": 106760
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 8.42919860108639e-06,
+      "loss": 0.2838,
+      "step": 106770
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 8.42622218915098e-06,
+      "loss": 0.2874,
+      "step": 106780
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 8.423245777215568e-06,
+      "loss": 0.276,
+      "step": 106790
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 8.420269365280156e-06,
+      "loss": 0.2753,
+      "step": 106800
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 8.417292953344743e-06,
+      "loss": 0.2836,
+      "step": 106810
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 8.414316541409332e-06,
+      "loss": 0.2801,
+      "step": 106820
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 8.411340129473919e-06,
+      "loss": 0.2762,
+      "step": 106830
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 8.408363717538507e-06,
+      "loss": 0.2885,
+      "step": 106840
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 8.405387305603096e-06,
+      "loss": 0.2717,
+      "step": 106850
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 8.402410893667685e-06,
+      "loss": 0.2854,
+      "step": 106860
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 8.399434481732273e-06,
+      "loss": 0.282,
+      "step": 106870
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 8.39645806979686e-06,
+      "loss": 0.2744,
+      "step": 106880
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 8.393481657861449e-06,
+      "loss": 0.2842,
+      "step": 106890
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 8.390505245926037e-06,
+      "loss": 0.2779,
+      "step": 106900
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 8.387528833990624e-06,
+      "loss": 0.287,
+      "step": 106910
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 8.384552422055213e-06,
+      "loss": 0.2772,
+      "step": 106920
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 8.381576010119801e-06,
+      "loss": 0.2886,
+      "step": 106930
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 8.37859959818439e-06,
+      "loss": 0.2867,
+      "step": 106940
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 8.375623186248979e-06,
+      "loss": 0.2882,
+      "step": 106950
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 8.372646774313565e-06,
+      "loss": 0.276,
+      "step": 106960
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 8.369670362378154e-06,
+      "loss": 0.2784,
+      "step": 106970
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 8.366693950442743e-06,
+      "loss": 0.2794,
+      "step": 106980
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 8.36371753850733e-06,
+      "loss": 0.2847,
+      "step": 106990
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 8.360741126571918e-06,
+      "loss": 0.2733,
+      "step": 107000
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 8.357764714636507e-06,
+      "loss": 0.306,
+      "step": 107010
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 8.354788302701095e-06,
+      "loss": 0.2817,
+      "step": 107020
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 8.351811890765684e-06,
+      "loss": 0.2866,
+      "step": 107030
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 8.34883547883027e-06,
+      "loss": 0.289,
+      "step": 107040
+    },
+    {
+      "epoch": 23.77,
+      "learning_rate": 8.34585906689486e-06,
+      "loss": 0.2978,
+      "step": 107050
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 8.342882654959446e-06,
+      "loss": 0.2863,
+      "step": 107060
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 8.339906243024035e-06,
+      "loss": 0.2816,
+      "step": 107070
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 8.336929831088623e-06,
+      "loss": 0.2837,
+      "step": 107080
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 8.333953419153212e-06,
+      "loss": 0.2752,
+      "step": 107090
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 8.3309770072178e-06,
+      "loss": 0.2831,
+      "step": 107100
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 8.328000595282388e-06,
+      "loss": 0.2864,
+      "step": 107110
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 8.325024183346976e-06,
+      "loss": 0.291,
+      "step": 107120
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 8.322047771411565e-06,
+      "loss": 0.3052,
+      "step": 107130
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 8.319071359476152e-06,
+      "loss": 0.2798,
+      "step": 107140
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 8.31609494754074e-06,
+      "loss": 0.2947,
+      "step": 107150
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 8.313118535605329e-06,
+      "loss": 0.2782,
+      "step": 107160
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 8.310142123669916e-06,
+      "loss": 0.3005,
+      "step": 107170
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 8.307165711734504e-06,
+      "loss": 0.2745,
+      "step": 107180
+    },
+    {
+      "epoch": 23.8,
+      "learning_rate": 8.304189299799093e-06,
+      "loss": 0.2926,
+      "step": 107190
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 8.301212887863681e-06,
+      "loss": 0.3027,
+      "step": 107200
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 8.29823647592827e-06,
+      "loss": 0.2883,
+      "step": 107210
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 8.295260063992857e-06,
+      "loss": 0.3006,
+      "step": 107220
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 8.292283652057446e-06,
+      "loss": 0.2897,
+      "step": 107230
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 8.289307240122032e-06,
+      "loss": 0.2794,
+      "step": 107240
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 8.286330828186621e-06,
+      "loss": 0.2875,
+      "step": 107250
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 8.28335441625121e-06,
+      "loss": 0.2853,
+      "step": 107260
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 8.280378004315798e-06,
+      "loss": 0.2778,
+      "step": 107270
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 8.277401592380387e-06,
+      "loss": 0.2843,
+      "step": 107280
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 8.274425180444974e-06,
+      "loss": 0.2838,
+      "step": 107290
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 8.271448768509562e-06,
+      "loss": 0.2812,
+      "step": 107300
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 8.268472356574151e-06,
+      "loss": 0.2872,
+      "step": 107310
+    },
+    {
+      "epoch": 23.83,
+      "learning_rate": 8.265495944638738e-06,
+      "loss": 0.2906,
+      "step": 107320
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 8.262519532703326e-06,
+      "loss": 0.2901,
+      "step": 107330
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 8.259543120767915e-06,
+      "loss": 0.2843,
+      "step": 107340
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 8.256566708832504e-06,
+      "loss": 0.2911,
+      "step": 107350
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 8.253590296897092e-06,
+      "loss": 0.3007,
+      "step": 107360
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 8.250613884961679e-06,
+      "loss": 0.2926,
+      "step": 107370
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 8.247637473026268e-06,
+      "loss": 0.2748,
+      "step": 107380
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 8.244661061090856e-06,
+      "loss": 0.2866,
+      "step": 107390
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 8.241684649155443e-06,
+      "loss": 0.2654,
+      "step": 107400
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 8.238708237220032e-06,
+      "loss": 0.2837,
+      "step": 107410
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 8.23573182528462e-06,
+      "loss": 0.293,
+      "step": 107420
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 8.232755413349209e-06,
+      "loss": 0.2824,
+      "step": 107430
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 8.229779001413798e-06,
+      "loss": 0.2883,
+      "step": 107440
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 8.226802589478384e-06,
+      "loss": 0.2948,
+      "step": 107450
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 8.223826177542973e-06,
+      "loss": 0.2842,
+      "step": 107460
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 8.22084976560756e-06,
+      "loss": 0.2876,
+      "step": 107470
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 8.217873353672149e-06,
+      "loss": 0.2848,
+      "step": 107480
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 8.214896941736737e-06,
+      "loss": 0.2961,
+      "step": 107490
+    },
+    {
+      "epoch": 23.87,
+      "learning_rate": 8.211920529801326e-06,
+      "loss": 0.2948,
+      "step": 107500
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 8.208944117865914e-06,
+      "loss": 0.28,
+      "step": 107510
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 8.205967705930501e-06,
+      "loss": 0.2772,
+      "step": 107520
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 8.20299129399509e-06,
+      "loss": 0.2728,
+      "step": 107530
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 8.200014882059678e-06,
+      "loss": 0.2932,
+      "step": 107540
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 8.197038470124265e-06,
+      "loss": 0.2966,
+      "step": 107550
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 8.194062058188854e-06,
+      "loss": 0.2853,
+      "step": 107560
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 8.19108564625344e-06,
+      "loss": 0.2946,
+      "step": 107570
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 8.18810923431803e-06,
+      "loss": 0.2945,
+      "step": 107580
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 8.185132822382618e-06,
+      "loss": 0.2872,
+      "step": 107590
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 8.182156410447207e-06,
+      "loss": 0.2928,
+      "step": 107600
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 8.179179998511795e-06,
+      "loss": 0.2919,
+      "step": 107610
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 8.176203586576384e-06,
+      "loss": 0.2933,
+      "step": 107620
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 8.17322717464097e-06,
+      "loss": 0.2985,
+      "step": 107630
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 8.17025076270556e-06,
+      "loss": 0.2796,
+      "step": 107640
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 8.167274350770146e-06,
+      "loss": 0.2922,
+      "step": 107650
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 8.164297938834735e-06,
+      "loss": 0.2862,
+      "step": 107660
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 8.161321526899323e-06,
+      "loss": 0.2785,
+      "step": 107670
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 8.158345114963912e-06,
+      "loss": 0.3014,
+      "step": 107680
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 8.1553687030285e-06,
+      "loss": 0.2929,
+      "step": 107690
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 8.152392291093087e-06,
+      "loss": 0.2827,
+      "step": 107700
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 8.149415879157676e-06,
+      "loss": 0.2984,
+      "step": 107710
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 8.146439467222265e-06,
+      "loss": 0.2918,
+      "step": 107720
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 8.143463055286851e-06,
+      "loss": 0.2861,
+      "step": 107730
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 8.14048664335144e-06,
+      "loss": 0.2839,
+      "step": 107740
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 8.137510231416029e-06,
+      "loss": 0.289,
+      "step": 107750
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 8.134533819480617e-06,
+      "loss": 0.279,
+      "step": 107760
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 8.131557407545206e-06,
+      "loss": 0.2948,
+      "step": 107770
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 8.128580995609793e-06,
+      "loss": 0.2824,
+      "step": 107780
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 8.125604583674381e-06,
+      "loss": 0.2748,
+      "step": 107790
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 8.12262817173897e-06,
+      "loss": 0.2854,
+      "step": 107800
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 8.119651759803557e-06,
+      "loss": 0.2913,
+      "step": 107810
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 8.116675347868145e-06,
+      "loss": 0.2965,
+      "step": 107820
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 8.113698935932734e-06,
+      "loss": 0.2879,
+      "step": 107830
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 8.110722523997323e-06,
+      "loss": 0.2907,
+      "step": 107840
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 8.107746112061911e-06,
+      "loss": 0.2888,
+      "step": 107850
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 8.104769700126498e-06,
+      "loss": 0.2898,
+      "step": 107860
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 8.101793288191087e-06,
+      "loss": 0.2986,
+      "step": 107870
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 8.098816876255674e-06,
+      "loss": 0.2922,
+      "step": 107880
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 8.095840464320262e-06,
+      "loss": 0.2778,
+      "step": 107890
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 8.09286405238485e-06,
+      "loss": 0.2939,
+      "step": 107900
+    },
+    {
+      "epoch": 23.96,
+      "learning_rate": 8.08988764044944e-06,
+      "loss": 0.29,
+      "step": 107910
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 8.086911228514028e-06,
+      "loss": 0.2939,
+      "step": 107920
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 8.083934816578615e-06,
+      "loss": 0.2788,
+      "step": 107930
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 8.080958404643203e-06,
+      "loss": 0.2814,
+      "step": 107940
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 8.077981992707792e-06,
+      "loss": 0.273,
+      "step": 107950
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 8.075005580772379e-06,
+      "loss": 0.2949,
+      "step": 107960
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 8.072029168836968e-06,
+      "loss": 0.2931,
+      "step": 107970
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 8.069052756901556e-06,
+      "loss": 0.279,
+      "step": 107980
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 8.066076344966145e-06,
+      "loss": 0.2987,
+      "step": 107990
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 8.063099933030732e-06,
+      "loss": 0.2845,
+      "step": 108000
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 8.06012352109532e-06,
+      "loss": 0.2929,
+      "step": 108010
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 8.057147109159909e-06,
+      "loss": 0.2861,
+      "step": 108020
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 8.054170697224497e-06,
+      "loss": 0.2799,
+      "step": 108030
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 8.051194285289084e-06,
+      "loss": 0.2878,
+      "step": 108040
+    },
+    {
+      "epoch": 23.99,
+      "eval_cer": 4.5405982905982905,
+      "eval_loss": 0.514737069606781,
+      "eval_runtime": 1.3006,
+      "eval_samples_per_second": 7.689,
+      "eval_steps_per_second": 0.769,
+      "eval_wer": 0.6493506493506493,
+      "step": 108048
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 8.048217873353673e-06,
+      "loss": 0.2929,
+      "step": 108050
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 8.04524146141826e-06,
+      "loss": 0.2794,
+      "step": 108060
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 8.042265049482848e-06,
+      "loss": 0.2779,
+      "step": 108070
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 8.039288637547437e-06,
+      "loss": 0.2879,
+      "step": 108080
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 8.036312225612026e-06,
+      "loss": 0.2925,
+      "step": 108090
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 8.033335813676614e-06,
+      "loss": 0.2821,
+      "step": 108100
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 8.030359401741201e-06,
+      "loss": 0.2859,
+      "step": 108110
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 8.02738298980579e-06,
+      "loss": 0.2781,
+      "step": 108120
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 8.024406577870378e-06,
+      "loss": 0.2892,
+      "step": 108130
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 8.021430165934965e-06,
+      "loss": 0.2833,
+      "step": 108140
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 8.018453753999554e-06,
+      "loss": 0.2649,
+      "step": 108150
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 8.015477342064142e-06,
+      "loss": 0.2863,
+      "step": 108160
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 8.012500930128731e-06,
+      "loss": 0.2757,
+      "step": 108170
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 8.00952451819332e-06,
+      "loss": 0.2794,
+      "step": 108180
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 8.006548106257906e-06,
+      "loss": 0.295,
+      "step": 108190
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 8.003571694322495e-06,
+      "loss": 0.273,
+      "step": 108200
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 8.000595282387082e-06,
+      "loss": 0.288,
+      "step": 108210
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 7.99761887045167e-06,
+      "loss": 0.275,
+      "step": 108220
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 7.994642458516259e-06,
+      "loss": 0.2715,
+      "step": 108230
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 7.991666046580848e-06,
+      "loss": 0.2878,
+      "step": 108240
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 7.988689634645436e-06,
+      "loss": 0.29,
+      "step": 108250
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 7.985713222710025e-06,
+      "loss": 0.271,
+      "step": 108260
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 7.982736810774612e-06,
+      "loss": 0.2792,
+      "step": 108270
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 7.9797603988392e-06,
+      "loss": 0.2898,
+      "step": 108280
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 7.976783986903787e-06,
+      "loss": 0.2821,
+      "step": 108290
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 7.973807574968376e-06,
+      "loss": 0.2854,
+      "step": 108300
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 7.970831163032964e-06,
+      "loss": 0.2749,
+      "step": 108310
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 7.967854751097553e-06,
+      "loss": 0.2833,
+      "step": 108320
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 7.964878339162142e-06,
+      "loss": 0.2866,
+      "step": 108330
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 7.961901927226729e-06,
+      "loss": 0.2823,
+      "step": 108340
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 7.958925515291317e-06,
+      "loss": 0.2788,
+      "step": 108350
+    },
+    {
+      "epoch": 24.06,
+      "learning_rate": 7.955949103355906e-06,
+      "loss": 0.2845,
+      "step": 108360
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 7.952972691420493e-06,
+      "loss": 0.2964,
+      "step": 108370
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 7.949996279485081e-06,
+      "loss": 0.2779,
+      "step": 108380
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 7.94701986754967e-06,
+      "loss": 0.2752,
+      "step": 108390
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 7.944043455614258e-06,
+      "loss": 0.293,
+      "step": 108400
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 7.941067043678847e-06,
+      "loss": 0.2664,
+      "step": 108410
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 7.938090631743434e-06,
+      "loss": 0.2847,
+      "step": 108420
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 7.935114219808022e-06,
+      "loss": 0.2711,
+      "step": 108430
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 7.93213780787261e-06,
+      "loss": 0.2809,
+      "step": 108440
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 7.929161395937198e-06,
+      "loss": 0.282,
+      "step": 108450
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 7.926184984001787e-06,
+      "loss": 0.2785,
+      "step": 108460
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 7.923208572066373e-06,
+      "loss": 0.2854,
+      "step": 108470
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 7.920232160130962e-06,
+      "loss": 0.2822,
+      "step": 108480
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 7.91725574819555e-06,
+      "loss": 0.275,
+      "step": 108490
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 7.91427933626014e-06,
+      "loss": 0.2719,
+      "step": 108500
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 7.911302924324728e-06,
+      "loss": 0.2935,
+      "step": 108510
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 7.908326512389315e-06,
+      "loss": 0.2819,
+      "step": 108520
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 7.905350100453903e-06,
+      "loss": 0.2695,
+      "step": 108530
+    },
+    {
+      "epoch": 24.1,
+      "learning_rate": 7.902373688518492e-06,
+      "loss": 0.2733,
+      "step": 108540
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 7.899397276583079e-06,
+      "loss": 0.2909,
+      "step": 108550
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 7.896420864647667e-06,
+      "loss": 0.2737,
+      "step": 108560
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 7.893444452712256e-06,
+      "loss": 0.2838,
+      "step": 108570
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 7.890468040776845e-06,
+      "loss": 0.2809,
+      "step": 108580
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 7.887491628841433e-06,
+      "loss": 0.2923,
+      "step": 108590
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 7.88451521690602e-06,
+      "loss": 0.2762,
+      "step": 108600
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 7.881538804970609e-06,
+      "loss": 0.2846,
+      "step": 108610
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 7.878562393035196e-06,
+      "loss": 0.2849,
+      "step": 108620
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 7.875585981099784e-06,
+      "loss": 0.2714,
+      "step": 108630
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 7.872609569164373e-06,
+      "loss": 0.2841,
+      "step": 108640
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 7.869633157228961e-06,
+      "loss": 0.2843,
+      "step": 108650
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 7.86665674529355e-06,
+      "loss": 0.2721,
+      "step": 108660
+    },
+    {
+      "epoch": 24.13,
+      "learning_rate": 7.863680333358139e-06,
+      "loss": 0.2708,
+      "step": 108670
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 7.860703921422725e-06,
+      "loss": 0.2783,
+      "step": 108680
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 7.857727509487314e-06,
+      "loss": 0.2843,
+      "step": 108690
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 7.854751097551901e-06,
+      "loss": 0.287,
+      "step": 108700
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 7.85177468561649e-06,
+      "loss": 0.2865,
+      "step": 108710
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 7.848798273681078e-06,
+      "loss": 0.2711,
+      "step": 108720
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 7.845821861745667e-06,
+      "loss": 0.2797,
+      "step": 108730
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 7.842845449810255e-06,
+      "loss": 0.2762,
+      "step": 108740
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 7.839869037874842e-06,
+      "loss": 0.2831,
+      "step": 108750
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 7.83689262593943e-06,
+      "loss": 0.2656,
+      "step": 108760
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 7.83391621400402e-06,
+      "loss": 0.2808,
+      "step": 108770
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 7.830939802068606e-06,
+      "loss": 0.2845,
+      "step": 108780
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 7.827963390133195e-06,
+      "loss": 0.2781,
+      "step": 108790
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 7.824986978197783e-06,
+      "loss": 0.2768,
+      "step": 108800
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 7.822010566262372e-06,
+      "loss": 0.2649,
+      "step": 108810
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.81903415432696e-06,
+      "loss": 0.2796,
+      "step": 108820
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.816057742391548e-06,
+      "loss": 0.2903,
+      "step": 108830
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.813081330456136e-06,
+      "loss": 0.2844,
+      "step": 108840
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.810104918520723e-06,
+      "loss": 0.2838,
+      "step": 108850
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 7.807128506585312e-06,
+      "loss": 0.2859,
+      "step": 108860
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 7.8041520946499e-06,
+      "loss": 0.2698,
+      "step": 108870
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 7.801175682714487e-06,
+      "loss": 0.2831,
+      "step": 108880
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 7.798199270779076e-06,
+      "loss": 0.2905,
+      "step": 108890
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 7.795222858843664e-06,
+      "loss": 0.2808,
+      "step": 108900
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 7.792246446908253e-06,
+      "loss": 0.2958,
+      "step": 108910
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 7.789270034972842e-06,
+      "loss": 0.285,
+      "step": 108920
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 7.786293623037428e-06,
+      "loss": 0.2874,
+      "step": 108930
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 7.783317211102017e-06,
+      "loss": 0.2707,
+      "step": 108940
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 7.780340799166606e-06,
+      "loss": 0.2842,
+      "step": 108950
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 7.777364387231193e-06,
+      "loss": 0.2842,
+      "step": 108960
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 7.774387975295781e-06,
+      "loss": 0.2819,
+      "step": 108970
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 7.77141156336037e-06,
+      "loss": 0.2742,
+      "step": 108980
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 7.768435151424958e-06,
+      "loss": 0.2775,
+      "step": 108990
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 7.765458739489547e-06,
+      "loss": 0.2566,
+      "step": 109000
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 7.762482327554134e-06,
+      "loss": 0.2805,
+      "step": 109010
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 7.759505915618722e-06,
+      "loss": 0.2922,
+      "step": 109020
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 7.75652950368331e-06,
+      "loss": 0.2779,
+      "step": 109030
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 7.753553091747898e-06,
+      "loss": 0.2702,
+      "step": 109040
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 7.750576679812486e-06,
+      "loss": 0.2778,
+      "step": 109050
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 7.747600267877075e-06,
+      "loss": 0.2724,
+      "step": 109060
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 7.744623855941664e-06,
+      "loss": 0.284,
+      "step": 109070
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 7.74164744400625e-06,
+      "loss": 0.277,
+      "step": 109080
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 7.738671032070839e-06,
+      "loss": 0.2867,
+      "step": 109090
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 7.735694620135428e-06,
+      "loss": 0.2876,
+      "step": 109100
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 7.732718208200015e-06,
+      "loss": 0.2673,
+      "step": 109110
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 7.729741796264603e-06,
+      "loss": 0.2825,
+      "step": 109120
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 7.726765384329192e-06,
+      "loss": 0.2725,
+      "step": 109130
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 7.72378897239378e-06,
+      "loss": 0.2901,
+      "step": 109140
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 7.720812560458369e-06,
+      "loss": 0.2762,
+      "step": 109150
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 7.717836148522956e-06,
+      "loss": 0.2688,
+      "step": 109160
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 7.714859736587544e-06,
+      "loss": 0.2841,
+      "step": 109170
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 7.711883324652133e-06,
+      "loss": 0.2843,
+      "step": 109180
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 7.70890691271672e-06,
+      "loss": 0.2825,
+      "step": 109190
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 7.705930500781309e-06,
+      "loss": 0.2774,
+      "step": 109200
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 7.702954088845897e-06,
+      "loss": 0.2809,
+      "step": 109210
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 7.699977676910486e-06,
+      "loss": 0.289,
+      "step": 109220
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 7.697001264975074e-06,
+      "loss": 0.2842,
+      "step": 109230
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 7.694024853039661e-06,
+      "loss": 0.2792,
+      "step": 109240
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 7.69104844110425e-06,
+      "loss": 0.2786,
+      "step": 109250
+    },
+    {
+      "epoch": 24.26,
+      "learning_rate": 7.688072029168837e-06,
+      "loss": 0.2775,
+      "step": 109260
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.685095617233425e-06,
+      "loss": 0.2748,
+      "step": 109270
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.682119205298014e-06,
+      "loss": 0.2762,
+      "step": 109280
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.679142793362603e-06,
+      "loss": 0.2886,
+      "step": 109290
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.676166381427191e-06,
+      "loss": 0.2807,
+      "step": 109300
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 7.673189969491778e-06,
+      "loss": 0.2912,
+      "step": 109310
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 7.670213557556367e-06,
+      "loss": 0.2755,
+      "step": 109320
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 7.667237145620955e-06,
+      "loss": 0.2755,
+      "step": 109330
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 7.664260733685542e-06,
+      "loss": 0.2878,
+      "step": 109340
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 7.66128432175013e-06,
+      "loss": 0.2732,
+      "step": 109350
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 7.65830790981472e-06,
+      "loss": 0.2872,
+      "step": 109360
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 7.655331497879306e-06,
+      "loss": 0.2712,
+      "step": 109370
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 7.652355085943895e-06,
+      "loss": 0.2807,
+      "step": 109380
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 7.649378674008483e-06,
+      "loss": 0.284,
+      "step": 109390
+    },
+    {
+      "epoch": 24.29,
+      "learning_rate": 7.646402262073072e-06,
+      "loss": 0.2873,
+      "step": 109400
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 7.64342585013766e-06,
+      "loss": 0.2879,
+      "step": 109410
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 7.640449438202247e-06,
+      "loss": 0.2694,
+      "step": 109420
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 7.637473026266836e-06,
+      "loss": 0.2804,
+      "step": 109430
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 7.634496614331423e-06,
+      "loss": 0.281,
+      "step": 109440
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 7.631520202396012e-06,
+      "loss": 0.2796,
+      "step": 109450
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 7.628543790460601e-06,
+      "loss": 0.2893,
+      "step": 109460
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 7.625567378525188e-06,
+      "loss": 0.284,
+      "step": 109470
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 7.6225909665897765e-06,
+      "loss": 0.28,
+      "step": 109480
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 7.619614554654364e-06,
+      "loss": 0.2932,
+      "step": 109490
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 7.616638142718953e-06,
+      "loss": 0.2959,
+      "step": 109500
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 7.613661730783541e-06,
+      "loss": 0.2744,
+      "step": 109510
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 7.610685318848129e-06,
+      "loss": 0.2832,
+      "step": 109520
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 7.607708906912718e-06,
+      "loss": 0.2791,
+      "step": 109530
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 7.604732494977306e-06,
+      "loss": 0.2781,
+      "step": 109540
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 7.601756083041893e-06,
+      "loss": 0.2872,
+      "step": 109550
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 7.598779671106482e-06,
+      "loss": 0.2904,
+      "step": 109560
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 7.5958032591710696e-06,
+      "loss": 0.2866,
+      "step": 109570
+    },
+    {
+      "epoch": 24.33,
+      "learning_rate": 7.592826847235658e-06,
+      "loss": 0.2784,
+      "step": 109580
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 7.589850435300247e-06,
+      "loss": 0.3049,
+      "step": 109590
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 7.5868740233648345e-06,
+      "loss": 0.2826,
+      "step": 109600
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 7.583897611429423e-06,
+      "loss": 0.2889,
+      "step": 109610
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 7.58092119949401e-06,
+      "loss": 0.2737,
+      "step": 109620
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 7.577944787558599e-06,
+      "loss": 0.2768,
+      "step": 109630
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 7.574968375623187e-06,
+      "loss": 0.285,
+      "step": 109640
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 7.571991963687775e-06,
+      "loss": 0.2861,
+      "step": 109650
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 7.5690155517523635e-06,
+      "loss": 0.2657,
+      "step": 109660
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 7.56603913981695e-06,
+      "loss": 0.2814,
+      "step": 109670
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 7.563062727881539e-06,
+      "loss": 0.2823,
+      "step": 109680
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 7.560086315946128e-06,
+      "loss": 0.2829,
+      "step": 109690
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 7.557109904010715e-06,
+      "loss": 0.27,
+      "step": 109700
+    },
+    {
+      "epoch": 24.36,
+      "learning_rate": 7.554133492075304e-06,
+      "loss": 0.2843,
+      "step": 109710
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 7.551157080139892e-06,
+      "loss": 0.2805,
+      "step": 109720
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 7.54818066820448e-06,
+      "loss": 0.2842,
+      "step": 109730
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 7.545204256269069e-06,
+      "loss": 0.2798,
+      "step": 109740
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 7.542227844333656e-06,
+      "loss": 0.2868,
+      "step": 109750
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 7.539251432398244e-06,
+      "loss": 0.292,
+      "step": 109760
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 7.536275020462833e-06,
+      "loss": 0.2951,
+      "step": 109770
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 7.533298608527421e-06,
+      "loss": 0.2755,
+      "step": 109780
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 7.530322196592009e-06,
+      "loss": 0.2842,
+      "step": 109790
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 7.527345784656597e-06,
+      "loss": 0.2851,
+      "step": 109800
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.524369372721186e-06,
+      "loss": 0.2726,
+      "step": 109810
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.521392960785774e-06,
+      "loss": 0.2857,
+      "step": 109820
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.518416548850361e-06,
+      "loss": 0.2876,
+      "step": 109830
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.51544013691495e-06,
+      "loss": 0.2774,
+      "step": 109840
+    },
+    {
+      "epoch": 24.39,
+      "learning_rate": 7.5124637249795375e-06,
+      "loss": 0.281,
+      "step": 109850
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 7.509487313044126e-06,
+      "loss": 0.294,
+      "step": 109860
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 7.506510901108715e-06,
+      "loss": 0.2799,
+      "step": 109870
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 7.5035344891733016e-06,
+      "loss": 0.2827,
+      "step": 109880
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 7.50055807723789e-06,
+      "loss": 0.2797,
+      "step": 109890
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 7.497581665302478e-06,
+      "loss": 0.284,
+      "step": 109900
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 7.4946052533670665e-06,
+      "loss": 0.2872,
+      "step": 109910
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 7.491628841431655e-06,
+      "loss": 0.2813,
+      "step": 109920
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 7.488652429496243e-06,
+      "loss": 0.2946,
+      "step": 109930
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 7.485676017560831e-06,
+      "loss": 0.2911,
+      "step": 109940
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 7.482699605625418e-06,
+      "loss": 0.2869,
+      "step": 109950
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 7.479723193690007e-06,
+      "loss": 0.2757,
+      "step": 109960
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 7.4767467817545955e-06,
+      "loss": 0.2902,
+      "step": 109970
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 7.473770369819183e-06,
+      "loss": 0.2811,
+      "step": 109980
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 7.470793957883772e-06,
+      "loss": 0.2857,
+      "step": 109990
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 7.4678175459483604e-06,
+      "loss": 0.2888,
+      "step": 110000
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 7.464841134012948e-06,
+      "loss": 0.2785,
+      "step": 110010
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 7.461864722077537e-06,
+      "loss": 0.2768,
+      "step": 110020
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 7.458888310142124e-06,
+      "loss": 0.2732,
+      "step": 110030
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 7.455911898206712e-06,
+      "loss": 0.2871,
+      "step": 110040
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 7.452935486271301e-06,
+      "loss": 0.2816,
+      "step": 110050
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 7.449959074335889e-06,
+      "loss": 0.2884,
+      "step": 110060
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 7.446982662400477e-06,
+      "loss": 0.2863,
+      "step": 110070
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 7.444006250465065e-06,
+      "loss": 0.2844,
+      "step": 110080
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 7.441029838529653e-06,
+      "loss": 0.2821,
+      "step": 110090
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 7.438053426594241e-06,
+      "loss": 0.2794,
+      "step": 110100
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 7.435077014658829e-06,
+      "loss": 0.2831,
+      "step": 110110
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 7.432100602723418e-06,
+      "loss": 0.2795,
+      "step": 110120
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 7.429124190788005e-06,
+      "loss": 0.2665,
+      "step": 110130
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 7.426147778852594e-06,
+      "loss": 0.2835,
+      "step": 110140
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 7.4231713669171826e-06,
+      "loss": 0.2756,
+      "step": 110150
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 7.4201949549817695e-06,
+      "loss": 0.2715,
+      "step": 110160
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.417218543046358e-06,
+      "loss": 0.2814,
+      "step": 110170
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.414242131110946e-06,
+      "loss": 0.2783,
+      "step": 110180
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.411265719175534e-06,
+      "loss": 0.2702,
+      "step": 110190
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.408289307240123e-06,
+      "loss": 0.2847,
+      "step": 110200
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 7.405312895304711e-06,
+      "loss": 0.277,
+      "step": 110210
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 7.402336483369299e-06,
+      "loss": 0.2514,
+      "step": 110220
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 7.399360071433888e-06,
+      "loss": 0.2747,
+      "step": 110230
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 7.396383659498475e-06,
+      "loss": 0.2851,
+      "step": 110240
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 7.393407247563063e-06,
+      "loss": 0.2776,
+      "step": 110250
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 7.390430835627651e-06,
+      "loss": 0.2806,
+      "step": 110260
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 7.38745442369224e-06,
+      "loss": 0.2792,
+      "step": 110270
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 7.384478011756828e-06,
+      "loss": 0.2941,
+      "step": 110280
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 7.381501599821416e-06,
+      "loss": 0.2773,
+      "step": 110290
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 7.378525187886005e-06,
+      "loss": 0.2763,
+      "step": 110300
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 7.3755487759505916e-06,
+      "loss": 0.2782,
+      "step": 110310
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 7.37257236401518e-06,
+      "loss": 0.2962,
+      "step": 110320
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 7.369595952079769e-06,
+      "loss": 0.2738,
+      "step": 110330
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 7.3666195401443565e-06,
+      "loss": 0.2933,
+      "step": 110340
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.363643128208945e-06,
+      "loss": 0.2707,
+      "step": 110350
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.360666716273532e-06,
+      "loss": 0.2737,
+      "step": 110360
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.357690304338121e-06,
+      "loss": 0.2735,
+      "step": 110370
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.354713892402709e-06,
+      "loss": 0.279,
+      "step": 110380
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 7.351737480467297e-06,
+      "loss": 0.2863,
+      "step": 110390
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 7.3487610685318855e-06,
+      "loss": 0.2817,
+      "step": 110400
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 7.345784656596474e-06,
+      "loss": 0.2798,
+      "step": 110410
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 7.342808244661062e-06,
+      "loss": 0.2908,
+      "step": 110420
+    },
+    {
+      "epoch": 24.52,
+      "learning_rate": 7.3398318327256504e-06,
+      "loss": 0.2956,
+      "step": 110430
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 7.336855420790237e-06,
+      "loss": 0.2875,
+      "step": 110440
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 7.333879008854826e-06,
+      "loss": 0.285,
+      "step": 110450
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 7.3309025969194145e-06,
+      "loss": 0.2788,
+      "step": 110460
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 7.327926184984002e-06,
+      "loss": 0.2924,
+      "step": 110470
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 7.324949773048591e-06,
+      "loss": 0.2832,
+      "step": 110480
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 7.321973361113179e-06,
+      "loss": 0.2742,
+      "step": 110490
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 7.318996949177767e-06,
+      "loss": 0.2766,
+      "step": 110500
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 7.316020537242356e-06,
+      "loss": 0.2927,
+      "step": 110510
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 7.313044125306943e-06,
+      "loss": 0.2898,
+      "step": 110520
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 7.310067713371531e-06,
+      "loss": 0.28,
+      "step": 110530
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 7.307091301436119e-06,
+      "loss": 0.2697,
+      "step": 110540
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 7.304114889500708e-06,
+      "loss": 0.2778,
+      "step": 110550
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 7.301138477565296e-06,
+      "loss": 0.2793,
+      "step": 110560
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 7.298162065629883e-06,
+      "loss": 0.291,
+      "step": 110570
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 7.295185653694472e-06,
+      "loss": 0.2887,
+      "step": 110580
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 7.2922092417590595e-06,
+      "loss": 0.2687,
+      "step": 110590
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 7.289232829823648e-06,
+      "loss": 0.2767,
+      "step": 110600
+    },
+    {
+      "epoch": 24.56,
+      "learning_rate": 7.286256417888237e-06,
+      "loss": 0.2864,
+      "step": 110610
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 7.283280005952824e-06,
+      "loss": 0.2942,
+      "step": 110620
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 7.280303594017413e-06,
+      "loss": 0.2932,
+      "step": 110630
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 7.277327182082002e-06,
+      "loss": 0.2688,
+      "step": 110640
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 7.2743507701465885e-06,
+      "loss": 0.2814,
+      "step": 110650
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 7.271374358211177e-06,
+      "loss": 0.2854,
+      "step": 110660
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 7.268397946275765e-06,
+      "loss": 0.28,
+      "step": 110670
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 7.265421534340353e-06,
+      "loss": 0.2751,
+      "step": 110680
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 7.262445122404942e-06,
+      "loss": 0.2729,
+      "step": 110690
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 7.25946871046953e-06,
+      "loss": 0.2903,
+      "step": 110700
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 7.256492298534118e-06,
+      "loss": 0.2894,
+      "step": 110710
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 7.253515886598705e-06,
+      "loss": 0.2928,
+      "step": 110720
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 7.250539474663294e-06,
+      "loss": 0.2824,
+      "step": 110730
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 7.2475630627278824e-06,
+      "loss": 0.2779,
+      "step": 110740
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 7.24458665079247e-06,
+      "loss": 0.2761,
+      "step": 110750
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 7.241610238857059e-06,
+      "loss": 0.2819,
+      "step": 110760
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 7.238633826921646e-06,
+      "loss": 0.2733,
+      "step": 110770
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 7.235657414986234e-06,
+      "loss": 0.2887,
+      "step": 110780
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 7.232681003050823e-06,
+      "loss": 0.2782,
+      "step": 110790
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 7.229704591115411e-06,
+      "loss": 0.2889,
+      "step": 110800
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 7.226728179179999e-06,
+      "loss": 0.2714,
+      "step": 110810
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 7.223751767244587e-06,
+      "loss": 0.2912,
+      "step": 110820
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 7.2207753553091755e-06,
+      "loss": 0.269,
+      "step": 110830
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 7.217798943373764e-06,
+      "loss": 0.2858,
+      "step": 110840
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 7.214822531438351e-06,
+      "loss": 0.2806,
+      "step": 110850
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 7.21184611950294e-06,
+      "loss": 0.2612,
+      "step": 110860
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 7.208869707567528e-06,
+      "loss": 0.2807,
+      "step": 110870
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 7.205893295632116e-06,
+      "loss": 0.2857,
+      "step": 110880
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 7.2029168836967045e-06,
+      "loss": 0.2853,
+      "step": 110890
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 7.199940471761292e-06,
+      "loss": 0.2886,
+      "step": 110900
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 7.196964059825881e-06,
+      "loss": 0.2807,
+      "step": 110910
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 7.1939876478904695e-06,
+      "loss": 0.3022,
+      "step": 110920
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 7.191011235955056e-06,
+      "loss": 0.2728,
+      "step": 110930
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 7.188034824019645e-06,
+      "loss": 0.2828,
+      "step": 110940
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 7.185058412084233e-06,
+      "loss": 0.2789,
+      "step": 110950
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 7.182082000148821e-06,
+      "loss": 0.2811,
+      "step": 110960
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 7.17910558821341e-06,
+      "loss": 0.2914,
+      "step": 110970
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 7.176129176277997e-06,
+      "loss": 0.2806,
+      "step": 110980
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 7.173152764342585e-06,
+      "loss": 0.2823,
+      "step": 110990
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 7.170176352407173e-06,
+      "loss": 0.2714,
+      "step": 111000
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 7.167199940471762e-06,
+      "loss": 0.2784,
+      "step": 111010
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 7.16422352853635e-06,
+      "loss": 0.286,
+      "step": 111020
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 7.161247116600938e-06,
+      "loss": 0.2859,
+      "step": 111030
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 7.158270704665527e-06,
+      "loss": 0.2795,
+      "step": 111040
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 7.1552942927301136e-06,
+      "loss": 0.27,
+      "step": 111050
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 7.152317880794702e-06,
+      "loss": 0.2689,
+      "step": 111060
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 7.149341468859291e-06,
+      "loss": 0.2715,
+      "step": 111070
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 7.1463650569238785e-06,
+      "loss": 0.2673,
+      "step": 111080
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 7.143388644988467e-06,
+      "loss": 0.282,
+      "step": 111090
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 7.140412233053056e-06,
+      "loss": 0.2768,
+      "step": 111100
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 7.1374358211176434e-06,
+      "loss": 0.2867,
+      "step": 111110
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 7.134459409182232e-06,
+      "loss": 0.2833,
+      "step": 111120
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 7.131482997246819e-06,
+      "loss": 0.2663,
+      "step": 111130
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 7.1285065853114075e-06,
+      "loss": 0.2829,
+      "step": 111140
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 7.125530173375996e-06,
+      "loss": 0.2781,
+      "step": 111150
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 7.122553761440584e-06,
+      "loss": 0.2809,
+      "step": 111160
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 7.1195773495051724e-06,
+      "loss": 0.2933,
+      "step": 111170
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 7.11660093756976e-06,
+      "loss": 0.2823,
+      "step": 111180
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 7.113624525634348e-06,
+      "loss": 0.2754,
+      "step": 111190
+    },
+    {
+      "epoch": 24.69,
+      "learning_rate": 7.1106481136989365e-06,
+      "loss": 0.2897,
+      "step": 111200
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 7.107671701763524e-06,
+      "loss": 0.292,
+      "step": 111210
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 7.104695289828113e-06,
+      "loss": 0.2882,
+      "step": 111220
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 7.101718877892701e-06,
+      "loss": 0.2769,
+      "step": 111230
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 7.098742465957289e-06,
+      "loss": 0.2819,
+      "step": 111240
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 7.095766054021878e-06,
+      "loss": 0.2767,
+      "step": 111250
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 7.092789642086465e-06,
+      "loss": 0.2671,
+      "step": 111260
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 7.089813230151053e-06,
+      "loss": 0.2775,
+      "step": 111270
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 7.086836818215642e-06,
+      "loss": 0.281,
+      "step": 111280
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 7.08386040628023e-06,
+      "loss": 0.2905,
+      "step": 111290
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 7.080883994344818e-06,
+      "loss": 0.2772,
+      "step": 111300
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 7.077907582409406e-06,
+      "loss": 0.2684,
+      "step": 111310
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 7.0749311704739946e-06,
+      "loss": 0.2896,
+      "step": 111320
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 7.071954758538583e-06,
+      "loss": 0.297,
+      "step": 111330
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 7.06897834660317e-06,
+      "loss": 0.2778,
+      "step": 111340
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 7.066001934667759e-06,
+      "loss": 0.2713,
+      "step": 111350
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 7.063025522732346e-06,
+      "loss": 0.2746,
+      "step": 111360
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 7.060049110796935e-06,
+      "loss": 0.2802,
+      "step": 111370
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 7.057072698861524e-06,
+      "loss": 0.2864,
+      "step": 111380
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 7.054096286926111e-06,
+      "loss": 0.286,
+      "step": 111390
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 7.0511198749907e-06,
+      "loss": 0.2829,
+      "step": 111400
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 7.048143463055287e-06,
+      "loss": 0.2798,
+      "step": 111410
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 7.045167051119875e-06,
+      "loss": 0.2741,
+      "step": 111420
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 7.042190639184464e-06,
+      "loss": 0.2775,
+      "step": 111430
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 7.039214227249052e-06,
+      "loss": 0.2677,
+      "step": 111440
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 7.03623781531364e-06,
+      "loss": 0.2823,
+      "step": 111450
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 7.033261403378227e-06,
+      "loss": 0.2939,
+      "step": 111460
+    },
+    {
+      "epoch": 24.75,
+      "learning_rate": 7.030284991442816e-06,
+      "loss": 0.2871,
+      "step": 111470
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 7.0273085795074044e-06,
+      "loss": 0.2647,
+      "step": 111480
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 7.024332167571992e-06,
+      "loss": 0.2724,
+      "step": 111490
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 7.021355755636581e-06,
+      "loss": 0.2702,
+      "step": 111500
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 7.018379343701169e-06,
+      "loss": 0.2771,
+      "step": 111510
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 7.015402931765757e-06,
+      "loss": 0.2876,
+      "step": 111520
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 7.012426519830346e-06,
+      "loss": 0.292,
+      "step": 111530
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 7.009450107894933e-06,
+      "loss": 0.2882,
+      "step": 111540
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 7.006473695959521e-06,
+      "loss": 0.2762,
+      "step": 111550
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 7.00349728402411e-06,
+      "loss": 0.2744,
+      "step": 111560
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 7.0005208720886975e-06,
+      "loss": 0.2776,
+      "step": 111570
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 6.997544460153286e-06,
+      "loss": 0.284,
+      "step": 111580
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 6.994568048217874e-06,
+      "loss": 0.2817,
+      "step": 111590
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 6.9915916362824625e-06,
+      "loss": 0.2841,
+      "step": 111600
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 6.988615224347051e-06,
+      "loss": 0.2861,
+      "step": 111610
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 6.985638812411638e-06,
+      "loss": 0.2748,
+      "step": 111620
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 6.9826624004762265e-06,
+      "loss": 0.2941,
+      "step": 111630
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 6.979685988540814e-06,
+      "loss": 0.2817,
+      "step": 111640
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 6.976709576605403e-06,
+      "loss": 0.2731,
+      "step": 111650
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 6.9737331646699915e-06,
+      "loss": 0.2772,
+      "step": 111660
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 6.970756752734578e-06,
+      "loss": 0.2886,
+      "step": 111670
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 6.967780340799167e-06,
+      "loss": 0.2737,
+      "step": 111680
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 6.964803928863755e-06,
+      "loss": 0.268,
+      "step": 111690
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 6.961827516928343e-06,
+      "loss": 0.2666,
+      "step": 111700
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 6.958851104992932e-06,
+      "loss": 0.2726,
+      "step": 111710
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 6.95587469305752e-06,
+      "loss": 0.2862,
+      "step": 111720
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 6.952898281122108e-06,
+      "loss": 0.2882,
+      "step": 111730
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 6.949921869186697e-06,
+      "loss": 0.2743,
+      "step": 111740
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 6.946945457251284e-06,
+      "loss": 0.2876,
+      "step": 111750
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 6.943969045315872e-06,
+      "loss": 0.28,
+      "step": 111760
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 6.94099263338046e-06,
+      "loss": 0.2818,
+      "step": 111770
+    },
+    {
+      "epoch": 24.82,
+      "learning_rate": 6.938016221445049e-06,
+      "loss": 0.2842,
+      "step": 111780
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 6.935039809509637e-06,
+      "loss": 0.2959,
+      "step": 111790
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 6.932063397574225e-06,
+      "loss": 0.2798,
+      "step": 111800
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 6.929086985638814e-06,
+      "loss": 0.2769,
+      "step": 111810
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 6.9261105737034005e-06,
+      "loss": 0.2733,
+      "step": 111820
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 6.923134161767989e-06,
+      "loss": 0.2863,
+      "step": 111830
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 6.920157749832578e-06,
+      "loss": 0.2787,
+      "step": 111840
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 6.917181337897165e-06,
+      "loss": 0.2813,
+      "step": 111850
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 6.914204925961754e-06,
+      "loss": 0.2809,
+      "step": 111860
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 6.911228514026341e-06,
+      "loss": 0.281,
+      "step": 111870
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 6.9082521020909295e-06,
+      "loss": 0.2727,
+      "step": 111880
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 6.905275690155518e-06,
+      "loss": 0.2888,
+      "step": 111890
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 6.902299278220106e-06,
+      "loss": 0.2843,
+      "step": 111900
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 6.8993228662846944e-06,
+      "loss": 0.2607,
+      "step": 111910
+    },
+    {
+      "epoch": 24.85,
+      "learning_rate": 6.896346454349282e-06,
+      "loss": 0.2922,
+      "step": 111920
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 6.893370042413871e-06,
+      "loss": 0.2759,
+      "step": 111930
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 6.890393630478459e-06,
+      "loss": 0.2805,
+      "step": 111940
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 6.887417218543046e-06,
+      "loss": 0.2748,
+      "step": 111950
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 6.884440806607635e-06,
+      "loss": 0.2854,
+      "step": 111960
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.8814643946722235e-06,
+      "loss": 0.2798,
+      "step": 111970
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.878487982736811e-06,
+      "loss": 0.2896,
+      "step": 111980
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.8755115708014e-06,
+      "loss": 0.2796,
+      "step": 111990
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.8725351588659875e-06,
+      "loss": 0.2783,
+      "step": 112000
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 6.869558746930576e-06,
+      "loss": 0.2796,
+      "step": 112010
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 6.866582334995165e-06,
+      "loss": 0.2935,
+      "step": 112020
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 6.863605923059752e-06,
+      "loss": 0.2925,
+      "step": 112030
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 6.86062951112434e-06,
+      "loss": 0.2723,
+      "step": 112040
+    },
+    {
+      "epoch": 24.88,
+      "learning_rate": 6.857653099188928e-06,
+      "loss": 0.2893,
+      "step": 112050
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 6.8546766872535166e-06,
+      "loss": 0.2769,
+      "step": 112060
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 6.851700275318105e-06,
+      "loss": 0.2747,
+      "step": 112070
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 6.848723863382692e-06,
+      "loss": 0.2772,
+      "step": 112080
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 6.845747451447281e-06,
+      "loss": 0.2862,
+      "step": 112090
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 6.842771039511868e-06,
+      "loss": 0.2732,
+      "step": 112100
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 6.839794627576457e-06,
+      "loss": 0.2807,
+      "step": 112110
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 6.8368182156410456e-06,
+      "loss": 0.2792,
+      "step": 112120
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 6.833841803705633e-06,
+      "loss": 0.2715,
+      "step": 112130
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 6.830865391770222e-06,
+      "loss": 0.2869,
+      "step": 112140
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 6.8278889798348105e-06,
+      "loss": 0.2853,
+      "step": 112150
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 6.824912567899397e-06,
+      "loss": 0.2822,
+      "step": 112160
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 6.821936155963986e-06,
+      "loss": 0.2967,
+      "step": 112170
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 6.818959744028574e-06,
+      "loss": 0.2837,
+      "step": 112180
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 6.815983332093162e-06,
+      "loss": 0.2815,
+      "step": 112190
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 6.813006920157751e-06,
+      "loss": 0.2801,
+      "step": 112200
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 6.810030508222339e-06,
+      "loss": 0.2854,
+      "step": 112210
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 6.807054096286927e-06,
+      "loss": 0.2845,
+      "step": 112220
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 6.804077684351514e-06,
+      "loss": 0.2872,
+      "step": 112230
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 6.801101272416103e-06,
+      "loss": 0.2855,
+      "step": 112240
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 6.798124860480691e-06,
+      "loss": 0.2705,
+      "step": 112250
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 6.795148448545279e-06,
+      "loss": 0.2793,
+      "step": 112260
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 6.792172036609868e-06,
+      "loss": 0.2863,
+      "step": 112270
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 6.7891956246744554e-06,
+      "loss": 0.2874,
+      "step": 112280
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 6.786219212739043e-06,
+      "loss": 0.272,
+      "step": 112290
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 6.783242800803632e-06,
+      "loss": 0.2948,
+      "step": 112300
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 6.7802663888682195e-06,
+      "loss": 0.2813,
+      "step": 112310
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 6.777289976932808e-06,
+      "loss": 0.2911,
+      "step": 112320
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 6.774313564997396e-06,
+      "loss": 0.2792,
+      "step": 112330
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 6.7713371530619845e-06,
+      "loss": 0.2865,
+      "step": 112340
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 6.768360741126573e-06,
+      "loss": 0.2894,
+      "step": 112350
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 6.76538432919116e-06,
+      "loss": 0.2678,
+      "step": 112360
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 6.7624079172557485e-06,
+      "loss": 0.2835,
+      "step": 112370
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 6.759431505320337e-06,
+      "loss": 0.2938,
+      "step": 112380
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 6.756455093384925e-06,
+      "loss": 0.286,
+      "step": 112390
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 6.7534786814495135e-06,
+      "loss": 0.28,
+      "step": 112400
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 6.750502269514101e-06,
+      "loss": 0.2833,
+      "step": 112410
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 6.74752585757869e-06,
+      "loss": 0.2846,
+      "step": 112420
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 6.744549445643278e-06,
+      "loss": 0.2839,
+      "step": 112430
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 6.741573033707865e-06,
+      "loss": 0.2704,
+      "step": 112440
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 6.738596621772454e-06,
+      "loss": 0.2774,
+      "step": 112450
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 6.735620209837042e-06,
+      "loss": 0.2835,
+      "step": 112460
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 6.73264379790163e-06,
+      "loss": 0.2843,
+      "step": 112470
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 6.729667385966219e-06,
+      "loss": 0.2792,
+      "step": 112480
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 6.7266909740308066e-06,
+      "loss": 0.2907,
+      "step": 112490
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 6.723714562095395e-06,
+      "loss": 0.281,
+      "step": 112500
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 6.720738150159982e-06,
+      "loss": 0.28,
+      "step": 112510
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 6.717761738224571e-06,
+      "loss": 0.2862,
+      "step": 112520
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 6.714785326289159e-06,
+      "loss": 0.2827,
+      "step": 112530
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 6.711808914353747e-06,
+      "loss": 0.2859,
+      "step": 112540
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 6.708832502418336e-06,
+      "loss": 0.2783,
+      "step": 112550
+    },
+    {
+      "epoch": 24.99,
+      "eval_cer": 4.5405982905982905,
+      "eval_loss": 0.5580922365188599,
+      "eval_runtime": 1.2535,
+      "eval_samples_per_second": 7.978,
+      "eval_steps_per_second": 0.798,
+      "eval_wer": 0.6103896103896104,
+      "step": 112550
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 6.7058560904829225e-06,
+      "loss": 0.2842,
+      "step": 112560
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 6.702879678547511e-06,
+      "loss": 0.2708,
+      "step": 112570
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 6.6999032666121e-06,
+      "loss": 0.2879,
+      "step": 112580
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 6.696926854676687e-06,
+      "loss": 0.2709,
+      "step": 112590
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 6.693950442741276e-06,
+      "loss": 0.2733,
+      "step": 112600
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 6.690974030805865e-06,
+      "loss": 0.2733,
+      "step": 112610
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 6.687997618870452e-06,
+      "loss": 0.3064,
+      "step": 112620
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 6.685021206935041e-06,
+      "loss": 0.2772,
+      "step": 112630
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 6.682044794999628e-06,
+      "loss": 0.2716,
+      "step": 112640
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 6.6790683830642164e-06,
+      "loss": 0.2614,
+      "step": 112650
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 6.676091971128805e-06,
+      "loss": 0.2696,
+      "step": 112660
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 6.673115559193393e-06,
+      "loss": 0.267,
+      "step": 112670
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 6.670139147257981e-06,
+      "loss": 0.2687,
+      "step": 112680
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 6.667162735322569e-06,
+      "loss": 0.2878,
+      "step": 112690
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 6.664186323387158e-06,
+      "loss": 0.2524,
+      "step": 112700
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 6.661209911451746e-06,
+      "loss": 0.2776,
+      "step": 112710
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 6.658233499516333e-06,
+      "loss": 0.2808,
+      "step": 112720
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 6.655257087580922e-06,
+      "loss": 0.2924,
+      "step": 112730
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 6.6522806756455095e-06,
+      "loss": 0.2882,
+      "step": 112740
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 6.649304263710098e-06,
+      "loss": 0.2726,
+      "step": 112750
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 6.646327851774687e-06,
+      "loss": 0.2724,
+      "step": 112760
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 6.643351439839274e-06,
+      "loss": 0.2755,
+      "step": 112770
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 6.640375027903862e-06,
+      "loss": 0.2597,
+      "step": 112780
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 6.63739861596845e-06,
+      "loss": 0.2907,
+      "step": 112790
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 6.6344222040330386e-06,
+      "loss": 0.2726,
+      "step": 112800
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 6.631445792097627e-06,
+      "loss": 0.2714,
+      "step": 112810
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 6.628469380162215e-06,
+      "loss": 0.2843,
+      "step": 112820
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 6.6254929682268035e-06,
+      "loss": 0.2818,
+      "step": 112830
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 6.622516556291392e-06,
+      "loss": 0.2795,
+      "step": 112840
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 6.619540144355979e-06,
+      "loss": 0.2875,
+      "step": 112850
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 6.6165637324205676e-06,
+      "loss": 0.2722,
+      "step": 112860
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 6.613587320485155e-06,
+      "loss": 0.2703,
+      "step": 112870
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 6.610610908549744e-06,
+      "loss": 0.2685,
+      "step": 112880
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 6.6076344966143325e-06,
+      "loss": 0.2833,
+      "step": 112890
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 6.60465808467892e-06,
+      "loss": 0.2675,
+      "step": 112900
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 6.601681672743509e-06,
+      "loss": 0.2758,
+      "step": 112910
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 6.598705260808096e-06,
+      "loss": 0.2836,
+      "step": 112920
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 6.595728848872684e-06,
+      "loss": 0.2808,
+      "step": 112930
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 6.592752436937273e-06,
+      "loss": 0.2807,
+      "step": 112940
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 6.589776025001861e-06,
+      "loss": 0.2732,
+      "step": 112950
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 6.586799613066449e-06,
+      "loss": 0.278,
+      "step": 112960
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 6.583823201131036e-06,
+      "loss": 0.2809,
+      "step": 112970
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 6.580846789195625e-06,
+      "loss": 0.2717,
+      "step": 112980
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 6.577870377260213e-06,
+      "loss": 0.2939,
+      "step": 112990
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 6.574893965324801e-06,
+      "loss": 0.2781,
+      "step": 113000
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 6.57191755338939e-06,
+      "loss": 0.2841,
+      "step": 113010
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 6.568941141453978e-06,
+      "loss": 0.2804,
+      "step": 113020
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 6.565964729518566e-06,
+      "loss": 0.2801,
+      "step": 113030
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 6.562988317583155e-06,
+      "loss": 0.2775,
+      "step": 113040
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 6.5600119056477415e-06,
+      "loss": 0.2818,
+      "step": 113050
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 6.55703549371233e-06,
+      "loss": 0.28,
+      "step": 113060
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 6.554059081776919e-06,
+      "loss": 0.2802,
+      "step": 113070
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 6.5510826698415064e-06,
+      "loss": 0.2753,
+      "step": 113080
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 6.548106257906095e-06,
+      "loss": 0.2704,
+      "step": 113090
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 6.545129845970683e-06,
+      "loss": 0.2896,
+      "step": 113100
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 6.542153434035271e-06,
+      "loss": 0.2755,
+      "step": 113110
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 6.53917702209986e-06,
+      "loss": 0.2765,
+      "step": 113120
+    },
+    {
+      "epoch": 25.12,
+      "learning_rate": 6.536200610164447e-06,
+      "loss": 0.2637,
+      "step": 113130
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.5332241982290355e-06,
+      "loss": 0.2935,
+      "step": 113140
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.530247786293623e-06,
+      "loss": 0.281,
+      "step": 113150
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.527271374358212e-06,
+      "loss": 0.2804,
+      "step": 113160
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.5242949624228e-06,
+      "loss": 0.273,
+      "step": 113170
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 6.521318550487387e-06,
+      "loss": 0.2729,
+      "step": 113180
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 6.518342138551976e-06,
+      "loss": 0.2886,
+      "step": 113190
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 6.515365726616564e-06,
+      "loss": 0.2912,
+      "step": 113200
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 6.512389314681152e-06,
+      "loss": 0.2807,
+      "step": 113210
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 6.509412902745741e-06,
+      "loss": 0.2824,
+      "step": 113220
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 6.5064364908103286e-06,
+      "loss": 0.2724,
+      "step": 113230
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 6.503460078874917e-06,
+      "loss": 0.2675,
+      "step": 113240
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 6.500483666939506e-06,
+      "loss": 0.2821,
+      "step": 113250
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 6.497507255004093e-06,
+      "loss": 0.284,
+      "step": 113260
+    },
+    {
+      "epoch": 25.15,
+      "learning_rate": 6.494530843068681e-06,
+      "loss": 0.2808,
+      "step": 113270
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 6.491554431133269e-06,
+      "loss": 0.2579,
+      "step": 113280
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 6.488578019197858e-06,
+      "loss": 0.2704,
+      "step": 113290
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 6.485601607262446e-06,
+      "loss": 0.2739,
+      "step": 113300
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 6.482625195327034e-06,
+      "loss": 0.282,
+      "step": 113310
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 6.4796487833916225e-06,
+      "loss": 0.2841,
+      "step": 113320
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 6.476672371456209e-06,
+      "loss": 0.2858,
+      "step": 113330
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 6.473695959520798e-06,
+      "loss": 0.2944,
+      "step": 113340
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 6.470719547585387e-06,
+      "loss": 0.2699,
+      "step": 113350
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 6.467743135649974e-06,
+      "loss": 0.2884,
+      "step": 113360
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 6.464766723714563e-06,
+      "loss": 0.2855,
+      "step": 113370
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 6.461790311779151e-06,
+      "loss": 0.2807,
+      "step": 113380
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 6.4588138998437384e-06,
+      "loss": 0.2613,
+      "step": 113390
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 6.455837487908327e-06,
+      "loss": 0.2709,
+      "step": 113400
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 6.452861075972915e-06,
+      "loss": 0.2758,
+      "step": 113410
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 6.449884664037503e-06,
+      "loss": 0.2878,
+      "step": 113420
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 6.446908252102091e-06,
+      "loss": 0.2771,
+      "step": 113430
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 6.44393184016668e-06,
+      "loss": 0.2924,
+      "step": 113440
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 6.440955428231268e-06,
+      "loss": 0.2778,
+      "step": 113450
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 6.437979016295855e-06,
+      "loss": 0.2704,
+      "step": 113460
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 6.435002604360444e-06,
+      "loss": 0.2788,
+      "step": 113470
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 6.432026192425032e-06,
+      "loss": 0.2707,
+      "step": 113480
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 6.42904978048962e-06,
+      "loss": 0.2722,
+      "step": 113490
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 6.426073368554209e-06,
+      "loss": 0.2781,
+      "step": 113500
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 6.4230969566187965e-06,
+      "loss": 0.287,
+      "step": 113510
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 6.420120544683385e-06,
+      "loss": 0.283,
+      "step": 113520
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 6.417144132747974e-06,
+      "loss": 0.2747,
+      "step": 113530
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 6.4141677208125605e-06,
+      "loss": 0.2658,
+      "step": 113540
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 6.411191308877149e-06,
+      "loss": 0.291,
+      "step": 113550
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 6.408214896941737e-06,
+      "loss": 0.2821,
+      "step": 113560
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 6.4052384850063255e-06,
+      "loss": 0.2723,
+      "step": 113570
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 6.402262073070914e-06,
+      "loss": 0.2716,
+      "step": 113580
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 6.399285661135502e-06,
+      "loss": 0.2927,
+      "step": 113590
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 6.39630924920009e-06,
+      "loss": 0.2773,
+      "step": 113600
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 6.393332837264677e-06,
+      "loss": 0.2744,
+      "step": 113610
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 6.390356425329266e-06,
+      "loss": 0.2819,
+      "step": 113620
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 6.3873800133938545e-06,
+      "loss": 0.2871,
+      "step": 113630
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 6.384403601458442e-06,
+      "loss": 0.2901,
+      "step": 113640
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 6.381427189523031e-06,
+      "loss": 0.2686,
+      "step": 113650
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 6.378450777587618e-06,
+      "loss": 0.2585,
+      "step": 113660
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 6.375474365652206e-06,
+      "loss": 0.2678,
+      "step": 113670
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 6.372497953716795e-06,
+      "loss": 0.2751,
+      "step": 113680
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 6.369521541781383e-06,
+      "loss": 0.2771,
+      "step": 113690
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 6.366545129845971e-06,
+      "loss": 0.2769,
+      "step": 113700
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 6.36356871791056e-06,
+      "loss": 0.2827,
+      "step": 113710
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 6.360592305975148e-06,
+      "loss": 0.2786,
+      "step": 113720
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 6.357615894039736e-06,
+      "loss": 0.2828,
+      "step": 113730
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 6.354639482104323e-06,
+      "loss": 0.2704,
+      "step": 113740
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 6.351663070168912e-06,
+      "loss": 0.2785,
+      "step": 113750
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 6.3486866582335e-06,
+      "loss": 0.2767,
+      "step": 113760
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 6.345710246298088e-06,
+      "loss": 0.2823,
+      "step": 113770
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 6.342733834362677e-06,
+      "loss": 0.2881,
+      "step": 113780
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 6.339757422427264e-06,
+      "loss": 0.2669,
+      "step": 113790
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 6.336781010491853e-06,
+      "loss": 0.2859,
+      "step": 113800
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 6.3338045985564415e-06,
+      "loss": 0.2742,
+      "step": 113810
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 6.3308281866210284e-06,
+      "loss": 0.2943,
+      "step": 113820
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 6.327851774685617e-06,
+      "loss": 0.2862,
+      "step": 113830
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 6.324875362750205e-06,
+      "loss": 0.2932,
+      "step": 113840
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 6.321898950814793e-06,
+      "loss": 0.2698,
+      "step": 113850
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 6.318922538879382e-06,
+      "loss": 0.2762,
+      "step": 113860
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 6.315946126943969e-06,
+      "loss": 0.2773,
+      "step": 113870
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 6.3129697150085575e-06,
+      "loss": 0.2753,
+      "step": 113880
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 6.309993303073146e-06,
+      "loss": 0.2951,
+      "step": 113890
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 6.307016891137734e-06,
+      "loss": 0.278,
+      "step": 113900
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 6.304040479202322e-06,
+      "loss": 0.2896,
+      "step": 113910
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 6.30106406726691e-06,
+      "loss": 0.2748,
+      "step": 113920
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 6.298087655331499e-06,
+      "loss": 0.2783,
+      "step": 113930
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 6.295111243396087e-06,
+      "loss": 0.2685,
+      "step": 113940
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.292134831460674e-06,
+      "loss": 0.2839,
+      "step": 113950
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.289158419525263e-06,
+      "loss": 0.278,
+      "step": 113960
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.2861820075898506e-06,
+      "loss": 0.2794,
+      "step": 113970
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.283205595654439e-06,
+      "loss": 0.2826,
+      "step": 113980
+    },
+    {
+      "epoch": 25.31,
+      "learning_rate": 6.280229183719028e-06,
+      "loss": 0.2665,
+      "step": 113990
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 6.2772527717836155e-06,
+      "loss": 0.2799,
+      "step": 114000
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 6.274276359848204e-06,
+      "loss": 0.2795,
+      "step": 114010
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 6.271299947912791e-06,
+      "loss": 0.2695,
+      "step": 114020
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 6.26832353597738e-06,
+      "loss": 0.2842,
+      "step": 114030
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 6.265347124041968e-06,
+      "loss": 0.2812,
+      "step": 114040
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 6.262370712106556e-06,
+      "loss": 0.2736,
+      "step": 114050
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 6.2593943001711445e-06,
+      "loss": 0.2692,
+      "step": 114060
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 6.256417888235731e-06,
+      "loss": 0.2779,
+      "step": 114070
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 6.25344147630032e-06,
+      "loss": 0.2785,
+      "step": 114080
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 6.250465064364909e-06,
+      "loss": 0.2717,
+      "step": 114090
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 6.247488652429496e-06,
+      "loss": 0.2641,
+      "step": 114100
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 6.244512240494085e-06,
+      "loss": 0.2687,
+      "step": 114110
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 6.2415358285586735e-06,
+      "loss": 0.2816,
+      "step": 114120
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 6.238559416623261e-06,
+      "loss": 0.268,
+      "step": 114130
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 6.23558300468785e-06,
+      "loss": 0.2684,
+      "step": 114140
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 6.232606592752437e-06,
+      "loss": 0.282,
+      "step": 114150
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 6.229630180817025e-06,
+      "loss": 0.2803,
+      "step": 114160
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 6.226653768881614e-06,
+      "loss": 0.2842,
+      "step": 114170
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 6.223677356946202e-06,
+      "loss": 0.2736,
+      "step": 114180
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 6.22070094501079e-06,
+      "loss": 0.2862,
+      "step": 114190
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 6.217724533075378e-06,
+      "loss": 0.265,
+      "step": 114200
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 6.214748121139967e-06,
+      "loss": 0.2628,
+      "step": 114210
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 6.211771709204555e-06,
+      "loss": 0.2698,
+      "step": 114220
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 6.208795297269142e-06,
+      "loss": 0.2674,
+      "step": 114230
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 6.205818885333731e-06,
+      "loss": 0.2724,
+      "step": 114240
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 6.2028424733983185e-06,
+      "loss": 0.2746,
+      "step": 114250
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 6.199866061462907e-06,
+      "loss": 0.2816,
+      "step": 114260
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 6.196889649527496e-06,
+      "loss": 0.2753,
+      "step": 114270
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 6.1939132375920825e-06,
+      "loss": 0.2789,
+      "step": 114280
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 6.190936825656671e-06,
+      "loss": 0.2691,
+      "step": 114290
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 6.187960413721259e-06,
+      "loss": 0.2805,
+      "step": 114300
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 6.1849840017858475e-06,
+      "loss": 0.2614,
+      "step": 114310
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 6.182007589850436e-06,
+      "loss": 0.2782,
+      "step": 114320
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 6.179031177915024e-06,
+      "loss": 0.2823,
+      "step": 114330
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 6.176054765979612e-06,
+      "loss": 0.2709,
+      "step": 114340
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 6.173078354044201e-06,
+      "loss": 0.2818,
+      "step": 114350
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 6.170101942108788e-06,
+      "loss": 0.2768,
+      "step": 114360
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 6.1671255301733765e-06,
+      "loss": 0.275,
+      "step": 114370
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 6.164149118237964e-06,
+      "loss": 0.289,
+      "step": 114380
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 6.161172706302553e-06,
+      "loss": 0.2731,
+      "step": 114390
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 6.1581962943671414e-06,
+      "loss": 0.2686,
+      "step": 114400
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 6.155219882431729e-06,
+      "loss": 0.2838,
+      "step": 114410
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 6.152243470496318e-06,
+      "loss": 0.2702,
+      "step": 114420
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 6.149267058560905e-06,
+      "loss": 0.2713,
+      "step": 114430
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 6.146290646625493e-06,
+      "loss": 0.2704,
+      "step": 114440
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 6.143314234690082e-06,
+      "loss": 0.2865,
+      "step": 114450
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 6.14033782275467e-06,
+      "loss": 0.274,
+      "step": 114460
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 6.137361410819258e-06,
+      "loss": 0.2697,
+      "step": 114470
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 6.134384998883846e-06,
+      "loss": 0.276,
+      "step": 114480
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 6.1314085869484345e-06,
+      "loss": 0.2893,
+      "step": 114490
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 6.128432175013022e-06,
+      "loss": 0.2888,
+      "step": 114500
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 6.12545576307761e-06,
+      "loss": 0.2719,
+      "step": 114510
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 6.122479351142199e-06,
+      "loss": 0.2746,
+      "step": 114520
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 6.119502939206786e-06,
+      "loss": 0.2828,
+      "step": 114530
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 6.116526527271375e-06,
+      "loss": 0.2789,
+      "step": 114540
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 6.1135501153359635e-06,
+      "loss": 0.2794,
+      "step": 114550
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 6.1105737034005504e-06,
+      "loss": 0.2947,
+      "step": 114560
+    },
+    {
+      "epoch": 25.44,
+      "learning_rate": 6.107597291465139e-06,
+      "loss": 0.2783,
+      "step": 114570
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 6.104620879529728e-06,
+      "loss": 0.2817,
+      "step": 114580
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 6.101644467594315e-06,
+      "loss": 0.2661,
+      "step": 114590
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 6.098668055658904e-06,
+      "loss": 0.2766,
+      "step": 114600
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 6.095691643723492e-06,
+      "loss": 0.2775,
+      "step": 114610
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 6.09271523178808e-06,
+      "loss": 0.2616,
+      "step": 114620
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 6.089738819852669e-06,
+      "loss": 0.277,
+      "step": 114630
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 6.086762407917256e-06,
+      "loss": 0.2713,
+      "step": 114640
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 6.083785995981844e-06,
+      "loss": 0.2812,
+      "step": 114650
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 6.080809584046432e-06,
+      "loss": 0.2638,
+      "step": 114660
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 6.077833172111021e-06,
+      "loss": 0.2867,
+      "step": 114670
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 6.074856760175609e-06,
+      "loss": 0.2739,
+      "step": 114680
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 6.071880348240197e-06,
+      "loss": 0.279,
+      "step": 114690
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 6.068903936304786e-06,
+      "loss": 0.2748,
+      "step": 114700
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 6.0659275243693726e-06,
+      "loss": 0.2765,
+      "step": 114710
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 6.062951112433961e-06,
+      "loss": 0.2811,
+      "step": 114720
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 6.05997470049855e-06,
+      "loss": 0.2756,
+      "step": 114730
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 6.0569982885631375e-06,
+      "loss": 0.2646,
+      "step": 114740
+    },
+    {
+      "epoch": 25.48,
+      "learning_rate": 6.054021876627726e-06,
+      "loss": 0.2765,
+      "step": 114750
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.051045464692315e-06,
+      "loss": 0.2792,
+      "step": 114760
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.0480690527569016e-06,
+      "loss": 0.2755,
+      "step": 114770
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.04509264082149e-06,
+      "loss": 0.2754,
+      "step": 114780
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.042116228886078e-06,
+      "loss": 0.2902,
+      "step": 114790
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 6.0391398169506665e-06,
+      "loss": 0.278,
+      "step": 114800
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 6.036163405015255e-06,
+      "loss": 0.2744,
+      "step": 114810
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 6.033186993079843e-06,
+      "loss": 0.2752,
+      "step": 114820
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 6.0302105811444314e-06,
+      "loss": 0.2756,
+      "step": 114830
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 6.027234169209018e-06,
+      "loss": 0.2794,
+      "step": 114840
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 6.024257757273607e-06,
+      "loss": 0.2899,
+      "step": 114850
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 6.0212813453381955e-06,
+      "loss": 0.272,
+      "step": 114860
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 6.018304933402783e-06,
+      "loss": 0.2998,
+      "step": 114870
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 6.015328521467372e-06,
+      "loss": 0.2812,
+      "step": 114880
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 6.01235210953196e-06,
+      "loss": 0.2726,
+      "step": 114890
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 6.009375697596548e-06,
+      "loss": 0.2754,
+      "step": 114900
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 6.006399285661137e-06,
+      "loss": 0.2719,
+      "step": 114910
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 6.003422873725724e-06,
+      "loss": 0.273,
+      "step": 114920
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 6.000446461790312e-06,
+      "loss": 0.2741,
+      "step": 114930
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 5.9974700498549e-06,
+      "loss": 0.2842,
+      "step": 114940
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 5.994493637919489e-06,
+      "loss": 0.27,
+      "step": 114950
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 5.991517225984077e-06,
+      "loss": 0.2669,
+      "step": 114960
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 5.988540814048664e-06,
+      "loss": 0.2654,
+      "step": 114970
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 5.985564402113253e-06,
+      "loss": 0.2697,
+      "step": 114980
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 5.982587990177841e-06,
+      "loss": 0.2984,
+      "step": 114990
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 5.979611578242429e-06,
+      "loss": 0.2753,
+      "step": 115000
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 5.976635166307018e-06,
+      "loss": 0.2746,
+      "step": 115010
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 5.973658754371605e-06,
+      "loss": 0.2663,
+      "step": 115020
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 5.970682342436194e-06,
+      "loss": 0.2821,
+      "step": 115030
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 5.9677059305007826e-06,
+      "loss": 0.2622,
+      "step": 115040
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 5.9647295185653695e-06,
+      "loss": 0.2647,
+      "step": 115050
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 5.961753106629958e-06,
+      "loss": 0.2543,
+      "step": 115060
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 5.958776694694546e-06,
+      "loss": 0.278,
+      "step": 115070
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 5.955800282759134e-06,
+      "loss": 0.2791,
+      "step": 115080
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 5.952823870823723e-06,
+      "loss": 0.2741,
+      "step": 115090
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 5.949847458888311e-06,
+      "loss": 0.2857,
+      "step": 115100
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 5.946871046952899e-06,
+      "loss": 0.28,
+      "step": 115110
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 5.943894635017486e-06,
+      "loss": 0.2736,
+      "step": 115120
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 5.940918223082075e-06,
+      "loss": 0.2853,
+      "step": 115130
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 5.937941811146663e-06,
+      "loss": 0.2704,
+      "step": 115140
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 5.934965399211251e-06,
+      "loss": 0.2725,
+      "step": 115150
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 5.93198898727584e-06,
+      "loss": 0.2859,
+      "step": 115160
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 5.929012575340427e-06,
+      "loss": 0.2716,
+      "step": 115170
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 5.926036163405015e-06,
+      "loss": 0.2762,
+      "step": 115180
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 5.923059751469604e-06,
+      "loss": 0.2774,
+      "step": 115190
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 5.920083339534192e-06,
+      "loss": 0.3022,
+      "step": 115200
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 5.91710692759878e-06,
+      "loss": 0.2811,
+      "step": 115210
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 5.914130515663369e-06,
+      "loss": 0.2847,
+      "step": 115220
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 5.9111541037279565e-06,
+      "loss": 0.2839,
+      "step": 115230
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 5.908177691792545e-06,
+      "loss": 0.2743,
+      "step": 115240
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 5.905201279857132e-06,
+      "loss": 0.2806,
+      "step": 115250
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 5.902224867921721e-06,
+      "loss": 0.262,
+      "step": 115260
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 5.899248455986309e-06,
+      "loss": 0.2756,
+      "step": 115270
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 5.896272044050897e-06,
+      "loss": 0.2789,
+      "step": 115280
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 5.8932956321154855e-06,
+      "loss": 0.2892,
+      "step": 115290
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 5.890319220180073e-06,
+      "loss": 0.2671,
+      "step": 115300
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 5.887342808244662e-06,
+      "loss": 0.2737,
+      "step": 115310
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 5.8843663963092505e-06,
+      "loss": 0.2705,
+      "step": 115320
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 5.881389984373837e-06,
+      "loss": 0.2757,
+      "step": 115330
+    },
+    {
+      "epoch": 25.61,
+      "learning_rate": 5.878413572438426e-06,
+      "loss": 0.2819,
+      "step": 115340
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 5.875437160503014e-06,
+      "loss": 0.2886,
+      "step": 115350
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 5.872460748567602e-06,
+      "loss": 0.2752,
+      "step": 115360
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 5.869484336632191e-06,
+      "loss": 0.256,
+      "step": 115370
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 5.866507924696778e-06,
+      "loss": 0.2691,
+      "step": 115380
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 5.863531512761366e-06,
+      "loss": 0.2743,
+      "step": 115390
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 5.860555100825954e-06,
+      "loss": 0.283,
+      "step": 115400
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 5.857578688890543e-06,
+      "loss": 0.2759,
+      "step": 115410
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 5.854602276955131e-06,
+      "loss": 0.292,
+      "step": 115420
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 5.851625865019719e-06,
+      "loss": 0.2962,
+      "step": 115430
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 5.848649453084308e-06,
+      "loss": 0.2841,
+      "step": 115440
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 5.845673041148896e-06,
+      "loss": 0.2765,
+      "step": 115450
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 5.842696629213483e-06,
+      "loss": 0.2741,
+      "step": 115460
+    },
+    {
+      "epoch": 25.64,
+      "learning_rate": 5.839720217278072e-06,
+      "loss": 0.2838,
+      "step": 115470
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 5.8367438053426595e-06,
+      "loss": 0.2829,
+      "step": 115480
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 5.833767393407248e-06,
+      "loss": 0.288,
+      "step": 115490
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 5.830790981471837e-06,
+      "loss": 0.2674,
+      "step": 115500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 5.827814569536424e-06,
+      "loss": 0.2877,
+      "step": 115510
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 5.824838157601013e-06,
+      "loss": 0.2679,
+      "step": 115520
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 5.8218617456656e-06,
+      "loss": 0.2714,
+      "step": 115530
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 5.8188853337301885e-06,
+      "loss": 0.2854,
+      "step": 115540
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 5.815908921794777e-06,
+      "loss": 0.2783,
+      "step": 115550
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 5.812932509859365e-06,
+      "loss": 0.2785,
+      "step": 115560
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 5.8099560979239534e-06,
+      "loss": 0.2823,
+      "step": 115570
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 5.806979685988541e-06,
+      "loss": 0.2739,
+      "step": 115580
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 5.80400327405313e-06,
+      "loss": 0.2714,
+      "step": 115590
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 5.8010268621177175e-06,
+      "loss": 0.281,
+      "step": 115600
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 5.798050450182305e-06,
+      "loss": 0.2742,
+      "step": 115610
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 5.795074038246894e-06,
+      "loss": 0.2781,
+      "step": 115620
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 5.7920976263114825e-06,
+      "loss": 0.2814,
+      "step": 115630
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 5.78912121437607e-06,
+      "loss": 0.2774,
+      "step": 115640
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 5.786144802440659e-06,
+      "loss": 0.2834,
+      "step": 115650
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 5.783168390505246e-06,
+      "loss": 0.2707,
+      "step": 115660
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 5.780191978569834e-06,
+      "loss": 0.2737,
+      "step": 115670
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 5.777215566634423e-06,
+      "loss": 0.2885,
+      "step": 115680
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 5.774239154699011e-06,
+      "loss": 0.2907,
+      "step": 115690
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 5.771262742763599e-06,
+      "loss": 0.2887,
+      "step": 115700
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 5.768286330828187e-06,
+      "loss": 0.2802,
+      "step": 115710
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 5.7653099188927756e-06,
+      "loss": 0.2753,
+      "step": 115720
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 5.762333506957364e-06,
+      "loss": 0.284,
+      "step": 115730
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 5.759357095021951e-06,
+      "loss": 0.2822,
+      "step": 115740
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 5.75638068308654e-06,
+      "loss": 0.2701,
+      "step": 115750
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 5.753404271151127e-06,
+      "loss": 0.2783,
+      "step": 115760
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 5.750427859215716e-06,
+      "loss": 0.2923,
+      "step": 115770
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 5.7474514472803046e-06,
+      "loss": 0.2774,
+      "step": 115780
+    },
+    {
+      "epoch": 25.71,
+      "learning_rate": 5.744475035344892e-06,
+      "loss": 0.2799,
+      "step": 115790
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 5.741498623409481e-06,
+      "loss": 0.2928,
+      "step": 115800
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 5.738522211474068e-06,
+      "loss": 0.278,
+      "step": 115810
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 5.735545799538656e-06,
+      "loss": 0.283,
+      "step": 115820
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 5.732569387603245e-06,
+      "loss": 0.2699,
+      "step": 115830
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 5.729592975667833e-06,
+      "loss": 0.2667,
+      "step": 115840
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 5.726616563732421e-06,
+      "loss": 0.2965,
+      "step": 115850
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 5.72364015179701e-06,
+      "loss": 0.2802,
+      "step": 115860
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 5.720663739861597e-06,
+      "loss": 0.2853,
+      "step": 115870
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 5.717687327926185e-06,
+      "loss": 0.2684,
+      "step": 115880
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 5.714710915990773e-06,
+      "loss": 0.2734,
+      "step": 115890
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 5.711734504055362e-06,
+      "loss": 0.2794,
+      "step": 115900
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 5.70875809211995e-06,
+      "loss": 0.2685,
+      "step": 115910
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 5.705781680184538e-06,
+      "loss": 0.2934,
+      "step": 115920
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 5.702805268249127e-06,
+      "loss": 0.2745,
+      "step": 115930
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 5.699828856313714e-06,
+      "loss": 0.2813,
+      "step": 115940
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 5.696852444378302e-06,
+      "loss": 0.2665,
+      "step": 115950
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 5.693876032442891e-06,
+      "loss": 0.2875,
+      "step": 115960
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 5.6908996205074785e-06,
+      "loss": 0.2833,
+      "step": 115970
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 5.687923208572067e-06,
+      "loss": 0.2783,
+      "step": 115980
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 5.684946796636655e-06,
+      "loss": 0.2817,
+      "step": 115990
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 5.6819703847012434e-06,
+      "loss": 0.2766,
+      "step": 116000
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 5.678993972765832e-06,
+      "loss": 0.2575,
+      "step": 116010
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 5.676017560830419e-06,
+      "loss": 0.2899,
+      "step": 116020
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 5.6730411488950075e-06,
+      "loss": 0.2787,
+      "step": 116030
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 5.670064736959595e-06,
+      "loss": 0.2896,
+      "step": 116040
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 5.667088325024184e-06,
+      "loss": 0.2745,
+      "step": 116050
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 5.6641119130887725e-06,
+      "loss": 0.2789,
+      "step": 116060
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 5.661135501153359e-06,
+      "loss": 0.2755,
+      "step": 116070
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 5.658159089217948e-06,
+      "loss": 0.2688,
+      "step": 116080
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 5.6551826772825366e-06,
+      "loss": 0.2808,
+      "step": 116090
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 5.652206265347124e-06,
+      "loss": 0.2871,
+      "step": 116100
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 5.649229853411713e-06,
+      "loss": 0.279,
+      "step": 116110
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 5.646253441476301e-06,
+      "loss": 0.2806,
+      "step": 116120
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 5.643277029540889e-06,
+      "loss": 0.277,
+      "step": 116130
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 5.640300617605478e-06,
+      "loss": 0.2842,
+      "step": 116140
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 5.637324205670065e-06,
+      "loss": 0.2728,
+      "step": 116150
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 5.634347793734653e-06,
+      "loss": 0.2852,
+      "step": 116160
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 5.631371381799241e-06,
+      "loss": 0.2755,
+      "step": 116170
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 5.62839496986383e-06,
+      "loss": 0.277,
+      "step": 116180
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 5.625418557928418e-06,
+      "loss": 0.2757,
+      "step": 116190
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 5.622442145993006e-06,
+      "loss": 0.2708,
+      "step": 116200
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 5.619465734057595e-06,
+      "loss": 0.2788,
+      "step": 116210
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 5.6164893221221815e-06,
+      "loss": 0.2683,
+      "step": 116220
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 5.61351291018677e-06,
+      "loss": 0.2727,
+      "step": 116230
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 5.610536498251359e-06,
+      "loss": 0.2811,
+      "step": 116240
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 5.607560086315946e-06,
+      "loss": 0.2818,
+      "step": 116250
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 5.604583674380535e-06,
+      "loss": 0.2821,
+      "step": 116260
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 5.601607262445122e-06,
+      "loss": 0.2806,
+      "step": 116270
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 5.5986308505097105e-06,
+      "loss": 0.282,
+      "step": 116280
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 5.595654438574299e-06,
+      "loss": 0.2598,
+      "step": 116290
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 5.592678026638887e-06,
+      "loss": 0.2797,
+      "step": 116300
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 5.5897016147034754e-06,
+      "loss": 0.289,
+      "step": 116310
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 5.586725202768064e-06,
+      "loss": 0.2801,
+      "step": 116320
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 5.583748790832652e-06,
+      "loss": 0.286,
+      "step": 116330
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 5.58077237889724e-06,
+      "loss": 0.2778,
+      "step": 116340
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 5.577795966961827e-06,
+      "loss": 0.2869,
+      "step": 116350
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 5.574819555026416e-06,
+      "loss": 0.2737,
+      "step": 116360
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 5.5718431430910044e-06,
+      "loss": 0.2731,
+      "step": 116370
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 5.568866731155592e-06,
+      "loss": 0.2732,
+      "step": 116380
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 5.565890319220181e-06,
+      "loss": 0.2919,
+      "step": 116390
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 5.5629139072847685e-06,
+      "loss": 0.2805,
+      "step": 116400
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 5.559937495349357e-06,
+      "loss": 0.2741,
+      "step": 116410
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 5.556961083413946e-06,
+      "loss": 0.273,
+      "step": 116420
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 5.553984671478533e-06,
+      "loss": 0.2886,
+      "step": 116430
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 5.551008259543121e-06,
+      "loss": 0.2726,
+      "step": 116440
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 5.548031847607709e-06,
+      "loss": 0.2793,
+      "step": 116450
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 5.5450554356722975e-06,
+      "loss": 0.2778,
+      "step": 116460
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 5.542079023736886e-06,
+      "loss": 0.2827,
+      "step": 116470
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 5.539102611801473e-06,
+      "loss": 0.2782,
+      "step": 116480
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 5.536126199866062e-06,
+      "loss": 0.2654,
+      "step": 116490
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 5.53314978793065e-06,
+      "loss": 0.2705,
+      "step": 116500
+    },
+    {
+      "epoch": 25.87,
+      "learning_rate": 5.530173375995238e-06,
+      "loss": 0.2882,
+      "step": 116510
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 5.5271969640598266e-06,
+      "loss": 0.2709,
+      "step": 116520
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 5.524220552124414e-06,
+      "loss": 0.2765,
+      "step": 116530
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 5.521244140189003e-06,
+      "loss": 0.2818,
+      "step": 116540
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 5.5182677282535915e-06,
+      "loss": 0.2785,
+      "step": 116550
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 5.515291316318178e-06,
+      "loss": 0.2628,
+      "step": 116560
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 5.512314904382767e-06,
+      "loss": 0.2757,
+      "step": 116570
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 5.509338492447355e-06,
+      "loss": 0.2672,
+      "step": 116580
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 5.506362080511943e-06,
+      "loss": 0.27,
+      "step": 116590
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 5.503385668576532e-06,
+      "loss": 0.2705,
+      "step": 116600
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 5.50040925664112e-06,
+      "loss": 0.2749,
+      "step": 116610
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 5.497432844705708e-06,
+      "loss": 0.2803,
+      "step": 116620
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 5.494456432770295e-06,
+      "loss": 0.2713,
+      "step": 116630
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 5.491480020834884e-06,
+      "loss": 0.2697,
+      "step": 116640
+    },
+    {
+      "epoch": 25.9,
+      "learning_rate": 5.488503608899472e-06,
+      "loss": 0.2711,
+      "step": 116650
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 5.48552719696406e-06,
+      "loss": 0.2884,
+      "step": 116660
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 5.482550785028649e-06,
+      "loss": 0.2749,
+      "step": 116670
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 5.4795743730932364e-06,
+      "loss": 0.2786,
+      "step": 116680
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 5.476597961157825e-06,
+      "loss": 0.2768,
+      "step": 116690
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 5.473621549222413e-06,
+      "loss": 0.2744,
+      "step": 116700
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 5.4706451372870005e-06,
+      "loss": 0.2762,
+      "step": 116710
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 5.467668725351589e-06,
+      "loss": 0.2827,
+      "step": 116720
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 5.464692313416178e-06,
+      "loss": 0.2744,
+      "step": 116730
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 5.4617159014807654e-06,
+      "loss": 0.2633,
+      "step": 116740
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 5.458739489545354e-06,
+      "loss": 0.2673,
+      "step": 116750
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 5.455763077609941e-06,
+      "loss": 0.2864,
+      "step": 116760
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 5.4527866656745295e-06,
+      "loss": 0.28,
+      "step": 116770
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 5.449810253739118e-06,
+      "loss": 0.2656,
+      "step": 116780
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 5.446833841803706e-06,
+      "loss": 0.2858,
+      "step": 116790
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 5.4438574298682945e-06,
+      "loss": 0.2769,
+      "step": 116800
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 5.440881017932882e-06,
+      "loss": 0.2798,
+      "step": 116810
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 5.437904605997471e-06,
+      "loss": 0.2726,
+      "step": 116820
+    },
+    {
+      "epoch": 25.94,
+      "learning_rate": 5.434928194062059e-06,
+      "loss": 0.2877,
+      "step": 116830
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 5.431951782126646e-06,
+      "loss": 0.2828,
+      "step": 116840
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 5.428975370191235e-06,
+      "loss": 0.2818,
+      "step": 116850
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 5.425998958255823e-06,
+      "loss": 0.2639,
+      "step": 116860
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 5.423022546320411e-06,
+      "loss": 0.2746,
+      "step": 116870
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 5.420046134385e-06,
+      "loss": 0.2779,
+      "step": 116880
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 5.4170697224495876e-06,
+      "loss": 0.2636,
+      "step": 116890
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 5.414093310514176e-06,
+      "loss": 0.2746,
+      "step": 116900
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 5.411116898578763e-06,
+      "loss": 0.2776,
+      "step": 116910
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 5.408140486643352e-06,
+      "loss": 0.2697,
+      "step": 116920
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 5.40516407470794e-06,
+      "loss": 0.2692,
+      "step": 116930
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 5.402187662772528e-06,
+      "loss": 0.2811,
+      "step": 116940
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 5.399211250837117e-06,
+      "loss": 0.2721,
+      "step": 116950
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 5.396234838901705e-06,
+      "loss": 0.2695,
+      "step": 116960
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 5.393258426966292e-06,
+      "loss": 0.28,
+      "step": 116970
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 5.390282015030881e-06,
+      "loss": 0.2894,
+      "step": 116980
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 5.387305603095468e-06,
+      "loss": 0.2919,
+      "step": 116990
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 5.384329191160057e-06,
+      "loss": 0.2865,
+      "step": 117000
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 5.381352779224646e-06,
+      "loss": 0.2741,
+      "step": 117010
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 5.378376367289233e-06,
+      "loss": 0.2862,
+      "step": 117020
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 5.375399955353822e-06,
+      "loss": 0.3024,
+      "step": 117030
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 5.372423543418409e-06,
+      "loss": 0.2648,
+      "step": 117040
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 5.3694471314829974e-06,
+      "loss": 0.2854,
+      "step": 117050
+    },
+    {
+      "epoch": 25.99,
+      "eval_cer": 4.534188034188034,
+      "eval_loss": 0.5745441913604736,
+      "eval_runtime": 1.3139,
+      "eval_samples_per_second": 7.611,
+      "eval_steps_per_second": 0.761,
+      "eval_wer": 0.6233766233766234,
+      "step": 117052
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 5.366470719547586e-06,
+      "loss": 0.2752,
+      "step": 117060
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 5.363494307612174e-06,
+      "loss": 0.2768,
+      "step": 117070
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 5.360517895676762e-06,
+      "loss": 0.2756,
+      "step": 117080
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 5.35754148374135e-06,
+      "loss": 0.2777,
+      "step": 117090
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 5.354565071805939e-06,
+      "loss": 0.2787,
+      "step": 117100
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 5.351588659870527e-06,
+      "loss": 0.2703,
+      "step": 117110
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 5.348612247935114e-06,
+      "loss": 0.273,
+      "step": 117120
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 5.345635835999703e-06,
+      "loss": 0.2683,
+      "step": 117130
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 5.3426594240642905e-06,
+      "loss": 0.2774,
+      "step": 117140
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 5.339683012128879e-06,
+      "loss": 0.2674,
+      "step": 117150
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 5.336706600193468e-06,
+      "loss": 0.285,
+      "step": 117160
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 5.333730188258055e-06,
+      "loss": 0.2648,
+      "step": 117170
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 5.330753776322643e-06,
+      "loss": 0.2801,
+      "step": 117180
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 5.327777364387232e-06,
+      "loss": 0.2744,
+      "step": 117190
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 5.3248009524518195e-06,
+      "loss": 0.2644,
+      "step": 117200
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 5.321824540516408e-06,
+      "loss": 0.2666,
+      "step": 117210
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 5.318848128580996e-06,
+      "loss": 0.2841,
+      "step": 117220
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 5.3158717166455845e-06,
+      "loss": 0.2681,
+      "step": 117230
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 5.312895304710173e-06,
+      "loss": 0.283,
+      "step": 117240
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 5.30991889277476e-06,
+      "loss": 0.2853,
+      "step": 117250
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 5.3069424808393486e-06,
+      "loss": 0.2702,
+      "step": 117260
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 5.303966068903936e-06,
+      "loss": 0.266,
+      "step": 117270
+    },
+    {
+      "epoch": 26.04,
+      "learning_rate": 5.300989656968525e-06,
+      "loss": 0.2754,
+      "step": 117280
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 5.2980132450331135e-06,
+      "loss": 0.2725,
+      "step": 117290
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 5.295036833097701e-06,
+      "loss": 0.2763,
+      "step": 117300
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 5.29206042116229e-06,
+      "loss": 0.2857,
+      "step": 117310
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 5.289084009226877e-06,
+      "loss": 0.2654,
+      "step": 117320
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 5.286107597291465e-06,
+      "loss": 0.2793,
+      "step": 117330
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 5.283131185356054e-06,
+      "loss": 0.2723,
+      "step": 117340
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 5.280154773420642e-06,
+      "loss": 0.2659,
+      "step": 117350
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 5.27717836148523e-06,
+      "loss": 0.2689,
+      "step": 117360
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 5.274201949549819e-06,
+      "loss": 0.2621,
+      "step": 117370
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 5.271225537614406e-06,
+      "loss": 0.2823,
+      "step": 117380
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 5.268249125678994e-06,
+      "loss": 0.2827,
+      "step": 117390
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 5.265272713743582e-06,
+      "loss": 0.2653,
+      "step": 117400
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 5.262296301808171e-06,
+      "loss": 0.2699,
+      "step": 117410
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 5.259319889872759e-06,
+      "loss": 0.2873,
+      "step": 117420
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 5.256343477937347e-06,
+      "loss": 0.2698,
+      "step": 117430
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 5.253367066001936e-06,
+      "loss": 0.2605,
+      "step": 117440
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 5.2503906540665225e-06,
+      "loss": 0.2646,
+      "step": 117450
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 5.247414242131111e-06,
+      "loss": 0.2839,
+      "step": 117460
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 5.2444378301957e-06,
+      "loss": 0.2877,
+      "step": 117470
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 5.2414614182602874e-06,
+      "loss": 0.2648,
+      "step": 117480
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 5.238485006324876e-06,
+      "loss": 0.2762,
+      "step": 117490
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 5.235508594389464e-06,
+      "loss": 0.2621,
+      "step": 117500
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 5.232532182454052e-06,
+      "loss": 0.2614,
+      "step": 117510
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 5.229555770518641e-06,
+      "loss": 0.2876,
+      "step": 117520
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 5.226579358583228e-06,
+      "loss": 0.2816,
+      "step": 117530
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 5.2236029466478165e-06,
+      "loss": 0.2726,
+      "step": 117540
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 5.220626534712404e-06,
+      "loss": 0.2659,
+      "step": 117550
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 5.217650122776993e-06,
+      "loss": 0.2746,
+      "step": 117560
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 5.214673710841581e-06,
+      "loss": 0.2845,
+      "step": 117570
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 5.211697298906168e-06,
+      "loss": 0.2659,
+      "step": 117580
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 5.208720886970757e-06,
+      "loss": 0.2793,
+      "step": 117590
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 5.2057444750353455e-06,
+      "loss": 0.2847,
+      "step": 117600
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 5.202768063099933e-06,
+      "loss": 0.2783,
+      "step": 117610
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 5.199791651164522e-06,
+      "loss": 0.2647,
+      "step": 117620
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 5.1968152392291096e-06,
+      "loss": 0.2714,
+      "step": 117630
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 5.193838827293698e-06,
+      "loss": 0.2757,
+      "step": 117640
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 5.190862415358287e-06,
+      "loss": 0.2757,
+      "step": 117650
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 5.187886003422874e-06,
+      "loss": 0.2734,
+      "step": 117660
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 5.184909591487462e-06,
+      "loss": 0.265,
+      "step": 117670
+    },
+    {
+      "epoch": 26.13,
+      "learning_rate": 5.18193317955205e-06,
+      "loss": 0.2799,
+      "step": 117680
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 5.1789567676166386e-06,
+      "loss": 0.2737,
+      "step": 117690
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 5.175980355681227e-06,
+      "loss": 0.2655,
+      "step": 117700
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 5.173003943745815e-06,
+      "loss": 0.2648,
+      "step": 117710
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 5.1700275318104035e-06,
+      "loss": 0.2728,
+      "step": 117720
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 5.16705111987499e-06,
+      "loss": 0.2693,
+      "step": 117730
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 5.164074707939579e-06,
+      "loss": 0.2798,
+      "step": 117740
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 5.161098296004168e-06,
+      "loss": 0.2676,
+      "step": 117750
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 5.158121884068755e-06,
+      "loss": 0.2822,
+      "step": 117760
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 5.155145472133344e-06,
+      "loss": 0.2827,
+      "step": 117770
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 5.152169060197932e-06,
+      "loss": 0.2774,
+      "step": 117780
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 5.14919264826252e-06,
+      "loss": 0.2886,
+      "step": 117790
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 5.146216236327108e-06,
+      "loss": 0.27,
+      "step": 117800
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 5.143239824391696e-06,
+      "loss": 0.2652,
+      "step": 117810
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 5.140263412456284e-06,
+      "loss": 0.2639,
+      "step": 117820
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 5.137287000520873e-06,
+      "loss": 0.2725,
+      "step": 117830
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 5.134310588585461e-06,
+      "loss": 0.2792,
+      "step": 117840
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 5.131334176650049e-06,
+      "loss": 0.2688,
+      "step": 117850
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 5.128357764714636e-06,
+      "loss": 0.2629,
+      "step": 117860
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 5.125381352779225e-06,
+      "loss": 0.267,
+      "step": 117870
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 5.122404940843813e-06,
+      "loss": 0.279,
+      "step": 117880
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 5.119428528908401e-06,
+      "loss": 0.272,
+      "step": 117890
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 5.11645211697299e-06,
+      "loss": 0.2714,
+      "step": 117900
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 5.1134757050375775e-06,
+      "loss": 0.2689,
+      "step": 117910
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 5.110499293102166e-06,
+      "loss": 0.2653,
+      "step": 117920
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 5.107522881166755e-06,
+      "loss": 0.2811,
+      "step": 117930
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 5.1045464692313415e-06,
+      "loss": 0.2701,
+      "step": 117940
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 5.10157005729593e-06,
+      "loss": 0.2579,
+      "step": 117950
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 5.098593645360518e-06,
+      "loss": 0.2655,
+      "step": 117960
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 5.0956172334251065e-06,
+      "loss": 0.2752,
+      "step": 117970
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 5.092640821489695e-06,
+      "loss": 0.2743,
+      "step": 117980
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 5.089664409554283e-06,
+      "loss": 0.2785,
+      "step": 117990
+    },
+    {
+      "epoch": 26.2,
+      "learning_rate": 5.086687997618871e-06,
+      "loss": 0.2804,
+      "step": 118000
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 5.083711585683458e-06,
+      "loss": 0.2811,
+      "step": 118010
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 5.080735173748047e-06,
+      "loss": 0.2644,
+      "step": 118020
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 5.0777587618126355e-06,
+      "loss": 0.2663,
+      "step": 118030
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 5.074782349877223e-06,
+      "loss": 0.2752,
+      "step": 118040
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 5.071805937941812e-06,
+      "loss": 0.2746,
+      "step": 118050
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 5.0688295260064e-06,
+      "loss": 0.265,
+      "step": 118060
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 5.065853114070987e-06,
+      "loss": 0.2821,
+      "step": 118070
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 5.062876702135576e-06,
+      "loss": 0.2738,
+      "step": 118080
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 5.059900290200164e-06,
+      "loss": 0.2667,
+      "step": 118090
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 5.056923878264752e-06,
+      "loss": 0.2639,
+      "step": 118100
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 5.053947466329341e-06,
+      "loss": 0.2848,
+      "step": 118110
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 5.050971054393929e-06,
+      "loss": 0.2672,
+      "step": 118120
+    },
+    {
+      "epoch": 26.23,
+      "learning_rate": 5.047994642458517e-06,
+      "loss": 0.2691,
+      "step": 118130
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 5.045018230523104e-06,
+      "loss": 0.2712,
+      "step": 118140
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 5.042041818587693e-06,
+      "loss": 0.2681,
+      "step": 118150
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 5.039065406652281e-06,
+      "loss": 0.2711,
+      "step": 118160
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 5.036088994716869e-06,
+      "loss": 0.2663,
+      "step": 118170
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 5.033112582781458e-06,
+      "loss": 0.2688,
+      "step": 118180
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 5.030136170846045e-06,
+      "loss": 0.2701,
+      "step": 118190
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 5.027159758910634e-06,
+      "loss": 0.2849,
+      "step": 118200
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 5.0241833469752225e-06,
+      "loss": 0.2943,
+      "step": 118210
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 5.0212069350398094e-06,
+      "loss": 0.2834,
+      "step": 118220
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 5.018230523104398e-06,
+      "loss": 0.2756,
+      "step": 118230
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 5.015254111168987e-06,
+      "loss": 0.2795,
+      "step": 118240
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 5.012277699233574e-06,
+      "loss": 0.2781,
+      "step": 118250
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 5.009301287298163e-06,
+      "loss": 0.268,
+      "step": 118260
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 5.00632487536275e-06,
+      "loss": 0.2675,
+      "step": 118270
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 5.0033484634273385e-06,
+      "loss": 0.2615,
+      "step": 118280
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 5.000372051491927e-06,
+      "loss": 0.2809,
+      "step": 118290
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 4.997395639556515e-06,
+      "loss": 0.2782,
+      "step": 118300
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 4.994419227621103e-06,
+      "loss": 0.2782,
+      "step": 118310
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.991442815685691e-06,
+      "loss": 0.2711,
+      "step": 118320
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.98846640375028e-06,
+      "loss": 0.277,
+      "step": 118330
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.9854899918148675e-06,
+      "loss": 0.2794,
+      "step": 118340
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.982513579879455e-06,
+      "loss": 0.2723,
+      "step": 118350
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 4.979537167944044e-06,
+      "loss": 0.2682,
+      "step": 118360
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.976560756008632e-06,
+      "loss": 0.2636,
+      "step": 118370
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.97358434407322e-06,
+      "loss": 0.2804,
+      "step": 118380
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.970607932137808e-06,
+      "loss": 0.2907,
+      "step": 118390
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 4.9676315202023965e-06,
+      "loss": 0.2878,
+      "step": 118400
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.964655108266985e-06,
+      "loss": 0.2716,
+      "step": 118410
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.961678696331573e-06,
+      "loss": 0.2787,
+      "step": 118420
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.9587022843961606e-06,
+      "loss": 0.278,
+      "step": 118430
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.955725872460749e-06,
+      "loss": 0.2701,
+      "step": 118440
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 4.952749460525338e-06,
+      "loss": 0.2642,
+      "step": 118450
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.9497730485899255e-06,
+      "loss": 0.2836,
+      "step": 118460
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.946796636654513e-06,
+      "loss": 0.2794,
+      "step": 118470
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.943820224719101e-06,
+      "loss": 0.2686,
+      "step": 118480
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 4.94084381278369e-06,
+      "loss": 0.2631,
+      "step": 118490
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.937867400848278e-06,
+      "loss": 0.2825,
+      "step": 118500
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.934890988912866e-06,
+      "loss": 0.2861,
+      "step": 118510
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.931914576977454e-06,
+      "loss": 0.2889,
+      "step": 118520
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.928938165042042e-06,
+      "loss": 0.2755,
+      "step": 118530
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 4.925961753106631e-06,
+      "loss": 0.2679,
+      "step": 118540
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.922985341171219e-06,
+      "loss": 0.2758,
+      "step": 118550
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.920008929235806e-06,
+      "loss": 0.2734,
+      "step": 118560
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.917032517300395e-06,
+      "loss": 0.2843,
+      "step": 118570
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 4.914056105364983e-06,
+      "loss": 0.2685,
+      "step": 118580
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.911079693429571e-06,
+      "loss": 0.2702,
+      "step": 118590
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.908103281494159e-06,
+      "loss": 0.2694,
+      "step": 118600
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.905126869558748e-06,
+      "loss": 0.2832,
+      "step": 118610
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.902150457623335e-06,
+      "loss": 0.2707,
+      "step": 118620
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 4.899174045687924e-06,
+      "loss": 0.2701,
+      "step": 118630
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.896197633752512e-06,
+      "loss": 0.2765,
+      "step": 118640
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.8932212218171e-06,
+      "loss": 0.2725,
+      "step": 118650
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.890244809881688e-06,
+      "loss": 0.2853,
+      "step": 118660
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 4.887268397946276e-06,
+      "loss": 0.2788,
+      "step": 118670
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.884291986010864e-06,
+      "loss": 0.2571,
+      "step": 118680
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.881315574075452e-06,
+      "loss": 0.2727,
+      "step": 118690
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.878339162140041e-06,
+      "loss": 0.2701,
+      "step": 118700
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.8753627502046285e-06,
+      "loss": 0.2683,
+      "step": 118710
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 4.872386338269216e-06,
+      "loss": 0.2615,
+      "step": 118720
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.869409926333805e-06,
+      "loss": 0.2661,
+      "step": 118730
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.866433514398393e-06,
+      "loss": 0.259,
+      "step": 118740
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.863457102462981e-06,
+      "loss": 0.2644,
+      "step": 118750
+    },
+    {
+      "epoch": 26.37,
+      "learning_rate": 4.860480690527569e-06,
+      "loss": 0.2718,
+      "step": 118760
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.8575042785921575e-06,
+      "loss": 0.2776,
+      "step": 118770
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.854527866656746e-06,
+      "loss": 0.275,
+      "step": 118780
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.851551454721334e-06,
+      "loss": 0.2692,
+      "step": 118790
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.8485750427859216e-06,
+      "loss": 0.2665,
+      "step": 118800
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 4.84559863085051e-06,
+      "loss": 0.2643,
+      "step": 118810
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.842622218915099e-06,
+      "loss": 0.2867,
+      "step": 118820
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.8396458069796865e-06,
+      "loss": 0.2773,
+      "step": 118830
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.836669395044274e-06,
+      "loss": 0.2702,
+      "step": 118840
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 4.833692983108863e-06,
+      "loss": 0.2692,
+      "step": 118850
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.8307165711734514e-06,
+      "loss": 0.2737,
+      "step": 118860
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.827740159238039e-06,
+      "loss": 0.2813,
+      "step": 118870
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.824763747302627e-06,
+      "loss": 0.2836,
+      "step": 118880
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.8217873353672155e-06,
+      "loss": 0.2698,
+      "step": 118890
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 4.818810923431803e-06,
+      "loss": 0.2623,
+      "step": 118900
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.815834511496392e-06,
+      "loss": 0.2879,
+      "step": 118910
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.81285809956098e-06,
+      "loss": 0.2758,
+      "step": 118920
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.809881687625567e-06,
+      "loss": 0.2715,
+      "step": 118930
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 4.806905275690156e-06,
+      "loss": 0.2614,
+      "step": 118940
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.8039288637547445e-06,
+      "loss": 0.2728,
+      "step": 118950
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.800952451819332e-06,
+      "loss": 0.281,
+      "step": 118960
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.79797603988392e-06,
+      "loss": 0.2744,
+      "step": 118970
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.794999627948509e-06,
+      "loss": 0.2748,
+      "step": 118980
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 4.792023216013096e-06,
+      "loss": 0.2752,
+      "step": 118990
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.789046804077685e-06,
+      "loss": 0.2684,
+      "step": 119000
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.786070392142273e-06,
+      "loss": 0.2719,
+      "step": 119010
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.783093980206861e-06,
+      "loss": 0.2614,
+      "step": 119020
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 4.780117568271449e-06,
+      "loss": 0.2961,
+      "step": 119030
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.777141156336037e-06,
+      "loss": 0.2712,
+      "step": 119040
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.774164744400625e-06,
+      "loss": 0.2665,
+      "step": 119050
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.771188332465214e-06,
+      "loss": 0.2737,
+      "step": 119060
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.768211920529802e-06,
+      "loss": 0.2756,
+      "step": 119070
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 4.7652355085943895e-06,
+      "loss": 0.2774,
+      "step": 119080
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.762259096658978e-06,
+      "loss": 0.2747,
+      "step": 119090
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.759282684723567e-06,
+      "loss": 0.2804,
+      "step": 119100
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.756306272788154e-06,
+      "loss": 0.2546,
+      "step": 119110
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 4.753329860852742e-06,
+      "loss": 0.2703,
+      "step": 119120
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.75035344891733e-06,
+      "loss": 0.2691,
+      "step": 119130
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.7473770369819185e-06,
+      "loss": 0.2789,
+      "step": 119140
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.744400625046507e-06,
+      "loss": 0.2796,
+      "step": 119150
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.741424213111095e-06,
+      "loss": 0.2774,
+      "step": 119160
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 4.7384478011756826e-06,
+      "loss": 0.2846,
+      "step": 119170
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.735471389240271e-06,
+      "loss": 0.2788,
+      "step": 119180
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.73249497730486e-06,
+      "loss": 0.2755,
+      "step": 119190
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.7295185653694475e-06,
+      "loss": 0.2727,
+      "step": 119200
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 4.726542153434035e-06,
+      "loss": 0.2777,
+      "step": 119210
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.723565741498624e-06,
+      "loss": 0.2761,
+      "step": 119220
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.7205893295632124e-06,
+      "loss": 0.2676,
+      "step": 119230
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.7176129176278e-06,
+      "loss": 0.2528,
+      "step": 119240
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.714636505692388e-06,
+      "loss": 0.2802,
+      "step": 119250
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 4.7116600937569765e-06,
+      "loss": 0.2747,
+      "step": 119260
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.708683681821565e-06,
+      "loss": 0.2673,
+      "step": 119270
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.705707269886153e-06,
+      "loss": 0.2761,
+      "step": 119280
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.702730857950741e-06,
+      "loss": 0.2882,
+      "step": 119290
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 4.699754446015329e-06,
+      "loss": 0.2787,
+      "step": 119300
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.696778034079917e-06,
+      "loss": 0.2753,
+      "step": 119310
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.6938016221445055e-06,
+      "loss": 0.2777,
+      "step": 119320
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.690825210209093e-06,
+      "loss": 0.2745,
+      "step": 119330
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.687848798273681e-06,
+      "loss": 0.2765,
+      "step": 119340
+    },
+    {
+      "epoch": 26.5,
+      "learning_rate": 4.68487238633827e-06,
+      "loss": 0.2791,
+      "step": 119350
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.681895974402857e-06,
+      "loss": 0.2801,
+      "step": 119360
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.678919562467446e-06,
+      "loss": 0.277,
+      "step": 119370
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.675943150532034e-06,
+      "loss": 0.2787,
+      "step": 119380
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 4.672966738596622e-06,
+      "loss": 0.2787,
+      "step": 119390
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.66999032666121e-06,
+      "loss": 0.2771,
+      "step": 119400
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.667013914725799e-06,
+      "loss": 0.2827,
+      "step": 119410
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.664037502790386e-06,
+      "loss": 0.2756,
+      "step": 119420
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.661061090854975e-06,
+      "loss": 0.2532,
+      "step": 119430
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 4.658084678919563e-06,
+      "loss": 0.2765,
+      "step": 119440
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.6551082669841505e-06,
+      "loss": 0.2662,
+      "step": 119450
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.652131855048739e-06,
+      "loss": 0.2689,
+      "step": 119460
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.649155443113328e-06,
+      "loss": 0.2689,
+      "step": 119470
+    },
+    {
+      "epoch": 26.53,
+      "learning_rate": 4.646179031177915e-06,
+      "loss": 0.2833,
+      "step": 119480
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.643202619242503e-06,
+      "loss": 0.2717,
+      "step": 119490
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.640226207307092e-06,
+      "loss": 0.2706,
+      "step": 119500
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.63724979537168e-06,
+      "loss": 0.2913,
+      "step": 119510
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.634273383436268e-06,
+      "loss": 0.2625,
+      "step": 119520
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 4.631296971500856e-06,
+      "loss": 0.2818,
+      "step": 119530
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.628320559565444e-06,
+      "loss": 0.2708,
+      "step": 119540
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.625344147630033e-06,
+      "loss": 0.2649,
+      "step": 119550
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.622367735694621e-06,
+      "loss": 0.2777,
+      "step": 119560
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 4.6193913237592085e-06,
+      "loss": 0.2852,
+      "step": 119570
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.616414911823796e-06,
+      "loss": 0.2843,
+      "step": 119580
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.613438499888385e-06,
+      "loss": 0.2853,
+      "step": 119590
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.6104620879529734e-06,
+      "loss": 0.2628,
+      "step": 119600
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.607485676017561e-06,
+      "loss": 0.2762,
+      "step": 119610
+    },
+    {
+      "epoch": 26.56,
+      "learning_rate": 4.604509264082149e-06,
+      "loss": 0.2756,
+      "step": 119620
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.6015328521467375e-06,
+      "loss": 0.2815,
+      "step": 119630
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.598556440211326e-06,
+      "loss": 0.2733,
+      "step": 119640
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.595580028275914e-06,
+      "loss": 0.2724,
+      "step": 119650
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 4.592603616340502e-06,
+      "loss": 0.2816,
+      "step": 119660
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.58962720440509e-06,
+      "loss": 0.2826,
+      "step": 119670
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.586650792469678e-06,
+      "loss": 0.282,
+      "step": 119680
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.5836743805342665e-06,
+      "loss": 0.2754,
+      "step": 119690
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.580697968598854e-06,
+      "loss": 0.2872,
+      "step": 119700
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 4.577721556663443e-06,
+      "loss": 0.2719,
+      "step": 119710
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.574745144728031e-06,
+      "loss": 0.2854,
+      "step": 119720
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.571768732792619e-06,
+      "loss": 0.2738,
+      "step": 119730
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.568792320857207e-06,
+      "loss": 0.2856,
+      "step": 119740
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 4.5658159089217955e-06,
+      "loss": 0.2732,
+      "step": 119750
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.562839496986383e-06,
+      "loss": 0.2675,
+      "step": 119760
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.559863085050971e-06,
+      "loss": 0.262,
+      "step": 119770
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.55688667311556e-06,
+      "loss": 0.2711,
+      "step": 119780
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.553910261180147e-06,
+      "loss": 0.2956,
+      "step": 119790
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 4.550933849244736e-06,
+      "loss": 0.2709,
+      "step": 119800
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.547957437309324e-06,
+      "loss": 0.277,
+      "step": 119810
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.544981025373912e-06,
+      "loss": 0.2823,
+      "step": 119820
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.5420046134385e-06,
+      "loss": 0.2559,
+      "step": 119830
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 4.539028201503089e-06,
+      "loss": 0.2762,
+      "step": 119840
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.536051789567676e-06,
+      "loss": 0.2767,
+      "step": 119850
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.533075377632264e-06,
+      "loss": 0.2784,
+      "step": 119860
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.530098965696853e-06,
+      "loss": 0.2718,
+      "step": 119870
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.527122553761441e-06,
+      "loss": 0.2736,
+      "step": 119880
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 4.524146141826029e-06,
+      "loss": 0.2773,
+      "step": 119890
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.521169729890617e-06,
+      "loss": 0.2762,
+      "step": 119900
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.518193317955205e-06,
+      "loss": 0.2721,
+      "step": 119910
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.515216906019794e-06,
+      "loss": 0.2774,
+      "step": 119920
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 4.512240494084382e-06,
+      "loss": 0.2765,
+      "step": 119930
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.5092640821489695e-06,
+      "loss": 0.2755,
+      "step": 119940
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.506287670213558e-06,
+      "loss": 0.2843,
+      "step": 119950
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.503311258278147e-06,
+      "loss": 0.2736,
+      "step": 119960
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.5003348463427344e-06,
+      "loss": 0.2786,
+      "step": 119970
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 4.497358434407322e-06,
+      "loss": 0.2734,
+      "step": 119980
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.494382022471911e-06,
+      "loss": 0.2774,
+      "step": 119990
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.4914056105364985e-06,
+      "loss": 0.2713,
+      "step": 120000
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.488429198601087e-06,
+      "loss": 0.2689,
+      "step": 120010
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 4.485452786665675e-06,
+      "loss": 0.2733,
+      "step": 120020
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.482476374730263e-06,
+      "loss": 0.2761,
+      "step": 120030
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.479499962794851e-06,
+      "loss": 0.2819,
+      "step": 120040
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.47652355085944e-06,
+      "loss": 0.2711,
+      "step": 120050
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.4735471389240275e-06,
+      "loss": 0.2736,
+      "step": 120060
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 4.470570726988615e-06,
+      "loss": 0.2868,
+      "step": 120070
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.467594315053204e-06,
+      "loss": 0.2689,
+      "step": 120080
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.464617903117792e-06,
+      "loss": 0.2592,
+      "step": 120090
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.46164149118238e-06,
+      "loss": 0.2869,
+      "step": 120100
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 4.458665079246968e-06,
+      "loss": 0.2775,
+      "step": 120110
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.4556886673115565e-06,
+      "loss": 0.2738,
+      "step": 120120
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.452712255376144e-06,
+      "loss": 0.2651,
+      "step": 120130
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.449735843440733e-06,
+      "loss": 0.2831,
+      "step": 120140
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.446759431505321e-06,
+      "loss": 0.2758,
+      "step": 120150
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 4.443783019569909e-06,
+      "loss": 0.2768,
+      "step": 120160
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.440806607634497e-06,
+      "loss": 0.2787,
+      "step": 120170
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.437830195699085e-06,
+      "loss": 0.2723,
+      "step": 120180
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.434853783763673e-06,
+      "loss": 0.2885,
+      "step": 120190
+    },
+    {
+      "epoch": 26.69,
+      "learning_rate": 4.431877371828262e-06,
+      "loss": 0.2678,
+      "step": 120200
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.42890095989285e-06,
+      "loss": 0.2813,
+      "step": 120210
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.425924547957437e-06,
+      "loss": 0.2754,
+      "step": 120220
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.422948136022025e-06,
+      "loss": 0.2576,
+      "step": 120230
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.419971724086614e-06,
+      "loss": 0.2789,
+      "step": 120240
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 4.416995312151202e-06,
+      "loss": 0.2755,
+      "step": 120250
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.41401890021579e-06,
+      "loss": 0.2748,
+      "step": 120260
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.411042488280378e-06,
+      "loss": 0.2782,
+      "step": 120270
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.408066076344966e-06,
+      "loss": 0.2533,
+      "step": 120280
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 4.405089664409555e-06,
+      "loss": 0.2723,
+      "step": 120290
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.402113252474143e-06,
+      "loss": 0.2774,
+      "step": 120300
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.3991368405387305e-06,
+      "loss": 0.2739,
+      "step": 120310
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.396160428603319e-06,
+      "loss": 0.2735,
+      "step": 120320
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.393184016667908e-06,
+      "loss": 0.2906,
+      "step": 120330
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 4.3902076047324954e-06,
+      "loss": 0.2714,
+      "step": 120340
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.387231192797083e-06,
+      "loss": 0.2772,
+      "step": 120350
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.384254780861672e-06,
+      "loss": 0.2764,
+      "step": 120360
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.38127836892626e-06,
+      "loss": 0.2821,
+      "step": 120370
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 4.378301956990848e-06,
+      "loss": 0.2757,
+      "step": 120380
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.375325545055436e-06,
+      "loss": 0.2857,
+      "step": 120390
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.3723491331200244e-06,
+      "loss": 0.265,
+      "step": 120400
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.369372721184612e-06,
+      "loss": 0.2666,
+      "step": 120410
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.366396309249201e-06,
+      "loss": 0.2636,
+      "step": 120420
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 4.3634198973137885e-06,
+      "loss": 0.2673,
+      "step": 120430
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.360443485378376e-06,
+      "loss": 0.2632,
+      "step": 120440
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.357467073442965e-06,
+      "loss": 0.2707,
+      "step": 120450
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.354490661507553e-06,
+      "loss": 0.2809,
+      "step": 120460
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 4.351514249572141e-06,
+      "loss": 0.2615,
+      "step": 120470
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.348537837636729e-06,
+      "loss": 0.2702,
+      "step": 120480
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.3455614257013175e-06,
+      "loss": 0.2647,
+      "step": 120490
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.342585013765905e-06,
+      "loss": 0.2694,
+      "step": 120500
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.339608601830494e-06,
+      "loss": 0.2846,
+      "step": 120510
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 4.336632189895082e-06,
+      "loss": 0.2804,
+      "step": 120520
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.33365577795967e-06,
+      "loss": 0.2707,
+      "step": 120530
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.330679366024258e-06,
+      "loss": 0.2661,
+      "step": 120540
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.327702954088846e-06,
+      "loss": 0.2767,
+      "step": 120550
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 4.324726542153434e-06,
+      "loss": 0.2685,
+      "step": 120560
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.321750130218023e-06,
+      "loss": 0.2712,
+      "step": 120570
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.318773718282611e-06,
+      "loss": 0.2598,
+      "step": 120580
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.315797306347198e-06,
+      "loss": 0.2819,
+      "step": 120590
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.312820894411787e-06,
+      "loss": 0.2816,
+      "step": 120600
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 4.3098444824763756e-06,
+      "loss": 0.2813,
+      "step": 120610
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.306868070540963e-06,
+      "loss": 0.2707,
+      "step": 120620
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.303891658605551e-06,
+      "loss": 0.2769,
+      "step": 120630
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.30091524667014e-06,
+      "loss": 0.2734,
+      "step": 120640
+    },
+    {
+      "epoch": 26.79,
+      "learning_rate": 4.297938834734728e-06,
+      "loss": 0.2811,
+      "step": 120650
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.294962422799316e-06,
+      "loss": 0.2913,
+      "step": 120660
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.291986010863904e-06,
+      "loss": 0.2702,
+      "step": 120670
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.2890095989284915e-06,
+      "loss": 0.262,
+      "step": 120680
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.28603318699308e-06,
+      "loss": 0.2805,
+      "step": 120690
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 4.283056775057669e-06,
+      "loss": 0.2692,
+      "step": 120700
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.280080363122256e-06,
+      "loss": 0.2801,
+      "step": 120710
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.277103951186844e-06,
+      "loss": 0.2755,
+      "step": 120720
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.274127539251433e-06,
+      "loss": 0.2776,
+      "step": 120730
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 4.271151127316021e-06,
+      "loss": 0.2879,
+      "step": 120740
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.268174715380609e-06,
+      "loss": 0.281,
+      "step": 120750
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.265198303445197e-06,
+      "loss": 0.2821,
+      "step": 120760
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.2622218915097854e-06,
+      "loss": 0.2736,
+      "step": 120770
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.259245479574373e-06,
+      "loss": 0.2776,
+      "step": 120780
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 4.256269067638962e-06,
+      "loss": 0.282,
+      "step": 120790
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.2532926557035495e-06,
+      "loss": 0.2848,
+      "step": 120800
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.250316243768138e-06,
+      "loss": 0.2784,
+      "step": 120810
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.247339831832726e-06,
+      "loss": 0.272,
+      "step": 120820
+    },
+    {
+      "epoch": 26.83,
+      "learning_rate": 4.2443634198973145e-06,
+      "loss": 0.2739,
+      "step": 120830
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.241387007961902e-06,
+      "loss": 0.2683,
+      "step": 120840
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.238410596026491e-06,
+      "loss": 0.2747,
+      "step": 120850
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.2354341840910785e-06,
+      "loss": 0.2758,
+      "step": 120860
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.232457772155666e-06,
+      "loss": 0.2773,
+      "step": 120870
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 4.229481360220255e-06,
+      "loss": 0.2797,
+      "step": 120880
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.226504948284843e-06,
+      "loss": 0.2723,
+      "step": 120890
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.223528536349431e-06,
+      "loss": 0.2725,
+      "step": 120900
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.220552124414019e-06,
+      "loss": 0.2713,
+      "step": 120910
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 4.2175757124786076e-06,
+      "loss": 0.2695,
+      "step": 120920
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.214599300543195e-06,
+      "loss": 0.2664,
+      "step": 120930
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.211622888607784e-06,
+      "loss": 0.261,
+      "step": 120940
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.208646476672372e-06,
+      "loss": 0.2938,
+      "step": 120950
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.205670064736959e-06,
+      "loss": 0.2739,
+      "step": 120960
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 4.202693652801548e-06,
+      "loss": 0.2704,
+      "step": 120970
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.1997172408661366e-06,
+      "loss": 0.2756,
+      "step": 120980
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.196740828930724e-06,
+      "loss": 0.2592,
+      "step": 120990
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.193764416995312e-06,
+      "loss": 0.2803,
+      "step": 121000
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 4.190788005059901e-06,
+      "loss": 0.2838,
+      "step": 121010
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.187811593124489e-06,
+      "loss": 0.279,
+      "step": 121020
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.184835181189077e-06,
+      "loss": 0.2789,
+      "step": 121030
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.181858769253665e-06,
+      "loss": 0.28,
+      "step": 121040
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.178882357318253e-06,
+      "loss": 0.2655,
+      "step": 121050
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 4.175905945382842e-06,
+      "loss": 0.2724,
+      "step": 121060
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.17292953344743e-06,
+      "loss": 0.271,
+      "step": 121070
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.169953121512017e-06,
+      "loss": 0.2903,
+      "step": 121080
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.166976709576606e-06,
+      "loss": 0.2595,
+      "step": 121090
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 4.164000297641194e-06,
+      "loss": 0.286,
+      "step": 121100
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.161023885705782e-06,
+      "loss": 0.2639,
+      "step": 121110
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.15804747377037e-06,
+      "loss": 0.2736,
+      "step": 121120
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.155071061834958e-06,
+      "loss": 0.2818,
+      "step": 121130
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.1520946498995464e-06,
+      "loss": 0.2764,
+      "step": 121140
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 4.149118237964135e-06,
+      "loss": 0.2776,
+      "step": 121150
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.146141826028723e-06,
+      "loss": 0.264,
+      "step": 121160
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.1431654140933105e-06,
+      "loss": 0.2762,
+      "step": 121170
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.140189002157899e-06,
+      "loss": 0.2634,
+      "step": 121180
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 4.137212590222487e-06,
+      "loss": 0.2674,
+      "step": 121190
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.1342361782870755e-06,
+      "loss": 0.2714,
+      "step": 121200
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.131259766351663e-06,
+      "loss": 0.2872,
+      "step": 121210
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.128283354416252e-06,
+      "loss": 0.2763,
+      "step": 121220
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.1253069424808395e-06,
+      "loss": 0.2714,
+      "step": 121230
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 4.122330530545428e-06,
+      "loss": 0.2676,
+      "step": 121240
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.119354118610016e-06,
+      "loss": 0.2648,
+      "step": 121250
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.1163777066746045e-06,
+      "loss": 0.2687,
+      "step": 121260
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.113401294739192e-06,
+      "loss": 0.258,
+      "step": 121270
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 4.11042488280378e-06,
+      "loss": 0.2853,
+      "step": 121280
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.1074484708683686e-06,
+      "loss": 0.2695,
+      "step": 121290
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.104472058932957e-06,
+      "loss": 0.2834,
+      "step": 121300
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.101495646997545e-06,
+      "loss": 0.2755,
+      "step": 121310
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.098519235062133e-06,
+      "loss": 0.2769,
+      "step": 121320
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 4.09554282312672e-06,
+      "loss": 0.2739,
+      "step": 121330
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.092566411191309e-06,
+      "loss": 0.2744,
+      "step": 121340
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.0895899992558976e-06,
+      "loss": 0.283,
+      "step": 121350
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.086613587320485e-06,
+      "loss": 0.2856,
+      "step": 121360
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 4.083637175385073e-06,
+      "loss": 0.2684,
+      "step": 121370
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.080660763449662e-06,
+      "loss": 0.282,
+      "step": 121380
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.07768435151425e-06,
+      "loss": 0.2723,
+      "step": 121390
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.074707939578838e-06,
+      "loss": 0.2828,
+      "step": 121400
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.071731527643426e-06,
+      "loss": 0.2747,
+      "step": 121410
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 4.068755115708014e-06,
+      "loss": 0.2778,
+      "step": 121420
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.065778703772603e-06,
+      "loss": 0.2852,
+      "step": 121430
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.062802291837191e-06,
+      "loss": 0.2834,
+      "step": 121440
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.059825879901778e-06,
+      "loss": 0.2788,
+      "step": 121450
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 4.056849467966367e-06,
+      "loss": 0.2666,
+      "step": 121460
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.053873056030956e-06,
+      "loss": 0.2786,
+      "step": 121470
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.050896644095543e-06,
+      "loss": 0.2686,
+      "step": 121480
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.047920232160131e-06,
+      "loss": 0.2842,
+      "step": 121490
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.04494382022472e-06,
+      "loss": 0.2805,
+      "step": 121500
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 4.0419674082893074e-06,
+      "loss": 0.2772,
+      "step": 121510
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.038990996353896e-06,
+      "loss": 0.2786,
+      "step": 121520
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.036014584418484e-06,
+      "loss": 0.28,
+      "step": 121530
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.033038172483072e-06,
+      "loss": 0.275,
+      "step": 121540
+    },
+    {
+      "epoch": 26.99,
+      "learning_rate": 4.03006176054766e-06,
+      "loss": 0.2635,
+      "step": 121550
+    },
+    {
+      "epoch": 26.99,
+      "eval_cer": 4.52991452991453,
+      "eval_loss": 0.5519577264785767,
+      "eval_runtime": 1.2557,
+      "eval_samples_per_second": 7.964,
+      "eval_steps_per_second": 0.796,
+      "eval_wer": 0.6233766233766234,
+      "step": 121554
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.027085348612249e-06,
+      "loss": 0.2751,
+      "step": 121560
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.0241089366768365e-06,
+      "loss": 0.2777,
+      "step": 121570
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.021132524741424e-06,
+      "loss": 0.2744,
+      "step": 121580
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.018156112806013e-06,
+      "loss": 0.2718,
+      "step": 121590
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 4.0151797008706005e-06,
+      "loss": 0.2812,
+      "step": 121600
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.012203288935189e-06,
+      "loss": 0.2796,
+      "step": 121610
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.009226876999777e-06,
+      "loss": 0.2699,
+      "step": 121620
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.0062504650643655e-06,
+      "loss": 0.2767,
+      "step": 121630
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 4.003274053128953e-06,
+      "loss": 0.2771,
+      "step": 121640
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 4.000297641193541e-06,
+      "loss": 0.2759,
+      "step": 121650
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.9973212292581296e-06,
+      "loss": 0.2726,
+      "step": 121660
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.994344817322718e-06,
+      "loss": 0.2681,
+      "step": 121670
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.991368405387306e-06,
+      "loss": 0.2739,
+      "step": 121680
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 3.988391993451894e-06,
+      "loss": 0.2869,
+      "step": 121690
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.985415581516482e-06,
+      "loss": 0.2738,
+      "step": 121700
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.982439169581071e-06,
+      "loss": 0.2772,
+      "step": 121710
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.9794627576456586e-06,
+      "loss": 0.2665,
+      "step": 121720
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.976486345710246e-06,
+      "loss": 0.2827,
+      "step": 121730
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.973509933774835e-06,
+      "loss": 0.2622,
+      "step": 121740
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.9705335218394235e-06,
+      "loss": 0.2719,
+      "step": 121750
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.967557109904011e-06,
+      "loss": 0.2693,
+      "step": 121760
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.964580697968599e-06,
+      "loss": 0.2715,
+      "step": 121770
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.961604286033187e-06,
+      "loss": 0.2805,
+      "step": 121780
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.958627874097775e-06,
+      "loss": 0.2873,
+      "step": 121790
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.955651462162364e-06,
+      "loss": 0.2773,
+      "step": 121800
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.952675050226952e-06,
+      "loss": 0.2783,
+      "step": 121810
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.949698638291539e-06,
+      "loss": 0.2556,
+      "step": 121820
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.946722226356128e-06,
+      "loss": 0.2681,
+      "step": 121830
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.943745814420717e-06,
+      "loss": 0.2672,
+      "step": 121840
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.940769402485304e-06,
+      "loss": 0.2812,
+      "step": 121850
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.937792990549892e-06,
+      "loss": 0.2758,
+      "step": 121860
+    },
+    {
+      "epoch": 27.06,
+      "learning_rate": 3.934816578614481e-06,
+      "loss": 0.283,
+      "step": 121870
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.931840166679069e-06,
+      "loss": 0.2682,
+      "step": 121880
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.928863754743657e-06,
+      "loss": 0.2714,
+      "step": 121890
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.925887342808245e-06,
+      "loss": 0.2744,
+      "step": 121900
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.922910930872833e-06,
+      "loss": 0.272,
+      "step": 121910
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.919934518937421e-06,
+      "loss": 0.2766,
+      "step": 121920
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.91695810700201e-06,
+      "loss": 0.2696,
+      "step": 121930
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.9139816950665974e-06,
+      "loss": 0.2769,
+      "step": 121940
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.911005283131186e-06,
+      "loss": 0.2733,
+      "step": 121950
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 3.908028871195774e-06,
+      "loss": 0.2801,
+      "step": 121960
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.9050524592603615e-06,
+      "loss": 0.2689,
+      "step": 121970
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.90207604732495e-06,
+      "loss": 0.2669,
+      "step": 121980
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.899099635389538e-06,
+      "loss": 0.2774,
+      "step": 121990
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 3.8961232234541265e-06,
+      "loss": 0.276,
+      "step": 122000
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.893146811518714e-06,
+      "loss": 0.2662,
+      "step": 122010
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.890170399583303e-06,
+      "loss": 0.27,
+      "step": 122020
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.8871939876478906e-06,
+      "loss": 0.2701,
+      "step": 122030
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.884217575712479e-06,
+      "loss": 0.2705,
+      "step": 122040
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 3.881241163777067e-06,
+      "loss": 0.2685,
+      "step": 122050
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.878264751841655e-06,
+      "loss": 0.2693,
+      "step": 122060
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.875288339906243e-06,
+      "loss": 0.2684,
+      "step": 122070
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.872311927970832e-06,
+      "loss": 0.2714,
+      "step": 122080
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 3.8693355160354196e-06,
+      "loss": 0.2818,
+      "step": 122090
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.866359104100007e-06,
+      "loss": 0.2869,
+      "step": 122100
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.863382692164596e-06,
+      "loss": 0.2714,
+      "step": 122110
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.8604062802291845e-06,
+      "loss": 0.2663,
+      "step": 122120
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.857429868293772e-06,
+      "loss": 0.2726,
+      "step": 122130
+    },
+    {
+      "epoch": 27.12,
+      "learning_rate": 3.85445345635836e-06,
+      "loss": 0.266,
+      "step": 122140
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.851477044422949e-06,
+      "loss": 0.2646,
+      "step": 122150
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.848500632487537e-06,
+      "loss": 0.2736,
+      "step": 122160
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.845524220552125e-06,
+      "loss": 0.2677,
+      "step": 122170
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 3.842547808616713e-06,
+      "loss": 0.2809,
+      "step": 122180
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.839571396681301e-06,
+      "loss": 0.2716,
+      "step": 122190
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.836594984745889e-06,
+      "loss": 0.2779,
+      "step": 122200
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.833618572810478e-06,
+      "loss": 0.2639,
+      "step": 122210
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.830642160875065e-06,
+      "loss": 0.2674,
+      "step": 122220
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 3.827665748939653e-06,
+      "loss": 0.2798,
+      "step": 122230
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.824689337004242e-06,
+      "loss": 0.2684,
+      "step": 122240
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.82171292506883e-06,
+      "loss": 0.263,
+      "step": 122250
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.818736513133418e-06,
+      "loss": 0.2593,
+      "step": 122260
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 3.815760101198006e-06,
+      "loss": 0.276,
+      "step": 122270
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.812783689262594e-06,
+      "loss": 0.2721,
+      "step": 122280
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.809807277327182e-06,
+      "loss": 0.2688,
+      "step": 122290
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.8068308653917707e-06,
+      "loss": 0.2901,
+      "step": 122300
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.803854453456359e-06,
+      "loss": 0.2748,
+      "step": 122310
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 3.8008780415209466e-06,
+      "loss": 0.267,
+      "step": 122320
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.7979016295855348e-06,
+      "loss": 0.2669,
+      "step": 122330
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.7949252176501234e-06,
+      "loss": 0.2596,
+      "step": 122340
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.7919488057147115e-06,
+      "loss": 0.2863,
+      "step": 122350
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 3.7889723937792993e-06,
+      "loss": 0.2895,
+      "step": 122360
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.7859959818438875e-06,
+      "loss": 0.2786,
+      "step": 122370
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.783019569908475e-06,
+      "loss": 0.2656,
+      "step": 122380
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.780043157973064e-06,
+      "loss": 0.2646,
+      "step": 122390
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.777066746037652e-06,
+      "loss": 0.2673,
+      "step": 122400
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 3.77409033410224e-06,
+      "loss": 0.2742,
+      "step": 122410
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.771113922166828e-06,
+      "loss": 0.269,
+      "step": 122420
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.7681375102314165e-06,
+      "loss": 0.2708,
+      "step": 122430
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.7651610982960046e-06,
+      "loss": 0.2779,
+      "step": 122440
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 3.762184686360593e-06,
+      "loss": 0.2682,
+      "step": 122450
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.7592082744251806e-06,
+      "loss": 0.2661,
+      "step": 122460
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.7562318624897687e-06,
+      "loss": 0.2701,
+      "step": 122470
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.7532554505543573e-06,
+      "loss": 0.2766,
+      "step": 122480
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.750279038618945e-06,
+      "loss": 0.2643,
+      "step": 122490
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 3.7473026266835332e-06,
+      "loss": 0.2759,
+      "step": 122500
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.7443262147481214e-06,
+      "loss": 0.2678,
+      "step": 122510
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.741349802812709e-06,
+      "loss": 0.2679,
+      "step": 122520
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.7383733908772978e-06,
+      "loss": 0.2678,
+      "step": 122530
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 3.735396978941886e-06,
+      "loss": 0.2805,
+      "step": 122540
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.732420567006474e-06,
+      "loss": 0.264,
+      "step": 122550
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.729444155071062e-06,
+      "loss": 0.2717,
+      "step": 122560
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.7264677431356504e-06,
+      "loss": 0.2872,
+      "step": 122570
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.7234913312002386e-06,
+      "loss": 0.2929,
+      "step": 122580
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 3.7205149192648263e-06,
+      "loss": 0.2802,
+      "step": 122590
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.7175385073294145e-06,
+      "loss": 0.2617,
+      "step": 122600
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.7145620953940027e-06,
+      "loss": 0.2797,
+      "step": 122610
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.7115856834585913e-06,
+      "loss": 0.2699,
+      "step": 122620
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 3.708609271523179e-06,
+      "loss": 0.2743,
+      "step": 122630
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.705632859587767e-06,
+      "loss": 0.2616,
+      "step": 122640
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.7026564476523554e-06,
+      "loss": 0.2698,
+      "step": 122650
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.699680035716944e-06,
+      "loss": 0.2672,
+      "step": 122660
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.6967036237815317e-06,
+      "loss": 0.2829,
+      "step": 122670
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 3.69372721184612e-06,
+      "loss": 0.2788,
+      "step": 122680
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.690750799910708e-06,
+      "loss": 0.2711,
+      "step": 122690
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.6877743879752958e-06,
+      "loss": 0.2686,
+      "step": 122700
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.6847979760398844e-06,
+      "loss": 0.2511,
+      "step": 122710
+    },
+    {
+      "epoch": 27.25,
+      "learning_rate": 3.6818215641044725e-06,
+      "loss": 0.2752,
+      "step": 122720
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.6788451521690603e-06,
+      "loss": 0.2789,
+      "step": 122730
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.6758687402336485e-06,
+      "loss": 0.2685,
+      "step": 122740
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.672892328298237e-06,
+      "loss": 0.2649,
+      "step": 122750
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.6699159163628252e-06,
+      "loss": 0.2769,
+      "step": 122760
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 3.666939504427413e-06,
+      "loss": 0.2713,
+      "step": 122770
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.663963092492001e-06,
+      "loss": 0.2774,
+      "step": 122780
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6609866805565893e-06,
+      "loss": 0.2685,
+      "step": 122790
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.658010268621178e-06,
+      "loss": 0.2782,
+      "step": 122800
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 3.6550338566857656e-06,
+      "loss": 0.2722,
+      "step": 122810
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.652057444750354e-06,
+      "loss": 0.2747,
+      "step": 122820
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6490810328149416e-06,
+      "loss": 0.2648,
+      "step": 122830
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6461046208795297e-06,
+      "loss": 0.2734,
+      "step": 122840
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6431282089441183e-06,
+      "loss": 0.2616,
+      "step": 122850
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 3.6401517970087065e-06,
+      "loss": 0.2745,
+      "step": 122860
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.6371753850732942e-06,
+      "loss": 0.2668,
+      "step": 122870
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.6341989731378824e-06,
+      "loss": 0.2739,
+      "step": 122880
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.631222561202471e-06,
+      "loss": 0.2712,
+      "step": 122890
+    },
+    {
+      "epoch": 27.29,
+      "learning_rate": 3.628246149267059e-06,
+      "loss": 0.261,
+      "step": 122900
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.625269737331647e-06,
+      "loss": 0.2868,
+      "step": 122910
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.622293325396235e-06,
+      "loss": 0.2784,
+      "step": 122920
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.619316913460823e-06,
+      "loss": 0.2704,
+      "step": 122930
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.6163405015254114e-06,
+      "loss": 0.2753,
+      "step": 122940
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 3.6133640895899996e-06,
+      "loss": 0.2688,
+      "step": 122950
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.6103876776545878e-06,
+      "loss": 0.2713,
+      "step": 122960
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.6074112657191755e-06,
+      "loss": 0.2765,
+      "step": 122970
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.604434853783764e-06,
+      "loss": 0.2778,
+      "step": 122980
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 3.6014584418483523e-06,
+      "loss": 0.2803,
+      "step": 122990
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.5984820299129404e-06,
+      "loss": 0.2605,
+      "step": 123000
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.595505617977528e-06,
+      "loss": 0.2669,
+      "step": 123010
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.5925292060421164e-06,
+      "loss": 0.2655,
+      "step": 123020
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.589552794106705e-06,
+      "loss": 0.2792,
+      "step": 123030
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 3.5865763821712927e-06,
+      "loss": 0.2795,
+      "step": 123040
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.583599970235881e-06,
+      "loss": 0.2553,
+      "step": 123050
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.580623558300469e-06,
+      "loss": 0.2645,
+      "step": 123060
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.5776471463650568e-06,
+      "loss": 0.2755,
+      "step": 123070
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 3.5746707344296454e-06,
+      "loss": 0.2653,
+      "step": 123080
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.5716943224942335e-06,
+      "loss": 0.2783,
+      "step": 123090
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.5687179105588217e-06,
+      "loss": 0.2631,
+      "step": 123100
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.5657414986234095e-06,
+      "loss": 0.2877,
+      "step": 123110
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.562765086687998e-06,
+      "loss": 0.2672,
+      "step": 123120
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 3.5597886747525862e-06,
+      "loss": 0.2758,
+      "step": 123130
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.556812262817174e-06,
+      "loss": 0.2644,
+      "step": 123140
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.553835850881762e-06,
+      "loss": 0.2765,
+      "step": 123150
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.5508594389463503e-06,
+      "loss": 0.2631,
+      "step": 123160
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 3.547883027010939e-06,
+      "loss": 0.2674,
+      "step": 123170
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.5449066150755266e-06,
+      "loss": 0.2648,
+      "step": 123180
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.541930203140115e-06,
+      "loss": 0.2763,
+      "step": 123190
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.538953791204703e-06,
+      "loss": 0.2738,
+      "step": 123200
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.5359773792692916e-06,
+      "loss": 0.2657,
+      "step": 123210
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 3.5330009673338793e-06,
+      "loss": 0.2793,
+      "step": 123220
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.5300245553984675e-06,
+      "loss": 0.2703,
+      "step": 123230
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.5270481434630557e-06,
+      "loss": 0.2793,
+      "step": 123240
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.5240717315276434e-06,
+      "loss": 0.274,
+      "step": 123250
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 3.521095319592232e-06,
+      "loss": 0.271,
+      "step": 123260
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.51811890765682e-06,
+      "loss": 0.2756,
+      "step": 123270
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.515142495721408e-06,
+      "loss": 0.2719,
+      "step": 123280
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.512166083785996e-06,
+      "loss": 0.2599,
+      "step": 123290
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.5091896718505847e-06,
+      "loss": 0.273,
+      "step": 123300
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 3.506213259915173e-06,
+      "loss": 0.2744,
+      "step": 123310
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.5032368479797606e-06,
+      "loss": 0.2848,
+      "step": 123320
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.5002604360443488e-06,
+      "loss": 0.2612,
+      "step": 123330
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.497284024108937e-06,
+      "loss": 0.2759,
+      "step": 123340
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 3.4943076121735255e-06,
+      "loss": 0.2916,
+      "step": 123350
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.4913312002381133e-06,
+      "loss": 0.2766,
+      "step": 123360
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.4883547883027014e-06,
+      "loss": 0.2834,
+      "step": 123370
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.485378376367289e-06,
+      "loss": 0.274,
+      "step": 123380
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.4824019644318774e-06,
+      "loss": 0.2602,
+      "step": 123390
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 3.479425552496466e-06,
+      "loss": 0.2564,
+      "step": 123400
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.476449140561054e-06,
+      "loss": 0.2794,
+      "step": 123410
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.473472728625642e-06,
+      "loss": 0.2641,
+      "step": 123420
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.47049631669023e-06,
+      "loss": 0.2625,
+      "step": 123430
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 3.4675199047548186e-06,
+      "loss": 0.2879,
+      "step": 123440
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.464543492819407e-06,
+      "loss": 0.269,
+      "step": 123450
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.4615670808839945e-06,
+      "loss": 0.2691,
+      "step": 123460
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.4585906689485827e-06,
+      "loss": 0.2709,
+      "step": 123470
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.4556142570131705e-06,
+      "loss": 0.2773,
+      "step": 123480
+    },
+    {
+      "epoch": 27.42,
+      "learning_rate": 3.452637845077759e-06,
+      "loss": 0.281,
+      "step": 123490
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.4496614331423472e-06,
+      "loss": 0.2799,
+      "step": 123500
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.4466850212069354e-06,
+      "loss": 0.2851,
+      "step": 123510
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.443708609271523e-06,
+      "loss": 0.2615,
+      "step": 123520
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 3.4407321973361117e-06,
+      "loss": 0.2739,
+      "step": 123530
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.4377557854007e-06,
+      "loss": 0.2763,
+      "step": 123540
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.434779373465288e-06,
+      "loss": 0.2663,
+      "step": 123550
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.431802961529876e-06,
+      "loss": 0.2652,
+      "step": 123560
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.428826549594464e-06,
+      "loss": 0.283,
+      "step": 123570
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 3.4258501376590526e-06,
+      "loss": 0.274,
+      "step": 123580
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.4228737257236403e-06,
+      "loss": 0.2844,
+      "step": 123590
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.4198973137882285e-06,
+      "loss": 0.2664,
+      "step": 123600
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.4169209018528167e-06,
+      "loss": 0.2686,
+      "step": 123610
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 3.4139444899174053e-06,
+      "loss": 0.2836,
+      "step": 123620
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.410968077981993e-06,
+      "loss": 0.2756,
+      "step": 123630
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.407991666046581e-06,
+      "loss": 0.2671,
+      "step": 123640
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.4050152541111693e-06,
+      "loss": 0.2842,
+      "step": 123650
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.402038842175757e-06,
+      "loss": 0.2771,
+      "step": 123660
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 3.3990624302403457e-06,
+      "loss": 0.2649,
+      "step": 123670
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.396086018304934e-06,
+      "loss": 0.2725,
+      "step": 123680
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.3931096063695216e-06,
+      "loss": 0.2722,
+      "step": 123690
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.3901331944341098e-06,
+      "loss": 0.2768,
+      "step": 123700
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 3.387156782498698e-06,
+      "loss": 0.2743,
+      "step": 123710
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.3841803705632865e-06,
+      "loss": 0.2623,
+      "step": 123720
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.3812039586278743e-06,
+      "loss": 0.2666,
+      "step": 123730
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.3782275466924624e-06,
+      "loss": 0.281,
+      "step": 123740
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.3752511347570506e-06,
+      "loss": 0.2776,
+      "step": 123750
+    },
+    {
+      "epoch": 27.48,
+      "learning_rate": 3.372274722821639e-06,
+      "loss": 0.2795,
+      "step": 123760
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.369298310886227e-06,
+      "loss": 0.2731,
+      "step": 123770
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.366321898950815e-06,
+      "loss": 0.2565,
+      "step": 123780
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.3633454870154033e-06,
+      "loss": 0.2691,
+      "step": 123790
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 3.360369075079991e-06,
+      "loss": 0.2726,
+      "step": 123800
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.3573926631445796e-06,
+      "loss": 0.2758,
+      "step": 123810
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.354416251209168e-06,
+      "loss": 0.2763,
+      "step": 123820
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.3514398392737555e-06,
+      "loss": 0.2683,
+      "step": 123830
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.3484634273383437e-06,
+      "loss": 0.2561,
+      "step": 123840
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 3.3454870154029323e-06,
+      "loss": 0.263,
+      "step": 123850
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.3425106034675205e-06,
+      "loss": 0.2755,
+      "step": 123860
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.3395341915321082e-06,
+      "loss": 0.2632,
+      "step": 123870
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.3365577795966964e-06,
+      "loss": 0.262,
+      "step": 123880
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.3335813676612846e-06,
+      "loss": 0.2696,
+      "step": 123890
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 3.330604955725873e-06,
+      "loss": 0.2655,
+      "step": 123900
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.327628543790461e-06,
+      "loss": 0.2704,
+      "step": 123910
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.324652131855049e-06,
+      "loss": 0.2569,
+      "step": 123920
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.321675719919637e-06,
+      "loss": 0.2684,
+      "step": 123930
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 3.318699307984225e-06,
+      "loss": 0.2744,
+      "step": 123940
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.3157228960488136e-06,
+      "loss": 0.2699,
+      "step": 123950
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.3127464841134017e-06,
+      "loss": 0.2754,
+      "step": 123960
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.3097700721779895e-06,
+      "loss": 0.2746,
+      "step": 123970
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.3067936602425777e-06,
+      "loss": 0.2681,
+      "step": 123980
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 3.3038172483071663e-06,
+      "loss": 0.2789,
+      "step": 123990
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.3008408363717544e-06,
+      "loss": 0.2716,
+      "step": 124000
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.297864424436342e-06,
+      "loss": 0.2758,
+      "step": 124010
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.2948880125009303e-06,
+      "loss": 0.2789,
+      "step": 124020
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 3.291911600565518e-06,
+      "loss": 0.2628,
+      "step": 124030
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.2889351886301067e-06,
+      "loss": 0.2594,
+      "step": 124040
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.285958776694695e-06,
+      "loss": 0.2747,
+      "step": 124050
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.282982364759283e-06,
+      "loss": 0.2621,
+      "step": 124060
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.2800059528238708e-06,
+      "loss": 0.2758,
+      "step": 124070
+    },
+    {
+      "epoch": 27.55,
+      "learning_rate": 3.2770295408884594e-06,
+      "loss": 0.2602,
+      "step": 124080
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.2740531289530475e-06,
+      "loss": 0.267,
+      "step": 124090
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.2710767170176357e-06,
+      "loss": 0.2643,
+      "step": 124100
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.2681003050822234e-06,
+      "loss": 0.2681,
+      "step": 124110
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 3.2651238931468116e-06,
+      "loss": 0.2596,
+      "step": 124120
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.2621474812114e-06,
+      "loss": 0.2909,
+      "step": 124130
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.259171069275988e-06,
+      "loss": 0.271,
+      "step": 124140
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.256194657340576e-06,
+      "loss": 0.2656,
+      "step": 124150
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.2532182454051643e-06,
+      "loss": 0.2701,
+      "step": 124160
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 3.250241833469753e-06,
+      "loss": 0.2621,
+      "step": 124170
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.2472654215343406e-06,
+      "loss": 0.2574,
+      "step": 124180
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.244289009598929e-06,
+      "loss": 0.276,
+      "step": 124190
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.241312597663517e-06,
+      "loss": 0.2702,
+      "step": 124200
+    },
+    {
+      "epoch": 27.58,
+      "learning_rate": 3.2383361857281047e-06,
+      "loss": 0.2725,
+      "step": 124210
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.2353597737926933e-06,
+      "loss": 0.2742,
+      "step": 124220
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.2323833618572815e-06,
+      "loss": 0.2718,
+      "step": 124230
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.2294069499218692e-06,
+      "loss": 0.264,
+      "step": 124240
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.2264305379864574e-06,
+      "loss": 0.2763,
+      "step": 124250
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 3.2234541260510456e-06,
+      "loss": 0.2612,
+      "step": 124260
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.220477714115634e-06,
+      "loss": 0.277,
+      "step": 124270
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.217501302180222e-06,
+      "loss": 0.2664,
+      "step": 124280
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.21452489024481e-06,
+      "loss": 0.2738,
+      "step": 124290
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 3.2115484783093982e-06,
+      "loss": 0.2793,
+      "step": 124300
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.208572066373987e-06,
+      "loss": 0.2807,
+      "step": 124310
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.2055956544385746e-06,
+      "loss": 0.2617,
+      "step": 124320
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.2026192425031627e-06,
+      "loss": 0.2749,
+      "step": 124330
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.199642830567751e-06,
+      "loss": 0.2703,
+      "step": 124340
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 3.1966664186323387e-06,
+      "loss": 0.2572,
+      "step": 124350
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.1936900066969272e-06,
+      "loss": 0.2793,
+      "step": 124360
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.1907135947615154e-06,
+      "loss": 0.2781,
+      "step": 124370
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.187737182826103e-06,
+      "loss": 0.2804,
+      "step": 124380
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 3.1847607708906913e-06,
+      "loss": 0.2651,
+      "step": 124390
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.18178435895528e-06,
+      "loss": 0.2791,
+      "step": 124400
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.178807947019868e-06,
+      "loss": 0.2831,
+      "step": 124410
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.175831535084456e-06,
+      "loss": 0.2617,
+      "step": 124420
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.172855123149044e-06,
+      "loss": 0.2734,
+      "step": 124430
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 3.169878711213632e-06,
+      "loss": 0.2647,
+      "step": 124440
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.1669022992782208e-06,
+      "loss": 0.2739,
+      "step": 124450
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.1639258873428085e-06,
+      "loss": 0.268,
+      "step": 124460
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.1609494754073967e-06,
+      "loss": 0.2725,
+      "step": 124470
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 3.1579730634719844e-06,
+      "loss": 0.2645,
+      "step": 124480
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.154996651536573e-06,
+      "loss": 0.2673,
+      "step": 124490
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.152020239601161e-06,
+      "loss": 0.2782,
+      "step": 124500
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.1490438276657494e-06,
+      "loss": 0.2682,
+      "step": 124510
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.146067415730337e-06,
+      "loss": 0.2746,
+      "step": 124520
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 3.1430910037949253e-06,
+      "loss": 0.2683,
+      "step": 124530
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.140114591859514e-06,
+      "loss": 0.271,
+      "step": 124540
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.137138179924102e-06,
+      "loss": 0.2632,
+      "step": 124550
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.13416176798869e-06,
+      "loss": 0.276,
+      "step": 124560
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 3.131185356053278e-06,
+      "loss": 0.275,
+      "step": 124570
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.1282089441178657e-06,
+      "loss": 0.2683,
+      "step": 124580
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.1252325321824543e-06,
+      "loss": 0.2724,
+      "step": 124590
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.1222561202470425e-06,
+      "loss": 0.2711,
+      "step": 124600
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.1192797083116306e-06,
+      "loss": 0.2663,
+      "step": 124610
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 3.1163032963762184e-06,
+      "loss": 0.2624,
+      "step": 124620
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.113326884440807e-06,
+      "loss": 0.2741,
+      "step": 124630
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.110350472505395e-06,
+      "loss": 0.2713,
+      "step": 124640
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.1073740605699833e-06,
+      "loss": 0.2817,
+      "step": 124650
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 3.104397648634571e-06,
+      "loss": 0.2633,
+      "step": 124660
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.1014212366991592e-06,
+      "loss": 0.2701,
+      "step": 124670
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.098444824763748e-06,
+      "loss": 0.2804,
+      "step": 124680
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.0954684128283356e-06,
+      "loss": 0.2644,
+      "step": 124690
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.0924920008929237e-06,
+      "loss": 0.2711,
+      "step": 124700
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 3.089515588957512e-06,
+      "loss": 0.2714,
+      "step": 124710
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.0865391770221005e-06,
+      "loss": 0.2784,
+      "step": 124720
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.0835627650866882e-06,
+      "loss": 0.2789,
+      "step": 124730
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.0805863531512764e-06,
+      "loss": 0.2825,
+      "step": 124740
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 3.0776099412158646e-06,
+      "loss": 0.2806,
+      "step": 124750
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.0746335292804523e-06,
+      "loss": 0.2717,
+      "step": 124760
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.071657117345041e-06,
+      "loss": 0.2837,
+      "step": 124770
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.068680705409629e-06,
+      "loss": 0.2749,
+      "step": 124780
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.0657042934742173e-06,
+      "loss": 0.2781,
+      "step": 124790
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 3.062727881538805e-06,
+      "loss": 0.262,
+      "step": 124800
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.059751469603393e-06,
+      "loss": 0.2701,
+      "step": 124810
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.0567750576679818e-06,
+      "loss": 0.2765,
+      "step": 124820
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.0537986457325695e-06,
+      "loss": 0.2784,
+      "step": 124830
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 3.0508222337971577e-06,
+      "loss": 0.2706,
+      "step": 124840
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.047845821861746e-06,
+      "loss": 0.2632,
+      "step": 124850
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.0448694099263344e-06,
+      "loss": 0.2605,
+      "step": 124860
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.041892997990922e-06,
+      "loss": 0.2684,
+      "step": 124870
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.0389165860555104e-06,
+      "loss": 0.2731,
+      "step": 124880
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 3.0359401741200985e-06,
+      "loss": 0.2646,
+      "step": 124890
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.0329637621846863e-06,
+      "loss": 0.2648,
+      "step": 124900
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.029987350249275e-06,
+      "loss": 0.2626,
+      "step": 124910
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.027010938313863e-06,
+      "loss": 0.2754,
+      "step": 124920
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 3.0240345263784508e-06,
+      "loss": 0.2797,
+      "step": 124930
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.021058114443039e-06,
+      "loss": 0.2644,
+      "step": 124940
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.0180817025076276e-06,
+      "loss": 0.2656,
+      "step": 124950
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.0151052905722157e-06,
+      "loss": 0.269,
+      "step": 124960
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.0121288786368035e-06,
+      "loss": 0.2725,
+      "step": 124970
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 3.0091524667013916e-06,
+      "loss": 0.2745,
+      "step": 124980
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.00617605476598e-06,
+      "loss": 0.276,
+      "step": 124990
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.0031996428305684e-06,
+      "loss": 0.274,
+      "step": 125000
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 3.000223230895156e-06,
+      "loss": 0.2578,
+      "step": 125010
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 2.9972468189597443e-06,
+      "loss": 0.2708,
+      "step": 125020
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.994270407024332e-06,
+      "loss": 0.2603,
+      "step": 125030
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.9912939950889207e-06,
+      "loss": 0.2891,
+      "step": 125040
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.988317583153509e-06,
+      "loss": 0.2548,
+      "step": 125050
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.985341171218097e-06,
+      "loss": 0.2738,
+      "step": 125060
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.9823647592826847e-06,
+      "loss": 0.2743,
+      "step": 125070
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 2.979388347347273e-06,
+      "loss": 0.2733,
+      "step": 125080
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 2.9764119354118615e-06,
+      "loss": 0.2777,
+      "step": 125090
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 2.9734355234764497e-06,
+      "loss": 0.2689,
+      "step": 125100
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 2.9704591115410374e-06,
+      "loss": 0.2704,
+      "step": 125110
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.9674826996056256e-06,
+      "loss": 0.2795,
+      "step": 125120
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.9645062876702133e-06,
+      "loss": 0.2731,
+      "step": 125130
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.961529875734802e-06,
+      "loss": 0.2701,
+      "step": 125140
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.95855346379939e-06,
+      "loss": 0.2873,
+      "step": 125150
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.9555770518639783e-06,
+      "loss": 0.2894,
+      "step": 125160
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 2.952600639928566e-06,
+      "loss": 0.2698,
+      "step": 125170
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 2.9496242279931546e-06,
+      "loss": 0.2846,
+      "step": 125180
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 2.9466478160577428e-06,
+      "loss": 0.2613,
+      "step": 125190
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 2.943671404122331e-06,
+      "loss": 0.2691,
+      "step": 125200
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 2.9406949921869187e-06,
+      "loss": 0.2678,
+      "step": 125210
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 2.937718580251507e-06,
+      "loss": 0.2692,
+      "step": 125220
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 2.9347421683160954e-06,
+      "loss": 0.2614,
+      "step": 125230
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 2.931765756380683e-06,
+      "loss": 0.2712,
+      "step": 125240
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 2.9287893444452714e-06,
+      "loss": 0.2743,
+      "step": 125250
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 2.9258129325098595e-06,
+      "loss": 0.2652,
+      "step": 125260
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 2.922836520574448e-06,
+      "loss": 0.2814,
+      "step": 125270
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 2.919860108639036e-06,
+      "loss": 0.2614,
+      "step": 125280
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 2.916883696703624e-06,
+      "loss": 0.2701,
+      "step": 125290
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.913907284768212e-06,
+      "loss": 0.2677,
+      "step": 125300
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.9109308728328e-06,
+      "loss": 0.2735,
+      "step": 125310
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.9079544608973885e-06,
+      "loss": 0.255,
+      "step": 125320
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.9049780489619767e-06,
+      "loss": 0.2781,
+      "step": 125330
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.902001637026565e-06,
+      "loss": 0.2559,
+      "step": 125340
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 2.8990252250911526e-06,
+      "loss": 0.2703,
+      "step": 125350
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 2.8960488131557412e-06,
+      "loss": 0.2761,
+      "step": 125360
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 2.8930724012203294e-06,
+      "loss": 0.2653,
+      "step": 125370
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 2.890095989284917e-06,
+      "loss": 0.2759,
+      "step": 125380
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 2.8871195773495053e-06,
+      "loss": 0.2812,
+      "step": 125390
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 2.8841431654140935e-06,
+      "loss": 0.2581,
+      "step": 125400
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 2.881166753478682e-06,
+      "loss": 0.2725,
+      "step": 125410
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 2.87819034154327e-06,
+      "loss": 0.2642,
+      "step": 125420
+    },
+    {
+      "epoch": 27.85,
+      "learning_rate": 2.875213929607858e-06,
+      "loss": 0.2701,
+      "step": 125430
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 2.872237517672446e-06,
+      "loss": 0.2627,
+      "step": 125440
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 2.869261105737034e-06,
+      "loss": 0.2573,
+      "step": 125450
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 2.8662846938016225e-06,
+      "loss": 0.2725,
+      "step": 125460
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 2.8633082818662107e-06,
+      "loss": 0.2763,
+      "step": 125470
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.8603318699307984e-06,
+      "loss": 0.2774,
+      "step": 125480
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.8573554579953866e-06,
+      "loss": 0.2663,
+      "step": 125490
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.854379046059975e-06,
+      "loss": 0.2727,
+      "step": 125500
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.8514026341245633e-06,
+      "loss": 0.278,
+      "step": 125510
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.848426222189151e-06,
+      "loss": 0.2781,
+      "step": 125520
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 2.8454498102537393e-06,
+      "loss": 0.2716,
+      "step": 125530
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 2.8424733983183274e-06,
+      "loss": 0.2668,
+      "step": 125540
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 2.839496986382916e-06,
+      "loss": 0.273,
+      "step": 125550
+    },
+    {
+      "epoch": 27.88,
+      "learning_rate": 2.8365205744475038e-06,
+      "loss": 0.2767,
+      "step": 125560
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.833544162512092e-06,
+      "loss": 0.2661,
+      "step": 125570
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.8305677505766797e-06,
+      "loss": 0.2714,
+      "step": 125580
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.8275913386412683e-06,
+      "loss": 0.2757,
+      "step": 125590
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.8246149267058564e-06,
+      "loss": 0.2793,
+      "step": 125600
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.8216385147704446e-06,
+      "loss": 0.2735,
+      "step": 125610
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 2.8186621028350324e-06,
+      "loss": 0.263,
+      "step": 125620
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 2.8156856908996205e-06,
+      "loss": 0.257,
+      "step": 125630
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 2.812709278964209e-06,
+      "loss": 0.2718,
+      "step": 125640
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 2.8097328670287973e-06,
+      "loss": 0.2619,
+      "step": 125650
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 2.806756455093385e-06,
+      "loss": 0.2816,
+      "step": 125660
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 2.803780043157973e-06,
+      "loss": 0.259,
+      "step": 125670
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 2.800803631222561e-06,
+      "loss": 0.2674,
+      "step": 125680
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 2.7978272192871495e-06,
+      "loss": 0.2596,
+      "step": 125690
+    },
+    {
+      "epoch": 27.91,
+      "learning_rate": 2.7948508073517377e-06,
+      "loss": 0.2648,
+      "step": 125700
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 2.791874395416326e-06,
+      "loss": 0.2691,
+      "step": 125710
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 2.7888979834809136e-06,
+      "loss": 0.2702,
+      "step": 125720
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 2.7859215715455022e-06,
+      "loss": 0.2746,
+      "step": 125730
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 2.7829451596100904e-06,
+      "loss": 0.2684,
+      "step": 125740
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.7799687476746786e-06,
+      "loss": 0.267,
+      "step": 125750
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.7769923357392663e-06,
+      "loss": 0.2636,
+      "step": 125760
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.7740159238038545e-06,
+      "loss": 0.2811,
+      "step": 125770
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.771039511868443e-06,
+      "loss": 0.2744,
+      "step": 125780
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.768063099933031e-06,
+      "loss": 0.2673,
+      "step": 125790
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 2.765086687997619e-06,
+      "loss": 0.2587,
+      "step": 125800
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 2.762110276062207e-06,
+      "loss": 0.2841,
+      "step": 125810
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 2.7591338641267957e-06,
+      "loss": 0.2664,
+      "step": 125820
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 2.7561574521913835e-06,
+      "loss": 0.2684,
+      "step": 125830
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.7531810402559717e-06,
+      "loss": 0.2619,
+      "step": 125840
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.75020462832056e-06,
+      "loss": 0.2697,
+      "step": 125850
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.7472282163851476e-06,
+      "loss": 0.2613,
+      "step": 125860
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.744251804449736e-06,
+      "loss": 0.2586,
+      "step": 125870
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.7412753925143243e-06,
+      "loss": 0.2675,
+      "step": 125880
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 2.7382989805789125e-06,
+      "loss": 0.2667,
+      "step": 125890
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 2.7353225686435003e-06,
+      "loss": 0.2753,
+      "step": 125900
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 2.732346156708089e-06,
+      "loss": 0.2702,
+      "step": 125910
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 2.729369744772677e-06,
+      "loss": 0.2665,
+      "step": 125920
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.7263933328372648e-06,
+      "loss": 0.2823,
+      "step": 125930
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.723416920901853e-06,
+      "loss": 0.2712,
+      "step": 125940
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.720440508966441e-06,
+      "loss": 0.2712,
+      "step": 125950
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.7174640970310297e-06,
+      "loss": 0.2725,
+      "step": 125960
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.7144876850956174e-06,
+      "loss": 0.2635,
+      "step": 125970
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 2.7115112731602056e-06,
+      "loss": 0.2859,
+      "step": 125980
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 2.7085348612247938e-06,
+      "loss": 0.2849,
+      "step": 125990
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 2.7055584492893815e-06,
+      "loss": 0.2667,
+      "step": 126000
+    },
+    {
+      "epoch": 27.98,
+      "learning_rate": 2.70258203735397e-06,
+      "loss": 0.2693,
+      "step": 126010
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.6996056254185583e-06,
+      "loss": 0.2701,
+      "step": 126020
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.696629213483146e-06,
+      "loss": 0.2708,
+      "step": 126030
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.693652801547734e-06,
+      "loss": 0.2716,
+      "step": 126040
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.690676389612323e-06,
+      "loss": 0.2688,
+      "step": 126050
+    },
+    {
+      "epoch": 27.99,
+      "eval_cer": 4.536324786324786,
+      "eval_loss": 0.5573694109916687,
+      "eval_runtime": 1.2718,
+      "eval_samples_per_second": 7.863,
+      "eval_steps_per_second": 0.786,
+      "eval_wer": 0.5974025974025974,
+      "step": 126056
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.687699977676911e-06,
+      "loss": 0.2786,
+      "step": 126060
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 2.6847235657414987e-06,
+      "loss": 0.2638,
+      "step": 126070
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 2.681747153806087e-06,
+      "loss": 0.2713,
+      "step": 126080
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 2.678770741870675e-06,
+      "loss": 0.281,
+      "step": 126090
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 2.6757943299352636e-06,
+      "loss": 0.2614,
+      "step": 126100
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 2.6728179179998514e-06,
+      "loss": 0.2611,
+      "step": 126110
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 2.6698415060644396e-06,
+      "loss": 0.269,
+      "step": 126120
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 2.6668650941290273e-06,
+      "loss": 0.2565,
+      "step": 126130
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 2.663888682193616e-06,
+      "loss": 0.276,
+      "step": 126140
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 2.660912270258204e-06,
+      "loss": 0.2684,
+      "step": 126150
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 2.6579358583227922e-06,
+      "loss": 0.2729,
+      "step": 126160
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 2.65495944638738e-06,
+      "loss": 0.2531,
+      "step": 126170
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 2.651983034451968e-06,
+      "loss": 0.2804,
+      "step": 126180
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 2.6490066225165567e-06,
+      "loss": 0.2722,
+      "step": 126190
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.646030210581145e-06,
+      "loss": 0.2766,
+      "step": 126200
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.6430537986457327e-06,
+      "loss": 0.2742,
+      "step": 126210
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.640077386710321e-06,
+      "loss": 0.2648,
+      "step": 126220
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.6371009747749094e-06,
+      "loss": 0.2614,
+      "step": 126230
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.634124562839497e-06,
+      "loss": 0.2805,
+      "step": 126240
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 2.6311481509040853e-06,
+      "loss": 0.2644,
+      "step": 126250
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 2.6281717389686735e-06,
+      "loss": 0.277,
+      "step": 126260
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 2.6251953270332613e-06,
+      "loss": 0.2672,
+      "step": 126270
+    },
+    {
+      "epoch": 28.04,
+      "learning_rate": 2.62221891509785e-06,
+      "loss": 0.2669,
+      "step": 126280
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.619242503162438e-06,
+      "loss": 0.2752,
+      "step": 126290
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.616266091227026e-06,
+      "loss": 0.2616,
+      "step": 126300
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.613289679291614e-06,
+      "loss": 0.2692,
+      "step": 126310
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.610313267356202e-06,
+      "loss": 0.2636,
+      "step": 126320
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.6073368554207907e-06,
+      "loss": 0.2546,
+      "step": 126330
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 2.6043604434853784e-06,
+      "loss": 0.2685,
+      "step": 126340
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 2.6013840315499666e-06,
+      "loss": 0.2784,
+      "step": 126350
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 2.5984076196145548e-06,
+      "loss": 0.2645,
+      "step": 126360
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 2.5954312076791434e-06,
+      "loss": 0.2546,
+      "step": 126370
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 2.592454795743731e-06,
+      "loss": 0.2704,
+      "step": 126380
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 2.5894783838083193e-06,
+      "loss": 0.2689,
+      "step": 126390
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 2.5865019718729075e-06,
+      "loss": 0.2579,
+      "step": 126400
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 2.583525559937495e-06,
+      "loss": 0.2662,
+      "step": 126410
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 2.580549148002084e-06,
+      "loss": 0.2663,
+      "step": 126420
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 2.577572736066672e-06,
+      "loss": 0.266,
+      "step": 126430
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 2.57459632413126e-06,
+      "loss": 0.2691,
+      "step": 126440
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 2.571619912195848e-06,
+      "loss": 0.2803,
+      "step": 126450
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 2.5686435002604365e-06,
+      "loss": 0.2827,
+      "step": 126460
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2.5656670883250246e-06,
+      "loss": 0.2692,
+      "step": 126470
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2.5626906763896124e-06,
+      "loss": 0.2699,
+      "step": 126480
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2.5597142644542006e-06,
+      "loss": 0.277,
+      "step": 126490
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2.5567378525187887e-06,
+      "loss": 0.2782,
+      "step": 126500
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 2.5537614405833773e-06,
+      "loss": 0.2771,
+      "step": 126510
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 2.550785028647965e-06,
+      "loss": 0.271,
+      "step": 126520
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 2.5478086167125532e-06,
+      "loss": 0.2703,
+      "step": 126530
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 2.5448322047771414e-06,
+      "loss": 0.2721,
+      "step": 126540
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 2.541855792841729e-06,
+      "loss": 0.2618,
+      "step": 126550
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 2.5388793809063177e-06,
+      "loss": 0.2728,
+      "step": 126560
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 2.535902968970906e-06,
+      "loss": 0.2756,
+      "step": 126570
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 2.5329265570354937e-06,
+      "loss": 0.2763,
+      "step": 126580
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 2.529950145100082e-06,
+      "loss": 0.2746,
+      "step": 126590
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 2.5269737331646704e-06,
+      "loss": 0.2678,
+      "step": 126600
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 2.5239973212292586e-06,
+      "loss": 0.2503,
+      "step": 126610
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 2.5210209092938463e-06,
+      "loss": 0.2645,
+      "step": 126620
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 2.5180444973584345e-06,
+      "loss": 0.2514,
+      "step": 126630
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 2.5150680854230227e-06,
+      "loss": 0.2759,
+      "step": 126640
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 2.5120916734876113e-06,
+      "loss": 0.2638,
+      "step": 126650
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 2.509115261552199e-06,
+      "loss": 0.2671,
+      "step": 126660
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 2.506138849616787e-06,
+      "loss": 0.261,
+      "step": 126670
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 2.503162437681375e-06,
+      "loss": 0.271,
+      "step": 126680
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 2.5001860257459635e-06,
+      "loss": 0.255,
+      "step": 126690
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 2.4972096138105517e-06,
+      "loss": 0.2665,
+      "step": 126700
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 2.49423320187514e-06,
+      "loss": 0.271,
+      "step": 126710
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 2.4912567899397276e-06,
+      "loss": 0.2819,
+      "step": 126720
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 2.488280378004316e-06,
+      "loss": 0.2601,
+      "step": 126730
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 2.485303966068904e-06,
+      "loss": 0.2678,
+      "step": 126740
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 2.4823275541334925e-06,
+      "loss": 0.2589,
+      "step": 126750
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 2.4793511421980803e-06,
+      "loss": 0.2862,
+      "step": 126760
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 2.476374730262669e-06,
+      "loss": 0.2649,
+      "step": 126770
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 2.4733983183272566e-06,
+      "loss": 0.2605,
+      "step": 126780
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 2.470421906391845e-06,
+      "loss": 0.2664,
+      "step": 126790
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 2.467445494456433e-06,
+      "loss": 0.2671,
+      "step": 126800
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 2.464469082521021e-06,
+      "loss": 0.2715,
+      "step": 126810
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 2.4614926705856093e-06,
+      "loss": 0.2765,
+      "step": 126820
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 2.4585162586501975e-06,
+      "loss": 0.2609,
+      "step": 126830
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 2.4555398467147856e-06,
+      "loss": 0.2627,
+      "step": 126840
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 2.452563434779374e-06,
+      "loss": 0.2714,
+      "step": 126850
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 2.449587022843962e-06,
+      "loss": 0.2773,
+      "step": 126860
+    },
+    {
+      "epoch": 28.17,
+      "learning_rate": 2.44661061090855e-06,
+      "loss": 0.2792,
+      "step": 126870
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 2.443634198973138e-06,
+      "loss": 0.2496,
+      "step": 126880
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 2.440657787037726e-06,
+      "loss": 0.2743,
+      "step": 126890
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 2.4376813751023142e-06,
+      "loss": 0.2826,
+      "step": 126900
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 2.4347049631669024e-06,
+      "loss": 0.2655,
+      "step": 126910
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 2.4317285512314906e-06,
+      "loss": 0.2705,
+      "step": 126920
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 2.4287521392960787e-06,
+      "loss": 0.2643,
+      "step": 126930
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 2.425775727360667e-06,
+      "loss": 0.2849,
+      "step": 126940
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 2.422799315425255e-06,
+      "loss": 0.2607,
+      "step": 126950
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 2.4198229034898433e-06,
+      "loss": 0.2668,
+      "step": 126960
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 2.4168464915544314e-06,
+      "loss": 0.2711,
+      "step": 126970
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 2.4138700796190196e-06,
+      "loss": 0.2773,
+      "step": 126980
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 2.4108936676836078e-06,
+      "loss": 0.2906,
+      "step": 126990
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 2.407917255748196e-06,
+      "loss": 0.2663,
+      "step": 127000
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 2.4049408438127837e-06,
+      "loss": 0.2634,
+      "step": 127010
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 2.4019644318773723e-06,
+      "loss": 0.2577,
+      "step": 127020
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 2.39898801994196e-06,
+      "loss": 0.2726,
+      "step": 127030
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 2.396011608006548e-06,
+      "loss": 0.2741,
+      "step": 127040
+    },
+    {
+      "epoch": 28.21,
+      "learning_rate": 2.3930351960711364e-06,
+      "loss": 0.2575,
+      "step": 127050
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 2.3900587841357245e-06,
+      "loss": 0.2577,
+      "step": 127060
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 2.3870823722003127e-06,
+      "loss": 0.2616,
+      "step": 127070
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 2.384105960264901e-06,
+      "loss": 0.2729,
+      "step": 127080
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 2.381129548329489e-06,
+      "loss": 0.2701,
+      "step": 127090
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 2.378153136394077e-06,
+      "loss": 0.2799,
+      "step": 127100
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 2.375176724458665e-06,
+      "loss": 0.2673,
+      "step": 127110
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 2.3722003125232535e-06,
+      "loss": 0.2691,
+      "step": 127120
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 2.3692239005878413e-06,
+      "loss": 0.2604,
+      "step": 127130
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 2.36624748865243e-06,
+      "loss": 0.2725,
+      "step": 127140
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 2.3632710767170176e-06,
+      "loss": 0.2643,
+      "step": 127150
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 2.3602946647816062e-06,
+      "loss": 0.2838,
+      "step": 127160
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 2.357318252846194e-06,
+      "loss": 0.2646,
+      "step": 127170
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 2.3543418409107826e-06,
+      "loss": 0.2692,
+      "step": 127180
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2.3513654289753703e-06,
+      "loss": 0.2747,
+      "step": 127190
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2.3483890170399585e-06,
+      "loss": 0.2672,
+      "step": 127200
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2.3454126051045466e-06,
+      "loss": 0.2605,
+      "step": 127210
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2.342436193169135e-06,
+      "loss": 0.2701,
+      "step": 127220
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 2.339459781233723e-06,
+      "loss": 0.2723,
+      "step": 127230
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 2.336483369298311e-06,
+      "loss": 0.2781,
+      "step": 127240
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 2.3335069573628993e-06,
+      "loss": 0.2758,
+      "step": 127250
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 2.3305305454274875e-06,
+      "loss": 0.2838,
+      "step": 127260
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 2.3275541334920752e-06,
+      "loss": 0.2838,
+      "step": 127270
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.324577721556664e-06,
+      "loss": 0.2673,
+      "step": 127280
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.3216013096212516e-06,
+      "loss": 0.2679,
+      "step": 127290
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.31862489768584e-06,
+      "loss": 0.2588,
+      "step": 127300
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.315648485750428e-06,
+      "loss": 0.281,
+      "step": 127310
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 2.3126720738150165e-06,
+      "loss": 0.2663,
+      "step": 127320
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 2.3096956618796042e-06,
+      "loss": 0.2814,
+      "step": 127330
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 2.3067192499441924e-06,
+      "loss": 0.2664,
+      "step": 127340
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 2.3037428380087806e-06,
+      "loss": 0.2629,
+      "step": 127350
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 2.3007664260733688e-06,
+      "loss": 0.2695,
+      "step": 127360
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 2.297790014137957e-06,
+      "loss": 0.2714,
+      "step": 127370
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 2.294813602202545e-06,
+      "loss": 0.266,
+      "step": 127380
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 2.2918371902671333e-06,
+      "loss": 0.2629,
+      "step": 127390
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 2.2888607783317214e-06,
+      "loss": 0.2657,
+      "step": 127400
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 2.2858843663963096e-06,
+      "loss": 0.2709,
+      "step": 127410
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 2.2829079544608978e-06,
+      "loss": 0.2644,
+      "step": 127420
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 2.2799315425254855e-06,
+      "loss": 0.2815,
+      "step": 127430
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 2.2769551305900737e-06,
+      "loss": 0.278,
+      "step": 127440
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 2.273978718654662e-06,
+      "loss": 0.2603,
+      "step": 127450
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 2.27100230671925e-06,
+      "loss": 0.2609,
+      "step": 127460
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 2.268025894783838e-06,
+      "loss": 0.2749,
+      "step": 127470
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 2.2650494828484264e-06,
+      "loss": 0.2615,
+      "step": 127480
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 2.2620730709130145e-06,
+      "loss": 0.2804,
+      "step": 127490
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 2.2590966589776027e-06,
+      "loss": 0.2679,
+      "step": 127500
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 2.256120247042191e-06,
+      "loss": 0.2774,
+      "step": 127510
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 2.253143835106779e-06,
+      "loss": 0.2703,
+      "step": 127520
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 2.2501674231713672e-06,
+      "loss": 0.2654,
+      "step": 127530
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 2.2471910112359554e-06,
+      "loss": 0.2659,
+      "step": 127540
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 2.2442145993005436e-06,
+      "loss": 0.2783,
+      "step": 127550
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 2.2412381873651313e-06,
+      "loss": 0.2715,
+      "step": 127560
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 2.23826177542972e-06,
+      "loss": 0.2612,
+      "step": 127570
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 2.2352853634943076e-06,
+      "loss": 0.2684,
+      "step": 127580
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 2.232308951558896e-06,
+      "loss": 0.2774,
+      "step": 127590
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 2.229332539623484e-06,
+      "loss": 0.269,
+      "step": 127600
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 2.226356127688072e-06,
+      "loss": 0.2565,
+      "step": 127610
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 2.2233797157526603e-06,
+      "loss": 0.2732,
+      "step": 127620
+    },
+    {
+      "epoch": 28.34,
+      "learning_rate": 2.2204033038172485e-06,
+      "loss": 0.2618,
+      "step": 127630
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2.2174268918818367e-06,
+      "loss": 0.262,
+      "step": 127640
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2.214450479946425e-06,
+      "loss": 0.2724,
+      "step": 127650
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2.2114740680110126e-06,
+      "loss": 0.2864,
+      "step": 127660
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2.208497656075601e-06,
+      "loss": 0.2597,
+      "step": 127670
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 2.205521244140189e-06,
+      "loss": 0.2799,
+      "step": 127680
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 2.2025448322047775e-06,
+      "loss": 0.2652,
+      "step": 127690
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 2.1995684202693652e-06,
+      "loss": 0.2757,
+      "step": 127700
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 2.196592008333954e-06,
+      "loss": 0.2698,
+      "step": 127710
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 2.1936155963985416e-06,
+      "loss": 0.2682,
+      "step": 127720
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 2.19063918446313e-06,
+      "loss": 0.2661,
+      "step": 127730
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 2.187662772527718e-06,
+      "loss": 0.278,
+      "step": 127740
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 2.184686360592306e-06,
+      "loss": 0.2678,
+      "step": 127750
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 2.1817099486568943e-06,
+      "loss": 0.2644,
+      "step": 127760
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 2.1787335367214824e-06,
+      "loss": 0.2626,
+      "step": 127770
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 2.1757571247860706e-06,
+      "loss": 0.2785,
+      "step": 127780
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 2.1727807128506588e-06,
+      "loss": 0.2835,
+      "step": 127790
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 2.169804300915247e-06,
+      "loss": 0.2704,
+      "step": 127800
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 2.166827888979835e-06,
+      "loss": 0.2614,
+      "step": 127810
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 2.163851477044423e-06,
+      "loss": 0.2618,
+      "step": 127820
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 2.1608750651090114e-06,
+      "loss": 0.2677,
+      "step": 127830
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 2.157898653173599e-06,
+      "loss": 0.2519,
+      "step": 127840
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 2.1549222412381878e-06,
+      "loss": 0.2726,
+      "step": 127850
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 2.1519458293027755e-06,
+      "loss": 0.2745,
+      "step": 127860
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2.148969417367364e-06,
+      "loss": 0.2739,
+      "step": 127870
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2.145993005431952e-06,
+      "loss": 0.2566,
+      "step": 127880
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2.14301659349654e-06,
+      "loss": 0.2691,
+      "step": 127890
+    },
+    {
+      "epoch": 28.4,
+      "learning_rate": 2.140040181561128e-06,
+      "loss": 0.2733,
+      "step": 127900
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 2.1370637696257164e-06,
+      "loss": 0.2561,
+      "step": 127910
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 2.1340873576903046e-06,
+      "loss": 0.2657,
+      "step": 127920
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 2.1311109457548927e-06,
+      "loss": 0.2723,
+      "step": 127930
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 2.128134533819481e-06,
+      "loss": 0.2665,
+      "step": 127940
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 2.125158121884069e-06,
+      "loss": 0.2632,
+      "step": 127950
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 2.1221817099486572e-06,
+      "loss": 0.2727,
+      "step": 127960
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 2.1192052980132454e-06,
+      "loss": 0.2696,
+      "step": 127970
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 2.116228886077833e-06,
+      "loss": 0.2804,
+      "step": 127980
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 2.1132524741424213e-06,
+      "loss": 0.2788,
+      "step": 127990
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 2.1102760622070095e-06,
+      "loss": 0.2809,
+      "step": 128000
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 2.1072996502715977e-06,
+      "loss": 0.2658,
+      "step": 128010
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 2.104323238336186e-06,
+      "loss": 0.2679,
+      "step": 128020
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 2.101346826400774e-06,
+      "loss": 0.2715,
+      "step": 128030
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 2.098370414465362e-06,
+      "loss": 0.2704,
+      "step": 128040
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 2.0953940025299503e-06,
+      "loss": 0.2655,
+      "step": 128050
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 2.0924175905945385e-06,
+      "loss": 0.2678,
+      "step": 128060
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 2.0894411786591267e-06,
+      "loss": 0.262,
+      "step": 128070
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 2.086464766723715e-06,
+      "loss": 0.2625,
+      "step": 128080
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2.083488354788303e-06,
+      "loss": 0.2676,
+      "step": 128090
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2.080511942852891e-06,
+      "loss": 0.2584,
+      "step": 128100
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2.077535530917479e-06,
+      "loss": 0.279,
+      "step": 128110
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2.0745591189820675e-06,
+      "loss": 0.2813,
+      "step": 128120
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 2.0715827070466553e-06,
+      "loss": 0.2785,
+      "step": 128130
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 2.0686062951112434e-06,
+      "loss": 0.2783,
+      "step": 128140
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 2.0656298831758316e-06,
+      "loss": 0.2723,
+      "step": 128150
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 2.0626534712404198e-06,
+      "loss": 0.2669,
+      "step": 128160
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 2.059677059305008e-06,
+      "loss": 0.2743,
+      "step": 128170
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 2.056700647369596e-06,
+      "loss": 0.2755,
+      "step": 128180
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 2.0537242354341843e-06,
+      "loss": 0.2689,
+      "step": 128190
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 2.0507478234987724e-06,
+      "loss": 0.2626,
+      "step": 128200
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 2.04777141156336e-06,
+      "loss": 0.26,
+      "step": 128210
+    },
+    {
+      "epoch": 28.47,
+      "learning_rate": 2.0447949996279488e-06,
+      "loss": 0.2619,
+      "step": 128220
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 2.0418185876925365e-06,
+      "loss": 0.2731,
+      "step": 128230
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 2.038842175757125e-06,
+      "loss": 0.2704,
+      "step": 128240
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 2.035865763821713e-06,
+      "loss": 0.2615,
+      "step": 128250
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 2.0328893518863015e-06,
+      "loss": 0.2697,
+      "step": 128260
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.029912939950889e-06,
+      "loss": 0.2607,
+      "step": 128270
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.026936528015478e-06,
+      "loss": 0.2705,
+      "step": 128280
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.0239601160800655e-06,
+      "loss": 0.2617,
+      "step": 128290
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.0209837041446537e-06,
+      "loss": 0.2611,
+      "step": 128300
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 2.018007292209242e-06,
+      "loss": 0.27,
+      "step": 128310
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 2.01503088027383e-06,
+      "loss": 0.2727,
+      "step": 128320
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 2.0120544683384182e-06,
+      "loss": 0.2739,
+      "step": 128330
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 2.0090780564030064e-06,
+      "loss": 0.2652,
+      "step": 128340
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 2.0061016444675946e-06,
+      "loss": 0.2762,
+      "step": 128350
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 2.0031252325321827e-06,
+      "loss": 0.2741,
+      "step": 128360
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 2.0001488205967705e-06,
+      "loss": 0.2697,
+      "step": 128370
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.997172408661359e-06,
+      "loss": 0.2698,
+      "step": 128380
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.994195996725947e-06,
+      "loss": 0.2773,
+      "step": 128390
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.9912195847905354e-06,
+      "loss": 0.2608,
+      "step": 128400
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.988243172855123e-06,
+      "loss": 0.263,
+      "step": 128410
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.9852667609197118e-06,
+      "loss": 0.2908,
+      "step": 128420
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.9822903489842995e-06,
+      "loss": 0.2614,
+      "step": 128430
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.9793139370488877e-06,
+      "loss": 0.2813,
+      "step": 128440
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.976337525113476e-06,
+      "loss": 0.2686,
+      "step": 128450
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.973361113178064e-06,
+      "loss": 0.27,
+      "step": 128460
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.970384701242652e-06,
+      "loss": 0.2791,
+      "step": 128470
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.9674082893072403e-06,
+      "loss": 0.2785,
+      "step": 128480
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.9644318773718285e-06,
+      "loss": 0.285,
+      "step": 128490
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.9614554654364167e-06,
+      "loss": 0.2658,
+      "step": 128500
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.958479053501005e-06,
+      "loss": 0.2688,
+      "step": 128510
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.955502641565593e-06,
+      "loss": 0.2755,
+      "step": 128520
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 1.9525262296301808e-06,
+      "loss": 0.2665,
+      "step": 128530
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.949549817694769e-06,
+      "loss": 0.263,
+      "step": 128540
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.946573405759357e-06,
+      "loss": 0.264,
+      "step": 128550
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.9435969938239453e-06,
+      "loss": 0.279,
+      "step": 128560
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.9406205818885334e-06,
+      "loss": 0.2738,
+      "step": 128570
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.9376441699531216e-06,
+      "loss": 0.2653,
+      "step": 128580
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.9346677580177098e-06,
+      "loss": 0.2538,
+      "step": 128590
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.931691346082298e-06,
+      "loss": 0.2831,
+      "step": 128600
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.928714934146886e-06,
+      "loss": 0.2853,
+      "step": 128610
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.9257385222114743e-06,
+      "loss": 0.2528,
+      "step": 128620
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.9227621102760625e-06,
+      "loss": 0.2683,
+      "step": 128630
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.9197856983406506e-06,
+      "loss": 0.2735,
+      "step": 128640
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.916809286405239e-06,
+      "loss": 0.2811,
+      "step": 128650
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.9138328744698265e-06,
+      "loss": 0.2689,
+      "step": 128660
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 1.910856462534415e-06,
+      "loss": 0.2806,
+      "step": 128670
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.907880050599003e-06,
+      "loss": 0.2589,
+      "step": 128680
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.904903638663591e-06,
+      "loss": 0.2728,
+      "step": 128690
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.9019272267281794e-06,
+      "loss": 0.2735,
+      "step": 128700
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.8989508147927674e-06,
+      "loss": 0.2702,
+      "step": 128710
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.8959744028573558e-06,
+      "loss": 0.2728,
+      "step": 128720
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.8929979909219437e-06,
+      "loss": 0.2708,
+      "step": 128730
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.890021578986532e-06,
+      "loss": 0.2753,
+      "step": 128740
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.88704516705112e-06,
+      "loss": 0.2808,
+      "step": 128750
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.8840687551157082e-06,
+      "loss": 0.2747,
+      "step": 128760
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.8810923431802964e-06,
+      "loss": 0.2799,
+      "step": 128770
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.8781159312448844e-06,
+      "loss": 0.2835,
+      "step": 128780
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.8751395193094725e-06,
+      "loss": 0.2648,
+      "step": 128790
+    },
+    {
+      "epoch": 28.6,
+      "learning_rate": 1.8721631073740607e-06,
+      "loss": 0.2627,
+      "step": 128800
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.8691866954386489e-06,
+      "loss": 0.2656,
+      "step": 128810
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.866210283503237e-06,
+      "loss": 0.2629,
+      "step": 128820
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.8632338715678252e-06,
+      "loss": 0.2716,
+      "step": 128830
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.8602574596324132e-06,
+      "loss": 0.2799,
+      "step": 128840
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.8572810476970013e-06,
+      "loss": 0.2624,
+      "step": 128850
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.8543046357615895e-06,
+      "loss": 0.2738,
+      "step": 128860
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.8513282238261777e-06,
+      "loss": 0.274,
+      "step": 128870
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.8483518118907658e-06,
+      "loss": 0.2808,
+      "step": 128880
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.845375399955354e-06,
+      "loss": 0.2766,
+      "step": 128890
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.8423989880199422e-06,
+      "loss": 0.2608,
+      "step": 128900
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.8394225760845301e-06,
+      "loss": 0.2711,
+      "step": 128910
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.8364461641491185e-06,
+      "loss": 0.259,
+      "step": 128920
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.8334697522137065e-06,
+      "loss": 0.2678,
+      "step": 128930
+    },
+    {
+      "epoch": 28.63,
+      "learning_rate": 1.8304933402782947e-06,
+      "loss": 0.2681,
+      "step": 128940
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.8275169283428828e-06,
+      "loss": 0.273,
+      "step": 128950
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.8245405164074708e-06,
+      "loss": 0.2796,
+      "step": 128960
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.8215641044720592e-06,
+      "loss": 0.2728,
+      "step": 128970
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.8185876925366471e-06,
+      "loss": 0.2612,
+      "step": 128980
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.8156112806012355e-06,
+      "loss": 0.2737,
+      "step": 128990
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.8126348686658235e-06,
+      "loss": 0.2792,
+      "step": 129000
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.8096584567304114e-06,
+      "loss": 0.2709,
+      "step": 129010
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.8066820447949998e-06,
+      "loss": 0.2655,
+      "step": 129020
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.8037056328595878e-06,
+      "loss": 0.2696,
+      "step": 129030
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.8007292209241761e-06,
+      "loss": 0.268,
+      "step": 129040
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.797752808988764e-06,
+      "loss": 0.2705,
+      "step": 129050
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.7947763970533525e-06,
+      "loss": 0.2682,
+      "step": 129060
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.7917999851179404e-06,
+      "loss": 0.2696,
+      "step": 129070
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.7888235731825284e-06,
+      "loss": 0.2588,
+      "step": 129080
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.7858471612471168e-06,
+      "loss": 0.2648,
+      "step": 129090
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.7828707493117047e-06,
+      "loss": 0.278,
+      "step": 129100
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.7798943373762931e-06,
+      "loss": 0.2865,
+      "step": 129110
+    },
+    {
+      "epoch": 28.67,
+      "learning_rate": 1.776917925440881e-06,
+      "loss": 0.2684,
+      "step": 129120
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.7739415135054694e-06,
+      "loss": 0.2645,
+      "step": 129130
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.7709651015700574e-06,
+      "loss": 0.2833,
+      "step": 129140
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.7679886896346458e-06,
+      "loss": 0.2689,
+      "step": 129150
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.7650122776992337e-06,
+      "loss": 0.267,
+      "step": 129160
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.7620358657638217e-06,
+      "loss": 0.2704,
+      "step": 129170
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.75905945382841e-06,
+      "loss": 0.2752,
+      "step": 129180
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.756083041892998e-06,
+      "loss": 0.2826,
+      "step": 129190
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.7531066299575864e-06,
+      "loss": 0.2736,
+      "step": 129200
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.7501302180221744e-06,
+      "loss": 0.2693,
+      "step": 129210
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.7471538060867628e-06,
+      "loss": 0.2633,
+      "step": 129220
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.7441773941513507e-06,
+      "loss": 0.2587,
+      "step": 129230
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.7412009822159387e-06,
+      "loss": 0.2956,
+      "step": 129240
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 1.738224570280527e-06,
+      "loss": 0.2736,
+      "step": 129250
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.735248158345115e-06,
+      "loss": 0.2691,
+      "step": 129260
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.7322717464097034e-06,
+      "loss": 0.2636,
+      "step": 129270
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.7292953344742914e-06,
+      "loss": 0.261,
+      "step": 129280
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.7263189225388795e-06,
+      "loss": 0.2474,
+      "step": 129290
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.7233425106034677e-06,
+      "loss": 0.2748,
+      "step": 129300
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.7203660986680559e-06,
+      "loss": 0.2724,
+      "step": 129310
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.717389686732644e-06,
+      "loss": 0.2711,
+      "step": 129320
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.714413274797232e-06,
+      "loss": 0.2769,
+      "step": 129330
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.7114368628618202e-06,
+      "loss": 0.2618,
+      "step": 129340
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.7084604509264083e-06,
+      "loss": 0.28,
+      "step": 129350
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.7054840389909965e-06,
+      "loss": 0.2681,
+      "step": 129360
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.7025076270555847e-06,
+      "loss": 0.2804,
+      "step": 129370
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.6995312151201728e-06,
+      "loss": 0.2685,
+      "step": 129380
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 1.6965548031847608e-06,
+      "loss": 0.2711,
+      "step": 129390
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.693578391249349e-06,
+      "loss": 0.2822,
+      "step": 129400
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.6906019793139371e-06,
+      "loss": 0.276,
+      "step": 129410
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.6876255673785253e-06,
+      "loss": 0.2567,
+      "step": 129420
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.6846491554431135e-06,
+      "loss": 0.2682,
+      "step": 129430
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.6816727435077016e-06,
+      "loss": 0.2683,
+      "step": 129440
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.6786963315722898e-06,
+      "loss": 0.2677,
+      "step": 129450
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.6757199196368778e-06,
+      "loss": 0.2538,
+      "step": 129460
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.6727435077014662e-06,
+      "loss": 0.2605,
+      "step": 129470
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.6697670957660541e-06,
+      "loss": 0.2652,
+      "step": 129480
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.6667906838306423e-06,
+      "loss": 0.275,
+      "step": 129490
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.6638142718952304e-06,
+      "loss": 0.2739,
+      "step": 129500
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.6608378599598184e-06,
+      "loss": 0.2725,
+      "step": 129510
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.6578614480244068e-06,
+      "loss": 0.2648,
+      "step": 129520
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.6548850360889947e-06,
+      "loss": 0.2688,
+      "step": 129530
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.6519086241535831e-06,
+      "loss": 0.258,
+      "step": 129540
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.648932212218171e-06,
+      "loss": 0.2765,
+      "step": 129550
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.645955800282759e-06,
+      "loss": 0.2712,
+      "step": 129560
+    },
+    {
+      "epoch": 28.77,
+      "learning_rate": 1.6429793883473474e-06,
+      "loss": 0.2555,
+      "step": 129570
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.6400029764119354e-06,
+      "loss": 0.2921,
+      "step": 129580
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.6370265644765238e-06,
+      "loss": 0.2654,
+      "step": 129590
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.6340501525411117e-06,
+      "loss": 0.2673,
+      "step": 129600
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.6310737406057e-06,
+      "loss": 0.2632,
+      "step": 129610
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.628097328670288e-06,
+      "loss": 0.2718,
+      "step": 129620
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.6251209167348764e-06,
+      "loss": 0.2626,
+      "step": 129630
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.6221445047994644e-06,
+      "loss": 0.2553,
+      "step": 129640
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.6191680928640524e-06,
+      "loss": 0.2723,
+      "step": 129650
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.6161916809286407e-06,
+      "loss": 0.2745,
+      "step": 129660
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.6132152689932287e-06,
+      "loss": 0.2645,
+      "step": 129670
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.610238857057817e-06,
+      "loss": 0.2744,
+      "step": 129680
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.607262445122405e-06,
+      "loss": 0.2629,
+      "step": 129690
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 1.6042860331869934e-06,
+      "loss": 0.2608,
+      "step": 129700
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.6013096212515814e-06,
+      "loss": 0.2605,
+      "step": 129710
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.5983332093161693e-06,
+      "loss": 0.2679,
+      "step": 129720
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.5953567973807577e-06,
+      "loss": 0.2665,
+      "step": 129730
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.5923803854453457e-06,
+      "loss": 0.2597,
+      "step": 129740
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.589403973509934e-06,
+      "loss": 0.2645,
+      "step": 129750
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.586427561574522e-06,
+      "loss": 0.27,
+      "step": 129760
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.5834511496391104e-06,
+      "loss": 0.256,
+      "step": 129770
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.5804747377036983e-06,
+      "loss": 0.276,
+      "step": 129780
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.5774983257682865e-06,
+      "loss": 0.2723,
+      "step": 129790
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.5745219138328747e-06,
+      "loss": 0.2761,
+      "step": 129800
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.5715455018974626e-06,
+      "loss": 0.2661,
+      "step": 129810
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.568569089962051e-06,
+      "loss": 0.2734,
+      "step": 129820
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.565592678026639e-06,
+      "loss": 0.2749,
+      "step": 129830
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 1.5626162660912271e-06,
+      "loss": 0.264,
+      "step": 129840
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.5596398541558153e-06,
+      "loss": 0.269,
+      "step": 129850
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.5566634422204035e-06,
+      "loss": 0.2795,
+      "step": 129860
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.5536870302849917e-06,
+      "loss": 0.269,
+      "step": 129870
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.5507106183495796e-06,
+      "loss": 0.2626,
+      "step": 129880
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.5477342064141678e-06,
+      "loss": 0.2693,
+      "step": 129890
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.544757794478756e-06,
+      "loss": 0.268,
+      "step": 129900
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.5417813825433441e-06,
+      "loss": 0.2602,
+      "step": 129910
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.5388049706079323e-06,
+      "loss": 0.2754,
+      "step": 129920
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.5358285586725205e-06,
+      "loss": 0.2507,
+      "step": 129930
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.5328521467371086e-06,
+      "loss": 0.2532,
+      "step": 129940
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.5298757348016966e-06,
+      "loss": 0.2747,
+      "step": 129950
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.5268993228662848e-06,
+      "loss": 0.2661,
+      "step": 129960
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.523922910930873e-06,
+      "loss": 0.2662,
+      "step": 129970
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.520946498995461e-06,
+      "loss": 0.2646,
+      "step": 129980
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.5179700870600493e-06,
+      "loss": 0.2607,
+      "step": 129990
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.5149936751246374e-06,
+      "loss": 0.2703,
+      "step": 130000
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.5120172631892254e-06,
+      "loss": 0.2583,
+      "step": 130010
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 1.5090408512538138e-06,
+      "loss": 0.2669,
+      "step": 130020
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.5060644393184017e-06,
+      "loss": 0.2613,
+      "step": 130030
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.50308802738299e-06,
+      "loss": 0.2703,
+      "step": 130040
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.500111615447578e-06,
+      "loss": 0.2654,
+      "step": 130050
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.497135203512166e-06,
+      "loss": 0.2805,
+      "step": 130060
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.4941587915767544e-06,
+      "loss": 0.2699,
+      "step": 130070
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.4911823796413424e-06,
+      "loss": 0.2672,
+      "step": 130080
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.4882059677059307e-06,
+      "loss": 0.2738,
+      "step": 130090
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.4852295557705187e-06,
+      "loss": 0.2804,
+      "step": 130100
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.4822531438351067e-06,
+      "loss": 0.2674,
+      "step": 130110
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.479276731899695e-06,
+      "loss": 0.2708,
+      "step": 130120
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.476300319964283e-06,
+      "loss": 0.2766,
+      "step": 130130
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.4733239080288714e-06,
+      "loss": 0.2668,
+      "step": 130140
+    },
+    {
+      "epoch": 28.9,
+      "learning_rate": 1.4703474960934593e-06,
+      "loss": 0.2623,
+      "step": 130150
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.4673710841580477e-06,
+      "loss": 0.272,
+      "step": 130160
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.4643946722226357e-06,
+      "loss": 0.2693,
+      "step": 130170
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.461418260287224e-06,
+      "loss": 0.2658,
+      "step": 130180
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.458441848351812e-06,
+      "loss": 0.2877,
+      "step": 130190
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.4554654364164e-06,
+      "loss": 0.2613,
+      "step": 130200
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.4524890244809884e-06,
+      "loss": 0.2772,
+      "step": 130210
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.4495126125455763e-06,
+      "loss": 0.2808,
+      "step": 130220
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.4465362006101647e-06,
+      "loss": 0.2675,
+      "step": 130230
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.4435597886747527e-06,
+      "loss": 0.2775,
+      "step": 130240
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.440583376739341e-06,
+      "loss": 0.2763,
+      "step": 130250
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.437606964803929e-06,
+      "loss": 0.2644,
+      "step": 130260
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.434630552868517e-06,
+      "loss": 0.2665,
+      "step": 130270
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.4316541409331053e-06,
+      "loss": 0.2677,
+      "step": 130280
+    },
+    {
+      "epoch": 28.93,
+      "learning_rate": 1.4286777289976933e-06,
+      "loss": 0.2858,
+      "step": 130290
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.4257013170622817e-06,
+      "loss": 0.2762,
+      "step": 130300
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.4227249051268696e-06,
+      "loss": 0.2634,
+      "step": 130310
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.419748493191458e-06,
+      "loss": 0.2723,
+      "step": 130320
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.416772081256046e-06,
+      "loss": 0.285,
+      "step": 130330
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.4137956693206341e-06,
+      "loss": 0.2758,
+      "step": 130340
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.4108192573852223e-06,
+      "loss": 0.2776,
+      "step": 130350
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.4078428454498103e-06,
+      "loss": 0.273,
+      "step": 130360
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.4048664335143986e-06,
+      "loss": 0.2518,
+      "step": 130370
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.4018900215789866e-06,
+      "loss": 0.2707,
+      "step": 130380
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.3989136096435748e-06,
+      "loss": 0.2661,
+      "step": 130390
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.395937197708163e-06,
+      "loss": 0.2819,
+      "step": 130400
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.3929607857727511e-06,
+      "loss": 0.2773,
+      "step": 130410
+    },
+    {
+      "epoch": 28.96,
+      "learning_rate": 1.3899843738373393e-06,
+      "loss": 0.2751,
+      "step": 130420
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.3870079619019272e-06,
+      "loss": 0.2645,
+      "step": 130430
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.3840315499665154e-06,
+      "loss": 0.2752,
+      "step": 130440
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.3810551380311036e-06,
+      "loss": 0.2776,
+      "step": 130450
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.3780787260956917e-06,
+      "loss": 0.2613,
+      "step": 130460
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.37510231416028e-06,
+      "loss": 0.2786,
+      "step": 130470
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.372125902224868e-06,
+      "loss": 0.2679,
+      "step": 130480
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.3691494902894563e-06,
+      "loss": 0.2432,
+      "step": 130490
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.3661730783540444e-06,
+      "loss": 0.277,
+      "step": 130500
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.3631966664186324e-06,
+      "loss": 0.2662,
+      "step": 130510
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.3602202544832206e-06,
+      "loss": 0.2665,
+      "step": 130520
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.3572438425478087e-06,
+      "loss": 0.2734,
+      "step": 130530
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.3542674306123969e-06,
+      "loss": 0.262,
+      "step": 130540
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.351291018676985e-06,
+      "loss": 0.278,
+      "step": 130550
+    },
+    {
+      "epoch": 28.99,
+      "eval_cer": 4.536324786324786,
+      "eval_loss": 0.5664880275726318,
+      "eval_runtime": 1.2781,
+      "eval_samples_per_second": 7.824,
+      "eval_steps_per_second": 0.782,
+      "eval_wer": 0.6103896103896104,
+      "step": 130558
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.348314606741573e-06,
+      "loss": 0.2681,
+      "step": 130560
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.3453381948061614e-06,
+      "loss": 0.2569,
+      "step": 130570
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.3423617828707494e-06,
+      "loss": 0.2802,
+      "step": 130580
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.3393853709353375e-06,
+      "loss": 0.2792,
+      "step": 130590
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 1.3364089589999257e-06,
+      "loss": 0.2777,
+      "step": 130600
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.3334325470645137e-06,
+      "loss": 0.2546,
+      "step": 130610
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.330456135129102e-06,
+      "loss": 0.2692,
+      "step": 130620
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.32747972319369e-06,
+      "loss": 0.2575,
+      "step": 130630
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.3245033112582784e-06,
+      "loss": 0.2578,
+      "step": 130640
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.3215268993228663e-06,
+      "loss": 0.2609,
+      "step": 130650
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.3185504873874547e-06,
+      "loss": 0.2757,
+      "step": 130660
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.3155740754520427e-06,
+      "loss": 0.2703,
+      "step": 130670
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.3125976635166306e-06,
+      "loss": 0.2771,
+      "step": 130680
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.309621251581219e-06,
+      "loss": 0.2651,
+      "step": 130690
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.306644839645807e-06,
+      "loss": 0.2631,
+      "step": 130700
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.3036684277103953e-06,
+      "loss": 0.2755,
+      "step": 130710
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.3006920157749833e-06,
+      "loss": 0.2699,
+      "step": 130720
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2977156038395717e-06,
+      "loss": 0.2689,
+      "step": 130730
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 1.2947391919041596e-06,
+      "loss": 0.2558,
+      "step": 130740
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2917627799687476e-06,
+      "loss": 0.2829,
+      "step": 130750
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.288786368033336e-06,
+      "loss": 0.2604,
+      "step": 130760
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.285809956097924e-06,
+      "loss": 0.2687,
+      "step": 130770
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 1.2828335441625123e-06,
+      "loss": 0.26,
+      "step": 130780
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2798571322271003e-06,
+      "loss": 0.2712,
+      "step": 130790
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2768807202916887e-06,
+      "loss": 0.2559,
+      "step": 130800
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2739043083562766e-06,
+      "loss": 0.2662,
+      "step": 130810
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.2709278964208646e-06,
+      "loss": 0.2754,
+      "step": 130820
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 1.267951484485453e-06,
+      "loss": 0.2696,
+      "step": 130830
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.264975072550041e-06,
+      "loss": 0.2659,
+      "step": 130840
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2619986606146293e-06,
+      "loss": 0.2571,
+      "step": 130850
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2590222486792173e-06,
+      "loss": 0.2826,
+      "step": 130860
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 1.2560458367438056e-06,
+      "loss": 0.2623,
+      "step": 130870
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2530694248083936e-06,
+      "loss": 0.2641,
+      "step": 130880
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2500930128729818e-06,
+      "loss": 0.2592,
+      "step": 130890
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.24711660093757e-06,
+      "loss": 0.2729,
+      "step": 130900
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.244140189002158e-06,
+      "loss": 0.2686,
+      "step": 130910
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 1.2411637770667463e-06,
+      "loss": 0.286,
+      "step": 130920
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2381873651313344e-06,
+      "loss": 0.2765,
+      "step": 130930
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2352109531959224e-06,
+      "loss": 0.2636,
+      "step": 130940
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2322345412605106e-06,
+      "loss": 0.2622,
+      "step": 130950
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 1.2292581293250987e-06,
+      "loss": 0.2776,
+      "step": 130960
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.226281717389687e-06,
+      "loss": 0.2562,
+      "step": 130970
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.223305305454275e-06,
+      "loss": 0.2727,
+      "step": 130980
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.220328893518863e-06,
+      "loss": 0.2784,
+      "step": 130990
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2173524815834512e-06,
+      "loss": 0.2502,
+      "step": 131000
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 1.2143760696480394e-06,
+      "loss": 0.2825,
+      "step": 131010
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2113996577126275e-06,
+      "loss": 0.274,
+      "step": 131020
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2084232457772157e-06,
+      "loss": 0.2668,
+      "step": 131030
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2054468338418039e-06,
+      "loss": 0.276,
+      "step": 131040
+    },
+    {
+      "epoch": 29.1,
+      "learning_rate": 1.2024704219063918e-06,
+      "loss": 0.257,
+      "step": 131050
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.19949400997098e-06,
+      "loss": 0.2753,
+      "step": 131060
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.1965175980355682e-06,
+      "loss": 0.2715,
+      "step": 131070
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.1935411861001563e-06,
+      "loss": 0.2763,
+      "step": 131080
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.1905647741647445e-06,
+      "loss": 0.2735,
+      "step": 131090
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 1.1875883622293325e-06,
+      "loss": 0.2764,
+      "step": 131100
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.1846119502939206e-06,
+      "loss": 0.2608,
+      "step": 131110
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.1816355383585088e-06,
+      "loss": 0.2898,
+      "step": 131120
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.178659126423097e-06,
+      "loss": 0.2707,
+      "step": 131130
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 1.1756827144876852e-06,
+      "loss": 0.2718,
+      "step": 131140
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.1727063025522733e-06,
+      "loss": 0.2649,
+      "step": 131150
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.1697298906168615e-06,
+      "loss": 0.26,
+      "step": 131160
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.1667534786814497e-06,
+      "loss": 0.2646,
+      "step": 131170
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.1637770667460376e-06,
+      "loss": 0.2537,
+      "step": 131180
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 1.1608006548106258e-06,
+      "loss": 0.256,
+      "step": 131190
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.157824242875214e-06,
+      "loss": 0.2644,
+      "step": 131200
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.1548478309398021e-06,
+      "loss": 0.2707,
+      "step": 131210
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.1518714190043903e-06,
+      "loss": 0.2641,
+      "step": 131220
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 1.1488950070689785e-06,
+      "loss": 0.2781,
+      "step": 131230
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.1459185951335666e-06,
+      "loss": 0.2769,
+      "step": 131240
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.1429421831981548e-06,
+      "loss": 0.2732,
+      "step": 131250
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.1399657712627428e-06,
+      "loss": 0.2674,
+      "step": 131260
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.136989359327331e-06,
+      "loss": 0.262,
+      "step": 131270
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 1.134012947391919e-06,
+      "loss": 0.2716,
+      "step": 131280
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.1310365354565073e-06,
+      "loss": 0.2617,
+      "step": 131290
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.1280601235210954e-06,
+      "loss": 0.2649,
+      "step": 131300
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.1250837115856836e-06,
+      "loss": 0.2689,
+      "step": 131310
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 1.1221072996502718e-06,
+      "loss": 0.275,
+      "step": 131320
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.11913088771486e-06,
+      "loss": 0.2846,
+      "step": 131330
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.116154475779448e-06,
+      "loss": 0.2724,
+      "step": 131340
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.113178063844036e-06,
+      "loss": 0.272,
+      "step": 131350
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.1102016519086242e-06,
+      "loss": 0.2774,
+      "step": 131360
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 1.1072252399732124e-06,
+      "loss": 0.2625,
+      "step": 131370
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.1042488280378006e-06,
+      "loss": 0.2818,
+      "step": 131380
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.1012724161023888e-06,
+      "loss": 0.2729,
+      "step": 131390
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.098296004166977e-06,
+      "loss": 0.2602,
+      "step": 131400
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.095319592231565e-06,
+      "loss": 0.2791,
+      "step": 131410
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 1.092343180296153e-06,
+      "loss": 0.2635,
+      "step": 131420
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.0893667683607412e-06,
+      "loss": 0.2645,
+      "step": 131430
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.0863903564253294e-06,
+      "loss": 0.2548,
+      "step": 131440
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.0834139444899176e-06,
+      "loss": 0.2598,
+      "step": 131450
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 1.0804375325545057e-06,
+      "loss": 0.265,
+      "step": 131460
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.0774611206190939e-06,
+      "loss": 0.2584,
+      "step": 131470
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.074484708683682e-06,
+      "loss": 0.2691,
+      "step": 131480
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.07150829674827e-06,
+      "loss": 0.2679,
+      "step": 131490
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.0685318848128582e-06,
+      "loss": 0.2567,
+      "step": 131500
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 1.0655554728774464e-06,
+      "loss": 0.2713,
+      "step": 131510
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.0625790609420345e-06,
+      "loss": 0.2711,
+      "step": 131520
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.0596026490066227e-06,
+      "loss": 0.2777,
+      "step": 131530
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.0566262370712107e-06,
+      "loss": 0.2666,
+      "step": 131540
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 1.0536498251357988e-06,
+      "loss": 0.269,
+      "step": 131550
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.050673413200387e-06,
+      "loss": 0.267,
+      "step": 131560
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.0476970012649752e-06,
+      "loss": 0.2777,
+      "step": 131570
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.0447205893295633e-06,
+      "loss": 0.2712,
+      "step": 131580
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.0417441773941515e-06,
+      "loss": 0.2668,
+      "step": 131590
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 1.0387677654587395e-06,
+      "loss": 0.2693,
+      "step": 131600
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.0357913535233276e-06,
+      "loss": 0.2674,
+      "step": 131610
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.0328149415879158e-06,
+      "loss": 0.2679,
+      "step": 131620
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.029838529652504e-06,
+      "loss": 0.2551,
+      "step": 131630
+    },
+    {
+      "epoch": 29.23,
+      "learning_rate": 1.0268621177170921e-06,
+      "loss": 0.2849,
+      "step": 131640
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.02388570578168e-06,
+      "loss": 0.2921,
+      "step": 131650
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.0209092938462683e-06,
+      "loss": 0.2666,
+      "step": 131660
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.0179328819108564e-06,
+      "loss": 0.2663,
+      "step": 131670
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.0149564699754446e-06,
+      "loss": 0.276,
+      "step": 131680
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 1.0119800580400328e-06,
+      "loss": 0.2723,
+      "step": 131690
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.009003646104621e-06,
+      "loss": 0.2805,
+      "step": 131700
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.0060272341692091e-06,
+      "loss": 0.2542,
+      "step": 131710
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.0030508222337973e-06,
+      "loss": 0.2759,
+      "step": 131720
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 1.0000744102983852e-06,
+      "loss": 0.2766,
+      "step": 131730
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 9.970979983629734e-07,
+      "loss": 0.2815,
+      "step": 131740
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 9.941215864275616e-07,
+      "loss": 0.2693,
+      "step": 131750
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 9.911451744921497e-07,
+      "loss": 0.2649,
+      "step": 131760
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 9.88168762556738e-07,
+      "loss": 0.2708,
+      "step": 131770
+    },
+    {
+      "epoch": 29.26,
+      "learning_rate": 9.85192350621326e-07,
+      "loss": 0.2828,
+      "step": 131780
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 9.822159386859143e-07,
+      "loss": 0.274,
+      "step": 131790
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 9.792395267505024e-07,
+      "loss": 0.253,
+      "step": 131800
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 9.762631148150904e-07,
+      "loss": 0.2701,
+      "step": 131810
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 9.732867028796786e-07,
+      "loss": 0.2742,
+      "step": 131820
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 9.703102909442667e-07,
+      "loss": 0.2675,
+      "step": 131830
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 9.673338790088549e-07,
+      "loss": 0.2769,
+      "step": 131840
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 9.64357467073443e-07,
+      "loss": 0.2634,
+      "step": 131850
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 9.613810551380312e-07,
+      "loss": 0.2772,
+      "step": 131860
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 9.584046432026194e-07,
+      "loss": 0.2713,
+      "step": 131870
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 9.554282312672076e-07,
+      "loss": 0.255,
+      "step": 131880
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 9.524518193317955e-07,
+      "loss": 0.2637,
+      "step": 131890
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 9.494754073963837e-07,
+      "loss": 0.2663,
+      "step": 131900
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 9.464989954609719e-07,
+      "loss": 0.2599,
+      "step": 131910
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 9.4352258352556e-07,
+      "loss": 0.2696,
+      "step": 131920
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 9.405461715901482e-07,
+      "loss": 0.2641,
+      "step": 131930
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 9.375697596547363e-07,
+      "loss": 0.2634,
+      "step": 131940
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 9.345933477193244e-07,
+      "loss": 0.2698,
+      "step": 131950
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 9.316169357839126e-07,
+      "loss": 0.2656,
+      "step": 131960
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 9.286405238485007e-07,
+      "loss": 0.2674,
+      "step": 131970
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 9.256641119130888e-07,
+      "loss": 0.2643,
+      "step": 131980
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 9.22687699977677e-07,
+      "loss": 0.2727,
+      "step": 131990
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 9.197112880422651e-07,
+      "loss": 0.2606,
+      "step": 132000
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 9.167348761068532e-07,
+      "loss": 0.2743,
+      "step": 132010
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 9.137584641714414e-07,
+      "loss": 0.2717,
+      "step": 132020
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 9.107820522360296e-07,
+      "loss": 0.2583,
+      "step": 132030
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 9.078056403006178e-07,
+      "loss": 0.2658,
+      "step": 132040
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 9.048292283652057e-07,
+      "loss": 0.2686,
+      "step": 132050
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 9.018528164297939e-07,
+      "loss": 0.2502,
+      "step": 132060
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 8.98876404494382e-07,
+      "loss": 0.2646,
+      "step": 132070
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 8.958999925589702e-07,
+      "loss": 0.2665,
+      "step": 132080
+    },
+    {
+      "epoch": 29.33,
+      "learning_rate": 8.929235806235584e-07,
+      "loss": 0.2682,
+      "step": 132090
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 8.899471686881466e-07,
+      "loss": 0.2754,
+      "step": 132100
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 8.869707567527347e-07,
+      "loss": 0.2807,
+      "step": 132110
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 8.839943448173229e-07,
+      "loss": 0.259,
+      "step": 132120
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 8.810179328819109e-07,
+      "loss": 0.2675,
+      "step": 132130
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 8.78041520946499e-07,
+      "loss": 0.2613,
+      "step": 132140
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 8.750651090110872e-07,
+      "loss": 0.2563,
+      "step": 132150
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 8.720886970756754e-07,
+      "loss": 0.2801,
+      "step": 132160
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 8.691122851402635e-07,
+      "loss": 0.2687,
+      "step": 132170
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 8.661358732048517e-07,
+      "loss": 0.2676,
+      "step": 132180
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 8.631594612694398e-07,
+      "loss": 0.2658,
+      "step": 132190
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 8.601830493340279e-07,
+      "loss": 0.2734,
+      "step": 132200
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 8.57206637398616e-07,
+      "loss": 0.2704,
+      "step": 132210
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 8.542302254632042e-07,
+      "loss": 0.2626,
+      "step": 132220
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 8.512538135277923e-07,
+      "loss": 0.2604,
+      "step": 132230
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 8.482774015923804e-07,
+      "loss": 0.268,
+      "step": 132240
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 8.453009896569686e-07,
+      "loss": 0.2793,
+      "step": 132250
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 8.423245777215567e-07,
+      "loss": 0.256,
+      "step": 132260
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 8.393481657861449e-07,
+      "loss": 0.2596,
+      "step": 132270
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 8.363717538507331e-07,
+      "loss": 0.2655,
+      "step": 132280
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 8.333953419153211e-07,
+      "loss": 0.2584,
+      "step": 132290
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 8.304189299799092e-07,
+      "loss": 0.2592,
+      "step": 132300
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 8.274425180444974e-07,
+      "loss": 0.267,
+      "step": 132310
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 8.244661061090855e-07,
+      "loss": 0.2663,
+      "step": 132320
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 8.214896941736737e-07,
+      "loss": 0.2752,
+      "step": 132330
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 8.185132822382619e-07,
+      "loss": 0.2762,
+      "step": 132340
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 8.1553687030285e-07,
+      "loss": 0.2792,
+      "step": 132350
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 8.125604583674382e-07,
+      "loss": 0.2662,
+      "step": 132360
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 8.095840464320262e-07,
+      "loss": 0.269,
+      "step": 132370
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 8.066076344966143e-07,
+      "loss": 0.2572,
+      "step": 132380
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 8.036312225612025e-07,
+      "loss": 0.2717,
+      "step": 132390
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 8.006548106257907e-07,
+      "loss": 0.2655,
+      "step": 132400
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 7.976783986903789e-07,
+      "loss": 0.2691,
+      "step": 132410
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 7.94701986754967e-07,
+      "loss": 0.2757,
+      "step": 132420
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 7.917255748195552e-07,
+      "loss": 0.2643,
+      "step": 132430
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 7.887491628841433e-07,
+      "loss": 0.2628,
+      "step": 132440
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 7.857727509487313e-07,
+      "loss": 0.2669,
+      "step": 132450
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 7.827963390133195e-07,
+      "loss": 0.2538,
+      "step": 132460
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 7.798199270779077e-07,
+      "loss": 0.2782,
+      "step": 132470
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 7.768435151424958e-07,
+      "loss": 0.2754,
+      "step": 132480
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 7.738671032070839e-07,
+      "loss": 0.2713,
+      "step": 132490
+    },
+    {
+      "epoch": 29.42,
+      "learning_rate": 7.708906912716721e-07,
+      "loss": 0.2601,
+      "step": 132500
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.679142793362602e-07,
+      "loss": 0.2597,
+      "step": 132510
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.649378674008483e-07,
+      "loss": 0.2686,
+      "step": 132520
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.619614554654365e-07,
+      "loss": 0.2531,
+      "step": 132530
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.589850435300246e-07,
+      "loss": 0.2567,
+      "step": 132540
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 7.560086315946127e-07,
+      "loss": 0.2654,
+      "step": 132550
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 7.530322196592009e-07,
+      "loss": 0.2599,
+      "step": 132560
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 7.50055807723789e-07,
+      "loss": 0.2714,
+      "step": 132570
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 7.470793957883772e-07,
+      "loss": 0.277,
+      "step": 132580
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 7.441029838529654e-07,
+      "loss": 0.2788,
+      "step": 132590
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 7.411265719175533e-07,
+      "loss": 0.265,
+      "step": 132600
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 7.381501599821415e-07,
+      "loss": 0.2798,
+      "step": 132610
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 7.351737480467297e-07,
+      "loss": 0.2546,
+      "step": 132620
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 7.321973361113178e-07,
+      "loss": 0.2607,
+      "step": 132630
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.29220924175906e-07,
+      "loss": 0.2585,
+      "step": 132640
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.262445122404942e-07,
+      "loss": 0.269,
+      "step": 132650
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.232681003050823e-07,
+      "loss": 0.2708,
+      "step": 132660
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.202916883696705e-07,
+      "loss": 0.2762,
+      "step": 132670
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 7.173152764342585e-07,
+      "loss": 0.2643,
+      "step": 132680
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 7.143388644988466e-07,
+      "loss": 0.2701,
+      "step": 132690
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 7.113624525634348e-07,
+      "loss": 0.2673,
+      "step": 132700
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 7.08386040628023e-07,
+      "loss": 0.2658,
+      "step": 132710
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 7.054096286926112e-07,
+      "loss": 0.2661,
+      "step": 132720
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 7.024332167571993e-07,
+      "loss": 0.2623,
+      "step": 132730
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 6.994568048217874e-07,
+      "loss": 0.2689,
+      "step": 132740
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 6.964803928863756e-07,
+      "loss": 0.2539,
+      "step": 132750
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 6.935039809509636e-07,
+      "loss": 0.2506,
+      "step": 132760
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 6.905275690155518e-07,
+      "loss": 0.2782,
+      "step": 132770
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 6.8755115708014e-07,
+      "loss": 0.2667,
+      "step": 132780
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 6.845747451447281e-07,
+      "loss": 0.2628,
+      "step": 132790
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 6.815983332093162e-07,
+      "loss": 0.2591,
+      "step": 132800
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 6.786219212739044e-07,
+      "loss": 0.2509,
+      "step": 132810
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 6.756455093384925e-07,
+      "loss": 0.2723,
+      "step": 132820
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 6.726690974030807e-07,
+      "loss": 0.2817,
+      "step": 132830
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 6.696926854676688e-07,
+      "loss": 0.2557,
+      "step": 132840
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 6.667162735322568e-07,
+      "loss": 0.2724,
+      "step": 132850
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 6.63739861596845e-07,
+      "loss": 0.2661,
+      "step": 132860
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 6.607634496614332e-07,
+      "loss": 0.2571,
+      "step": 132870
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 6.577870377260213e-07,
+      "loss": 0.2749,
+      "step": 132880
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 6.548106257906095e-07,
+      "loss": 0.2783,
+      "step": 132890
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 6.518342138551977e-07,
+      "loss": 0.2694,
+      "step": 132900
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 6.488578019197858e-07,
+      "loss": 0.2729,
+      "step": 132910
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 6.458813899843738e-07,
+      "loss": 0.2657,
+      "step": 132920
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 6.42904978048962e-07,
+      "loss": 0.2777,
+      "step": 132930
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 6.399285661135501e-07,
+      "loss": 0.2574,
+      "step": 132940
+    },
+    {
+      "epoch": 29.52,
+      "learning_rate": 6.369521541781383e-07,
+      "loss": 0.2894,
+      "step": 132950
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 6.339757422427265e-07,
+      "loss": 0.2797,
+      "step": 132960
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 6.309993303073146e-07,
+      "loss": 0.2697,
+      "step": 132970
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 6.280229183719028e-07,
+      "loss": 0.2672,
+      "step": 132980
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 6.250465064364909e-07,
+      "loss": 0.262,
+      "step": 132990
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 6.22070094501079e-07,
+      "loss": 0.2621,
+      "step": 133000
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 6.190936825656672e-07,
+      "loss": 0.2606,
+      "step": 133010
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 6.161172706302553e-07,
+      "loss": 0.2748,
+      "step": 133020
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 6.131408586948435e-07,
+      "loss": 0.2595,
+      "step": 133030
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 6.101644467594315e-07,
+      "loss": 0.2596,
+      "step": 133040
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 6.071880348240197e-07,
+      "loss": 0.2645,
+      "step": 133050
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 6.042116228886079e-07,
+      "loss": 0.2604,
+      "step": 133060
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 6.012352109531959e-07,
+      "loss": 0.2633,
+      "step": 133070
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 5.982587990177841e-07,
+      "loss": 0.2518,
+      "step": 133080
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 5.952823870823723e-07,
+      "loss": 0.2678,
+      "step": 133090
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 5.923059751469603e-07,
+      "loss": 0.2731,
+      "step": 133100
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 5.893295632115485e-07,
+      "loss": 0.2626,
+      "step": 133110
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 5.863531512761367e-07,
+      "loss": 0.2553,
+      "step": 133120
+    },
+    {
+      "epoch": 29.56,
+      "learning_rate": 5.833767393407248e-07,
+      "loss": 0.2678,
+      "step": 133130
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 5.804003274053129e-07,
+      "loss": 0.2712,
+      "step": 133140
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 5.774239154699011e-07,
+      "loss": 0.2832,
+      "step": 133150
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 5.744475035344892e-07,
+      "loss": 0.2467,
+      "step": 133160
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 5.714710915990774e-07,
+      "loss": 0.263,
+      "step": 133170
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 5.684946796636655e-07,
+      "loss": 0.2597,
+      "step": 133180
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 5.655182677282536e-07,
+      "loss": 0.269,
+      "step": 133190
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 5.625418557928418e-07,
+      "loss": 0.253,
+      "step": 133200
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 5.5956544385743e-07,
+      "loss": 0.268,
+      "step": 133210
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 5.56589031922018e-07,
+      "loss": 0.2786,
+      "step": 133220
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 5.536126199866062e-07,
+      "loss": 0.279,
+      "step": 133230
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 5.506362080511944e-07,
+      "loss": 0.2622,
+      "step": 133240
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 5.476597961157825e-07,
+      "loss": 0.2715,
+      "step": 133250
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 5.446833841803706e-07,
+      "loss": 0.2765,
+      "step": 133260
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 5.417069722449588e-07,
+      "loss": 0.2618,
+      "step": 133270
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 5.387305603095469e-07,
+      "loss": 0.2734,
+      "step": 133280
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 5.35754148374135e-07,
+      "loss": 0.2675,
+      "step": 133290
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 5.327777364387232e-07,
+      "loss": 0.2771,
+      "step": 133300
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 5.298013245033113e-07,
+      "loss": 0.2688,
+      "step": 133310
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 5.268249125678994e-07,
+      "loss": 0.2728,
+      "step": 133320
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 5.238485006324876e-07,
+      "loss": 0.2743,
+      "step": 133330
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 5.208720886970758e-07,
+      "loss": 0.2818,
+      "step": 133340
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 5.178956767616638e-07,
+      "loss": 0.2709,
+      "step": 133350
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 5.14919264826252e-07,
+      "loss": 0.2704,
+      "step": 133360
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 5.1194285289084e-07,
+      "loss": 0.2673,
+      "step": 133370
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 5.089664409554282e-07,
+      "loss": 0.2746,
+      "step": 133380
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 5.059900290200164e-07,
+      "loss": 0.2722,
+      "step": 133390
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 5.030136170846046e-07,
+      "loss": 0.2739,
+      "step": 133400
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 5.000372051491926e-07,
+      "loss": 0.2599,
+      "step": 133410
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.970607932137808e-07,
+      "loss": 0.2681,
+      "step": 133420
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.94084381278369e-07,
+      "loss": 0.2763,
+      "step": 133430
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 4.911079693429571e-07,
+      "loss": 0.2637,
+      "step": 133440
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.881315574075452e-07,
+      "loss": 0.2695,
+      "step": 133450
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.851551454721334e-07,
+      "loss": 0.2654,
+      "step": 133460
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.821787335367215e-07,
+      "loss": 0.2666,
+      "step": 133470
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.792023216013097e-07,
+      "loss": 0.2685,
+      "step": 133480
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 4.7622590966589776e-07,
+      "loss": 0.2691,
+      "step": 133490
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.7324949773048593e-07,
+      "loss": 0.2583,
+      "step": 133500
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.702730857950741e-07,
+      "loss": 0.2686,
+      "step": 133510
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.672966738596622e-07,
+      "loss": 0.2546,
+      "step": 133520
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 4.6432026192425034e-07,
+      "loss": 0.2672,
+      "step": 133530
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.613438499888385e-07,
+      "loss": 0.2597,
+      "step": 133540
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.583674380534266e-07,
+      "loss": 0.266,
+      "step": 133550
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.553910261180148e-07,
+      "loss": 0.2615,
+      "step": 133560
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.5241461418260285e-07,
+      "loss": 0.2723,
+      "step": 133570
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 4.49438202247191e-07,
+      "loss": 0.2753,
+      "step": 133580
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.464617903117792e-07,
+      "loss": 0.2592,
+      "step": 133590
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.4348537837636736e-07,
+      "loss": 0.2459,
+      "step": 133600
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.405089664409554e-07,
+      "loss": 0.28,
+      "step": 133610
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 4.375325545055436e-07,
+      "loss": 0.266,
+      "step": 133620
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.3455614257013176e-07,
+      "loss": 0.2748,
+      "step": 133630
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.315797306347199e-07,
+      "loss": 0.2617,
+      "step": 133640
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.28603318699308e-07,
+      "loss": 0.2521,
+      "step": 133650
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.2562690676389617e-07,
+      "loss": 0.2816,
+      "step": 133660
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 4.226504948284843e-07,
+      "loss": 0.2711,
+      "step": 133670
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.1967408289307245e-07,
+      "loss": 0.2843,
+      "step": 133680
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.1669767095766057e-07,
+      "loss": 0.2583,
+      "step": 133690
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.137212590222487e-07,
+      "loss": 0.2704,
+      "step": 133700
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 4.1074484708683686e-07,
+      "loss": 0.2591,
+      "step": 133710
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 4.07768435151425e-07,
+      "loss": 0.2588,
+      "step": 133720
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 4.047920232160131e-07,
+      "loss": 0.2609,
+      "step": 133730
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 4.0181561128060126e-07,
+      "loss": 0.26,
+      "step": 133740
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.9883919934518943e-07,
+      "loss": 0.259,
+      "step": 133750
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.958627874097776e-07,
+      "loss": 0.2736,
+      "step": 133760
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.9288637547436566e-07,
+      "loss": 0.2721,
+      "step": 133770
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.8990996353895383e-07,
+      "loss": 0.2635,
+      "step": 133780
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.8693355160354195e-07,
+      "loss": 0.2699,
+      "step": 133790
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.839571396681301e-07,
+      "loss": 0.2877,
+      "step": 133800
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.8098072773271823e-07,
+      "loss": 0.2623,
+      "step": 133810
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.7800431579730635e-07,
+      "loss": 0.2625,
+      "step": 133820
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.750279038618945e-07,
+      "loss": 0.2735,
+      "step": 133830
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.720514919264827e-07,
+      "loss": 0.2751,
+      "step": 133840
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 3.6907507999107075e-07,
+      "loss": 0.2777,
+      "step": 133850
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.660986680556589e-07,
+      "loss": 0.2794,
+      "step": 133860
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.631222561202471e-07,
+      "loss": 0.2597,
+      "step": 133870
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.6014584418483526e-07,
+      "loss": 0.2622,
+      "step": 133880
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 3.571694322494233e-07,
+      "loss": 0.2792,
+      "step": 133890
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.541930203140115e-07,
+      "loss": 0.2701,
+      "step": 133900
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.5121660837859966e-07,
+      "loss": 0.2532,
+      "step": 133910
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.482401964431878e-07,
+      "loss": 0.2729,
+      "step": 133920
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.452637845077759e-07,
+      "loss": 0.2638,
+      "step": 133930
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 3.4228737257236406e-07,
+      "loss": 0.2643,
+      "step": 133940
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.393109606369522e-07,
+      "loss": 0.2771,
+      "step": 133950
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.3633454870154035e-07,
+      "loss": 0.2585,
+      "step": 133960
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.333581367661284e-07,
+      "loss": 0.265,
+      "step": 133970
+    },
+    {
+      "epoch": 29.75,
+      "learning_rate": 3.303817248307166e-07,
+      "loss": 0.2762,
+      "step": 133980
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.2740531289530475e-07,
+      "loss": 0.2641,
+      "step": 133990
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.244289009598929e-07,
+      "loss": 0.2764,
+      "step": 134000
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.21452489024481e-07,
+      "loss": 0.2699,
+      "step": 134010
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.1847607708906915e-07,
+      "loss": 0.2615,
+      "step": 134020
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 3.154996651536573e-07,
+      "loss": 0.2836,
+      "step": 134030
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.1252325321824544e-07,
+      "loss": 0.2693,
+      "step": 134040
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.095468412828336e-07,
+      "loss": 0.271,
+      "step": 134050
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.065704293474217e-07,
+      "loss": 0.2651,
+      "step": 134060
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 3.0359401741200984e-07,
+      "loss": 0.2767,
+      "step": 134070
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 3.0061760547659796e-07,
+      "loss": 0.2741,
+      "step": 134080
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 2.9764119354118613e-07,
+      "loss": 0.2674,
+      "step": 134090
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 2.9466478160577425e-07,
+      "loss": 0.2706,
+      "step": 134100
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 2.916883696703624e-07,
+      "loss": 0.2761,
+      "step": 134110
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 2.8871195773495053e-07,
+      "loss": 0.2607,
+      "step": 134120
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2.857355457995387e-07,
+      "loss": 0.2641,
+      "step": 134130
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2.827591338641268e-07,
+      "loss": 0.2691,
+      "step": 134140
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2.79782721928715e-07,
+      "loss": 0.2684,
+      "step": 134150
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 2.768063099933031e-07,
+      "loss": 0.2749,
+      "step": 134160
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.7382989805789127e-07,
+      "loss": 0.2826,
+      "step": 134170
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.708534861224794e-07,
+      "loss": 0.2662,
+      "step": 134180
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.678770741870675e-07,
+      "loss": 0.2733,
+      "step": 134190
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.649006622516557e-07,
+      "loss": 0.271,
+      "step": 134200
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.619242503162438e-07,
+      "loss": 0.2756,
+      "step": 134210
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.589478383808319e-07,
+      "loss": 0.2623,
+      "step": 134220
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.5597142644542e-07,
+      "loss": 0.269,
+      "step": 134230
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.529950145100082e-07,
+      "loss": 0.2687,
+      "step": 134240
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.500186025745963e-07,
+      "loss": 0.2626,
+      "step": 134250
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 2.470421906391845e-07,
+      "loss": 0.2609,
+      "step": 134260
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 2.440657787037726e-07,
+      "loss": 0.2618,
+      "step": 134270
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 2.4108936676836077e-07,
+      "loss": 0.2658,
+      "step": 134280
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 2.3811295483294888e-07,
+      "loss": 0.2624,
+      "step": 134290
+    },
+    {
+      "epoch": 29.82,
+      "learning_rate": 2.3513654289753705e-07,
+      "loss": 0.2617,
+      "step": 134300
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 2.3216013096212517e-07,
+      "loss": 0.283,
+      "step": 134310
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 2.291837190267133e-07,
+      "loss": 0.271,
+      "step": 134320
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 2.2620730709130143e-07,
+      "loss": 0.2701,
+      "step": 134330
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 2.232308951558896e-07,
+      "loss": 0.2578,
+      "step": 134340
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2.202544832204777e-07,
+      "loss": 0.2774,
+      "step": 134350
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2.1727807128506588e-07,
+      "loss": 0.2603,
+      "step": 134360
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2.14301659349654e-07,
+      "loss": 0.2647,
+      "step": 134370
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2.1132524741424214e-07,
+      "loss": 0.2763,
+      "step": 134380
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 2.0834883547883028e-07,
+      "loss": 0.2822,
+      "step": 134390
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 2.0537242354341843e-07,
+      "loss": 0.2645,
+      "step": 134400
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 2.0239601160800654e-07,
+      "loss": 0.2621,
+      "step": 134410
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.9941959967259471e-07,
+      "loss": 0.2647,
+      "step": 134420
+    },
+    {
+      "epoch": 29.85,
+      "learning_rate": 1.9644318773718283e-07,
+      "loss": 0.2649,
+      "step": 134430
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.9346677580177097e-07,
+      "loss": 0.2886,
+      "step": 134440
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.9049036386635912e-07,
+      "loss": 0.2825,
+      "step": 134450
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.8751395193094726e-07,
+      "loss": 0.2621,
+      "step": 134460
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.8453753999553538e-07,
+      "loss": 0.2762,
+      "step": 134470
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.8156112806012354e-07,
+      "loss": 0.2685,
+      "step": 134480
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.7858471612471166e-07,
+      "loss": 0.2828,
+      "step": 134490
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.7560830418929983e-07,
+      "loss": 0.2628,
+      "step": 134500
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.7263189225388795e-07,
+      "loss": 0.278,
+      "step": 134510
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.696554803184761e-07,
+      "loss": 0.2682,
+      "step": 134520
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.666790683830642e-07,
+      "loss": 0.2847,
+      "step": 134530
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.6370265644765238e-07,
+      "loss": 0.2524,
+      "step": 134540
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.607262445122405e-07,
+      "loss": 0.2742,
+      "step": 134550
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.5774983257682866e-07,
+      "loss": 0.275,
+      "step": 134560
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 1.547734206414168e-07,
+      "loss": 0.2704,
+      "step": 134570
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.5179700870600492e-07,
+      "loss": 0.2739,
+      "step": 134580
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.4882059677059306e-07,
+      "loss": 0.259,
+      "step": 134590
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.458441848351812e-07,
+      "loss": 0.2687,
+      "step": 134600
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.4286777289976935e-07,
+      "loss": 0.2701,
+      "step": 134610
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.398913609643575e-07,
+      "loss": 0.2667,
+      "step": 134620
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.3691494902894564e-07,
+      "loss": 0.2689,
+      "step": 134630
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.3393853709353375e-07,
+      "loss": 0.2657,
+      "step": 134640
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.309621251581219e-07,
+      "loss": 0.2643,
+      "step": 134650
+    },
+    {
+      "epoch": 29.9,
+      "learning_rate": 1.2798571322271e-07,
+      "loss": 0.2744,
+      "step": 134660
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.2500930128729815e-07,
+      "loss": 0.2629,
+      "step": 134670
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.220328893518863e-07,
+      "loss": 0.2722,
+      "step": 134680
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.1905647741647444e-07,
+      "loss": 0.2733,
+      "step": 134690
+    },
+    {
+      "epoch": 29.91,
+      "learning_rate": 1.1608006548106258e-07,
+      "loss": 0.2556,
+      "step": 134700
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.1310365354565071e-07,
+      "loss": 0.2631,
+      "step": 134710
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.1012724161023886e-07,
+      "loss": 0.2722,
+      "step": 134720
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.07150829674827e-07,
+      "loss": 0.2723,
+      "step": 134730
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.0417441773941514e-07,
+      "loss": 0.2704,
+      "step": 134740
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 1.0119800580400327e-07,
+      "loss": 0.2634,
+      "step": 134750
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 9.822159386859142e-08,
+      "loss": 0.2601,
+      "step": 134760
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 9.524518193317956e-08,
+      "loss": 0.2578,
+      "step": 134770
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 9.226876999776769e-08,
+      "loss": 0.2518,
+      "step": 134780
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 8.929235806235583e-08,
+      "loss": 0.2615,
+      "step": 134790
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 8.631594612694397e-08,
+      "loss": 0.2606,
+      "step": 134800
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 8.33395341915321e-08,
+      "loss": 0.271,
+      "step": 134810
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 8.036312225612025e-08,
+      "loss": 0.2778,
+      "step": 134820
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 7.73867103207084e-08,
+      "loss": 0.2696,
+      "step": 134830
+    },
+    {
+      "epoch": 29.94,
+      "learning_rate": 7.441029838529653e-08,
+      "loss": 0.2659,
+      "step": 134840
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 7.143388644988468e-08,
+      "loss": 0.2658,
+      "step": 134850
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 6.845747451447282e-08,
+      "loss": 0.2681,
+      "step": 134860
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 6.548106257906095e-08,
+      "loss": 0.2625,
+      "step": 134870
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 6.250465064364908e-08,
+      "loss": 0.2682,
+      "step": 134880
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 5.952823870823722e-08,
+      "loss": 0.2677,
+      "step": 134890
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 5.655182677282536e-08,
+      "loss": 0.2862,
+      "step": 134900
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 5.35754148374135e-08,
+      "loss": 0.2691,
+      "step": 134910
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 5.0599002902001636e-08,
+      "loss": 0.2654,
+      "step": 134920
+    },
+    {
+      "epoch": 29.96,
+      "learning_rate": 4.762259096658978e-08,
+      "loss": 0.2752,
+      "step": 134930
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 4.4646179031177915e-08,
+      "loss": 0.2585,
+      "step": 134940
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 4.166976709576605e-08,
+      "loss": 0.2584,
+      "step": 134950
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.86933551603542e-08,
+      "loss": 0.291,
+      "step": 134960
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 3.571694322494234e-08,
+      "loss": 0.2555,
+      "step": 134970
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 3.2740531289530474e-08,
+      "loss": 0.2705,
+      "step": 134980
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 2.976411935411861e-08,
+      "loss": 0.2726,
+      "step": 134990
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 2.678770741870675e-08,
+      "loss": 0.2758,
+      "step": 135000
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 2.381129548329489e-08,
+      "loss": 0.2649,
+      "step": 135010
+    },
+    {
+      "epoch": 29.98,
+      "learning_rate": 2.0834883547883026e-08,
+      "loss": 0.2719,
+      "step": 135020
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.785847161247117e-08,
+      "loss": 0.2711,
+      "step": 135030
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.4882059677059305e-08,
+      "loss": 0.2688,
+      "step": 135040
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 1.1905647741647445e-08,
+      "loss": 0.2592,
+      "step": 135050
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 8.929235806235584e-09,
+      "loss": 0.2573,
+      "step": 135060
+    },
+    {
+      "epoch": 29.99,
+      "eval_cer": 4.534188034188034,
+      "eval_loss": 0.5672242641448975,
+      "eval_runtime": 1.2676,
+      "eval_samples_per_second": 7.889,
+      "eval_steps_per_second": 0.789,
+      "eval_wer": 0.5974025974025974,
+      "step": 135060
+    }
+  ],
+  "max_steps": 135090,
+  "num_train_epochs": 30,
+  "total_flos": 3.9846332642287524e+21,
+  "trial_name": null,
+  "trial_params": null
+}