diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,34225 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.994739610731195,
+  "eval_steps": 500,
+  "global_step": 57000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99982465369104e-06,
+      "loss": 8.0177,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.998246536910398e-06,
+      "loss": 7.7882,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.996493073820796e-06,
+      "loss": 7.2987,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.994739610731195e-06,
+      "loss": 7.1048,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.992986147641594e-06,
+      "loss": 6.7477,
+      "step": 40
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.991232684551992e-06,
+      "loss": 6.4061,
+      "step": 50
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.989479221462389e-06,
+      "loss": 6.2727,
+      "step": 60
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.987725758372787e-06,
+      "loss": 5.9566,
+      "step": 70
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985972295283186e-06,
+      "loss": 5.5651,
+      "step": 80
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.984218832193583e-06,
+      "loss": 5.3653,
+      "step": 90
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98246536910398e-06,
+      "loss": 5.0674,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.980711906014378e-06,
+      "loss": 4.8475,
+      "step": 110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.978958442924777e-06,
+      "loss": 4.7907,
+      "step": 120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.977204979835175e-06,
+      "loss": 4.6305,
+      "step": 130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.975451516745574e-06,
+      "loss": 4.4261,
+      "step": 140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973698053655971e-06,
+      "loss": 4.2855,
+      "step": 150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.971944590566369e-06,
+      "loss": 4.232,
+      "step": 160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.970191127476766e-06,
+      "loss": 3.8801,
+      "step": 170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.968437664387166e-06,
+      "loss": 3.9224,
+      "step": 180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.966684201297565e-06,
+      "loss": 3.9685,
+      "step": 190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964930738207962e-06,
+      "loss": 3.7678,
+      "step": 200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96317727511836e-06,
+      "loss": 3.8441,
+      "step": 210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961423812028757e-06,
+      "loss": 3.5948,
+      "step": 220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959670348939155e-06,
+      "loss": 3.5949,
+      "step": 230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957916885849554e-06,
+      "loss": 3.7051,
+      "step": 240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956163422759951e-06,
+      "loss": 3.5148,
+      "step": 250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.95440995967035e-06,
+      "loss": 3.6728,
+      "step": 260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952656496580748e-06,
+      "loss": 3.4876,
+      "step": 270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950903033491145e-06,
+      "loss": 3.5185,
+      "step": 280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949149570401545e-06,
+      "loss": 3.9175,
+      "step": 290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947396107311942e-06,
+      "loss": 3.5419,
+      "step": 300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94564264422234e-06,
+      "loss": 3.4683,
+      "step": 310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.943889181132737e-06,
+      "loss": 3.2562,
+      "step": 320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.942135718043136e-06,
+      "loss": 3.5485,
+      "step": 330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.940382254953534e-06,
+      "loss": 3.5317,
+      "step": 340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.938628791863933e-06,
+      "loss": 3.4571,
+      "step": 350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.93687532877433e-06,
+      "loss": 3.4896,
+      "step": 360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.935121865684728e-06,
+      "loss": 3.4214,
+      "step": 370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.933368402595125e-06,
+      "loss": 3.2299,
+      "step": 380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.931614939505525e-06,
+      "loss": 3.311,
+      "step": 390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.929861476415922e-06,
+      "loss": 3.3407,
+      "step": 400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.928108013326321e-06,
+      "loss": 3.2792,
+      "step": 410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.926354550236719e-06,
+      "loss": 3.1829,
+      "step": 420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.924601087147116e-06,
+      "loss": 3.1426,
+      "step": 430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.922847624057515e-06,
+      "loss": 3.3389,
+      "step": 440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.921094160967913e-06,
+      "loss": 3.2891,
+      "step": 450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.91934069787831e-06,
+      "loss": 3.248,
+      "step": 460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.917587234788708e-06,
+      "loss": 3.2948,
+      "step": 470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.915833771699105e-06,
+      "loss": 3.3955,
+      "step": 480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.914080308609504e-06,
+      "loss": 3.5032,
+      "step": 490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.912326845519904e-06,
+      "loss": 3.2219,
+      "step": 500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.910573382430301e-06,
+      "loss": 3.1888,
+      "step": 510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.908819919340699e-06,
+      "loss": 3.1657,
+      "step": 520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.907066456251096e-06,
+      "loss": 2.9912,
+      "step": 530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.905312993161495e-06,
+      "loss": 3.1808,
+      "step": 540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.903559530071893e-06,
+      "loss": 3.1986,
+      "step": 550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.901806066982292e-06,
+      "loss": 3.2659,
+      "step": 560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.90005260389269e-06,
+      "loss": 3.1635,
+      "step": 570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.898299140803087e-06,
+      "loss": 3.26,
+      "step": 580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.896545677713484e-06,
+      "loss": 3.1953,
+      "step": 590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.894792214623883e-06,
+      "loss": 3.2359,
+      "step": 600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.893038751534281e-06,
+      "loss": 3.1948,
+      "step": 610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.891285288444678e-06,
+      "loss": 3.1147,
+      "step": 620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.889531825355078e-06,
+      "loss": 2.9777,
+      "step": 630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.887778362265475e-06,
+      "loss": 3.2594,
+      "step": 640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.886024899175874e-06,
+      "loss": 3.0564,
+      "step": 650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.884271436086272e-06,
+      "loss": 3.0078,
+      "step": 660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.88251797299667e-06,
+      "loss": 3.0686,
+      "step": 670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.880764509907067e-06,
+      "loss": 3.085,
+      "step": 680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.879011046817464e-06,
+      "loss": 3.1512,
+      "step": 690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.877257583727863e-06,
+      "loss": 3.207,
+      "step": 700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.875504120638262e-06,
+      "loss": 3.0796,
+      "step": 710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.87375065754866e-06,
+      "loss": 2.9466,
+      "step": 720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.871997194459057e-06,
+      "loss": 3.2069,
+      "step": 730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.870243731369455e-06,
+      "loss": 3.0494,
+      "step": 740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.868490268279854e-06,
+      "loss": 2.8207,
+      "step": 750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.866736805190252e-06,
+      "loss": 2.9201,
+      "step": 760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.864983342100649e-06,
+      "loss": 3.1829,
+      "step": 770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.863229879011048e-06,
+      "loss": 3.0087,
+      "step": 780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.861476415921446e-06,
+      "loss": 2.9208,
+      "step": 790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.859722952831843e-06,
+      "loss": 3.11,
+      "step": 800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.857969489742242e-06,
+      "loss": 3.0263,
+      "step": 810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.85621602665264e-06,
+      "loss": 3.0204,
+      "step": 820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.854462563563037e-06,
+      "loss": 2.8355,
+      "step": 830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.852709100473435e-06,
+      "loss": 3.2564,
+      "step": 840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.850955637383834e-06,
+      "loss": 2.7172,
+      "step": 850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.849202174294233e-06,
+      "loss": 3.1269,
+      "step": 860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.84744871120463e-06,
+      "loss": 2.8452,
+      "step": 870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.845695248115028e-06,
+      "loss": 3.0133,
+      "step": 880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.843941785025426e-06,
+      "loss": 2.7922,
+      "step": 890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.842188321935823e-06,
+      "loss": 3.0022,
+      "step": 900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.840434858846222e-06,
+      "loss": 3.0219,
+      "step": 910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.83868139575662e-06,
+      "loss": 2.7571,
+      "step": 920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.836927932667019e-06,
+      "loss": 3.0023,
+      "step": 930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.835174469577416e-06,
+      "loss": 2.7702,
+      "step": 940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.833421006487814e-06,
+      "loss": 2.8619,
+      "step": 950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.831667543398213e-06,
+      "loss": 2.9508,
+      "step": 960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.82991408030861e-06,
+      "loss": 2.8996,
+      "step": 970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.828160617219008e-06,
+      "loss": 2.748,
+      "step": 980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.826407154129405e-06,
+      "loss": 2.8687,
+      "step": 990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.824653691039805e-06,
+      "loss": 2.7547,
+      "step": 1000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.822900227950202e-06,
+      "loss": 2.9394,
+      "step": 1010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.821146764860601e-06,
+      "loss": 3.0444,
+      "step": 1020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.819393301770999e-06,
+      "loss": 2.7471,
+      "step": 1030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.817639838681396e-06,
+      "loss": 2.8771,
+      "step": 1040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.815886375591794e-06,
+      "loss": 2.8859,
+      "step": 1050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.814132912502193e-06,
+      "loss": 2.7986,
+      "step": 1060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.81237944941259e-06,
+      "loss": 2.792,
+      "step": 1070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.81062598632299e-06,
+      "loss": 2.9498,
+      "step": 1080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.808872523233387e-06,
+      "loss": 2.7372,
+      "step": 1090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.807119060143784e-06,
+      "loss": 2.7424,
+      "step": 1100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.805365597054182e-06,
+      "loss": 2.7027,
+      "step": 1110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.803612133964581e-06,
+      "loss": 2.6847,
+      "step": 1120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.801858670874979e-06,
+      "loss": 2.7594,
+      "step": 1130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.800105207785376e-06,
+      "loss": 2.7884,
+      "step": 1140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.798351744695775e-06,
+      "loss": 2.8724,
+      "step": 1150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.796598281606173e-06,
+      "loss": 2.944,
+      "step": 1160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.794844818516572e-06,
+      "loss": 2.8604,
+      "step": 1170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.79309135542697e-06,
+      "loss": 2.7796,
+      "step": 1180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.791337892337367e-06,
+      "loss": 2.7401,
+      "step": 1190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.789584429247764e-06,
+      "loss": 2.8694,
+      "step": 1200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.787830966158164e-06,
+      "loss": 2.8161,
+      "step": 1210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.786077503068561e-06,
+      "loss": 2.7323,
+      "step": 1220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.78432403997896e-06,
+      "loss": 2.7068,
+      "step": 1230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.782570576889358e-06,
+      "loss": 2.9801,
+      "step": 1240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.780817113799755e-06,
+      "loss": 2.7526,
+      "step": 1250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.779063650710153e-06,
+      "loss": 2.8807,
+      "step": 1260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.777310187620552e-06,
+      "loss": 2.6007,
+      "step": 1270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.77555672453095e-06,
+      "loss": 2.6662,
+      "step": 1280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.773803261441347e-06,
+      "loss": 2.7437,
+      "step": 1290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.772049798351746e-06,
+      "loss": 2.8427,
+      "step": 1300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.770296335262143e-06,
+      "loss": 2.8134,
+      "step": 1310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.768542872172543e-06,
+      "loss": 3.3729,
+      "step": 1320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.76678940908294e-06,
+      "loss": 2.6442,
+      "step": 1330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.765035945993338e-06,
+      "loss": 2.7481,
+      "step": 1340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.763282482903735e-06,
+      "loss": 2.6837,
+      "step": 1350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.761529019814132e-06,
+      "loss": 2.6141,
+      "step": 1360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.759775556724532e-06,
+      "loss": 2.7657,
+      "step": 1370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.75802209363493e-06,
+      "loss": 2.8052,
+      "step": 1380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.756268630545328e-06,
+      "loss": 2.6371,
+      "step": 1390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.754515167455726e-06,
+      "loss": 2.5978,
+      "step": 1400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.752761704366123e-06,
+      "loss": 2.8843,
+      "step": 1410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.751008241276522e-06,
+      "loss": 2.7276,
+      "step": 1420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.74925477818692e-06,
+      "loss": 2.7263,
+      "step": 1430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.747501315097317e-06,
+      "loss": 2.8338,
+      "step": 1440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.745747852007717e-06,
+      "loss": 2.6814,
+      "step": 1450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.743994388918114e-06,
+      "loss": 2.6611,
+      "step": 1460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.742240925828512e-06,
+      "loss": 2.7185,
+      "step": 1470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.74048746273891e-06,
+      "loss": 2.6073,
+      "step": 1480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.738733999649308e-06,
+      "loss": 2.8844,
+      "step": 1490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.736980536559706e-06,
+      "loss": 2.9444,
+      "step": 1500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.735227073470103e-06,
+      "loss": 2.7275,
+      "step": 1510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.733473610380502e-06,
+      "loss": 2.6465,
+      "step": 1520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.731720147290901e-06,
+      "loss": 2.6635,
+      "step": 1530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.729966684201299e-06,
+      "loss": 2.5966,
+      "step": 1540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.728213221111696e-06,
+      "loss": 2.7412,
+      "step": 1550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.726459758022094e-06,
+      "loss": 2.583,
+      "step": 1560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.724706294932491e-06,
+      "loss": 2.7463,
+      "step": 1570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.72295283184289e-06,
+      "loss": 2.835,
+      "step": 1580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.72119936875329e-06,
+      "loss": 2.5939,
+      "step": 1590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.719445905663687e-06,
+      "loss": 2.7028,
+      "step": 1600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.717692442574085e-06,
+      "loss": 2.6235,
+      "step": 1610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.715938979484482e-06,
+      "loss": 2.6455,
+      "step": 1620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.714185516394881e-06,
+      "loss": 2.6201,
+      "step": 1630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.712432053305279e-06,
+      "loss": 2.6363,
+      "step": 1640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.710678590215676e-06,
+      "loss": 2.5928,
+      "step": 1650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.708925127126074e-06,
+      "loss": 2.4115,
+      "step": 1660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.707171664036473e-06,
+      "loss": 2.7525,
+      "step": 1670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.70541820094687e-06,
+      "loss": 2.8015,
+      "step": 1680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.70366473785727e-06,
+      "loss": 2.6711,
+      "step": 1690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.701911274767667e-06,
+      "loss": 2.717,
+      "step": 1700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.700157811678065e-06,
+      "loss": 2.6318,
+      "step": 1710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.698404348588462e-06,
+      "loss": 2.6789,
+      "step": 1720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.696650885498861e-06,
+      "loss": 2.8015,
+      "step": 1730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.69489742240926e-06,
+      "loss": 2.5995,
+      "step": 1740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.693143959319658e-06,
+      "loss": 2.6995,
+      "step": 1750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.691390496230055e-06,
+      "loss": 2.7188,
+      "step": 1760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.689637033140453e-06,
+      "loss": 2.6059,
+      "step": 1770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.68788357005085e-06,
+      "loss": 2.5704,
+      "step": 1780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.68613010696125e-06,
+      "loss": 2.7837,
+      "step": 1790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.684376643871647e-06,
+      "loss": 2.7151,
+      "step": 1800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.682623180782046e-06,
+      "loss": 2.6826,
+      "step": 1810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.680869717692444e-06,
+      "loss": 2.6582,
+      "step": 1820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.679116254602841e-06,
+      "loss": 2.7197,
+      "step": 1830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.67736279151324e-06,
+      "loss": 2.6871,
+      "step": 1840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.675609328423638e-06,
+      "loss": 2.4124,
+      "step": 1850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.673855865334035e-06,
+      "loss": 2.5675,
+      "step": 1860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.672102402244433e-06,
+      "loss": 2.6552,
+      "step": 1870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.670348939154832e-06,
+      "loss": 2.7088,
+      "step": 1880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.66859547606523e-06,
+      "loss": 2.8753,
+      "step": 1890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.666842012975629e-06,
+      "loss": 2.5528,
+      "step": 1900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.665088549886026e-06,
+      "loss": 2.5209,
+      "step": 1910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.663335086796423e-06,
+      "loss": 2.5817,
+      "step": 1920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.661581623706821e-06,
+      "loss": 2.5113,
+      "step": 1930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.65982816061722e-06,
+      "loss": 2.5738,
+      "step": 1940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.658074697527618e-06,
+      "loss": 2.6545,
+      "step": 1950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.656321234438017e-06,
+      "loss": 2.5496,
+      "step": 1960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.654567771348414e-06,
+      "loss": 2.5111,
+      "step": 1970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.652814308258812e-06,
+      "loss": 2.4898,
+      "step": 1980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.651060845169211e-06,
+      "loss": 2.4928,
+      "step": 1990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.649307382079608e-06,
+      "loss": 2.4647,
+      "step": 2000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.647553918990006e-06,
+      "loss": 2.4033,
+      "step": 2010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.645800455900403e-06,
+      "loss": 2.5318,
+      "step": 2020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.6440469928108e-06,
+      "loss": 2.412,
+      "step": 2030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.6422935297212e-06,
+      "loss": 2.507,
+      "step": 2040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.640540066631599e-06,
+      "loss": 2.6522,
+      "step": 2050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.638786603541997e-06,
+      "loss": 2.5087,
+      "step": 2060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.637033140452394e-06,
+      "loss": 2.5347,
+      "step": 2070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.635279677362792e-06,
+      "loss": 2.4953,
+      "step": 2080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.63352621427319e-06,
+      "loss": 2.5055,
+      "step": 2090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.631772751183588e-06,
+      "loss": 2.54,
+      "step": 2100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.630019288093987e-06,
+      "loss": 2.4943,
+      "step": 2110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.628265825004385e-06,
+      "loss": 2.6013,
+      "step": 2120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.626512361914782e-06,
+      "loss": 2.4517,
+      "step": 2130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.62475889882518e-06,
+      "loss": 2.4576,
+      "step": 2140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.623005435735579e-06,
+      "loss": 2.5155,
+      "step": 2150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.621251972645976e-06,
+      "loss": 2.8095,
+      "step": 2160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.619498509556374e-06,
+      "loss": 2.578,
+      "step": 2170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.617745046466773e-06,
+      "loss": 2.4727,
+      "step": 2180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.61599158337717e-06,
+      "loss": 2.4369,
+      "step": 2190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.61423812028757e-06,
+      "loss": 2.414,
+      "step": 2200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.612484657197967e-06,
+      "loss": 2.3512,
+      "step": 2210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.610731194108365e-06,
+      "loss": 2.6469,
+      "step": 2220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.608977731018762e-06,
+      "loss": 2.5604,
+      "step": 2230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.60722426792916e-06,
+      "loss": 2.5682,
+      "step": 2240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.605470804839559e-06,
+      "loss": 2.4492,
+      "step": 2250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.603717341749958e-06,
+      "loss": 2.4746,
+      "step": 2260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.601963878660356e-06,
+      "loss": 2.4148,
+      "step": 2270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.600210415570753e-06,
+      "loss": 2.4039,
+      "step": 2280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.59845695248115e-06,
+      "loss": 2.5797,
+      "step": 2290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.59670348939155e-06,
+      "loss": 2.7855,
+      "step": 2300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.594950026301947e-06,
+      "loss": 2.5806,
+      "step": 2310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.593196563212345e-06,
+      "loss": 2.3264,
+      "step": 2320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.591443100122744e-06,
+      "loss": 2.4011,
+      "step": 2330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.589689637033141e-06,
+      "loss": 2.4197,
+      "step": 2340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.587936173943539e-06,
+      "loss": 2.6,
+      "step": 2350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.586182710853938e-06,
+      "loss": 2.4424,
+      "step": 2360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.584429247764335e-06,
+      "loss": 2.5037,
+      "step": 2370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.582675784674733e-06,
+      "loss": 2.2881,
+      "step": 2380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.58092232158513e-06,
+      "loss": 2.3594,
+      "step": 2390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.57916885849553e-06,
+      "loss": 2.417,
+      "step": 2400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.577415395405929e-06,
+      "loss": 2.3677,
+      "step": 2410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.575661932316326e-06,
+      "loss": 2.4908,
+      "step": 2420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.573908469226724e-06,
+      "loss": 2.3145,
+      "step": 2430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.572155006137121e-06,
+      "loss": 2.3709,
+      "step": 2440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.570401543047519e-06,
+      "loss": 2.3841,
+      "step": 2450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.568648079957918e-06,
+      "loss": 2.3462,
+      "step": 2460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.566894616868315e-06,
+      "loss": 2.4069,
+      "step": 2470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.565141153778714e-06,
+      "loss": 2.596,
+      "step": 2480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.563387690689112e-06,
+      "loss": 2.5217,
+      "step": 2490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.56163422759951e-06,
+      "loss": 2.2776,
+      "step": 2500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.559880764509909e-06,
+      "loss": 2.491,
+      "step": 2510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.558127301420306e-06,
+      "loss": 2.6675,
+      "step": 2520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.556373838330704e-06,
+      "loss": 2.4835,
+      "step": 2530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.554620375241101e-06,
+      "loss": 2.4233,
+      "step": 2540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.5528669121515e-06,
+      "loss": 2.4219,
+      "step": 2550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.551113449061898e-06,
+      "loss": 2.4222,
+      "step": 2560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.549359985972297e-06,
+      "loss": 2.3844,
+      "step": 2570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.547606522882694e-06,
+      "loss": 2.1749,
+      "step": 2580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.545853059793092e-06,
+      "loss": 2.5028,
+      "step": 2590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.54409959670349e-06,
+      "loss": 2.4831,
+      "step": 2600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.542346133613888e-06,
+      "loss": 2.5916,
+      "step": 2610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.540592670524286e-06,
+      "loss": 2.4177,
+      "step": 2620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.538839207434685e-06,
+      "loss": 2.288,
+      "step": 2630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.537085744345083e-06,
+      "loss": 2.4141,
+      "step": 2640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.53533228125548e-06,
+      "loss": 2.4218,
+      "step": 2650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.533578818165878e-06,
+      "loss": 2.4864,
+      "step": 2660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.531825355076277e-06,
+      "loss": 2.5332,
+      "step": 2670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.530071891986674e-06,
+      "loss": 2.5455,
+      "step": 2680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.528318428897072e-06,
+      "loss": 2.4063,
+      "step": 2690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.52656496580747e-06,
+      "loss": 2.2459,
+      "step": 2700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.524811502717868e-06,
+      "loss": 2.4605,
+      "step": 2710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.523058039628267e-06,
+      "loss": 2.2808,
+      "step": 2720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.521304576538665e-06,
+      "loss": 2.4057,
+      "step": 2730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.519551113449062e-06,
+      "loss": 2.1542,
+      "step": 2740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.51779765035946e-06,
+      "loss": 2.4931,
+      "step": 2750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.516044187269859e-06,
+      "loss": 2.544,
+      "step": 2760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.514290724180257e-06,
+      "loss": 2.4762,
+      "step": 2770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.512537261090656e-06,
+      "loss": 2.4087,
+      "step": 2780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.510783798001053e-06,
+      "loss": 2.4316,
+      "step": 2790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.50903033491145e-06,
+      "loss": 2.3531,
+      "step": 2800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.507276871821848e-06,
+      "loss": 2.3086,
+      "step": 2810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.505523408732247e-06,
+      "loss": 2.317,
+      "step": 2820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.503769945642645e-06,
+      "loss": 2.258,
+      "step": 2830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.502016482553042e-06,
+      "loss": 2.49,
+      "step": 2840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.500263019463441e-06,
+      "loss": 2.6021,
+      "step": 2850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.498509556373839e-06,
+      "loss": 2.3718,
+      "step": 2860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.496756093284238e-06,
+      "loss": 2.4477,
+      "step": 2870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.495002630194636e-06,
+      "loss": 2.4307,
+      "step": 2880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.493249167105033e-06,
+      "loss": 2.3134,
+      "step": 2890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.49149570401543e-06,
+      "loss": 2.515,
+      "step": 2900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.489742240925828e-06,
+      "loss": 2.2921,
+      "step": 2910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.487988777836227e-06,
+      "loss": 2.2276,
+      "step": 2920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.486235314746626e-06,
+      "loss": 2.4116,
+      "step": 2930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.484481851657024e-06,
+      "loss": 1.9966,
+      "step": 2940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.482728388567421e-06,
+      "loss": 2.4265,
+      "step": 2950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.480974925477819e-06,
+      "loss": 2.2709,
+      "step": 2960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.479221462388218e-06,
+      "loss": 2.4179,
+      "step": 2970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.477467999298615e-06,
+      "loss": 2.1645,
+      "step": 2980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.475714536209013e-06,
+      "loss": 2.626,
+      "step": 2990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.473961073119412e-06,
+      "loss": 2.3627,
+      "step": 3000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47220761002981e-06,
+      "loss": 2.1947,
+      "step": 3010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470454146940207e-06,
+      "loss": 2.4313,
+      "step": 3020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468700683850606e-06,
+      "loss": 2.3694,
+      "step": 3030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466947220761004e-06,
+      "loss": 2.5647,
+      "step": 3040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465193757671401e-06,
+      "loss": 2.4725,
+      "step": 3050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.463440294581799e-06,
+      "loss": 2.2314,
+      "step": 3060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.461686831492198e-06,
+      "loss": 2.3894,
+      "step": 3070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.459933368402597e-06,
+      "loss": 2.1846,
+      "step": 3080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.458179905312995e-06,
+      "loss": 2.3701,
+      "step": 3090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.456426442223392e-06,
+      "loss": 2.4204,
+      "step": 3100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.45467297913379e-06,
+      "loss": 2.7345,
+      "step": 3110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.452919516044187e-06,
+      "loss": 2.2564,
+      "step": 3120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.451166052954586e-06,
+      "loss": 2.417,
+      "step": 3130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.449412589864985e-06,
+      "loss": 2.4262,
+      "step": 3140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.447659126775383e-06,
+      "loss": 2.344,
+      "step": 3150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.44590566368578e-06,
+      "loss": 2.502,
+      "step": 3160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.444152200596178e-06,
+      "loss": 2.1648,
+      "step": 3170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.442398737506577e-06,
+      "loss": 2.2945,
+      "step": 3180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.440645274416974e-06,
+      "loss": 2.2183,
+      "step": 3190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.438891811327372e-06,
+      "loss": 2.3549,
+      "step": 3200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.43713834823777e-06,
+      "loss": 2.3084,
+      "step": 3210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.435384885148169e-06,
+      "loss": 2.219,
+      "step": 3220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.433631422058566e-06,
+      "loss": 2.4011,
+      "step": 3230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.431877958968965e-06,
+      "loss": 2.098,
+      "step": 3240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.430124495879363e-06,
+      "loss": 2.3893,
+      "step": 3250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.42837103278976e-06,
+      "loss": 2.3517,
+      "step": 3260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.426617569700158e-06,
+      "loss": 2.2879,
+      "step": 3270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.424864106610557e-06,
+      "loss": 2.2792,
+      "step": 3280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.423110643520956e-06,
+      "loss": 2.4755,
+      "step": 3290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.421357180431353e-06,
+      "loss": 2.3828,
+      "step": 3300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.419603717341751e-06,
+      "loss": 2.3954,
+      "step": 3310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.417850254252148e-06,
+      "loss": 2.3523,
+      "step": 3320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.416096791162546e-06,
+      "loss": 2.3405,
+      "step": 3330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.414343328072945e-06,
+      "loss": 2.3062,
+      "step": 3340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.412589864983343e-06,
+      "loss": 2.2781,
+      "step": 3350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.41083640189374e-06,
+      "loss": 2.3968,
+      "step": 3360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.40908293880414e-06,
+      "loss": 2.1498,
+      "step": 3370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.407329475714537e-06,
+      "loss": 2.2984,
+      "step": 3380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.405576012624936e-06,
+      "loss": 2.4012,
+      "step": 3390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.403822549535333e-06,
+      "loss": 2.3081,
+      "step": 3400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.40206908644573e-06,
+      "loss": 2.3302,
+      "step": 3410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.400315623356128e-06,
+      "loss": 2.2146,
+      "step": 3420
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.398562160266527e-06,
+      "loss": 2.468,
+      "step": 3430
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.396808697176925e-06,
+      "loss": 2.2726,
+      "step": 3440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.395055234087324e-06,
+      "loss": 2.2769,
+      "step": 3450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.393301770997722e-06,
+      "loss": 2.2363,
+      "step": 3460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.391548307908119e-06,
+      "loss": 2.3426,
+      "step": 3470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.389794844818517e-06,
+      "loss": 2.2352,
+      "step": 3480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.388041381728916e-06,
+      "loss": 2.4611,
+      "step": 3490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.386287918639313e-06,
+      "loss": 2.346,
+      "step": 3500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.384534455549712e-06,
+      "loss": 2.4333,
+      "step": 3510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.38278099246011e-06,
+      "loss": 2.2281,
+      "step": 3520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.381027529370507e-06,
+      "loss": 2.2974,
+      "step": 3530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.379274066280906e-06,
+      "loss": 2.1574,
+      "step": 3540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.377520603191304e-06,
+      "loss": 2.4937,
+      "step": 3550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.375767140101701e-06,
+      "loss": 2.2883,
+      "step": 3560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.374013677012099e-06,
+      "loss": 2.0432,
+      "step": 3570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.372260213922496e-06,
+      "loss": 2.222,
+      "step": 3580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.370506750832896e-06,
+      "loss": 2.3632,
+      "step": 3590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.368753287743295e-06,
+      "loss": 2.2782,
+      "step": 3600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.366999824653692e-06,
+      "loss": 2.2872,
+      "step": 3610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.36524636156409e-06,
+      "loss": 2.6353,
+      "step": 3620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.363492898474487e-06,
+      "loss": 2.4413,
+      "step": 3630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.361739435384886e-06,
+      "loss": 2.2579,
+      "step": 3640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.359985972295284e-06,
+      "loss": 2.2705,
+      "step": 3650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.358232509205683e-06,
+      "loss": 2.4356,
+      "step": 3660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.35647904611608e-06,
+      "loss": 2.3122,
+      "step": 3670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.354725583026478e-06,
+      "loss": 2.4049,
+      "step": 3680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.352972119936875e-06,
+      "loss": 2.3069,
+      "step": 3690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.351218656847275e-06,
+      "loss": 2.4374,
+      "step": 3700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.349465193757672e-06,
+      "loss": 2.4365,
+      "step": 3710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.34771173066807e-06,
+      "loss": 2.2283,
+      "step": 3720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.345958267578467e-06,
+      "loss": 2.5062,
+      "step": 3730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.344204804488866e-06,
+      "loss": 2.4388,
+      "step": 3740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.342451341399265e-06,
+      "loss": 2.4626,
+      "step": 3750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.340697878309663e-06,
+      "loss": 2.2904,
+      "step": 3760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.33894441522006e-06,
+      "loss": 2.2895,
+      "step": 3770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.337190952130458e-06,
+      "loss": 2.217,
+      "step": 3780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.335437489040855e-06,
+      "loss": 2.3075,
+      "step": 3790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.333684025951254e-06,
+      "loss": 2.2582,
+      "step": 3800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.331930562861654e-06,
+      "loss": 2.1267,
+      "step": 3810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.330177099772051e-06,
+      "loss": 2.4391,
+      "step": 3820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.328423636682449e-06,
+      "loss": 2.3733,
+      "step": 3830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.326670173592846e-06,
+      "loss": 2.2349,
+      "step": 3840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.324916710503245e-06,
+      "loss": 2.3483,
+      "step": 3850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.323163247413643e-06,
+      "loss": 2.2351,
+      "step": 3860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.32140978432404e-06,
+      "loss": 2.2335,
+      "step": 3870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.31965632123444e-06,
+      "loss": 2.198,
+      "step": 3880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.317902858144837e-06,
+      "loss": 2.4781,
+      "step": 3890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.316149395055234e-06,
+      "loss": 2.183,
+      "step": 3900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.314395931965634e-06,
+      "loss": 2.1565,
+      "step": 3910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.312642468876031e-06,
+      "loss": 2.2096,
+      "step": 3920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.310889005786428e-06,
+      "loss": 2.0826,
+      "step": 3930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.309135542696826e-06,
+      "loss": 2.2183,
+      "step": 3940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.307382079607225e-06,
+      "loss": 2.1272,
+      "step": 3950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.305628616517624e-06,
+      "loss": 2.3105,
+      "step": 3960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.303875153428022e-06,
+      "loss": 2.2885,
+      "step": 3970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.30212169033842e-06,
+      "loss": 2.175,
+      "step": 3980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.300368227248817e-06,
+      "loss": 2.2774,
+      "step": 3990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.298614764159214e-06,
+      "loss": 2.1772,
+      "step": 4000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.296861301069613e-06,
+      "loss": 2.5513,
+      "step": 4010
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.295107837980011e-06,
+      "loss": 2.3535,
+      "step": 4020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.29335437489041e-06,
+      "loss": 2.3053,
+      "step": 4030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.291600911800808e-06,
+      "loss": 2.2043,
+      "step": 4040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.289847448711205e-06,
+      "loss": 2.2956,
+      "step": 4050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.288093985621604e-06,
+      "loss": 2.5142,
+      "step": 4060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.286340522532002e-06,
+      "loss": 2.3574,
+      "step": 4070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.284587059442399e-06,
+      "loss": 2.4091,
+      "step": 4080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.282833596352797e-06,
+      "loss": 2.3349,
+      "step": 4090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.281080133263196e-06,
+      "loss": 2.2412,
+      "step": 4100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.279326670173593e-06,
+      "loss": 2.0851,
+      "step": 4110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.277573207083992e-06,
+      "loss": 2.156,
+      "step": 4120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.27581974399439e-06,
+      "loss": 2.3909,
+      "step": 4130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.274066280904787e-06,
+      "loss": 2.201,
+      "step": 4140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.272312817815185e-06,
+      "loss": 2.2402,
+      "step": 4150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.270559354725584e-06,
+      "loss": 2.2349,
+      "step": 4160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.268805891635982e-06,
+      "loss": 2.1729,
+      "step": 4170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.26705242854638e-06,
+      "loss": 2.0962,
+      "step": 4180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.265298965456778e-06,
+      "loss": 2.0997,
+      "step": 4190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.263545502367176e-06,
+      "loss": 2.1656,
+      "step": 4200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.261792039277573e-06,
+      "loss": 2.3744,
+      "step": 4210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.260038576187972e-06,
+      "loss": 2.2758,
+      "step": 4220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.25828511309837e-06,
+      "loss": 2.2037,
+      "step": 4230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.256531650008767e-06,
+      "loss": 2.3912,
+      "step": 4240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.254778186919166e-06,
+      "loss": 2.3023,
+      "step": 4250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.253024723829564e-06,
+      "loss": 2.3445,
+      "step": 4260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.251271260739963e-06,
+      "loss": 2.251,
+      "step": 4270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.24951779765036e-06,
+      "loss": 2.1929,
+      "step": 4280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.247764334560758e-06,
+      "loss": 2.0806,
+      "step": 4290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.246010871471156e-06,
+      "loss": 2.258,
+      "step": 4300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.244257408381555e-06,
+      "loss": 2.2252,
+      "step": 4310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.242503945291952e-06,
+      "loss": 2.2906,
+      "step": 4320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.240750482202351e-06,
+      "loss": 2.2018,
+      "step": 4330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.238997019112749e-06,
+      "loss": 2.2472,
+      "step": 4340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.237243556023146e-06,
+      "loss": 2.2988,
+      "step": 4350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.235490092933544e-06,
+      "loss": 2.0677,
+      "step": 4360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.233736629843943e-06,
+      "loss": 2.0756,
+      "step": 4370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.23198316675434e-06,
+      "loss": 2.2818,
+      "step": 4380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.230229703664738e-06,
+      "loss": 2.433,
+      "step": 4390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.228476240575137e-06,
+      "loss": 2.336,
+      "step": 4400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.226722777485535e-06,
+      "loss": 2.4276,
+      "step": 4410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.224969314395934e-06,
+      "loss": 2.1697,
+      "step": 4420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.223215851306331e-06,
+      "loss": 2.0642,
+      "step": 4430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.221462388216729e-06,
+      "loss": 2.23,
+      "step": 4440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.219708925127126e-06,
+      "loss": 2.1703,
+      "step": 4450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.217955462037524e-06,
+      "loss": 2.0771,
+      "step": 4460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.216201998947923e-06,
+      "loss": 2.2217,
+      "step": 4470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.214448535858322e-06,
+      "loss": 1.9788,
+      "step": 4480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.21269507276872e-06,
+      "loss": 2.1382,
+      "step": 4490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.210941609679117e-06,
+      "loss": 2.2412,
+      "step": 4500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.209188146589514e-06,
+      "loss": 2.0613,
+      "step": 4510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.207434683499914e-06,
+      "loss": 2.1652,
+      "step": 4520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.205681220410311e-06,
+      "loss": 2.1491,
+      "step": 4530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.203927757320709e-06,
+      "loss": 2.2028,
+      "step": 4540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.202174294231108e-06,
+      "loss": 2.1681,
+      "step": 4550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.200420831141505e-06,
+      "loss": 2.3093,
+      "step": 4560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.198667368051903e-06,
+      "loss": 2.1754,
+      "step": 4570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.196913904962302e-06,
+      "loss": 2.0744,
+      "step": 4580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.1951604418727e-06,
+      "loss": 2.1654,
+      "step": 4590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.193406978783097e-06,
+      "loss": 2.3086,
+      "step": 4600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.191653515693494e-06,
+      "loss": 2.1916,
+      "step": 4610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.189900052603893e-06,
+      "loss": 2.3129,
+      "step": 4620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.188146589514293e-06,
+      "loss": 2.2434,
+      "step": 4630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.18639312642469e-06,
+      "loss": 2.0726,
+      "step": 4640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.184639663335088e-06,
+      "loss": 2.2226,
+      "step": 4650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.182886200245485e-06,
+      "loss": 2.0556,
+      "step": 4660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.181132737155883e-06,
+      "loss": 2.2813,
+      "step": 4670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.179379274066282e-06,
+      "loss": 2.3198,
+      "step": 4680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.17762581097668e-06,
+      "loss": 2.3953,
+      "step": 4690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.175872347887078e-06,
+      "loss": 2.38,
+      "step": 4700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.174118884797476e-06,
+      "loss": 2.1563,
+      "step": 4710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.172365421707873e-06,
+      "loss": 2.0642,
+      "step": 4720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.170611958618273e-06,
+      "loss": 2.322,
+      "step": 4730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.16885849552867e-06,
+      "loss": 2.2775,
+      "step": 4740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.167105032439067e-06,
+      "loss": 2.3036,
+      "step": 4750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.165351569349465e-06,
+      "loss": 2.1699,
+      "step": 4760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.163598106259864e-06,
+      "loss": 2.0745,
+      "step": 4770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.161844643170262e-06,
+      "loss": 2.195,
+      "step": 4780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.16009118008066e-06,
+      "loss": 2.2779,
+      "step": 4790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.158337716991058e-06,
+      "loss": 2.2684,
+      "step": 4800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.156584253901456e-06,
+      "loss": 1.9256,
+      "step": 4810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.154830790811853e-06,
+      "loss": 2.3165,
+      "step": 4820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.153077327722252e-06,
+      "loss": 1.9181,
+      "step": 4830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.151323864632652e-06,
+      "loss": 2.0031,
+      "step": 4840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.149570401543049e-06,
+      "loss": 2.0968,
+      "step": 4850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.147816938453447e-06,
+      "loss": 2.0997,
+      "step": 4860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.146063475363844e-06,
+      "loss": 2.0481,
+      "step": 4870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.144310012274241e-06,
+      "loss": 2.2254,
+      "step": 4880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.14255654918464e-06,
+      "loss": 2.1065,
+      "step": 4890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.140803086095038e-06,
+      "loss": 2.0723,
+      "step": 4900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.139049623005436e-06,
+      "loss": 1.8847,
+      "step": 4910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.137296159915835e-06,
+      "loss": 2.0991,
+      "step": 4920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.135542696826232e-06,
+      "loss": 2.0728,
+      "step": 4930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.133789233736631e-06,
+      "loss": 2.0512,
+      "step": 4940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.132035770647029e-06,
+      "loss": 2.118,
+      "step": 4950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.130282307557426e-06,
+      "loss": 2.0086,
+      "step": 4960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.128528844467824e-06,
+      "loss": 2.2291,
+      "step": 4970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.126775381378223e-06,
+      "loss": 2.1701,
+      "step": 4980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.12502191828862e-06,
+      "loss": 2.1501,
+      "step": 4990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.12326845519902e-06,
+      "loss": 2.1895,
+      "step": 5000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.121514992109417e-06,
+      "loss": 2.0718,
+      "step": 5010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.119761529019815e-06,
+      "loss": 2.1742,
+      "step": 5020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.118008065930212e-06,
+      "loss": 2.198,
+      "step": 5030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.116254602840611e-06,
+      "loss": 2.2668,
+      "step": 5040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.114501139751009e-06,
+      "loss": 2.0705,
+      "step": 5050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.112747676661406e-06,
+      "loss": 2.3068,
+      "step": 5060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.110994213571805e-06,
+      "loss": 2.1745,
+      "step": 5070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.109240750482203e-06,
+      "loss": 2.2511,
+      "step": 5080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.1074872873926e-06,
+      "loss": 2.227,
+      "step": 5090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.105733824303e-06,
+      "loss": 2.1808,
+      "step": 5100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.103980361213397e-06,
+      "loss": 2.1228,
+      "step": 5110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.102226898123795e-06,
+      "loss": 2.2807,
+      "step": 5120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.100473435034192e-06,
+      "loss": 2.1627,
+      "step": 5130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.098719971944591e-06,
+      "loss": 2.2982,
+      "step": 5140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.09696650885499e-06,
+      "loss": 2.122,
+      "step": 5150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.095213045765388e-06,
+      "loss": 2.1312,
+      "step": 5160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.093459582675785e-06,
+      "loss": 2.0652,
+      "step": 5170
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.091706119586183e-06,
+      "loss": 2.0698,
+      "step": 5180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.089952656496582e-06,
+      "loss": 2.2353,
+      "step": 5190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.08819919340698e-06,
+      "loss": 2.157,
+      "step": 5200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.086445730317379e-06,
+      "loss": 2.0618,
+      "step": 5210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.084692267227776e-06,
+      "loss": 2.0435,
+      "step": 5220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.082938804138174e-06,
+      "loss": 2.1194,
+      "step": 5230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.081185341048571e-06,
+      "loss": 2.1898,
+      "step": 5240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.07943187795897e-06,
+      "loss": 2.0503,
+      "step": 5250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.077678414869368e-06,
+      "loss": 2.1526,
+      "step": 5260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.075924951779765e-06,
+      "loss": 2.1508,
+      "step": 5270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.074171488690163e-06,
+      "loss": 2.194,
+      "step": 5280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.072418025600562e-06,
+      "loss": 2.0932,
+      "step": 5290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.070664562510961e-06,
+      "loss": 2.0567,
+      "step": 5300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.068911099421358e-06,
+      "loss": 2.0332,
+      "step": 5310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.067157636331756e-06,
+      "loss": 2.0265,
+      "step": 5320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.065404173242153e-06,
+      "loss": 2.0524,
+      "step": 5330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.063650710152551e-06,
+      "loss": 2.0769,
+      "step": 5340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.06189724706295e-06,
+      "loss": 1.9945,
+      "step": 5350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.06014378397335e-06,
+      "loss": 2.1984,
+      "step": 5360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.058390320883747e-06,
+      "loss": 2.1733,
+      "step": 5370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.056636857794144e-06,
+      "loss": 2.0702,
+      "step": 5380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.054883394704542e-06,
+      "loss": 2.1214,
+      "step": 5390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.053129931614941e-06,
+      "loss": 1.941,
+      "step": 5400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.051376468525338e-06,
+      "loss": 2.1955,
+      "step": 5410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.049623005435736e-06,
+      "loss": 2.3949,
+      "step": 5420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.047869542346135e-06,
+      "loss": 2.1562,
+      "step": 5430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.046116079256532e-06,
+      "loss": 2.2308,
+      "step": 5440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.04436261616693e-06,
+      "loss": 2.2329,
+      "step": 5450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.042609153077329e-06,
+      "loss": 1.9858,
+      "step": 5460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.040855689987727e-06,
+      "loss": 2.0552,
+      "step": 5470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.039102226898124e-06,
+      "loss": 2.0772,
+      "step": 5480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.037348763808522e-06,
+      "loss": 2.1852,
+      "step": 5490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.03559530071892e-06,
+      "loss": 2.3069,
+      "step": 5500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.03384183762932e-06,
+      "loss": 2.2356,
+      "step": 5510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.032088374539717e-06,
+      "loss": 2.1729,
+      "step": 5520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.030334911450115e-06,
+      "loss": 2.1609,
+      "step": 5530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.028581448360512e-06,
+      "loss": 2.1841,
+      "step": 5540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.02682798527091e-06,
+      "loss": 2.1166,
+      "step": 5550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.025074522181309e-06,
+      "loss": 1.9916,
+      "step": 5560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.023321059091706e-06,
+      "loss": 2.1319,
+      "step": 5570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.021567596002106e-06,
+      "loss": 2.269,
+      "step": 5580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.019814132912503e-06,
+      "loss": 2.1639,
+      "step": 5590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.0180606698229e-06,
+      "loss": 2.0776,
+      "step": 5600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.0163072067333e-06,
+      "loss": 2.1873,
+      "step": 5610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.014553743643697e-06,
+      "loss": 2.1391,
+      "step": 5620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.012800280554095e-06,
+      "loss": 2.0905,
+      "step": 5630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.011046817464492e-06,
+      "loss": 2.0698,
+      "step": 5640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.009293354374891e-06,
+      "loss": 2.1197,
+      "step": 5650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.007539891285289e-06,
+      "loss": 2.3111,
+      "step": 5660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.005786428195688e-06,
+      "loss": 2.1282,
+      "step": 5670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.004032965106086e-06,
+      "loss": 2.1536,
+      "step": 5680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.002279502016483e-06,
+      "loss": 2.1526,
+      "step": 5690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.00052603892688e-06,
+      "loss": 2.1441,
+      "step": 5700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.99877257583728e-06,
+      "loss": 2.1939,
+      "step": 5710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.997019112747677e-06,
+      "loss": 2.0605,
+      "step": 5720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.995265649658076e-06,
+      "loss": 2.1177,
+      "step": 5730
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 8.993512186568474e-06,
+      "loss": 2.0728,
+      "step": 5740
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 8.991758723478871e-06,
+      "loss": 1.9477,
+      "step": 5750
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 8.990005260389269e-06,
+      "loss": 2.0901,
+      "step": 5760
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 8.988251797299668e-06,
+      "loss": 2.2447,
+      "step": 5770
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 8.986498334210065e-06,
+      "loss": 1.9208,
+      "step": 5780
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.984744871120463e-06,
+      "loss": 2.0831,
+      "step": 5790
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.982991408030862e-06,
+      "loss": 2.3926,
+      "step": 5800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.98123794494126e-06,
+      "loss": 2.0827,
+      "step": 5810
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.979484481851659e-06,
+      "loss": 2.0788,
+      "step": 5820
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.977731018762056e-06,
+      "loss": 2.1964,
+      "step": 5830
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 8.975977555672454e-06,
+      "loss": 2.188,
+      "step": 5840
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.974224092582851e-06,
+      "loss": 2.1584,
+      "step": 5850
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.97247062949325e-06,
+      "loss": 2.123,
+      "step": 5860
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.970717166403648e-06,
+      "loss": 2.0259,
+      "step": 5870
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.968963703314047e-06,
+      "loss": 2.0544,
+      "step": 5880
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.967210240224444e-06,
+      "loss": 2.057,
+      "step": 5890
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 8.965456777134842e-06,
+      "loss": 1.8251,
+      "step": 5900
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.96370331404524e-06,
+      "loss": 1.9718,
+      "step": 5910
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.961949850955639e-06,
+      "loss": 1.8774,
+      "step": 5920
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.960196387866036e-06,
+      "loss": 2.14,
+      "step": 5930
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.958442924776434e-06,
+      "loss": 1.9823,
+      "step": 5940
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 8.956689461686833e-06,
+      "loss": 2.237,
+      "step": 5950
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.95493599859723e-06,
+      "loss": 2.1228,
+      "step": 5960
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.95318253550763e-06,
+      "loss": 1.9662,
+      "step": 5970
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.951429072418027e-06,
+      "loss": 2.2064,
+      "step": 5980
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.949675609328424e-06,
+      "loss": 1.9929,
+      "step": 5990
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.947922146238822e-06,
+      "loss": 2.1435,
+      "step": 6000
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 8.94616868314922e-06,
+      "loss": 2.0758,
+      "step": 6010
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.944415220059618e-06,
+      "loss": 2.1253,
+      "step": 6020
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.942661756970018e-06,
+      "loss": 2.1229,
+      "step": 6030
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.940908293880415e-06,
+      "loss": 2.0281,
+      "step": 6040
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.939154830790813e-06,
+      "loss": 1.9592,
+      "step": 6050
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.93740136770121e-06,
+      "loss": 1.9891,
+      "step": 6060
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 8.93564790461161e-06,
+      "loss": 2.2361,
+      "step": 6070
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.933894441522007e-06,
+      "loss": 2.0054,
+      "step": 6080
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.932140978432404e-06,
+      "loss": 2.0427,
+      "step": 6090
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.930387515342803e-06,
+      "loss": 1.9342,
+      "step": 6100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.9286340522532e-06,
+      "loss": 2.0739,
+      "step": 6110
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.926880589163598e-06,
+      "loss": 1.9577,
+      "step": 6120
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 8.925127126073997e-06,
+      "loss": 1.9935,
+      "step": 6130
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.923373662984395e-06,
+      "loss": 2.1148,
+      "step": 6140
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.921620199894792e-06,
+      "loss": 2.0652,
+      "step": 6150
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.91986673680519e-06,
+      "loss": 1.9819,
+      "step": 6160
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.918113273715589e-06,
+      "loss": 2.2294,
+      "step": 6170
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 8.916359810625988e-06,
+      "loss": 1.991,
+      "step": 6180
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.914606347536386e-06,
+      "loss": 2.0409,
+      "step": 6190
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.912852884446783e-06,
+      "loss": 1.8692,
+      "step": 6200
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.91109942135718e-06,
+      "loss": 2.2048,
+      "step": 6210
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.909345958267578e-06,
+      "loss": 2.1289,
+      "step": 6220
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.907592495177977e-06,
+      "loss": 2.0174,
+      "step": 6230
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 8.905839032088375e-06,
+      "loss": 2.0823,
+      "step": 6240
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.904085568998774e-06,
+      "loss": 1.8747,
+      "step": 6250
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.902332105909171e-06,
+      "loss": 2.0248,
+      "step": 6260
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.900578642819569e-06,
+      "loss": 2.0564,
+      "step": 6270
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.898825179729968e-06,
+      "loss": 2.1538,
+      "step": 6280
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.897071716640366e-06,
+      "loss": 2.2712,
+      "step": 6290
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 8.895318253550763e-06,
+      "loss": 2.0554,
+      "step": 6300
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89356479046116e-06,
+      "loss": 1.9753,
+      "step": 6310
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.89181132737156e-06,
+      "loss": 2.1354,
+      "step": 6320
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.890057864281957e-06,
+      "loss": 1.9283,
+      "step": 6330
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.888304401192356e-06,
+      "loss": 2.0628,
+      "step": 6340
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 8.886550938102754e-06,
+      "loss": 1.9338,
+      "step": 6350
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.884797475013151e-06,
+      "loss": 2.0819,
+      "step": 6360
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.883044011923549e-06,
+      "loss": 2.1737,
+      "step": 6370
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.881290548833948e-06,
+      "loss": 1.9829,
+      "step": 6380
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.879537085744345e-06,
+      "loss": 1.9941,
+      "step": 6390
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.877783622654745e-06,
+      "loss": 2.0321,
+      "step": 6400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 8.876030159565142e-06,
+      "loss": 1.9913,
+      "step": 6410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.87427669647554e-06,
+      "loss": 1.893,
+      "step": 6420
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.872523233385937e-06,
+      "loss": 1.9422,
+      "step": 6430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.870769770296336e-06,
+      "loss": 1.968,
+      "step": 6440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.869016307206734e-06,
+      "loss": 2.0178,
+      "step": 6450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.867262844117131e-06,
+      "loss": 2.0007,
+      "step": 6460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.86550938102753e-06,
+      "loss": 2.0872,
+      "step": 6470
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.863755917937928e-06,
+      "loss": 2.1649,
+      "step": 6480
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.862002454848327e-06,
+      "loss": 2.0749,
+      "step": 6490
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.860248991758725e-06,
+      "loss": 2.0655,
+      "step": 6500
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.858495528669122e-06,
+      "loss": 2.0575,
+      "step": 6510
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.85674206557952e-06,
+      "loss": 2.0497,
+      "step": 6520
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.854988602489919e-06,
+      "loss": 2.138,
+      "step": 6530
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.853235139400316e-06,
+      "loss": 2.0597,
+      "step": 6540
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.851481676310715e-06,
+      "loss": 1.9994,
+      "step": 6550
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.849728213221113e-06,
+      "loss": 1.9549,
+      "step": 6560
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.84797475013151e-06,
+      "loss": 2.3336,
+      "step": 6570
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.846221287041908e-06,
+      "loss": 2.026,
+      "step": 6580
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.844467823952307e-06,
+      "loss": 2.0599,
+      "step": 6590
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.842714360862704e-06,
+      "loss": 1.8892,
+      "step": 6600
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.840960897773102e-06,
+      "loss": 2.0428,
+      "step": 6610
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.839207434683501e-06,
+      "loss": 2.1208,
+      "step": 6620
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.837453971593899e-06,
+      "loss": 2.1453,
+      "step": 6630
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.835700508504296e-06,
+      "loss": 2.0037,
+      "step": 6640
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.833947045414695e-06,
+      "loss": 1.9911,
+      "step": 6650
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.832193582325093e-06,
+      "loss": 2.0452,
+      "step": 6660
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.83044011923549e-06,
+      "loss": 1.8799,
+      "step": 6670
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.828686656145888e-06,
+      "loss": 2.0513,
+      "step": 6680
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.826933193056287e-06,
+      "loss": 1.8061,
+      "step": 6690
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.825179729966686e-06,
+      "loss": 2.1391,
+      "step": 6700
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.823426266877083e-06,
+      "loss": 2.0935,
+      "step": 6710
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.821672803787481e-06,
+      "loss": 2.0646,
+      "step": 6720
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.819919340697878e-06,
+      "loss": 2.146,
+      "step": 6730
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.818165877608278e-06,
+      "loss": 1.9651,
+      "step": 6740
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.816412414518675e-06,
+      "loss": 2.0127,
+      "step": 6750
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.814658951429074e-06,
+      "loss": 1.9121,
+      "step": 6760
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.812905488339472e-06,
+      "loss": 2.0953,
+      "step": 6770
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.81115202524987e-06,
+      "loss": 1.9628,
+      "step": 6780
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.809398562160267e-06,
+      "loss": 2.2837,
+      "step": 6790
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.807645099070666e-06,
+      "loss": 2.0474,
+      "step": 6800
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.805891635981063e-06,
+      "loss": 1.8321,
+      "step": 6810
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.80413817289146e-06,
+      "loss": 1.8997,
+      "step": 6820
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.802384709801858e-06,
+      "loss": 2.1622,
+      "step": 6830
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.800631246712257e-06,
+      "loss": 2.0263,
+      "step": 6840
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.798877783622657e-06,
+      "loss": 1.904,
+      "step": 6850
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.797124320533054e-06,
+      "loss": 1.9165,
+      "step": 6860
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.795370857443452e-06,
+      "loss": 1.9329,
+      "step": 6870
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.793617394353849e-06,
+      "loss": 2.0968,
+      "step": 6880
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.791863931264247e-06,
+      "loss": 2.0522,
+      "step": 6890
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.790110468174646e-06,
+      "loss": 2.261,
+      "step": 6900
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.788357005085045e-06,
+      "loss": 2.265,
+      "step": 6910
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.786603541995442e-06,
+      "loss": 2.0506,
+      "step": 6920
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.78485007890584e-06,
+      "loss": 2.0013,
+      "step": 6930
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.783096615816237e-06,
+      "loss": 2.0103,
+      "step": 6940
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.781343152726636e-06,
+      "loss": 2.1846,
+      "step": 6950
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.779589689637034e-06,
+      "loss": 1.8876,
+      "step": 6960
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.777836226547431e-06,
+      "loss": 1.9929,
+      "step": 6970
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.776082763457829e-06,
+      "loss": 1.9183,
+      "step": 6980
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.774329300368228e-06,
+      "loss": 1.9269,
+      "step": 6990
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.772575837278626e-06,
+      "loss": 1.8781,
+      "step": 7000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.770822374189025e-06,
+      "loss": 2.1732,
+      "step": 7010
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.769068911099422e-06,
+      "loss": 1.9491,
+      "step": 7020
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.76731544800982e-06,
+      "loss": 1.9168,
+      "step": 7030
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 8.765561984920217e-06,
+      "loss": 2.2144,
+      "step": 7040
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.763808521830616e-06,
+      "loss": 2.0088,
+      "step": 7050
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.762055058741016e-06,
+      "loss": 2.0934,
+      "step": 7060
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.760301595651413e-06,
+      "loss": 2.0554,
+      "step": 7070
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.75854813256181e-06,
+      "loss": 1.9514,
+      "step": 7080
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.756794669472208e-06,
+      "loss": 2.0145,
+      "step": 7090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.755041206382605e-06,
+      "loss": 1.9482,
+      "step": 7100
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.753287743293005e-06,
+      "loss": 2.1827,
+      "step": 7110
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.751534280203402e-06,
+      "loss": 2.1314,
+      "step": 7120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.749780817113801e-06,
+      "loss": 1.8251,
+      "step": 7130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.748027354024199e-06,
+      "loss": 1.8696,
+      "step": 7140
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.746273890934596e-06,
+      "loss": 2.1607,
+      "step": 7150
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.744520427844995e-06,
+      "loss": 2.1328,
+      "step": 7160
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.742766964755393e-06,
+      "loss": 2.0899,
+      "step": 7170
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.74101350166579e-06,
+      "loss": 2.0573,
+      "step": 7180
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.739260038576188e-06,
+      "loss": 1.9968,
+      "step": 7190
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.737506575486587e-06,
+      "loss": 1.991,
+      "step": 7200
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.735753112396984e-06,
+      "loss": 2.2269,
+      "step": 7210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.733999649307384e-06,
+      "loss": 2.1294,
+      "step": 7220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.732246186217781e-06,
+      "loss": 1.9176,
+      "step": 7230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.730492723128179e-06,
+      "loss": 1.8168,
+      "step": 7240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.728739260038576e-06,
+      "loss": 1.8621,
+      "step": 7250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.726985796948975e-06,
+      "loss": 1.8164,
+      "step": 7260
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.725232333859373e-06,
+      "loss": 1.8586,
+      "step": 7270
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.723478870769772e-06,
+      "loss": 1.9511,
+      "step": 7280
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.72172540768017e-06,
+      "loss": 2.3752,
+      "step": 7290
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.719971944590567e-06,
+      "loss": 1.9227,
+      "step": 7300
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.718218481500964e-06,
+      "loss": 2.0172,
+      "step": 7310
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 8.716465018411364e-06,
+      "loss": 2.0428,
+      "step": 7320
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.714711555321761e-06,
+      "loss": 1.9362,
+      "step": 7330
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.712958092232158e-06,
+      "loss": 2.0874,
+      "step": 7340
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.711204629142556e-06,
+      "loss": 1.8527,
+      "step": 7350
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.709451166052955e-06,
+      "loss": 2.2299,
+      "step": 7360
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.707697702963354e-06,
+      "loss": 2.036,
+      "step": 7370
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 8.705944239873752e-06,
+      "loss": 1.862,
+      "step": 7380
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.70419077678415e-06,
+      "loss": 2.0204,
+      "step": 7390
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.702437313694547e-06,
+      "loss": 2.1152,
+      "step": 7400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.700683850604946e-06,
+      "loss": 2.0088,
+      "step": 7410
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.698930387515343e-06,
+      "loss": 2.1824,
+      "step": 7420
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.697176924425743e-06,
+      "loss": 1.9612,
+      "step": 7430
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 8.69542346133614e-06,
+      "loss": 2.0992,
+      "step": 7440
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.693669998246538e-06,
+      "loss": 2.1787,
+      "step": 7450
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.691916535156935e-06,
+      "loss": 2.0621,
+      "step": 7460
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.690163072067334e-06,
+      "loss": 1.9243,
+      "step": 7470
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.688409608977732e-06,
+      "loss": 1.8972,
+      "step": 7480
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 8.686656145888129e-06,
+      "loss": 2.0974,
+      "step": 7490
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.684902682798528e-06,
+      "loss": 1.8727,
+      "step": 7500
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.683149219708926e-06,
+      "loss": 1.9229,
+      "step": 7510
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.681395756619325e-06,
+      "loss": 1.8004,
+      "step": 7520
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.679642293529722e-06,
+      "loss": 1.915,
+      "step": 7530
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.67788883044012e-06,
+      "loss": 1.8445,
+      "step": 7540
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 8.676135367350517e-06,
+      "loss": 2.1407,
+      "step": 7550
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.674381904260915e-06,
+      "loss": 2.0058,
+      "step": 7560
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.672628441171314e-06,
+      "loss": 1.9014,
+      "step": 7570
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.670874978081713e-06,
+      "loss": 1.948,
+      "step": 7580
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.66912151499211e-06,
+      "loss": 2.0255,
+      "step": 7590
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.667368051902508e-06,
+      "loss": 2.0289,
+      "step": 7600
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 8.665614588812906e-06,
+      "loss": 2.0317,
+      "step": 7610
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.663861125723305e-06,
+      "loss": 1.9867,
+      "step": 7620
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.662107662633702e-06,
+      "loss": 2.1904,
+      "step": 7630
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.6603541995441e-06,
+      "loss": 1.8195,
+      "step": 7640
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.658600736454499e-06,
+      "loss": 2.0896,
+      "step": 7650
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.656847273364896e-06,
+      "loss": 2.0193,
+      "step": 7660
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 8.655093810275294e-06,
+      "loss": 2.0471,
+      "step": 7670
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.653340347185693e-06,
+      "loss": 1.9884,
+      "step": 7680
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.65158688409609e-06,
+      "loss": 1.8363,
+      "step": 7690
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.649833421006488e-06,
+      "loss": 2.0123,
+      "step": 7700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.648079957916886e-06,
+      "loss": 2.0866,
+      "step": 7710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 8.646326494827285e-06,
+      "loss": 1.8461,
+      "step": 7720
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.644573031737684e-06,
+      "loss": 1.9943,
+      "step": 7730
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.642819568648081e-06,
+      "loss": 2.1395,
+      "step": 7740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.641066105558479e-06,
+      "loss": 1.9925,
+      "step": 7750
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.639312642468876e-06,
+      "loss": 1.9898,
+      "step": 7760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.637559179379274e-06,
+      "loss": 2.0501,
+      "step": 7770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 8.635805716289673e-06,
+      "loss": 1.8729,
+      "step": 7780
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.63405225320007e-06,
+      "loss": 1.8684,
+      "step": 7790
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.63229879011047e-06,
+      "loss": 1.9667,
+      "step": 7800
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.630545327020867e-06,
+      "loss": 2.1698,
+      "step": 7810
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.628791863931265e-06,
+      "loss": 2.1606,
+      "step": 7820
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.627038400841664e-06,
+      "loss": 1.9659,
+      "step": 7830
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 8.625284937752061e-06,
+      "loss": 1.9593,
+      "step": 7840
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.623531474662459e-06,
+      "loss": 1.9702,
+      "step": 7850
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.621778011572856e-06,
+      "loss": 2.1453,
+      "step": 7860
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.620024548483255e-06,
+      "loss": 1.9599,
+      "step": 7870
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.618271085393653e-06,
+      "loss": 2.0417,
+      "step": 7880
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 8.616517622304052e-06,
+      "loss": 2.0459,
+      "step": 7890
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.61476415921445e-06,
+      "loss": 2.0787,
+      "step": 7900
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.613010696124847e-06,
+      "loss": 1.9023,
+      "step": 7910
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.611257233035244e-06,
+      "loss": 1.9731,
+      "step": 7920
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.609503769945644e-06,
+      "loss": 2.1286,
+      "step": 7930
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.607750306856041e-06,
+      "loss": 1.9603,
+      "step": 7940
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 8.60599684376644e-06,
+      "loss": 1.9755,
+      "step": 7950
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.604243380676838e-06,
+      "loss": 1.8862,
+      "step": 7960
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.602489917587235e-06,
+      "loss": 2.1383,
+      "step": 7970
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.600736454497633e-06,
+      "loss": 2.1113,
+      "step": 7980
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.598982991408032e-06,
+      "loss": 2.0204,
+      "step": 7990
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.59722952831843e-06,
+      "loss": 1.8182,
+      "step": 8000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 8.595476065228827e-06,
+      "loss": 1.918,
+      "step": 8010
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.593722602139226e-06,
+      "loss": 2.0219,
+      "step": 8020
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.591969139049623e-06,
+      "loss": 2.1181,
+      "step": 8030
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.590215675960023e-06,
+      "loss": 1.9452,
+      "step": 8040
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.58846221287042e-06,
+      "loss": 1.8707,
+      "step": 8050
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 8.586708749780818e-06,
+      "loss": 2.0175,
+      "step": 8060
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.584955286691215e-06,
+      "loss": 1.8622,
+      "step": 8070
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.583201823601614e-06,
+      "loss": 2.1894,
+      "step": 8080
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.581448360512012e-06,
+      "loss": 1.969,
+      "step": 8090
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.579694897422411e-06,
+      "loss": 1.9867,
+      "step": 8100
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.577941434332808e-06,
+      "loss": 1.5976,
+      "step": 8110
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 8.576187971243206e-06,
+      "loss": 1.7773,
+      "step": 8120
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.574434508153603e-06,
+      "loss": 1.9372,
+      "step": 8130
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.572681045064003e-06,
+      "loss": 1.8554,
+      "step": 8140
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.5709275819744e-06,
+      "loss": 1.7028,
+      "step": 8150
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.569174118884797e-06,
+      "loss": 1.9102,
+      "step": 8160
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.567420655795197e-06,
+      "loss": 2.0164,
+      "step": 8170
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 8.565667192705594e-06,
+      "loss": 1.9751,
+      "step": 8180
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.563913729615992e-06,
+      "loss": 2.037,
+      "step": 8190
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.56216026652639e-06,
+      "loss": 2.0332,
+      "step": 8200
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.560406803436788e-06,
+      "loss": 1.9074,
+      "step": 8210
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.558653340347186e-06,
+      "loss": 1.8129,
+      "step": 8220
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.556899877257583e-06,
+      "loss": 1.9539,
+      "step": 8230
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 8.555146414167982e-06,
+      "loss": 1.9856,
+      "step": 8240
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.553392951078382e-06,
+      "loss": 2.0746,
+      "step": 8250
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.551639487988779e-06,
+      "loss": 2.0352,
+      "step": 8260
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.549886024899177e-06,
+      "loss": 1.8535,
+      "step": 8270
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.548132561809574e-06,
+      "loss": 1.8868,
+      "step": 8280
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 8.546379098719973e-06,
+      "loss": 1.9253,
+      "step": 8290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.54462563563037e-06,
+      "loss": 1.8915,
+      "step": 8300
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.542872172540768e-06,
+      "loss": 1.943,
+      "step": 8310
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.541118709451167e-06,
+      "loss": 1.8794,
+      "step": 8320
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.539365246361565e-06,
+      "loss": 2.0541,
+      "step": 8330
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.537611783271962e-06,
+      "loss": 1.9305,
+      "step": 8340
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 8.535858320182361e-06,
+      "loss": 2.0028,
+      "step": 8350
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.534104857092759e-06,
+      "loss": 1.927,
+      "step": 8360
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.532351394003156e-06,
+      "loss": 2.1076,
+      "step": 8370
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.530597930913554e-06,
+      "loss": 2.0641,
+      "step": 8380
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.528844467823953e-06,
+      "loss": 2.0961,
+      "step": 8390
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.527091004734352e-06,
+      "loss": 1.8663,
+      "step": 8400
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 8.52533754164475e-06,
+      "loss": 1.8887,
+      "step": 8410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.523584078555147e-06,
+      "loss": 2.0467,
+      "step": 8420
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.521830615465545e-06,
+      "loss": 2.3173,
+      "step": 8430
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.520077152375942e-06,
+      "loss": 1.9275,
+      "step": 8440
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.518323689286341e-06,
+      "loss": 1.7625,
+      "step": 8450
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.51657022619674e-06,
+      "loss": 1.9257,
+      "step": 8460
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.514816763107138e-06,
+      "loss": 1.9631,
+      "step": 8470
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.513063300017535e-06,
+      "loss": 1.9798,
+      "step": 8480
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.511309836927933e-06,
+      "loss": 2.0029,
+      "step": 8490
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.509556373838332e-06,
+      "loss": 2.053,
+      "step": 8500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.50780291074873e-06,
+      "loss": 2.0634,
+      "step": 8510
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 8.506049447659127e-06,
+      "loss": 2.0267,
+      "step": 8520
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.504295984569525e-06,
+      "loss": 1.8538,
+      "step": 8530
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.502542521479924e-06,
+      "loss": 1.8943,
+      "step": 8540
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.500789058390321e-06,
+      "loss": 1.9018,
+      "step": 8550
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.49903559530072e-06,
+      "loss": 2.0128,
+      "step": 8560
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.497282132211118e-06,
+      "loss": 1.9665,
+      "step": 8570
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.495528669121515e-06,
+      "loss": 1.9322,
+      "step": 8580
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.493775206031913e-06,
+      "loss": 1.9474,
+      "step": 8590
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.492021742942312e-06,
+      "loss": 1.9059,
+      "step": 8600
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.490268279852711e-06,
+      "loss": 1.9152,
+      "step": 8610
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.488514816763109e-06,
+      "loss": 1.978,
+      "step": 8620
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.486761353673506e-06,
+      "loss": 1.8913,
+      "step": 8630
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 8.485007890583904e-06,
+      "loss": 1.8965,
+      "step": 8640
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.483254427494301e-06,
+      "loss": 1.7409,
+      "step": 8650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.4815009644047e-06,
+      "loss": 2.0136,
+      "step": 8660
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.479747501315098e-06,
+      "loss": 2.0647,
+      "step": 8670
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.477994038225495e-06,
+      "loss": 2.0028,
+      "step": 8680
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 8.476240575135894e-06,
+      "loss": 1.9845,
+      "step": 8690
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.474487112046292e-06,
+      "loss": 1.8874,
+      "step": 8700
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.472733648956691e-06,
+      "loss": 1.745,
+      "step": 8710
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.470980185867088e-06,
+      "loss": 1.9102,
+      "step": 8720
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.469226722777486e-06,
+      "loss": 1.7143,
+      "step": 8730
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.467473259687883e-06,
+      "loss": 1.8992,
+      "step": 8740
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 8.465719796598283e-06,
+      "loss": 1.7833,
+      "step": 8750
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.46396633350868e-06,
+      "loss": 1.9334,
+      "step": 8760
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.46221287041908e-06,
+      "loss": 1.7448,
+      "step": 8770
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.460459407329477e-06,
+      "loss": 2.0539,
+      "step": 8780
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.458705944239874e-06,
+      "loss": 1.9234,
+      "step": 8790
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.456952481150272e-06,
+      "loss": 2.0793,
+      "step": 8800
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 8.45519901806067e-06,
+      "loss": 2.0877,
+      "step": 8810
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.453445554971068e-06,
+      "loss": 1.7543,
+      "step": 8820
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.451692091881467e-06,
+      "loss": 1.8062,
+      "step": 8830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.449938628791865e-06,
+      "loss": 1.8212,
+      "step": 8840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.448185165702262e-06,
+      "loss": 2.0078,
+      "step": 8850
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 8.44643170261266e-06,
+      "loss": 1.8253,
+      "step": 8860
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.444678239523059e-06,
+      "loss": 2.1439,
+      "step": 8870
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.442924776433457e-06,
+      "loss": 1.9554,
+      "step": 8880
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.441171313343854e-06,
+      "loss": 2.1399,
+      "step": 8890
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.439417850254252e-06,
+      "loss": 1.9319,
+      "step": 8900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.43766438716465e-06,
+      "loss": 1.8686,
+      "step": 8910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 8.43591092407505e-06,
+      "loss": 1.8591,
+      "step": 8920
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.434157460985447e-06,
+      "loss": 2.0404,
+      "step": 8930
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.432403997895845e-06,
+      "loss": 2.0929,
+      "step": 8940
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.430650534806242e-06,
+      "loss": 2.3782,
+      "step": 8950
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.428897071716641e-06,
+      "loss": 2.2402,
+      "step": 8960
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.427143608627039e-06,
+      "loss": 1.9254,
+      "step": 8970
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 8.425390145537438e-06,
+      "loss": 2.003,
+      "step": 8980
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.423636682447836e-06,
+      "loss": 1.8802,
+      "step": 8990
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.421883219358233e-06,
+      "loss": 1.8555,
+      "step": 9000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.42012975626863e-06,
+      "loss": 1.9109,
+      "step": 9010
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.41837629317903e-06,
+      "loss": 2.2475,
+      "step": 9020
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 8.416622830089427e-06,
+      "loss": 1.8998,
+      "step": 9030
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.414869366999825e-06,
+      "loss": 1.95,
+      "step": 9040
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.413115903910224e-06,
+      "loss": 1.9381,
+      "step": 9050
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.411362440820621e-06,
+      "loss": 1.9799,
+      "step": 9060
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.40960897773102e-06,
+      "loss": 1.9678,
+      "step": 9070
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.407855514641418e-06,
+      "loss": 2.0006,
+      "step": 9080
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 8.406102051551815e-06,
+      "loss": 1.7843,
+      "step": 9090
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.404348588462213e-06,
+      "loss": 1.6794,
+      "step": 9100
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.40259512537261e-06,
+      "loss": 1.8309,
+      "step": 9110
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.40084166228301e-06,
+      "loss": 1.8703,
+      "step": 9120
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.399088199193409e-06,
+      "loss": 1.8652,
+      "step": 9130
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.397334736103806e-06,
+      "loss": 2.0208,
+      "step": 9140
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 8.395581273014204e-06,
+      "loss": 1.8659,
+      "step": 9150
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.393827809924601e-06,
+      "loss": 1.946,
+      "step": 9160
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.392074346835e-06,
+      "loss": 1.9618,
+      "step": 9170
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.390320883745398e-06,
+      "loss": 1.994,
+      "step": 9180
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.388567420655795e-06,
+      "loss": 1.708,
+      "step": 9190
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.386813957566195e-06,
+      "loss": 2.1473,
+      "step": 9200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 8.385060494476592e-06,
+      "loss": 1.9425,
+      "step": 9210
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.38330703138699e-06,
+      "loss": 1.9913,
+      "step": 9220
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.381553568297389e-06,
+      "loss": 1.97,
+      "step": 9230
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.379800105207786e-06,
+      "loss": 1.8667,
+      "step": 9240
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.378046642118184e-06,
+      "loss": 1.876,
+      "step": 9250
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 8.376293179028581e-06,
+      "loss": 1.8153,
+      "step": 9260
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.37453971593898e-06,
+      "loss": 1.8742,
+      "step": 9270
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.37278625284938e-06,
+      "loss": 1.945,
+      "step": 9280
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.371032789759777e-06,
+      "loss": 1.9081,
+      "step": 9290
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.369279326670174e-06,
+      "loss": 2.1266,
+      "step": 9300
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.367525863580572e-06,
+      "loss": 1.9848,
+      "step": 9310
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 8.36577240049097e-06,
+      "loss": 2.0075,
+      "step": 9320
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.364018937401369e-06,
+      "loss": 2.0973,
+      "step": 9330
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.362265474311766e-06,
+      "loss": 2.0852,
+      "step": 9340
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.360512011222165e-06,
+      "loss": 1.9711,
+      "step": 9350
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.358758548132563e-06,
+      "loss": 1.6906,
+      "step": 9360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.35700508504296e-06,
+      "loss": 2.2158,
+      "step": 9370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 8.35525162195336e-06,
+      "loss": 1.8902,
+      "step": 9380
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.353498158863757e-06,
+      "loss": 1.9982,
+      "step": 9390
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.351744695774154e-06,
+      "loss": 1.795,
+      "step": 9400
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.349991232684552e-06,
+      "loss": 1.9619,
+      "step": 9410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.348237769594951e-06,
+      "loss": 1.9581,
+      "step": 9420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.346484306505348e-06,
+      "loss": 1.8957,
+      "step": 9430
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.344730843415748e-06,
+      "loss": 1.7884,
+      "step": 9440
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.342977380326145e-06,
+      "loss": 1.9992,
+      "step": 9450
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.341223917236543e-06,
+      "loss": 1.8318,
+      "step": 9460
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.33947045414694e-06,
+      "loss": 1.8552,
+      "step": 9470
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.33771699105734e-06,
+      "loss": 1.903,
+      "step": 9480
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.335963527967737e-06,
+      "loss": 1.7742,
+      "step": 9490
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.334210064878136e-06,
+      "loss": 1.9196,
+      "step": 9500
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.332456601788533e-06,
+      "loss": 2.2706,
+      "step": 9510
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.33070313869893e-06,
+      "loss": 1.9618,
+      "step": 9520
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.328949675609328e-06,
+      "loss": 1.9339,
+      "step": 9530
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.327196212519727e-06,
+      "loss": 1.8614,
+      "step": 9540
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.325442749430125e-06,
+      "loss": 1.8085,
+      "step": 9550
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.323689286340522e-06,
+      "loss": 1.8628,
+      "step": 9560
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.321935823250922e-06,
+      "loss": 2.0486,
+      "step": 9570
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.320182360161319e-06,
+      "loss": 1.7742,
+      "step": 9580
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.318428897071718e-06,
+      "loss": 2.1517,
+      "step": 9590
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.316675433982116e-06,
+      "loss": 2.0303,
+      "step": 9600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.314921970892513e-06,
+      "loss": 1.9003,
+      "step": 9610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.31316850780291e-06,
+      "loss": 1.931,
+      "step": 9620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.31141504471331e-06,
+      "loss": 2.0205,
+      "step": 9630
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.309661581623707e-06,
+      "loss": 2.0019,
+      "step": 9640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.307908118534106e-06,
+      "loss": 1.9155,
+      "step": 9650
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.306154655444504e-06,
+      "loss": 1.853,
+      "step": 9660
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.304401192354901e-06,
+      "loss": 2.2928,
+      "step": 9670
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.302647729265299e-06,
+      "loss": 1.8948,
+      "step": 9680
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.300894266175698e-06,
+      "loss": 1.9152,
+      "step": 9690
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.299140803086096e-06,
+      "loss": 1.9365,
+      "step": 9700
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.297387339996493e-06,
+      "loss": 1.7576,
+      "step": 9710
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.295633876906892e-06,
+      "loss": 1.78,
+      "step": 9720
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.29388041381729e-06,
+      "loss": 1.886,
+      "step": 9730
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.292126950727687e-06,
+      "loss": 1.8823,
+      "step": 9740
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.290373487638086e-06,
+      "loss": 1.8384,
+      "step": 9750
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.288620024548484e-06,
+      "loss": 1.8768,
+      "step": 9760
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.286866561458881e-06,
+      "loss": 2.1253,
+      "step": 9770
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.285113098369279e-06,
+      "loss": 2.0191,
+      "step": 9780
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.283359635279678e-06,
+      "loss": 1.8313,
+      "step": 9790
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.281606172190077e-06,
+      "loss": 1.8763,
+      "step": 9800
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.279852709100475e-06,
+      "loss": 1.7839,
+      "step": 9810
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.278099246010872e-06,
+      "loss": 1.7735,
+      "step": 9820
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.27634578292127e-06,
+      "loss": 1.9695,
+      "step": 9830
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.274592319831669e-06,
+      "loss": 1.8045,
+      "step": 9840
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.272838856742066e-06,
+      "loss": 1.9127,
+      "step": 9850
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.271085393652464e-06,
+      "loss": 1.8685,
+      "step": 9860
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.269331930562863e-06,
+      "loss": 1.8558,
+      "step": 9870
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.26757846747326e-06,
+      "loss": 1.9203,
+      "step": 9880
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.265825004383658e-06,
+      "loss": 1.9515,
+      "step": 9890
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.264071541294057e-06,
+      "loss": 1.8098,
+      "step": 9900
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.262318078204454e-06,
+      "loss": 1.8519,
+      "step": 9910
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.260564615114852e-06,
+      "loss": 1.9481,
+      "step": 9920
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.25881115202525e-06,
+      "loss": 1.9001,
+      "step": 9930
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.257057688935649e-06,
+      "loss": 1.7545,
+      "step": 9940
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.255304225846048e-06,
+      "loss": 2.0513,
+      "step": 9950
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.253550762756445e-06,
+      "loss": 2.0191,
+      "step": 9960
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.251797299666843e-06,
+      "loss": 1.8285,
+      "step": 9970
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.25004383657724e-06,
+      "loss": 1.9072,
+      "step": 9980
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.248290373487638e-06,
+      "loss": 1.932,
+      "step": 9990
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.246536910398037e-06,
+      "loss": 1.8423,
+      "step": 10000
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.244783447308436e-06,
+      "loss": 1.976,
+      "step": 10010
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.243029984218834e-06,
+      "loss": 1.982,
+      "step": 10020
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.241276521129231e-06,
+      "loss": 1.9788,
+      "step": 10030
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.239523058039628e-06,
+      "loss": 1.9688,
+      "step": 10040
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.237769594950028e-06,
+      "loss": 1.9353,
+      "step": 10050
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.236016131860425e-06,
+      "loss": 1.7583,
+      "step": 10060
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.234262668770823e-06,
+      "loss": 1.7947,
+      "step": 10070
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.23250920568122e-06,
+      "loss": 2.0301,
+      "step": 10080
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.23075574259162e-06,
+      "loss": 1.875,
+      "step": 10090
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.229002279502017e-06,
+      "loss": 1.8854,
+      "step": 10100
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.227248816412416e-06,
+      "loss": 1.9947,
+      "step": 10110
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.225495353322813e-06,
+      "loss": 1.898,
+      "step": 10120
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.223741890233211e-06,
+      "loss": 1.8292,
+      "step": 10130
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.221988427143608e-06,
+      "loss": 1.9051,
+      "step": 10140
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.220234964054008e-06,
+      "loss": 1.9823,
+      "step": 10150
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.218481500964407e-06,
+      "loss": 2.0169,
+      "step": 10160
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.216728037874804e-06,
+      "loss": 1.7472,
+      "step": 10170
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.214974574785202e-06,
+      "loss": 2.0488,
+      "step": 10180
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.213221111695599e-06,
+      "loss": 1.8345,
+      "step": 10190
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.211467648605997e-06,
+      "loss": 1.8641,
+      "step": 10200
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.209714185516396e-06,
+      "loss": 1.9216,
+      "step": 10210
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.207960722426793e-06,
+      "loss": 2.2013,
+      "step": 10220
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.20620725933719e-06,
+      "loss": 2.0157,
+      "step": 10230
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.20445379624759e-06,
+      "loss": 1.8661,
+      "step": 10240
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.202700333157987e-06,
+      "loss": 2.1425,
+      "step": 10250
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.200946870068387e-06,
+      "loss": 2.1275,
+      "step": 10260
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.199193406978784e-06,
+      "loss": 1.8265,
+      "step": 10270
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.197439943889182e-06,
+      "loss": 2.088,
+      "step": 10280
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.195686480799579e-06,
+      "loss": 1.7487,
+      "step": 10290
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.193933017709978e-06,
+      "loss": 1.8734,
+      "step": 10300
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.192179554620376e-06,
+      "loss": 1.8312,
+      "step": 10310
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.190426091530775e-06,
+      "loss": 1.8073,
+      "step": 10320
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.188672628441172e-06,
+      "loss": 1.837,
+      "step": 10330
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.18691916535157e-06,
+      "loss": 1.903,
+      "step": 10340
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 8.185165702261967e-06,
+      "loss": 1.8039,
+      "step": 10350
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.183412239172366e-06,
+      "loss": 2.0665,
+      "step": 10360
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.181658776082764e-06,
+      "loss": 1.9399,
+      "step": 10370
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.179905312993163e-06,
+      "loss": 1.7886,
+      "step": 10380
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.17815184990356e-06,
+      "loss": 1.76,
+      "step": 10390
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 8.176398386813958e-06,
+      "loss": 1.8434,
+      "step": 10400
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.174644923724356e-06,
+      "loss": 1.8586,
+      "step": 10410
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.172891460634755e-06,
+      "loss": 1.8745,
+      "step": 10420
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.171137997545152e-06,
+      "loss": 1.8868,
+      "step": 10430
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.16938453445555e-06,
+      "loss": 1.8632,
+      "step": 10440
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.167631071365947e-06,
+      "loss": 1.8393,
+      "step": 10450
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 8.165877608276346e-06,
+      "loss": 1.7994,
+      "step": 10460
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.164124145186745e-06,
+      "loss": 1.9666,
+      "step": 10470
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.162370682097143e-06,
+      "loss": 1.8533,
+      "step": 10480
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.16061721900754e-06,
+      "loss": 1.9341,
+      "step": 10490
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.158863755917938e-06,
+      "loss": 1.8166,
+      "step": 10500
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.157110292828337e-06,
+      "loss": 1.9075,
+      "step": 10510
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 8.155356829738735e-06,
+      "loss": 2.0828,
+      "step": 10520
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.153603366649134e-06,
+      "loss": 2.017,
+      "step": 10530
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.151849903559531e-06,
+      "loss": 1.7744,
+      "step": 10540
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.150096440469929e-06,
+      "loss": 1.7367,
+      "step": 10550
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.148342977380326e-06,
+      "loss": 1.8515,
+      "step": 10560
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 8.146589514290725e-06,
+      "loss": 1.8424,
+      "step": 10570
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.144836051201123e-06,
+      "loss": 1.707,
+      "step": 10580
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.14308258811152e-06,
+      "loss": 1.7851,
+      "step": 10590
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.141329125021918e-06,
+      "loss": 1.7744,
+      "step": 10600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.139575661932317e-06,
+      "loss": 2.0111,
+      "step": 10610
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.137822198842716e-06,
+      "loss": 1.9731,
+      "step": 10620
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 8.136068735753114e-06,
+      "loss": 1.82,
+      "step": 10630
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.134315272663511e-06,
+      "loss": 1.7537,
+      "step": 10640
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.132561809573909e-06,
+      "loss": 1.9348,
+      "step": 10650
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.130808346484306e-06,
+      "loss": 1.747,
+      "step": 10660
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.129054883394705e-06,
+      "loss": 1.9937,
+      "step": 10670
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.127301420305104e-06,
+      "loss": 1.7707,
+      "step": 10680
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 8.125547957215502e-06,
+      "loss": 1.8422,
+      "step": 10690
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.1237944941259e-06,
+      "loss": 1.7335,
+      "step": 10700
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.122041031036297e-06,
+      "loss": 1.9627,
+      "step": 10710
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.120287567946696e-06,
+      "loss": 1.8911,
+      "step": 10720
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.118534104857093e-06,
+      "loss": 1.8421,
+      "step": 10730
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.116780641767491e-06,
+      "loss": 1.8757,
+      "step": 10740
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 8.11502717867789e-06,
+      "loss": 1.9267,
+      "step": 10750
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.113273715588288e-06,
+      "loss": 1.859,
+      "step": 10760
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.111520252498685e-06,
+      "loss": 1.7302,
+      "step": 10770
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.109766789409084e-06,
+      "loss": 1.702,
+      "step": 10780
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.108013326319482e-06,
+      "loss": 2.0137,
+      "step": 10790
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 8.10625986322988e-06,
+      "loss": 1.8068,
+      "step": 10800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.104506400140277e-06,
+      "loss": 1.8201,
+      "step": 10810
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.102752937050676e-06,
+      "loss": 1.9747,
+      "step": 10820
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.100999473961075e-06,
+      "loss": 1.9243,
+      "step": 10830
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.099246010871473e-06,
+      "loss": 1.8543,
+      "step": 10840
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.09749254778187e-06,
+      "loss": 2.0614,
+      "step": 10850
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 8.095739084692267e-06,
+      "loss": 1.7591,
+      "step": 10860
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.093985621602665e-06,
+      "loss": 2.0379,
+      "step": 10870
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.092232158513064e-06,
+      "loss": 1.8163,
+      "step": 10880
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.090478695423462e-06,
+      "loss": 1.9022,
+      "step": 10890
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.08872523233386e-06,
+      "loss": 2.1186,
+      "step": 10900
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.086971769244258e-06,
+      "loss": 2.3047,
+      "step": 10910
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.085218306154656e-06,
+      "loss": 1.8662,
+      "step": 10920
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.083464843065055e-06,
+      "loss": 1.7395,
+      "step": 10930
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.081711379975452e-06,
+      "loss": 1.7932,
+      "step": 10940
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.07995791688585e-06,
+      "loss": 1.8792,
+      "step": 10950
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.078204453796247e-06,
+      "loss": 2.0018,
+      "step": 10960
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.076450990706647e-06,
+      "loss": 1.8819,
+      "step": 10970
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.074697527617044e-06,
+      "loss": 1.8711,
+      "step": 10980
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.072944064527443e-06,
+      "loss": 1.9117,
+      "step": 10990
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.07119060143784e-06,
+      "loss": 1.8709,
+      "step": 11000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.069437138348238e-06,
+      "loss": 1.6542,
+      "step": 11010
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.067683675258636e-06,
+      "loss": 1.9611,
+      "step": 11020
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 8.065930212169035e-06,
+      "loss": 1.834,
+      "step": 11030
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.064176749079432e-06,
+      "loss": 1.8074,
+      "step": 11040
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.062423285989831e-06,
+      "loss": 1.8347,
+      "step": 11050
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.060669822900229e-06,
+      "loss": 1.8276,
+      "step": 11060
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.058916359810626e-06,
+      "loss": 2.0004,
+      "step": 11070
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.057162896721024e-06,
+      "loss": 1.8985,
+      "step": 11080
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 8.055409433631423e-06,
+      "loss": 1.7755,
+      "step": 11090
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.05365597054182e-06,
+      "loss": 1.8523,
+      "step": 11100
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.051902507452218e-06,
+      "loss": 1.9431,
+      "step": 11110
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.050149044362617e-06,
+      "loss": 1.904,
+      "step": 11120
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.048395581273015e-06,
+      "loss": 1.6649,
+      "step": 11130
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 8.046642118183414e-06,
+      "loss": 1.7846,
+      "step": 11140
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.044888655093811e-06,
+      "loss": 1.8322,
+      "step": 11150
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.043135192004209e-06,
+      "loss": 2.0217,
+      "step": 11160
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.041381728914606e-06,
+      "loss": 1.809,
+      "step": 11170
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.039628265825005e-06,
+      "loss": 1.7709,
+      "step": 11180
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.037874802735403e-06,
+      "loss": 1.9532,
+      "step": 11190
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 8.036121339645802e-06,
+      "loss": 1.901,
+      "step": 11200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.0343678765562e-06,
+      "loss": 1.8917,
+      "step": 11210
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.032614413466597e-06,
+      "loss": 1.844,
+      "step": 11220
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.030860950376995e-06,
+      "loss": 1.9656,
+      "step": 11230
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.029107487287394e-06,
+      "loss": 1.8356,
+      "step": 11240
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.027354024197791e-06,
+      "loss": 1.7783,
+      "step": 11250
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 8.025600561108189e-06,
+      "loss": 1.9177,
+      "step": 11260
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.023847098018588e-06,
+      "loss": 1.9108,
+      "step": 11270
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.022093634928985e-06,
+      "loss": 2.2531,
+      "step": 11280
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.020340171839383e-06,
+      "loss": 1.7363,
+      "step": 11290
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.018586708749782e-06,
+      "loss": 1.7134,
+      "step": 11300
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.01683324566018e-06,
+      "loss": 1.9232,
+      "step": 11310
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 8.015079782570577e-06,
+      "loss": 1.8327,
+      "step": 11320
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.013326319480974e-06,
+      "loss": 1.7914,
+      "step": 11330
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.011572856391374e-06,
+      "loss": 1.8052,
+      "step": 11340
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.009819393301773e-06,
+      "loss": 1.7882,
+      "step": 11350
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.00806593021217e-06,
+      "loss": 1.9097,
+      "step": 11360
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 8.006312467122568e-06,
+      "loss": 1.8309,
+      "step": 11370
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.004559004032965e-06,
+      "loss": 2.0826,
+      "step": 11380
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.002805540943364e-06,
+      "loss": 1.8474,
+      "step": 11390
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 8.001052077853762e-06,
+      "loss": 1.8953,
+      "step": 11400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.99929861476416e-06,
+      "loss": 1.9536,
+      "step": 11410
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.997545151674558e-06,
+      "loss": 2.0664,
+      "step": 11420
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.995791688584956e-06,
+      "loss": 1.7157,
+      "step": 11430
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.994038225495353e-06,
+      "loss": 1.9595,
+      "step": 11440
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.992284762405753e-06,
+      "loss": 1.6362,
+      "step": 11450
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.99053129931615e-06,
+      "loss": 1.7906,
+      "step": 11460
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.988777836226548e-06,
+      "loss": 1.8785,
+      "step": 11470
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.987024373136945e-06,
+      "loss": 1.7122,
+      "step": 11480
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 7.985270910047344e-06,
+      "loss": 1.7888,
+      "step": 11490
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.983517446957743e-06,
+      "loss": 1.8291,
+      "step": 11500
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.981763983868141e-06,
+      "loss": 1.9154,
+      "step": 11510
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.980010520778538e-06,
+      "loss": 1.9213,
+      "step": 11520
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.978257057688936e-06,
+      "loss": 1.7712,
+      "step": 11530
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 7.976503594599333e-06,
+      "loss": 1.8536,
+      "step": 11540
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.974750131509732e-06,
+      "loss": 1.6643,
+      "step": 11550
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.97299666842013e-06,
+      "loss": 1.8473,
+      "step": 11560
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.971243205330529e-06,
+      "loss": 2.1296,
+      "step": 11570
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.969489742240927e-06,
+      "loss": 1.9098,
+      "step": 11580
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.967736279151324e-06,
+      "loss": 1.7273,
+      "step": 11590
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.965982816061723e-06,
+      "loss": 1.9107,
+      "step": 11600
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.96422935297212e-06,
+      "loss": 1.711,
+      "step": 11610
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.962475889882518e-06,
+      "loss": 1.733,
+      "step": 11620
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.960722426792916e-06,
+      "loss": 1.781,
+      "step": 11630
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.958968963703315e-06,
+      "loss": 1.6836,
+      "step": 11640
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.957215500613712e-06,
+      "loss": 1.7972,
+      "step": 11650
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 7.955462037524112e-06,
+      "loss": 1.8459,
+      "step": 11660
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.953708574434509e-06,
+      "loss": 1.6663,
+      "step": 11670
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.951955111344906e-06,
+      "loss": 1.8542,
+      "step": 11680
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.950201648255304e-06,
+      "loss": 1.8906,
+      "step": 11690
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.948448185165703e-06,
+      "loss": 1.7059,
+      "step": 11700
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.946694722076102e-06,
+      "loss": 1.8149,
+      "step": 11710
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.9449412589865e-06,
+      "loss": 1.6809,
+      "step": 11720
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.943187795896897e-06,
+      "loss": 1.7548,
+      "step": 11730
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.941434332807295e-06,
+      "loss": 1.7966,
+      "step": 11740
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.939680869717692e-06,
+      "loss": 1.9276,
+      "step": 11750
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.937927406628091e-06,
+      "loss": 1.7986,
+      "step": 11760
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 7.936173943538489e-06,
+      "loss": 1.8181,
+      "step": 11770
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.934420480448886e-06,
+      "loss": 1.7895,
+      "step": 11780
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.932667017359286e-06,
+      "loss": 1.7416,
+      "step": 11790
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.930913554269683e-06,
+      "loss": 2.02,
+      "step": 11800
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.929160091180082e-06,
+      "loss": 2.1845,
+      "step": 11810
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.92740662809048e-06,
+      "loss": 2.0006,
+      "step": 11820
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 7.925653165000877e-06,
+      "loss": 1.746,
+      "step": 11830
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.923899701911275e-06,
+      "loss": 1.8519,
+      "step": 11840
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.922146238821674e-06,
+      "loss": 1.7632,
+      "step": 11850
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.920392775732071e-06,
+      "loss": 1.9966,
+      "step": 11860
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.91863931264247e-06,
+      "loss": 1.8129,
+      "step": 11870
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.916885849552868e-06,
+      "loss": 1.9709,
+      "step": 11880
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.915132386463265e-06,
+      "loss": 1.8516,
+      "step": 11890
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.913378923373663e-06,
+      "loss": 1.7642,
+      "step": 11900
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.911625460284062e-06,
+      "loss": 1.8255,
+      "step": 11910
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.90987199719446e-06,
+      "loss": 1.8623,
+      "step": 11920
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.908118534104857e-06,
+      "loss": 1.851,
+      "step": 11930
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 7.906365071015256e-06,
+      "loss": 1.7857,
+      "step": 11940
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.904611607925654e-06,
+      "loss": 1.7843,
+      "step": 11950
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.902858144836051e-06,
+      "loss": 2.0436,
+      "step": 11960
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.90110468174645e-06,
+      "loss": 1.9318,
+      "step": 11970
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.899351218656848e-06,
+      "loss": 1.9273,
+      "step": 11980
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.897597755567245e-06,
+      "loss": 1.7848,
+      "step": 11990
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.895844292477643e-06,
+      "loss": 1.9228,
+      "step": 12000
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.894090829388042e-06,
+      "loss": 1.873,
+      "step": 12010
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.892337366298441e-06,
+      "loss": 1.8616,
+      "step": 12020
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.890583903208839e-06,
+      "loss": 1.8775,
+      "step": 12030
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.888830440119236e-06,
+      "loss": 1.7623,
+      "step": 12040
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.887076977029634e-06,
+      "loss": 1.7119,
+      "step": 12050
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.885323513940033e-06,
+      "loss": 1.7343,
+      "step": 12060
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.88357005085043e-06,
+      "loss": 2.2161,
+      "step": 12070
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.88181658776083e-06,
+      "loss": 1.8154,
+      "step": 12080
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.880063124671227e-06,
+      "loss": 1.9461,
+      "step": 12090
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.878309661581624e-06,
+      "loss": 1.8316,
+      "step": 12100
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.876556198492022e-06,
+      "loss": 1.6284,
+      "step": 12110
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.874802735402421e-06,
+      "loss": 1.859,
+      "step": 12120
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.873049272312818e-06,
+      "loss": 2.0095,
+      "step": 12130
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.871295809223216e-06,
+      "loss": 1.7983,
+      "step": 12140
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.869542346133613e-06,
+      "loss": 1.8029,
+      "step": 12150
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.867788883044013e-06,
+      "loss": 1.6885,
+      "step": 12160
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 7.866035419954412e-06,
+      "loss": 2.2343,
+      "step": 12170
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.86428195686481e-06,
+      "loss": 1.6879,
+      "step": 12180
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.862528493775207e-06,
+      "loss": 2.0421,
+      "step": 12190
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.860775030685604e-06,
+      "loss": 1.8959,
+      "step": 12200
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.859021567596002e-06,
+      "loss": 1.9173,
+      "step": 12210
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.8572681045064e-06,
+      "loss": 1.751,
+      "step": 12220
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.8555146414168e-06,
+      "loss": 1.6287,
+      "step": 12230
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.853761178327197e-06,
+      "loss": 1.621,
+      "step": 12240
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.852007715237595e-06,
+      "loss": 1.5844,
+      "step": 12250
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.850254252147992e-06,
+      "loss": 1.699,
+      "step": 12260
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.848500789058392e-06,
+      "loss": 1.687,
+      "step": 12270
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.846747325968789e-06,
+      "loss": 1.8342,
+      "step": 12280
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.844993862879187e-06,
+      "loss": 2.0125,
+      "step": 12290
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.843240399789586e-06,
+      "loss": 1.6449,
+      "step": 12300
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.841486936699983e-06,
+      "loss": 1.7311,
+      "step": 12310
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.83973347361038e-06,
+      "loss": 1.7764,
+      "step": 12320
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.83798001052078e-06,
+      "loss": 1.7264,
+      "step": 12330
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.836226547431177e-06,
+      "loss": 1.9664,
+      "step": 12340
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.834473084341575e-06,
+      "loss": 1.9273,
+      "step": 12350
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.832719621251972e-06,
+      "loss": 1.6697,
+      "step": 12360
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.830966158162371e-06,
+      "loss": 1.6833,
+      "step": 12370
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.82921269507277e-06,
+      "loss": 1.8938,
+      "step": 12380
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.827459231983168e-06,
+      "loss": 1.8463,
+      "step": 12390
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 7.825705768893566e-06,
+      "loss": 1.945,
+      "step": 12400
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.823952305803963e-06,
+      "loss": 1.7941,
+      "step": 12410
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.82219884271436e-06,
+      "loss": 1.9824,
+      "step": 12420
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.82044537962476e-06,
+      "loss": 1.8737,
+      "step": 12430
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.818691916535157e-06,
+      "loss": 1.9103,
+      "step": 12440
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.816938453445556e-06,
+      "loss": 1.8686,
+      "step": 12450
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 7.815184990355954e-06,
+      "loss": 1.8224,
+      "step": 12460
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.813431527266351e-06,
+      "loss": 2.1603,
+      "step": 12470
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.81167806417675e-06,
+      "loss": 1.9301,
+      "step": 12480
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.809924601087148e-06,
+      "loss": 1.7398,
+      "step": 12490
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.808171137997545e-06,
+      "loss": 1.8025,
+      "step": 12500
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.806417674907943e-06,
+      "loss": 1.9076,
+      "step": 12510
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.804664211818342e-06,
+      "loss": 1.8228,
+      "step": 12520
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.80291074872874e-06,
+      "loss": 1.8824,
+      "step": 12530
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.801157285639139e-06,
+      "loss": 1.7691,
+      "step": 12540
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.799403822549536e-06,
+      "loss": 1.7672,
+      "step": 12550
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.797650359459934e-06,
+      "loss": 1.6306,
+      "step": 12560
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 7.795896896370331e-06,
+      "loss": 1.8269,
+      "step": 12570
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.79414343328073e-06,
+      "loss": 1.8036,
+      "step": 12580
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.792389970191128e-06,
+      "loss": 1.9066,
+      "step": 12590
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.790636507101527e-06,
+      "loss": 1.8864,
+      "step": 12600
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.788883044011925e-06,
+      "loss": 1.6457,
+      "step": 12610
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.787129580922322e-06,
+      "loss": 1.6535,
+      "step": 12620
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 7.78537611783272e-06,
+      "loss": 1.6868,
+      "step": 12630
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.783622654743119e-06,
+      "loss": 1.9189,
+      "step": 12640
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.781869191653516e-06,
+      "loss": 1.8265,
+      "step": 12650
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.780115728563914e-06,
+      "loss": 1.8908,
+      "step": 12660
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.778362265474313e-06,
+      "loss": 1.8056,
+      "step": 12670
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.77660880238471e-06,
+      "loss": 1.7606,
+      "step": 12680
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.77485533929511e-06,
+      "loss": 1.7242,
+      "step": 12690
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.773101876205507e-06,
+      "loss": 1.9335,
+      "step": 12700
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.771348413115904e-06,
+      "loss": 1.8838,
+      "step": 12710
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.769594950026302e-06,
+      "loss": 1.7456,
+      "step": 12720
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.767841486936701e-06,
+      "loss": 1.9107,
+      "step": 12730
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 7.766088023847099e-06,
+      "loss": 1.6641,
+      "step": 12740
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.764334560757498e-06,
+      "loss": 1.7643,
+      "step": 12750
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.762581097667895e-06,
+      "loss": 1.7466,
+      "step": 12760
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.760827634578293e-06,
+      "loss": 1.888,
+      "step": 12770
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.75907417148869e-06,
+      "loss": 1.9414,
+      "step": 12780
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.75732070839909e-06,
+      "loss": 1.6188,
+      "step": 12790
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 7.755567245309487e-06,
+      "loss": 1.8258,
+      "step": 12800
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.753813782219884e-06,
+      "loss": 1.9648,
+      "step": 12810
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.752060319130283e-06,
+      "loss": 1.7147,
+      "step": 12820
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.750306856040681e-06,
+      "loss": 1.8296,
+      "step": 12830
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.748553392951078e-06,
+      "loss": 2.0227,
+      "step": 12840
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.746799929861478e-06,
+      "loss": 1.9037,
+      "step": 12850
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 7.745046466771875e-06,
+      "loss": 1.7033,
+      "step": 12860
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.743293003682273e-06,
+      "loss": 1.8833,
+      "step": 12870
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.74153954059267e-06,
+      "loss": 1.8115,
+      "step": 12880
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.73978607750307e-06,
+      "loss": 2.1681,
+      "step": 12890
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.738032614413468e-06,
+      "loss": 2.053,
+      "step": 12900
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 7.736279151323866e-06,
+      "loss": 1.8253,
+      "step": 12910
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.734525688234263e-06,
+      "loss": 1.809,
+      "step": 12920
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.73277222514466e-06,
+      "loss": 1.67,
+      "step": 12930
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.73101876205506e-06,
+      "loss": 1.8496,
+      "step": 12940
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.729265298965457e-06,
+      "loss": 1.8131,
+      "step": 12950
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.727511835875855e-06,
+      "loss": 1.7704,
+      "step": 12960
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 7.725758372786254e-06,
+      "loss": 1.7849,
+      "step": 12970
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.724004909696652e-06,
+      "loss": 1.7009,
+      "step": 12980
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.722251446607049e-06,
+      "loss": 1.787,
+      "step": 12990
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.720497983517448e-06,
+      "loss": 1.8763,
+      "step": 13000
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.718744520427846e-06,
+      "loss": 1.682,
+      "step": 13010
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.716991057338243e-06,
+      "loss": 1.8758,
+      "step": 13020
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 7.71523759424864e-06,
+      "loss": 1.7936,
+      "step": 13030
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.71348413115904e-06,
+      "loss": 1.7944,
+      "step": 13040
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.711730668069439e-06,
+      "loss": 1.7941,
+      "step": 13050
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.709977204979836e-06,
+      "loss": 1.6941,
+      "step": 13060
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.708223741890234e-06,
+      "loss": 1.6552,
+      "step": 13070
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 7.706470278800631e-06,
+      "loss": 1.776,
+      "step": 13080
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.704716815711029e-06,
+      "loss": 1.8536,
+      "step": 13090
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.702963352621428e-06,
+      "loss": 1.754,
+      "step": 13100
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.701209889531826e-06,
+      "loss": 1.7029,
+      "step": 13110
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.699456426442225e-06,
+      "loss": 1.816,
+      "step": 13120
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.697702963352622e-06,
+      "loss": 1.888,
+      "step": 13130
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 7.69594950026302e-06,
+      "loss": 1.7072,
+      "step": 13140
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.694196037173419e-06,
+      "loss": 1.7477,
+      "step": 13150
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.692442574083816e-06,
+      "loss": 1.7162,
+      "step": 13160
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.690689110994214e-06,
+      "loss": 1.7856,
+      "step": 13170
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.688935647904611e-06,
+      "loss": 1.7761,
+      "step": 13180
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.68718218481501e-06,
+      "loss": 1.7534,
+      "step": 13190
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 7.685428721725408e-06,
+      "loss": 1.6927,
+      "step": 13200
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.683675258635807e-06,
+      "loss": 1.7289,
+      "step": 13210
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.681921795546205e-06,
+      "loss": 1.7594,
+      "step": 13220
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.680168332456602e-06,
+      "loss": 1.9356,
+      "step": 13230
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.678414869367e-06,
+      "loss": 1.7595,
+      "step": 13240
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 7.676661406277399e-06,
+      "loss": 1.9998,
+      "step": 13250
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.674907943187796e-06,
+      "loss": 1.8808,
+      "step": 13260
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.673154480098195e-06,
+      "loss": 1.5751,
+      "step": 13270
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.671401017008593e-06,
+      "loss": 1.9622,
+      "step": 13280
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.66964755391899e-06,
+      "loss": 1.8688,
+      "step": 13290
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.667894090829388e-06,
+      "loss": 1.7383,
+      "step": 13300
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 7.666140627739787e-06,
+      "loss": 1.8158,
+      "step": 13310
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.664387164650184e-06,
+      "loss": 2.0285,
+      "step": 13320
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.662633701560582e-06,
+      "loss": 1.9864,
+      "step": 13330
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.660880238470981e-06,
+      "loss": 1.6351,
+      "step": 13340
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.659126775381379e-06,
+      "loss": 1.7594,
+      "step": 13350
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.657373312291778e-06,
+      "loss": 1.8523,
+      "step": 13360
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 7.655619849202175e-06,
+      "loss": 1.9988,
+      "step": 13370
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.653866386112573e-06,
+      "loss": 1.932,
+      "step": 13380
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.65211292302297e-06,
+      "loss": 1.8219,
+      "step": 13390
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.65035945993337e-06,
+      "loss": 1.8245,
+      "step": 13400
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.648605996843767e-06,
+      "loss": 1.7931,
+      "step": 13410
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.646852533754166e-06,
+      "loss": 1.6874,
+      "step": 13420
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 7.645099070664564e-06,
+      "loss": 1.7678,
+      "step": 13430
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.643345607574961e-06,
+      "loss": 1.836,
+      "step": 13440
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.641592144485358e-06,
+      "loss": 1.8064,
+      "step": 13450
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.639838681395758e-06,
+      "loss": 1.9287,
+      "step": 13460
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.638085218306155e-06,
+      "loss": 1.8624,
+      "step": 13470
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 7.636331755216553e-06,
+      "loss": 1.7724,
+      "step": 13480
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.634578292126952e-06,
+      "loss": 2.1205,
+      "step": 13490
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.63282482903735e-06,
+      "loss": 1.849,
+      "step": 13500
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.631071365947747e-06,
+      "loss": 1.6962,
+      "step": 13510
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.629317902858145e-06,
+      "loss": 1.7845,
+      "step": 13520
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.627564439768543e-06,
+      "loss": 1.6643,
+      "step": 13530
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 7.625810976678941e-06,
+      "loss": 1.8373,
+      "step": 13540
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.624057513589339e-06,
+      "loss": 1.7243,
+      "step": 13550
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.622304050499738e-06,
+      "loss": 1.7297,
+      "step": 13560
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.620550587410136e-06,
+      "loss": 1.7004,
+      "step": 13570
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.618797124320534e-06,
+      "loss": 1.8522,
+      "step": 13580
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.617043661230932e-06,
+      "loss": 1.6293,
+      "step": 13590
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 7.61529019814133e-06,
+      "loss": 1.8768,
+      "step": 13600
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.6135367350517274e-06,
+      "loss": 1.5947,
+      "step": 13610
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.611783271962126e-06,
+      "loss": 1.9046,
+      "step": 13620
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.610029808872524e-06,
+      "loss": 1.7384,
+      "step": 13630
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.608276345782922e-06,
+      "loss": 1.96,
+      "step": 13640
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 7.60652288269332e-06,
+      "loss": 1.9216,
+      "step": 13650
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.604769419603718e-06,
+      "loss": 1.8663,
+      "step": 13660
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.603015956514116e-06,
+      "loss": 1.7196,
+      "step": 13670
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.601262493424514e-06,
+      "loss": 1.6422,
+      "step": 13680
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.5995090303349115e-06,
+      "loss": 1.7454,
+      "step": 13690
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.59775556724531e-06,
+      "loss": 1.9267,
+      "step": 13700
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 7.596002104155709e-06,
+      "loss": 1.6736,
+      "step": 13710
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.5942486410661065e-06,
+      "loss": 2.0302,
+      "step": 13720
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.592495177976504e-06,
+      "loss": 1.8016,
+      "step": 13730
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.590741714886902e-06,
+      "loss": 1.7177,
+      "step": 13740
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.5889882517973e-06,
+      "loss": 1.8043,
+      "step": 13750
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.587234788707698e-06,
+      "loss": 1.7935,
+      "step": 13760
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 7.5854813256180956e-06,
+      "loss": 1.9451,
+      "step": 13770
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.583727862528495e-06,
+      "loss": 1.7675,
+      "step": 13780
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.581974399438893e-06,
+      "loss": 1.8109,
+      "step": 13790
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.5802209363492906e-06,
+      "loss": 1.8297,
+      "step": 13800
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.578467473259689e-06,
+      "loss": 1.922,
+      "step": 13810
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 7.576714010170086e-06,
+      "loss": 1.7459,
+      "step": 13820
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.574960547080485e-06,
+      "loss": 1.626,
+      "step": 13830
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.573207083990882e-06,
+      "loss": 1.9014,
+      "step": 13840
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.5714536209012805e-06,
+      "loss": 2.0172,
+      "step": 13850
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.569700157811679e-06,
+      "loss": 1.759,
+      "step": 13860
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.567946694722077e-06,
+      "loss": 1.8082,
+      "step": 13870
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 7.566193231632475e-06,
+      "loss": 1.8415,
+      "step": 13880
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.564439768542873e-06,
+      "loss": 1.8263,
+      "step": 13890
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.56268630545327e-06,
+      "loss": 1.7729,
+      "step": 13900
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.560932842363669e-06,
+      "loss": 1.7517,
+      "step": 13910
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.559179379274066e-06,
+      "loss": 1.7453,
+      "step": 13920
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.557425916184465e-06,
+      "loss": 1.805,
+      "step": 13930
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 7.555672453094864e-06,
+      "loss": 1.724,
+      "step": 13940
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.553918990005261e-06,
+      "loss": 1.7959,
+      "step": 13950
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.552165526915659e-06,
+      "loss": 1.7262,
+      "step": 13960
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.550412063826057e-06,
+      "loss": 1.9568,
+      "step": 13970
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.5486586007364545e-06,
+      "loss": 1.7033,
+      "step": 13980
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.546905137646853e-06,
+      "loss": 1.6677,
+      "step": 13990
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 7.545151674557252e-06,
+      "loss": 1.747,
+      "step": 14000
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.5433982114676495e-06,
+      "loss": 1.7557,
+      "step": 14010
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.541644748378048e-06,
+      "loss": 1.734,
+      "step": 14020
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.539891285288445e-06,
+      "loss": 1.8904,
+      "step": 14030
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.538137822198844e-06,
+      "loss": 1.902,
+      "step": 14040
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 7.536384359109241e-06,
+      "loss": 1.6249,
+      "step": 14050
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.534630896019639e-06,
+      "loss": 1.9354,
+      "step": 14060
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.532877432930037e-06,
+      "loss": 1.8792,
+      "step": 14070
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.531123969840436e-06,
+      "loss": 1.8047,
+      "step": 14080
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.5293705067508335e-06,
+      "loss": 1.7666,
+      "step": 14090
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.527617043661232e-06,
+      "loss": 1.8408,
+      "step": 14100
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 7.525863580571629e-06,
+      "loss": 1.865,
+      "step": 14110
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.524110117482028e-06,
+      "loss": 1.8534,
+      "step": 14120
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.522356654392425e-06,
+      "loss": 1.7888,
+      "step": 14130
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.5206031913028235e-06,
+      "loss": 1.8283,
+      "step": 14140
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.518849728213223e-06,
+      "loss": 1.7089,
+      "step": 14150
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.51709626512362e-06,
+      "loss": 1.7673,
+      "step": 14160
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 7.515342802034018e-06,
+      "loss": 1.8626,
+      "step": 14170
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.513589338944416e-06,
+      "loss": 1.7429,
+      "step": 14180
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.511835875854813e-06,
+      "loss": 1.726,
+      "step": 14190
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.510082412765212e-06,
+      "loss": 1.7765,
+      "step": 14200
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.508328949675609e-06,
+      "loss": 1.9756,
+      "step": 14210
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 7.5065754865860075e-06,
+      "loss": 1.9395,
+      "step": 14220
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.504822023496407e-06,
+      "loss": 1.5938,
+      "step": 14230
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.503068560406804e-06,
+      "loss": 1.7557,
+      "step": 14240
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.5013150973172025e-06,
+      "loss": 1.9978,
+      "step": 14250
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.4995616342276e-06,
+      "loss": 1.7908,
+      "step": 14260
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.497808171137998e-06,
+      "loss": 1.797,
+      "step": 14270
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 7.496054708048396e-06,
+      "loss": 1.7206,
+      "step": 14280
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.494301244958794e-06,
+      "loss": 1.7381,
+      "step": 14290
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.4925477818691924e-06,
+      "loss": 1.7343,
+      "step": 14300
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.490794318779591e-06,
+      "loss": 1.7146,
+      "step": 14310
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.489040855689988e-06,
+      "loss": 1.8966,
+      "step": 14320
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.4872873926003866e-06,
+      "loss": 1.8773,
+      "step": 14330
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 7.485533929510784e-06,
+      "loss": 1.7078,
+      "step": 14340
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.483780466421182e-06,
+      "loss": 1.7718,
+      "step": 14350
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.48202700333158e-06,
+      "loss": 1.9557,
+      "step": 14360
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.480273540241979e-06,
+      "loss": 1.7657,
+      "step": 14370
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.478520077152377e-06,
+      "loss": 1.9263,
+      "step": 14380
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.476766614062775e-06,
+      "loss": 1.7082,
+      "step": 14390
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 7.475013150973172e-06,
+      "loss": 1.6216,
+      "step": 14400
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.473259687883571e-06,
+      "loss": 1.6599,
+      "step": 14410
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.471506224793968e-06,
+      "loss": 1.5185,
+      "step": 14420
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.4697527617043664e-06,
+      "loss": 1.6954,
+      "step": 14430
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.467999298614764e-06,
+      "loss": 1.9517,
+      "step": 14440
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 7.466245835525163e-06,
+      "loss": 1.728,
+      "step": 14450
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.464492372435561e-06,
+      "loss": 1.8844,
+      "step": 14460
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.462738909345959e-06,
+      "loss": 1.8956,
+      "step": 14470
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.460985446256357e-06,
+      "loss": 1.8501,
+      "step": 14480
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.459231983166755e-06,
+      "loss": 2.0468,
+      "step": 14490
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.457478520077153e-06,
+      "loss": 1.8476,
+      "step": 14500
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 7.4557250569875505e-06,
+      "loss": 1.7897,
+      "step": 14510
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.45397159389795e-06,
+      "loss": 1.8213,
+      "step": 14520
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.452218130808347e-06,
+      "loss": 1.7118,
+      "step": 14530
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.4504646677187455e-06,
+      "loss": 1.7043,
+      "step": 14540
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.448711204629143e-06,
+      "loss": 1.6525,
+      "step": 14550
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.446957741539541e-06,
+      "loss": 1.6368,
+      "step": 14560
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 7.445204278449939e-06,
+      "loss": 1.8015,
+      "step": 14570
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.443450815360337e-06,
+      "loss": 1.8991,
+      "step": 14580
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.441697352270736e-06,
+      "loss": 1.6808,
+      "step": 14590
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.439943889181134e-06,
+      "loss": 1.7441,
+      "step": 14600
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.438190426091531e-06,
+      "loss": 1.9061,
+      "step": 14610
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 7.4364369630019295e-06,
+      "loss": 1.7761,
+      "step": 14620
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.434683499912327e-06,
+      "loss": 1.7722,
+      "step": 14630
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.432930036822725e-06,
+      "loss": 1.696,
+      "step": 14640
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.431176573733123e-06,
+      "loss": 1.6828,
+      "step": 14650
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.429423110643521e-06,
+      "loss": 1.8127,
+      "step": 14660
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.42766964755392e-06,
+      "loss": 1.7007,
+      "step": 14670
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 7.425916184464318e-06,
+      "loss": 1.7722,
+      "step": 14680
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.424162721374716e-06,
+      "loss": 1.7491,
+      "step": 14690
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.422409258285114e-06,
+      "loss": 1.6452,
+      "step": 14700
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.420655795195512e-06,
+      "loss": 1.7903,
+      "step": 14710
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.418902332105909e-06,
+      "loss": 1.8359,
+      "step": 14720
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.417148869016308e-06,
+      "loss": 1.6293,
+      "step": 14730
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 7.415395405926706e-06,
+      "loss": 1.6849,
+      "step": 14740
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.413641942837104e-06,
+      "loss": 1.7987,
+      "step": 14750
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.411888479747502e-06,
+      "loss": 1.7444,
+      "step": 14760
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.4101350166579e-06,
+      "loss": 1.7246,
+      "step": 14770
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.408381553568298e-06,
+      "loss": 1.8903,
+      "step": 14780
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 7.406628090478696e-06,
+      "loss": 1.7117,
+      "step": 14790
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.4048746273890935e-06,
+      "loss": 1.8499,
+      "step": 14800
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.403121164299492e-06,
+      "loss": 1.747,
+      "step": 14810
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.401367701209891e-06,
+      "loss": 1.8911,
+      "step": 14820
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.3996142381202885e-06,
+      "loss": 1.5993,
+      "step": 14830
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.397860775030686e-06,
+      "loss": 1.7534,
+      "step": 14840
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 7.396107311941084e-06,
+      "loss": 1.6845,
+      "step": 14850
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.394353848851482e-06,
+      "loss": 1.5986,
+      "step": 14860
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.39260038576188e-06,
+      "loss": 1.7468,
+      "step": 14870
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.3908469226722775e-06,
+      "loss": 1.8479,
+      "step": 14880
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.389093459582677e-06,
+      "loss": 1.7921,
+      "step": 14890
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.387339996493075e-06,
+      "loss": 1.8092,
+      "step": 14900
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 7.3855865334034725e-06,
+      "loss": 1.779,
+      "step": 14910
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.383833070313871e-06,
+      "loss": 1.778,
+      "step": 14920
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.382079607224268e-06,
+      "loss": 1.7634,
+      "step": 14930
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.380326144134667e-06,
+      "loss": 1.6841,
+      "step": 14940
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.378572681045064e-06,
+      "loss": 1.9134,
+      "step": 14950
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.376819217955463e-06,
+      "loss": 1.8361,
+      "step": 14960
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 7.375065754865861e-06,
+      "loss": 1.9049,
+      "step": 14970
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.373312291776259e-06,
+      "loss": 1.8173,
+      "step": 14980
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.371558828686657e-06,
+      "loss": 1.9787,
+      "step": 14990
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.369805365597055e-06,
+      "loss": 2.0249,
+      "step": 15000
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.368051902507452e-06,
+      "loss": 1.8173,
+      "step": 15010
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 7.366298439417851e-06,
+      "loss": 1.7675,
+      "step": 15020
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.364544976328248e-06,
+      "loss": 1.665,
+      "step": 15030
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.362791513238647e-06,
+      "loss": 1.7102,
+      "step": 15040
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.361038050149045e-06,
+      "loss": 1.6858,
+      "step": 15050
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.359284587059443e-06,
+      "loss": 1.7137,
+      "step": 15060
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.357531123969841e-06,
+      "loss": 1.835,
+      "step": 15070
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 7.355777660880239e-06,
+      "loss": 1.6256,
+      "step": 15080
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.3540241977906365e-06,
+      "loss": 1.7287,
+      "step": 15090
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.352270734701035e-06,
+      "loss": 1.7351,
+      "step": 15100
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.350517271611434e-06,
+      "loss": 1.7337,
+      "step": 15110
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.3487638085218314e-06,
+      "loss": 1.7038,
+      "step": 15120
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.34701034543223e-06,
+      "loss": 1.8254,
+      "step": 15130
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 7.345256882342627e-06,
+      "loss": 1.5932,
+      "step": 15140
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.3435034192530256e-06,
+      "loss": 1.7462,
+      "step": 15150
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.341749956163423e-06,
+      "loss": 1.8713,
+      "step": 15160
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.339996493073821e-06,
+      "loss": 1.6707,
+      "step": 15170
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.338243029984219e-06,
+      "loss": 1.7786,
+      "step": 15180
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 7.336489566894618e-06,
+      "loss": 1.7464,
+      "step": 15190
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.3347361038050155e-06,
+      "loss": 1.7423,
+      "step": 15200
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.332982640715414e-06,
+      "loss": 1.7348,
+      "step": 15210
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.331229177625811e-06,
+      "loss": 1.8394,
+      "step": 15220
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.32947571453621e-06,
+      "loss": 1.731,
+      "step": 15230
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.327722251446607e-06,
+      "loss": 1.7515,
+      "step": 15240
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 7.3259687883570054e-06,
+      "loss": 1.7395,
+      "step": 15250
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.324215325267405e-06,
+      "loss": 1.806,
+      "step": 15260
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.322461862177802e-06,
+      "loss": 1.6191,
+      "step": 15270
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.3207083990881996e-06,
+      "loss": 1.8357,
+      "step": 15280
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.318954935998598e-06,
+      "loss": 1.8826,
+      "step": 15290
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.317201472908995e-06,
+      "loss": 1.846,
+      "step": 15300
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 7.315448009819394e-06,
+      "loss": 1.7033,
+      "step": 15310
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.313694546729791e-06,
+      "loss": 1.8298,
+      "step": 15320
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.31194108364019e-06,
+      "loss": 1.8092,
+      "step": 15330
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.310187620550589e-06,
+      "loss": 1.6802,
+      "step": 15340
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.308434157460986e-06,
+      "loss": 1.8485,
+      "step": 15350
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 7.3066806943713845e-06,
+      "loss": 1.8029,
+      "step": 15360
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.304927231281782e-06,
+      "loss": 1.719,
+      "step": 15370
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.30317376819218e-06,
+      "loss": 1.6665,
+      "step": 15380
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.301420305102578e-06,
+      "loss": 1.9031,
+      "step": 15390
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.299666842012976e-06,
+      "loss": 1.6099,
+      "step": 15400
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.297913378923374e-06,
+      "loss": 1.6879,
+      "step": 15410
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 7.296159915833773e-06,
+      "loss": 1.9254,
+      "step": 15420
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.29440645274417e-06,
+      "loss": 1.7492,
+      "step": 15430
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.2926529896545685e-06,
+      "loss": 1.8584,
+      "step": 15440
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.290899526564966e-06,
+      "loss": 1.8338,
+      "step": 15450
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.289146063475364e-06,
+      "loss": 1.8058,
+      "step": 15460
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.287392600385762e-06,
+      "loss": 1.6067,
+      "step": 15470
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 7.285639137296161e-06,
+      "loss": 1.7478,
+      "step": 15480
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.283885674206559e-06,
+      "loss": 1.7681,
+      "step": 15490
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.282132211116957e-06,
+      "loss": 1.6946,
+      "step": 15500
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.280378748027354e-06,
+      "loss": 1.6292,
+      "step": 15510
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.278625284937753e-06,
+      "loss": 1.7086,
+      "step": 15520
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.27687182184815e-06,
+      "loss": 1.8205,
+      "step": 15530
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 7.275118358758548e-06,
+      "loss": 1.8646,
+      "step": 15540
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.273364895668946e-06,
+      "loss": 1.8226,
+      "step": 15550
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.271611432579345e-06,
+      "loss": 1.7703,
+      "step": 15560
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.269857969489743e-06,
+      "loss": 1.724,
+      "step": 15570
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.268104506400141e-06,
+      "loss": 1.8938,
+      "step": 15580
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 7.266351043310539e-06,
+      "loss": 1.7003,
+      "step": 15590
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.264597580220937e-06,
+      "loss": 1.7656,
+      "step": 15600
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.262844117131335e-06,
+      "loss": 1.8592,
+      "step": 15610
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.2610906540417325e-06,
+      "loss": 1.5586,
+      "step": 15620
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.259337190952132e-06,
+      "loss": 1.7107,
+      "step": 15630
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.257583727862529e-06,
+      "loss": 1.6405,
+      "step": 15640
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 7.2558302647729275e-06,
+      "loss": 1.6943,
+      "step": 15650
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.254076801683325e-06,
+      "loss": 1.8367,
+      "step": 15660
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.252323338593723e-06,
+      "loss": 1.8278,
+      "step": 15670
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.250569875504121e-06,
+      "loss": 1.9294,
+      "step": 15680
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.248816412414519e-06,
+      "loss": 1.7274,
+      "step": 15690
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.247062949324918e-06,
+      "loss": 1.7423,
+      "step": 15700
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 7.245309486235316e-06,
+      "loss": 1.8382,
+      "step": 15710
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.243556023145713e-06,
+      "loss": 1.7647,
+      "step": 15720
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.2418025600561115e-06,
+      "loss": 1.7139,
+      "step": 15730
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.240049096966509e-06,
+      "loss": 1.8781,
+      "step": 15740
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.238295633876907e-06,
+      "loss": 1.4914,
+      "step": 15750
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 7.236542170787305e-06,
+      "loss": 1.7385,
+      "step": 15760
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.234788707697703e-06,
+      "loss": 1.6579,
+      "step": 15770
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.233035244608102e-06,
+      "loss": 1.729,
+      "step": 15780
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.2312817815185e-06,
+      "loss": 1.9178,
+      "step": 15790
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.229528318428898e-06,
+      "loss": 1.8805,
+      "step": 15800
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.227774855339296e-06,
+      "loss": 1.7521,
+      "step": 15810
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 7.226021392249694e-06,
+      "loss": 1.658,
+      "step": 15820
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.224267929160091e-06,
+      "loss": 1.937,
+      "step": 15830
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.22251446607049e-06,
+      "loss": 1.802,
+      "step": 15840
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.220761002980888e-06,
+      "loss": 1.6891,
+      "step": 15850
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.219007539891286e-06,
+      "loss": 1.5953,
+      "step": 15860
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.217254076801684e-06,
+      "loss": 1.6722,
+      "step": 15870
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 7.215500613712082e-06,
+      "loss": 1.7269,
+      "step": 15880
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.21374715062248e-06,
+      "loss": 1.7088,
+      "step": 15890
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.211993687532878e-06,
+      "loss": 1.8087,
+      "step": 15900
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.2102402244432755e-06,
+      "loss": 1.659,
+      "step": 15910
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.208486761353675e-06,
+      "loss": 1.6689,
+      "step": 15920
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 7.206733298264073e-06,
+      "loss": 1.9698,
+      "step": 15930
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.2049798351744704e-06,
+      "loss": 1.6079,
+      "step": 15940
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.203226372084868e-06,
+      "loss": 1.6906,
+      "step": 15950
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.201472908995266e-06,
+      "loss": 1.7156,
+      "step": 15960
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.199719445905664e-06,
+      "loss": 1.617,
+      "step": 15970
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.197965982816062e-06,
+      "loss": 1.7256,
+      "step": 15980
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 7.1962125197264595e-06,
+      "loss": 1.7592,
+      "step": 15990
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.194459056636859e-06,
+      "loss": 1.6377,
+      "step": 16000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.192705593547257e-06,
+      "loss": 1.7695,
+      "step": 16010
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.1909521304576545e-06,
+      "loss": 1.7235,
+      "step": 16020
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.189198667368053e-06,
+      "loss": 1.7895,
+      "step": 16030
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.18744520427845e-06,
+      "loss": 1.756,
+      "step": 16040
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 7.185691741188849e-06,
+      "loss": 1.7966,
+      "step": 16050
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.183938278099246e-06,
+      "loss": 1.8866,
+      "step": 16060
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.182184815009645e-06,
+      "loss": 1.7126,
+      "step": 16070
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.180431351920043e-06,
+      "loss": 1.8344,
+      "step": 16080
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.178677888830441e-06,
+      "loss": 1.645,
+      "step": 16090
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.1769244257408386e-06,
+      "loss": 1.7515,
+      "step": 16100
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 7.175170962651237e-06,
+      "loss": 1.776,
+      "step": 16110
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.173417499561634e-06,
+      "loss": 1.7266,
+      "step": 16120
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.171664036472033e-06,
+      "loss": 1.704,
+      "step": 16130
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.16991057338243e-06,
+      "loss": 1.6979,
+      "step": 16140
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.168157110292829e-06,
+      "loss": 1.6311,
+      "step": 16150
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 7.166403647203227e-06,
+      "loss": 1.5911,
+      "step": 16160
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.164650184113625e-06,
+      "loss": 1.6438,
+      "step": 16170
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.162896721024023e-06,
+      "loss": 1.8946,
+      "step": 16180
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.161143257934421e-06,
+      "loss": 1.7663,
+      "step": 16190
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.1593897948448184e-06,
+      "loss": 1.7166,
+      "step": 16200
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.157636331755217e-06,
+      "loss": 1.6438,
+      "step": 16210
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 7.155882868665616e-06,
+      "loss": 1.7118,
+      "step": 16220
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.154129405576013e-06,
+      "loss": 1.759,
+      "step": 16230
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.152375942486412e-06,
+      "loss": 1.6375,
+      "step": 16240
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.150622479396809e-06,
+      "loss": 1.5803,
+      "step": 16250
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.1488690163072075e-06,
+      "loss": 1.7685,
+      "step": 16260
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.147115553217605e-06,
+      "loss": 1.7441,
+      "step": 16270
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 7.145362090128003e-06,
+      "loss": 1.7356,
+      "step": 16280
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.143608627038402e-06,
+      "loss": 1.8478,
+      "step": 16290
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.1418551639488e-06,
+      "loss": 1.7865,
+      "step": 16300
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.1401017008591975e-06,
+      "loss": 1.7401,
+      "step": 16310
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.138348237769596e-06,
+      "loss": 1.5318,
+      "step": 16320
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 7.136594774679993e-06,
+      "loss": 1.5942,
+      "step": 16330
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.134841311590392e-06,
+      "loss": 1.6936,
+      "step": 16340
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.133087848500789e-06,
+      "loss": 1.7066,
+      "step": 16350
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.131334385411187e-06,
+      "loss": 1.739,
+      "step": 16360
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.129580922321587e-06,
+      "loss": 1.6848,
+      "step": 16370
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.127827459231984e-06,
+      "loss": 1.6103,
+      "step": 16380
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 7.1260739961423815e-06,
+      "loss": 1.6129,
+      "step": 16390
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.12432053305278e-06,
+      "loss": 1.8478,
+      "step": 16400
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.122567069963177e-06,
+      "loss": 1.7495,
+      "step": 16410
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.120813606873576e-06,
+      "loss": 1.7424,
+      "step": 16420
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.119060143783973e-06,
+      "loss": 1.8873,
+      "step": 16430
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.117306680694372e-06,
+      "loss": 1.7719,
+      "step": 16440
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 7.115553217604771e-06,
+      "loss": 1.6918,
+      "step": 16450
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.113799754515168e-06,
+      "loss": 1.7314,
+      "step": 16460
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.1120462914255664e-06,
+      "loss": 1.6067,
+      "step": 16470
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.110292828335964e-06,
+      "loss": 1.8838,
+      "step": 16480
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.108539365246362e-06,
+      "loss": 1.6685,
+      "step": 16490
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.10678590215676e-06,
+      "loss": 1.7442,
+      "step": 16500
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.105032439067158e-06,
+      "loss": 1.6672,
+      "step": 16510
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.103278975977556e-06,
+      "loss": 1.7055,
+      "step": 16520
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.101525512887955e-06,
+      "loss": 1.7834,
+      "step": 16530
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.099772049798352e-06,
+      "loss": 1.6,
+      "step": 16540
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.0980185867087505e-06,
+      "loss": 1.6671,
+      "step": 16550
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 7.096265123619148e-06,
+      "loss": 1.8788,
+      "step": 16560
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.094511660529546e-06,
+      "loss": 1.6939,
+      "step": 16570
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.092758197439944e-06,
+      "loss": 1.5918,
+      "step": 16580
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.091004734350343e-06,
+      "loss": 1.6631,
+      "step": 16590
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.0892512712607404e-06,
+      "loss": 1.8833,
+      "step": 16600
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.087497808171139e-06,
+      "loss": 1.718,
+      "step": 16610
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 7.085744345081536e-06,
+      "loss": 1.7257,
+      "step": 16620
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.0839908819919346e-06,
+      "loss": 1.5866,
+      "step": 16630
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.082237418902332e-06,
+      "loss": 2.0318,
+      "step": 16640
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.08048395581273e-06,
+      "loss": 1.7672,
+      "step": 16650
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.0787304927231296e-06,
+      "loss": 1.6017,
+      "step": 16660
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.076977029633527e-06,
+      "loss": 1.8829,
+      "step": 16670
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 7.075223566543925e-06,
+      "loss": 1.6536,
+      "step": 16680
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.073470103454323e-06,
+      "loss": 1.8304,
+      "step": 16690
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.071716640364721e-06,
+      "loss": 1.7781,
+      "step": 16700
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.069963177275119e-06,
+      "loss": 1.6314,
+      "step": 16710
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.068209714185517e-06,
+      "loss": 1.6779,
+      "step": 16720
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 7.0664562510959144e-06,
+      "loss": 1.5029,
+      "step": 16730
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.064702788006314e-06,
+      "loss": 1.754,
+      "step": 16740
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.062949324916711e-06,
+      "loss": 1.653,
+      "step": 16750
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.061195861827109e-06,
+      "loss": 1.7793,
+      "step": 16760
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.059442398737507e-06,
+      "loss": 1.7573,
+      "step": 16770
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.057688935647905e-06,
+      "loss": 1.5961,
+      "step": 16780
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 7.055935472558303e-06,
+      "loss": 1.8458,
+      "step": 16790
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.054182009468701e-06,
+      "loss": 1.6938,
+      "step": 16800
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0524285463791e-06,
+      "loss": 1.6944,
+      "step": 16810
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.050675083289498e-06,
+      "loss": 1.7915,
+      "step": 16820
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.048921620199895e-06,
+      "loss": 1.6066,
+      "step": 16830
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.0471681571102935e-06,
+      "loss": 1.5091,
+      "step": 16840
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 7.045414694020691e-06,
+      "loss": 1.8254,
+      "step": 16850
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.043661230931089e-06,
+      "loss": 1.6598,
+      "step": 16860
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.041907767841487e-06,
+      "loss": 1.6694,
+      "step": 16870
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.040154304751886e-06,
+      "loss": 1.9253,
+      "step": 16880
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.038400841662284e-06,
+      "loss": 1.6857,
+      "step": 16890
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 7.036647378572682e-06,
+      "loss": 1.7603,
+      "step": 16900
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.03489391548308e-06,
+      "loss": 1.6563,
+      "step": 16910
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.0331404523934776e-06,
+      "loss": 1.858,
+      "step": 16920
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.031386989303876e-06,
+      "loss": 1.7014,
+      "step": 16930
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.029633526214273e-06,
+      "loss": 1.6669,
+      "step": 16940
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.027880063124672e-06,
+      "loss": 1.8076,
+      "step": 16950
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 7.02612660003507e-06,
+      "loss": 1.733,
+      "step": 16960
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.024373136945468e-06,
+      "loss": 1.8408,
+      "step": 16970
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.022619673855866e-06,
+      "loss": 1.8135,
+      "step": 16980
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.020866210766264e-06,
+      "loss": 1.8848,
+      "step": 16990
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.019112747676662e-06,
+      "loss": 1.7066,
+      "step": 17000
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.01735928458706e-06,
+      "loss": 1.832,
+      "step": 17010
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 7.015605821497457e-06,
+      "loss": 1.602,
+      "step": 17020
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.013852358407857e-06,
+      "loss": 1.7419,
+      "step": 17030
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.012098895318254e-06,
+      "loss": 1.6395,
+      "step": 17040
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.010345432228652e-06,
+      "loss": 1.7314,
+      "step": 17050
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.00859196913905e-06,
+      "loss": 1.7174,
+      "step": 17060
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.006838506049448e-06,
+      "loss": 1.7515,
+      "step": 17070
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 7.005085042959846e-06,
+      "loss": 1.5936,
+      "step": 17080
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.003331579870244e-06,
+      "loss": 1.6643,
+      "step": 17090
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 7.0015781167806415e-06,
+      "loss": 2.1187,
+      "step": 17100
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.999824653691041e-06,
+      "loss": 1.8061,
+      "step": 17110
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.998071190601439e-06,
+      "loss": 1.5931,
+      "step": 17120
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 6.9963177275118365e-06,
+      "loss": 1.7101,
+      "step": 17130
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.994564264422235e-06,
+      "loss": 2.1276,
+      "step": 17140
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.992810801332632e-06,
+      "loss": 1.5761,
+      "step": 17150
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.991057338243031e-06,
+      "loss": 1.6742,
+      "step": 17160
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.989303875153428e-06,
+      "loss": 1.7765,
+      "step": 17170
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.987550412063827e-06,
+      "loss": 1.5549,
+      "step": 17180
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 6.985796948974225e-06,
+      "loss": 1.7864,
+      "step": 17190
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.984043485884623e-06,
+      "loss": 1.834,
+      "step": 17200
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.9822900227950205e-06,
+      "loss": 1.8516,
+      "step": 17210
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.980536559705419e-06,
+      "loss": 1.7273,
+      "step": 17220
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.978783096615816e-06,
+      "loss": 1.6218,
+      "step": 17230
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.977029633526215e-06,
+      "loss": 1.7905,
+      "step": 17240
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 6.975276170436614e-06,
+      "loss": 1.6243,
+      "step": 17250
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.973522707347011e-06,
+      "loss": 1.6288,
+      "step": 17260
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.971769244257409e-06,
+      "loss": 1.6385,
+      "step": 17270
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.970015781167807e-06,
+      "loss": 1.7326,
+      "step": 17280
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.968262318078205e-06,
+      "loss": 1.7675,
+      "step": 17290
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 6.966508854988603e-06,
+      "loss": 1.7388,
+      "step": 17300
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.964755391899e-06,
+      "loss": 1.6819,
+      "step": 17310
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.963001928809399e-06,
+      "loss": 1.6272,
+      "step": 17320
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.961248465719798e-06,
+      "loss": 1.7894,
+      "step": 17330
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.959495002630195e-06,
+      "loss": 1.9504,
+      "step": 17340
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.957741539540594e-06,
+      "loss": 1.5337,
+      "step": 17350
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 6.955988076450991e-06,
+      "loss": 1.806,
+      "step": 17360
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9542346133613895e-06,
+      "loss": 1.6582,
+      "step": 17370
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.952481150271787e-06,
+      "loss": 1.6977,
+      "step": 17380
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.950727687182185e-06,
+      "loss": 1.7161,
+      "step": 17390
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.948974224092584e-06,
+      "loss": 1.7536,
+      "step": 17400
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.947220761002982e-06,
+      "loss": 1.6956,
+      "step": 17410
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 6.9454672979133794e-06,
+      "loss": 1.8633,
+      "step": 17420
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.943713834823778e-06,
+      "loss": 1.6237,
+      "step": 17430
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.941960371734175e-06,
+      "loss": 1.665,
+      "step": 17440
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.9402069086445736e-06,
+      "loss": 1.7932,
+      "step": 17450
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.938453445554971e-06,
+      "loss": 1.6671,
+      "step": 17460
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 6.936699982465369e-06,
+      "loss": 1.7175,
+      "step": 17470
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9349465193757685e-06,
+      "loss": 1.6868,
+      "step": 17480
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.933193056286166e-06,
+      "loss": 1.672,
+      "step": 17490
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.9314395931965635e-06,
+      "loss": 1.606,
+      "step": 17500
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.929686130106962e-06,
+      "loss": 1.767,
+      "step": 17510
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.927932667017359e-06,
+      "loss": 1.5884,
+      "step": 17520
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 6.926179203927758e-06,
+      "loss": 1.6212,
+      "step": 17530
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.924425740838155e-06,
+      "loss": 1.7161,
+      "step": 17540
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.922672277748554e-06,
+      "loss": 1.6481,
+      "step": 17550
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.920918814658953e-06,
+      "loss": 1.5393,
+      "step": 17560
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.91916535156935e-06,
+      "loss": 1.7452,
+      "step": 17570
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.917411888479748e-06,
+      "loss": 1.5269,
+      "step": 17580
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 6.915658425390146e-06,
+      "loss": 1.6223,
+      "step": 17590
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.913904962300544e-06,
+      "loss": 1.7545,
+      "step": 17600
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.912151499210942e-06,
+      "loss": 1.8403,
+      "step": 17610
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.910398036121341e-06,
+      "loss": 1.627,
+      "step": 17620
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.908644573031738e-06,
+      "loss": 1.6855,
+      "step": 17630
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.906891109942137e-06,
+      "loss": 1.7026,
+      "step": 17640
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 6.905137646852534e-06,
+      "loss": 1.6032,
+      "step": 17650
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.9033841837629325e-06,
+      "loss": 1.6653,
+      "step": 17660
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.90163072067333e-06,
+      "loss": 1.6498,
+      "step": 17670
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.899877257583728e-06,
+      "loss": 1.5799,
+      "step": 17680
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.898123794494126e-06,
+      "loss": 1.7281,
+      "step": 17690
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 6.896370331404525e-06,
+      "loss": 1.7886,
+      "step": 17700
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.894616868314922e-06,
+      "loss": 1.6021,
+      "step": 17710
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.892863405225321e-06,
+      "loss": 1.6628,
+      "step": 17720
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.891109942135718e-06,
+      "loss": 1.9435,
+      "step": 17730
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.8893564790461165e-06,
+      "loss": 1.8514,
+      "step": 17740
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.887603015956514e-06,
+      "loss": 1.7322,
+      "step": 17750
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 6.885849552866912e-06,
+      "loss": 1.6206,
+      "step": 17760
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.8840960897773115e-06,
+      "loss": 1.6817,
+      "step": 17770
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.882342626687709e-06,
+      "loss": 1.8623,
+      "step": 17780
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.880589163598107e-06,
+      "loss": 1.6825,
+      "step": 17790
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.878835700508505e-06,
+      "loss": 1.6541,
+      "step": 17800
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.877082237418903e-06,
+      "loss": 1.6445,
+      "step": 17810
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 6.875328774329301e-06,
+      "loss": 1.7434,
+      "step": 17820
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.873575311239699e-06,
+      "loss": 1.5815,
+      "step": 17830
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.871821848150096e-06,
+      "loss": 1.9869,
+      "step": 17840
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.870068385060496e-06,
+      "loss": 1.7372,
+      "step": 17850
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.868314921970893e-06,
+      "loss": 1.7056,
+      "step": 17860
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 6.866561458881291e-06,
+      "loss": 1.7598,
+      "step": 17870
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.864807995791689e-06,
+      "loss": 1.7207,
+      "step": 17880
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.863054532702087e-06,
+      "loss": 1.7716,
+      "step": 17890
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.861301069612485e-06,
+      "loss": 1.6479,
+      "step": 17900
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.859547606522883e-06,
+      "loss": 1.7977,
+      "step": 17910
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.857794143433282e-06,
+      "loss": 1.737,
+      "step": 17920
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 6.85604068034368e-06,
+      "loss": 1.713,
+      "step": 17930
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.854287217254077e-06,
+      "loss": 1.8284,
+      "step": 17940
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.8525337541644755e-06,
+      "loss": 1.7747,
+      "step": 17950
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.850780291074873e-06,
+      "loss": 1.7569,
+      "step": 17960
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.849026827985271e-06,
+      "loss": 1.6349,
+      "step": 17970
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.847273364895669e-06,
+      "loss": 1.7761,
+      "step": 17980
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 6.845519901806068e-06,
+      "loss": 1.8345,
+      "step": 17990
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.843766438716466e-06,
+      "loss": 1.8238,
+      "step": 18000
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.842012975626864e-06,
+      "loss": 1.827,
+      "step": 18010
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.840259512537262e-06,
+      "loss": 1.8177,
+      "step": 18020
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.8385060494476595e-06,
+      "loss": 1.5954,
+      "step": 18030
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 6.836752586358058e-06,
+      "loss": 1.6558,
+      "step": 18040
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.834999123268455e-06,
+      "loss": 1.5904,
+      "step": 18050
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.833245660178854e-06,
+      "loss": 1.6747,
+      "step": 18060
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.831492197089252e-06,
+      "loss": 1.7606,
+      "step": 18070
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.82973873399965e-06,
+      "loss": 1.8921,
+      "step": 18080
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.827985270910048e-06,
+      "loss": 1.6982,
+      "step": 18090
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 6.826231807820446e-06,
+      "loss": 1.7036,
+      "step": 18100
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.824478344730844e-06,
+      "loss": 1.6819,
+      "step": 18110
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.822724881641242e-06,
+      "loss": 1.7046,
+      "step": 18120
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.820971418551639e-06,
+      "loss": 1.7223,
+      "step": 18130
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.8192179554620386e-06,
+      "loss": 1.6362,
+      "step": 18140
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.817464492372436e-06,
+      "loss": 1.7749,
+      "step": 18150
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 6.815711029282834e-06,
+      "loss": 1.8443,
+      "step": 18160
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.813957566193232e-06,
+      "loss": 1.7992,
+      "step": 18170
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.81220410310363e-06,
+      "loss": 1.5747,
+      "step": 18180
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.810450640014028e-06,
+      "loss": 1.656,
+      "step": 18190
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.808697176924426e-06,
+      "loss": 1.617,
+      "step": 18200
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.806943713834825e-06,
+      "loss": 1.7054,
+      "step": 18210
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 6.805190250745223e-06,
+      "loss": 1.6403,
+      "step": 18220
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.803436787655621e-06,
+      "loss": 1.6879,
+      "step": 18230
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.8016833245660184e-06,
+      "loss": 1.5362,
+      "step": 18240
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.799929861476417e-06,
+      "loss": 1.7089,
+      "step": 18250
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.798176398386814e-06,
+      "loss": 1.724,
+      "step": 18260
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 6.7964229352972126e-06,
+      "loss": 1.5994,
+      "step": 18270
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.79466947220761e-06,
+      "loss": 1.7352,
+      "step": 18280
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.792916009118009e-06,
+      "loss": 1.6922,
+      "step": 18290
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.791162546028407e-06,
+      "loss": 1.6138,
+      "step": 18300
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.789409082938805e-06,
+      "loss": 1.648,
+      "step": 18310
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.7876556198492025e-06,
+      "loss": 1.545,
+      "step": 18320
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 6.785902156759601e-06,
+      "loss": 1.6807,
+      "step": 18330
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.784148693669998e-06,
+      "loss": 1.7569,
+      "step": 18340
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.782395230580397e-06,
+      "loss": 1.5773,
+      "step": 18350
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.780641767490796e-06,
+      "loss": 1.5246,
+      "step": 18360
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.778888304401193e-06,
+      "loss": 1.6827,
+      "step": 18370
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.777134841311591e-06,
+      "loss": 1.6516,
+      "step": 18380
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 6.775381378221989e-06,
+      "loss": 1.9075,
+      "step": 18390
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.7736279151323866e-06,
+      "loss": 1.701,
+      "step": 18400
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.771874452042785e-06,
+      "loss": 1.5893,
+      "step": 18410
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.770120988953182e-06,
+      "loss": 1.7151,
+      "step": 18420
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.768367525863581e-06,
+      "loss": 1.6583,
+      "step": 18430
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 6.76661406277398e-06,
+      "loss": 1.5747,
+      "step": 18440
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.764860599684377e-06,
+      "loss": 1.6329,
+      "step": 18450
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.763107136594776e-06,
+      "loss": 1.695,
+      "step": 18460
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.761353673505173e-06,
+      "loss": 1.5948,
+      "step": 18470
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.7596002104155715e-06,
+      "loss": 1.5252,
+      "step": 18480
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.757846747325969e-06,
+      "loss": 1.5979,
+      "step": 18490
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 6.756093284236367e-06,
+      "loss": 1.6813,
+      "step": 18500
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.754339821146766e-06,
+      "loss": 1.6701,
+      "step": 18510
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.752586358057164e-06,
+      "loss": 1.6762,
+      "step": 18520
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.750832894967561e-06,
+      "loss": 1.7475,
+      "step": 18530
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.74907943187796e-06,
+      "loss": 1.7754,
+      "step": 18540
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.747325968788357e-06,
+      "loss": 1.7532,
+      "step": 18550
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 6.7455725056987555e-06,
+      "loss": 1.9617,
+      "step": 18560
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.743819042609153e-06,
+      "loss": 1.715,
+      "step": 18570
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.742065579519552e-06,
+      "loss": 1.7033,
+      "step": 18580
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.74031211642995e-06,
+      "loss": 1.7191,
+      "step": 18590
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.738558653340348e-06,
+      "loss": 1.6609,
+      "step": 18600
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.7368051902507455e-06,
+      "loss": 1.5635,
+      "step": 18610
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 6.735051727161144e-06,
+      "loss": 1.8071,
+      "step": 18620
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.733298264071541e-06,
+      "loss": 1.6084,
+      "step": 18630
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.73154480098194e-06,
+      "loss": 1.7121,
+      "step": 18640
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.729791337892337e-06,
+      "loss": 1.7934,
+      "step": 18650
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.728037874802736e-06,
+      "loss": 1.8046,
+      "step": 18660
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 6.726284411713135e-06,
+      "loss": 1.4032,
+      "step": 18670
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.724530948623532e-06,
+      "loss": 1.62,
+      "step": 18680
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.72277748553393e-06,
+      "loss": 1.7595,
+      "step": 18690
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.721024022444328e-06,
+      "loss": 1.8531,
+      "step": 18700
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.719270559354726e-06,
+      "loss": 1.4982,
+      "step": 18710
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.717517096265124e-06,
+      "loss": 1.6061,
+      "step": 18720
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 6.715763633175523e-06,
+      "loss": 1.4729,
+      "step": 18730
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.71401017008592e-06,
+      "loss": 1.6487,
+      "step": 18740
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.712256706996319e-06,
+      "loss": 1.5282,
+      "step": 18750
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.710503243906716e-06,
+      "loss": 1.6188,
+      "step": 18760
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.7087497808171145e-06,
+      "loss": 1.5058,
+      "step": 18770
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.706996317727512e-06,
+      "loss": 1.6336,
+      "step": 18780
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 6.70524285463791e-06,
+      "loss": 1.9261,
+      "step": 18790
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.703489391548308e-06,
+      "loss": 1.9018,
+      "step": 18800
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.701735928458707e-06,
+      "loss": 1.6227,
+      "step": 18810
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.699982465369104e-06,
+      "loss": 1.7751,
+      "step": 18820
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.698229002279503e-06,
+      "loss": 1.6554,
+      "step": 18830
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 6.6964755391899e-06,
+      "loss": 1.6247,
+      "step": 18840
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.6947220761002985e-06,
+      "loss": 1.6806,
+      "step": 18850
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.692968613010696e-06,
+      "loss": 1.712,
+      "step": 18860
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.691215149921094e-06,
+      "loss": 1.4966,
+      "step": 18870
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.6894616868314935e-06,
+      "loss": 1.8652,
+      "step": 18880
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.687708223741891e-06,
+      "loss": 1.5226,
+      "step": 18890
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 6.685954760652289e-06,
+      "loss": 1.6803,
+      "step": 18900
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.684201297562687e-06,
+      "loss": 1.6192,
+      "step": 18910
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.682447834473085e-06,
+      "loss": 1.6572,
+      "step": 18920
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.680694371383483e-06,
+      "loss": 1.5413,
+      "step": 18930
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.678940908293881e-06,
+      "loss": 1.6,
+      "step": 18940
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.677187445204279e-06,
+      "loss": 1.6882,
+      "step": 18950
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 6.6754339821146776e-06,
+      "loss": 1.9096,
+      "step": 18960
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.673680519025075e-06,
+      "loss": 1.5617,
+      "step": 18970
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.671927055935473e-06,
+      "loss": 1.7167,
+      "step": 18980
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.670173592845871e-06,
+      "loss": 1.6592,
+      "step": 18990
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.668420129756269e-06,
+      "loss": 1.5047,
+      "step": 19000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.683,
+      "step": 19010
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.664913203577065e-06,
+      "loss": 1.7125,
+      "step": 19020
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.663159740487464e-06,
+      "loss": 1.5049,
+      "step": 19030
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.661406277397862e-06,
+      "loss": 1.85,
+      "step": 19040
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.659652814308259e-06,
+      "loss": 1.7132,
+      "step": 19050
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.6578993512186574e-06,
+      "loss": 1.7389,
+      "step": 19060
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 6.656145888129055e-06,
+      "loss": 1.552,
+      "step": 19070
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.654392425039453e-06,
+      "loss": 1.5415,
+      "step": 19080
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.652638961949851e-06,
+      "loss": 1.798,
+      "step": 19090
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.65088549886025e-06,
+      "loss": 1.5831,
+      "step": 19100
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.649132035770648e-06,
+      "loss": 1.7881,
+      "step": 19110
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.647378572681046e-06,
+      "loss": 1.8442,
+      "step": 19120
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 6.645625109591444e-06,
+      "loss": 1.8095,
+      "step": 19130
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.6438716465018415e-06,
+      "loss": 1.7129,
+      "step": 19140
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.64211818341224e-06,
+      "loss": 1.8166,
+      "step": 19150
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.640364720322637e-06,
+      "loss": 1.6939,
+      "step": 19160
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.638611257233036e-06,
+      "loss": 1.5992,
+      "step": 19170
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.636857794143434e-06,
+      "loss": 1.6481,
+      "step": 19180
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 6.635104331053832e-06,
+      "loss": 1.6497,
+      "step": 19190
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.63335086796423e-06,
+      "loss": 1.6756,
+      "step": 19200
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.631597404874628e-06,
+      "loss": 1.715,
+      "step": 19210
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.6298439417850256e-06,
+      "loss": 1.6715,
+      "step": 19220
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.628090478695424e-06,
+      "loss": 1.6655,
+      "step": 19230
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 6.626337015605821e-06,
+      "loss": 1.5722,
+      "step": 19240
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.6245835525162205e-06,
+      "loss": 1.796,
+      "step": 19250
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.622830089426618e-06,
+      "loss": 1.6701,
+      "step": 19260
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.621076626337016e-06,
+      "loss": 1.6448,
+      "step": 19270
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.619323163247414e-06,
+      "loss": 1.674,
+      "step": 19280
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.617569700157812e-06,
+      "loss": 1.5919,
+      "step": 19290
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 6.61581623706821e-06,
+      "loss": 1.5365,
+      "step": 19300
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.614062773978608e-06,
+      "loss": 1.8251,
+      "step": 19310
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.612309310889007e-06,
+      "loss": 1.8588,
+      "step": 19320
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.610555847799405e-06,
+      "loss": 1.6982,
+      "step": 19330
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.608802384709803e-06,
+      "loss": 1.7045,
+      "step": 19340
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.6070489216202e-06,
+      "loss": 1.6391,
+      "step": 19350
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 6.605295458530599e-06,
+      "loss": 1.7037,
+      "step": 19360
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.603541995440996e-06,
+      "loss": 1.5936,
+      "step": 19370
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.6017885323513945e-06,
+      "loss": 1.7769,
+      "step": 19380
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.600035069261792e-06,
+      "loss": 1.6068,
+      "step": 19390
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.598281606172191e-06,
+      "loss": 1.6565,
+      "step": 19400
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 6.596528143082589e-06,
+      "loss": 1.6981,
+      "step": 19410
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.594774679992987e-06,
+      "loss": 1.6652,
+      "step": 19420
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.5930212169033845e-06,
+      "loss": 1.5547,
+      "step": 19430
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.591267753813783e-06,
+      "loss": 1.7238,
+      "step": 19440
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.58951429072418e-06,
+      "loss": 1.7191,
+      "step": 19450
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.587760827634579e-06,
+      "loss": 1.5815,
+      "step": 19460
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 6.586007364544978e-06,
+      "loss": 1.6233,
+      "step": 19470
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.584253901455375e-06,
+      "loss": 1.677,
+      "step": 19480
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.582500438365773e-06,
+      "loss": 1.6966,
+      "step": 19490
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.580746975276171e-06,
+      "loss": 1.707,
+      "step": 19500
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.5789935121865685e-06,
+      "loss": 1.6062,
+      "step": 19510
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.577240049096967e-06,
+      "loss": 1.8445,
+      "step": 19520
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 6.575486586007364e-06,
+      "loss": 1.7025,
+      "step": 19530
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.5737331229177635e-06,
+      "loss": 1.6965,
+      "step": 19540
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.571979659828162e-06,
+      "loss": 1.7458,
+      "step": 19550
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.570226196738559e-06,
+      "loss": 1.6596,
+      "step": 19560
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.568472733648958e-06,
+      "loss": 1.6628,
+      "step": 19570
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 6.566719270559355e-06,
+      "loss": 2.0629,
+      "step": 19580
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.5649658074697534e-06,
+      "loss": 1.6355,
+      "step": 19590
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.563212344380151e-06,
+      "loss": 1.7344,
+      "step": 19600
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.561458881290549e-06,
+      "loss": 1.6324,
+      "step": 19610
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.559705418200948e-06,
+      "loss": 1.9835,
+      "step": 19620
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.557951955111346e-06,
+      "loss": 1.7162,
+      "step": 19630
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 6.556198492021743e-06,
+      "loss": 1.5926,
+      "step": 19640
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.554445028932142e-06,
+      "loss": 1.8371,
+      "step": 19650
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.552691565842539e-06,
+      "loss": 1.6316,
+      "step": 19660
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.5509381027529375e-06,
+      "loss": 1.9405,
+      "step": 19670
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.549184639663335e-06,
+      "loss": 1.5229,
+      "step": 19680
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.547431176573734e-06,
+      "loss": 1.7459,
+      "step": 19690
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 6.545677713484132e-06,
+      "loss": 1.887,
+      "step": 19700
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.54392425039453e-06,
+      "loss": 1.6624,
+      "step": 19710
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.5421707873049274e-06,
+      "loss": 1.6193,
+      "step": 19720
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.540417324215326e-06,
+      "loss": 1.6889,
+      "step": 19730
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.538663861125723e-06,
+      "loss": 1.7709,
+      "step": 19740
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.536910398036122e-06,
+      "loss": 1.7502,
+      "step": 19750
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 6.535156934946519e-06,
+      "loss": 1.8655,
+      "step": 19760
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.533403471856918e-06,
+      "loss": 1.5918,
+      "step": 19770
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.5316500087673166e-06,
+      "loss": 1.6995,
+      "step": 19780
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.529896545677714e-06,
+      "loss": 1.7271,
+      "step": 19790
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.528143082588112e-06,
+      "loss": 1.9522,
+      "step": 19800
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 6.52638961949851e-06,
+      "loss": 1.6665,
+      "step": 19810
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.524636156408908e-06,
+      "loss": 1.5873,
+      "step": 19820
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.522882693319306e-06,
+      "loss": 1.607,
+      "step": 19830
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.521129230229705e-06,
+      "loss": 1.5412,
+      "step": 19840
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.519375767140102e-06,
+      "loss": 1.7616,
+      "step": 19850
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.517622304050501e-06,
+      "loss": 1.6809,
+      "step": 19860
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 6.515868840960898e-06,
+      "loss": 1.5975,
+      "step": 19870
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.514115377871296e-06,
+      "loss": 1.6624,
+      "step": 19880
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.512361914781694e-06,
+      "loss": 1.5951,
+      "step": 19890
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.510608451692092e-06,
+      "loss": 1.5557,
+      "step": 19900
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.508854988602491e-06,
+      "loss": 1.6458,
+      "step": 19910
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.507101525512889e-06,
+      "loss": 1.8276,
+      "step": 19920
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 6.505348062423286e-06,
+      "loss": 1.6954,
+      "step": 19930
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.503594599333685e-06,
+      "loss": 1.8568,
+      "step": 19940
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.501841136244082e-06,
+      "loss": 1.6765,
+      "step": 19950
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.5000876731544805e-06,
+      "loss": 1.725,
+      "step": 19960
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.498334210064878e-06,
+      "loss": 1.6068,
+      "step": 19970
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 6.496580746975276e-06,
+      "loss": 1.5584,
+      "step": 19980
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.4948272838856755e-06,
+      "loss": 1.6247,
+      "step": 19990
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.493073820796073e-06,
+      "loss": 1.7621,
+      "step": 20000
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.491320357706471e-06,
+      "loss": 1.8063,
+      "step": 20010
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.489566894616869e-06,
+      "loss": 1.6019,
+      "step": 20020
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.487813431527267e-06,
+      "loss": 1.6838,
+      "step": 20030
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 6.4860599684376646e-06,
+      "loss": 1.7143,
+      "step": 20040
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.484306505348063e-06,
+      "loss": 1.6654,
+      "step": 20050
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.482553042258461e-06,
+      "loss": 1.794,
+      "step": 20060
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.4807995791688595e-06,
+      "loss": 1.7265,
+      "step": 20070
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.479046116079257e-06,
+      "loss": 1.682,
+      "step": 20080
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.477292652989655e-06,
+      "loss": 1.6397,
+      "step": 20090
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 6.475539189900053e-06,
+      "loss": 1.6531,
+      "step": 20100
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.473785726810451e-06,
+      "loss": 1.4973,
+      "step": 20110
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.472032263720849e-06,
+      "loss": 1.7023,
+      "step": 20120
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.470278800631247e-06,
+      "loss": 1.7776,
+      "step": 20130
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.468525337541645e-06,
+      "loss": 1.5404,
+      "step": 20140
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.466771874452044e-06,
+      "loss": 1.7565,
+      "step": 20150
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 6.465018411362441e-06,
+      "loss": 1.7596,
+      "step": 20160
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.463264948272839e-06,
+      "loss": 1.6828,
+      "step": 20170
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.461511485183237e-06,
+      "loss": 1.6402,
+      "step": 20180
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.459758022093635e-06,
+      "loss": 1.6644,
+      "step": 20190
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.458004559004033e-06,
+      "loss": 1.7584,
+      "step": 20200
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 6.456251095914432e-06,
+      "loss": 1.515,
+      "step": 20210
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.45449763282483e-06,
+      "loss": 1.5802,
+      "step": 20220
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.452744169735228e-06,
+      "loss": 1.8021,
+      "step": 20230
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.450990706645626e-06,
+      "loss": 1.6874,
+      "step": 20240
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.4492372435560235e-06,
+      "loss": 1.6734,
+      "step": 20250
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.447483780466422e-06,
+      "loss": 1.6061,
+      "step": 20260
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 6.445730317376819e-06,
+      "loss": 1.6151,
+      "step": 20270
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.4439768542872184e-06,
+      "loss": 1.7378,
+      "step": 20280
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.442223391197616e-06,
+      "loss": 1.7177,
+      "step": 20290
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.440469928108014e-06,
+      "loss": 1.6332,
+      "step": 20300
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.438716465018412e-06,
+      "loss": 1.6894,
+      "step": 20310
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.43696300192881e-06,
+      "loss": 1.6461,
+      "step": 20320
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 6.4352095388392075e-06,
+      "loss": 1.5732,
+      "step": 20330
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.433456075749606e-06,
+      "loss": 1.6442,
+      "step": 20340
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.431702612660003e-06,
+      "loss": 1.6534,
+      "step": 20350
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.4299491495704025e-06,
+      "loss": 1.5502,
+      "step": 20360
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.4281956864808e-06,
+      "loss": 1.6126,
+      "step": 20370
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 6.426442223391198e-06,
+      "loss": 1.6102,
+      "step": 20380
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.424688760301596e-06,
+      "loss": 1.4931,
+      "step": 20390
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.422935297211994e-06,
+      "loss": 1.7497,
+      "step": 20400
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.421181834122392e-06,
+      "loss": 1.6019,
+      "step": 20410
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.41942837103279e-06,
+      "loss": 1.7002,
+      "step": 20420
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.417674907943189e-06,
+      "loss": 1.7407,
+      "step": 20430
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 6.4159214448535866e-06,
+      "loss": 1.5857,
+      "step": 20440
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.414167981763985e-06,
+      "loss": 1.7686,
+      "step": 20450
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.412414518674382e-06,
+      "loss": 1.7172,
+      "step": 20460
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.410661055584781e-06,
+      "loss": 1.6329,
+      "step": 20470
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.408907592495178e-06,
+      "loss": 1.5091,
+      "step": 20480
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.4071541294055765e-06,
+      "loss": 1.6452,
+      "step": 20490
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.405400666315975e-06,
+      "loss": 1.5853,
+      "step": 20500
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.403647203226373e-06,
+      "loss": 1.5134,
+      "step": 20510
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.401893740136771e-06,
+      "loss": 1.602,
+      "step": 20520
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.400140277047169e-06,
+      "loss": 1.5143,
+      "step": 20530
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.3983868139575664e-06,
+      "loss": 1.6875,
+      "step": 20540
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 6.396633350867965e-06,
+      "loss": 1.585,
+      "step": 20550
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.394879887778362e-06,
+      "loss": 1.703,
+      "step": 20560
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.3931264246887606e-06,
+      "loss": 1.7589,
+      "step": 20570
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.391372961599159e-06,
+      "loss": 1.6978,
+      "step": 20580
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.389619498509557e-06,
+      "loss": 1.4896,
+      "step": 20590
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.387866035419955e-06,
+      "loss": 1.5789,
+      "step": 20600
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.386112572330353e-06,
+      "loss": 1.5342,
+      "step": 20610
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.3843591092407505e-06,
+      "loss": 1.499,
+      "step": 20620
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.382605646151149e-06,
+      "loss": 1.6787,
+      "step": 20630
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.380852183061546e-06,
+      "loss": 1.8626,
+      "step": 20640
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.3790987199719455e-06,
+      "loss": 1.5965,
+      "step": 20650
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.377345256882344e-06,
+      "loss": 1.5087,
+      "step": 20660
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.375591793792741e-06,
+      "loss": 1.6831,
+      "step": 20670
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.37383833070314e-06,
+      "loss": 1.5521,
+      "step": 20680
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.372084867613537e-06,
+      "loss": 1.6879,
+      "step": 20690
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.370331404523935e-06,
+      "loss": 1.5881,
+      "step": 20700
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.368577941434333e-06,
+      "loss": 1.5335,
+      "step": 20710
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.366824478344731e-06,
+      "loss": 1.8092,
+      "step": 20720
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 6.3650710152551295e-06,
+      "loss": 1.6677,
+      "step": 20730
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.363317552165528e-06,
+      "loss": 1.8414,
+      "step": 20740
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.361564089075925e-06,
+      "loss": 1.745,
+      "step": 20750
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.359810625986324e-06,
+      "loss": 1.919,
+      "step": 20760
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.358057162896721e-06,
+      "loss": 1.8182,
+      "step": 20770
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 6.3563036998071195e-06,
+      "loss": 1.746,
+      "step": 20780
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.354550236717517e-06,
+      "loss": 1.6005,
+      "step": 20790
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.352796773627916e-06,
+      "loss": 1.7569,
+      "step": 20800
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.351043310538314e-06,
+      "loss": 1.7826,
+      "step": 20810
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.349289847448712e-06,
+      "loss": 1.6532,
+      "step": 20820
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.347536384359109e-06,
+      "loss": 1.6096,
+      "step": 20830
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 6.345782921269508e-06,
+      "loss": 1.6813,
+      "step": 20840
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.344029458179905e-06,
+      "loss": 1.7375,
+      "step": 20850
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.3422759950903035e-06,
+      "loss": 1.6832,
+      "step": 20860
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.340522532000703e-06,
+      "loss": 1.6425,
+      "step": 20870
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.3387690689111e-06,
+      "loss": 1.4821,
+      "step": 20880
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.3370156058214985e-06,
+      "loss": 1.6386,
+      "step": 20890
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 6.335262142731896e-06,
+      "loss": 1.7637,
+      "step": 20900
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.333508679642294e-06,
+      "loss": 1.576,
+      "step": 20910
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.331755216552692e-06,
+      "loss": 1.4946,
+      "step": 20920
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.33000175346309e-06,
+      "loss": 1.6015,
+      "step": 20930
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.328248290373488e-06,
+      "loss": 1.6751,
+      "step": 20940
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 6.326494827283887e-06,
+      "loss": 1.8371,
+      "step": 20950
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.324741364194284e-06,
+      "loss": 1.646,
+      "step": 20960
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.322987901104683e-06,
+      "loss": 1.5164,
+      "step": 20970
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.32123443801508e-06,
+      "loss": 1.7619,
+      "step": 20980
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.319480974925478e-06,
+      "loss": 1.7087,
+      "step": 20990
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.317727511835876e-06,
+      "loss": 1.7645,
+      "step": 21000
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 6.315974048746274e-06,
+      "loss": 1.6186,
+      "step": 21010
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.314220585656673e-06,
+      "loss": 1.8573,
+      "step": 21020
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.312467122567071e-06,
+      "loss": 1.5355,
+      "step": 21030
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.310713659477468e-06,
+      "loss": 1.4969,
+      "step": 21040
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.308960196387867e-06,
+      "loss": 1.7371,
+      "step": 21050
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.307206733298264e-06,
+      "loss": 1.5866,
+      "step": 21060
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 6.3054532702086625e-06,
+      "loss": 1.7841,
+      "step": 21070
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.30369980711906e-06,
+      "loss": 1.8672,
+      "step": 21080
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.301946344029458e-06,
+      "loss": 1.6374,
+      "step": 21090
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.3001928809398574e-06,
+      "loss": 1.7639,
+      "step": 21100
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.298439417850255e-06,
+      "loss": 1.6039,
+      "step": 21110
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 6.296685954760653e-06,
+      "loss": 1.5898,
+      "step": 21120
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.294932491671051e-06,
+      "loss": 1.7627,
+      "step": 21130
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.293179028581449e-06,
+      "loss": 1.7706,
+      "step": 21140
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.2914255654918465e-06,
+      "loss": 1.613,
+      "step": 21150
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.289672102402245e-06,
+      "loss": 1.5772,
+      "step": 21160
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.287918639312643e-06,
+      "loss": 1.4731,
+      "step": 21170
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 6.2861651762230415e-06,
+      "loss": 1.8445,
+      "step": 21180
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.284411713133439e-06,
+      "loss": 1.5867,
+      "step": 21190
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.282658250043837e-06,
+      "loss": 1.5108,
+      "step": 21200
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.280904786954235e-06,
+      "loss": 1.4657,
+      "step": 21210
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.279151323864633e-06,
+      "loss": 1.6645,
+      "step": 21220
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.277397860775031e-06,
+      "loss": 1.6175,
+      "step": 21230
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 6.27564439768543e-06,
+      "loss": 1.5807,
+      "step": 21240
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.273890934595827e-06,
+      "loss": 1.6869,
+      "step": 21250
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.2721374715062256e-06,
+      "loss": 1.6681,
+      "step": 21260
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.270384008416623e-06,
+      "loss": 1.7705,
+      "step": 21270
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.268630545327021e-06,
+      "loss": 1.5908,
+      "step": 21280
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.266877082237419e-06,
+      "loss": 1.5956,
+      "step": 21290
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 6.265123619147817e-06,
+      "loss": 1.693,
+      "step": 21300
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.263370156058215e-06,
+      "loss": 1.9131,
+      "step": 21310
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.261616692968614e-06,
+      "loss": 1.5826,
+      "step": 21320
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.259863229879012e-06,
+      "loss": 1.7576,
+      "step": 21330
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.25810976678941e-06,
+      "loss": 1.6575,
+      "step": 21340
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 6.256356303699808e-06,
+      "loss": 1.6885,
+      "step": 21350
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.2546028406102054e-06,
+      "loss": 1.5871,
+      "step": 21360
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.252849377520604e-06,
+      "loss": 1.8073,
+      "step": 21370
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.251095914431001e-06,
+      "loss": 1.6795,
+      "step": 21380
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.2493424513414e-06,
+      "loss": 1.5491,
+      "step": 21390
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.247588988251798e-06,
+      "loss": 1.8934,
+      "step": 21400
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 6.245835525162196e-06,
+      "loss": 1.7535,
+      "step": 21410
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.244082062072594e-06,
+      "loss": 1.8098,
+      "step": 21420
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.242328598982992e-06,
+      "loss": 1.615,
+      "step": 21430
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.2405751358933895e-06,
+      "loss": 1.8232,
+      "step": 21440
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.238821672803788e-06,
+      "loss": 1.5875,
+      "step": 21450
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.237068209714185e-06,
+      "loss": 1.6513,
+      "step": 21460
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 6.2353147466245845e-06,
+      "loss": 1.5792,
+      "step": 21470
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.233561283534982e-06,
+      "loss": 1.5908,
+      "step": 21480
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.23180782044538e-06,
+      "loss": 1.5656,
+      "step": 21490
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.230054357355778e-06,
+      "loss": 1.5914,
+      "step": 21500
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.228300894266176e-06,
+      "loss": 1.5018,
+      "step": 21510
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 6.2265474311765736e-06,
+      "loss": 1.6647,
+      "step": 21520
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.224793968086972e-06,
+      "loss": 1.5867,
+      "step": 21530
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.223040504997371e-06,
+      "loss": 1.5231,
+      "step": 21540
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.2212870419077685e-06,
+      "loss": 1.7193,
+      "step": 21550
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.219533578818167e-06,
+      "loss": 1.6993,
+      "step": 21560
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.217780115728564e-06,
+      "loss": 1.5411,
+      "step": 21570
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 6.216026652638963e-06,
+      "loss": 1.7324,
+      "step": 21580
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.21427318954936e-06,
+      "loss": 1.5322,
+      "step": 21590
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.2125197264597585e-06,
+      "loss": 1.902,
+      "step": 21600
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.210766263370157e-06,
+      "loss": 1.6927,
+      "step": 21610
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.209012800280555e-06,
+      "loss": 1.6522,
+      "step": 21620
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.207259337190953e-06,
+      "loss": 2.0485,
+      "step": 21630
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 6.205505874101351e-06,
+      "loss": 1.7328,
+      "step": 21640
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.203752411011748e-06,
+      "loss": 1.6748,
+      "step": 21650
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.201998947922147e-06,
+      "loss": 1.6455,
+      "step": 21660
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.200245484832544e-06,
+      "loss": 1.6091,
+      "step": 21670
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.1984920217429425e-06,
+      "loss": 1.6241,
+      "step": 21680
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 6.196738558653341e-06,
+      "loss": 1.6819,
+      "step": 21690
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.194985095563739e-06,
+      "loss": 1.6307,
+      "step": 21700
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.193231632474137e-06,
+      "loss": 1.6679,
+      "step": 21710
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.191478169384535e-06,
+      "loss": 1.8705,
+      "step": 21720
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.1897247062949325e-06,
+      "loss": 1.5966,
+      "step": 21730
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.187971243205331e-06,
+      "loss": 1.6129,
+      "step": 21740
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 6.186217780115728e-06,
+      "loss": 1.6951,
+      "step": 21750
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.1844643170261275e-06,
+      "loss": 1.867,
+      "step": 21760
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.182710853936526e-06,
+      "loss": 1.714,
+      "step": 21770
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.180957390846923e-06,
+      "loss": 1.6992,
+      "step": 21780
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.179203927757322e-06,
+      "loss": 1.641,
+      "step": 21790
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.177450464667719e-06,
+      "loss": 1.6091,
+      "step": 21800
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 6.175697001578117e-06,
+      "loss": 1.5441,
+      "step": 21810
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.173943538488515e-06,
+      "loss": 1.737,
+      "step": 21820
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.172190075398914e-06,
+      "loss": 1.6071,
+      "step": 21830
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.1704366123093115e-06,
+      "loss": 1.5721,
+      "step": 21840
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.16868314921971e-06,
+      "loss": 1.6271,
+      "step": 21850
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.166929686130107e-06,
+      "loss": 1.7817,
+      "step": 21860
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 6.165176223040506e-06,
+      "loss": 1.6104,
+      "step": 21870
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.163422759950903e-06,
+      "loss": 1.5856,
+      "step": 21880
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.1616692968613015e-06,
+      "loss": 1.6971,
+      "step": 21890
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.159915833771699e-06,
+      "loss": 1.5579,
+      "step": 21900
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.158162370682098e-06,
+      "loss": 1.546,
+      "step": 21910
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 6.156408907592496e-06,
+      "loss": 1.5955,
+      "step": 21920
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.154655444502894e-06,
+      "loss": 1.626,
+      "step": 21930
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.152901981413291e-06,
+      "loss": 1.5496,
+      "step": 21940
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.15114851832369e-06,
+      "loss": 1.533,
+      "step": 21950
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.149395055234087e-06,
+      "loss": 1.7958,
+      "step": 21960
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.1476415921444855e-06,
+      "loss": 1.5653,
+      "step": 21970
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 6.145888129054885e-06,
+      "loss": 1.9646,
+      "step": 21980
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.144134665965282e-06,
+      "loss": 1.7286,
+      "step": 21990
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.1423812028756805e-06,
+      "loss": 1.7461,
+      "step": 22000
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.140627739786078e-06,
+      "loss": 1.5231,
+      "step": 22010
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.138874276696476e-06,
+      "loss": 1.5648,
+      "step": 22020
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.137120813606874e-06,
+      "loss": 1.8033,
+      "step": 22030
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 6.135367350517272e-06,
+      "loss": 1.7259,
+      "step": 22040
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.13361388742767e-06,
+      "loss": 1.666,
+      "step": 22050
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.131860424338069e-06,
+      "loss": 1.5328,
+      "step": 22060
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.130106961248466e-06,
+      "loss": 1.6303,
+      "step": 22070
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.1283534981588646e-06,
+      "loss": 1.8127,
+      "step": 22080
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 6.126600035069262e-06,
+      "loss": 1.5124,
+      "step": 22090
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.12484657197966e-06,
+      "loss": 1.4758,
+      "step": 22100
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.123093108890058e-06,
+      "loss": 1.6448,
+      "step": 22110
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.121339645800456e-06,
+      "loss": 1.5328,
+      "step": 22120
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.1195861827108545e-06,
+      "loss": 1.5284,
+      "step": 22130
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.117832719621253e-06,
+      "loss": 1.6041,
+      "step": 22140
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 6.11607925653165e-06,
+      "loss": 1.6256,
+      "step": 22150
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.114325793442049e-06,
+      "loss": 1.7044,
+      "step": 22160
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.112572330352446e-06,
+      "loss": 1.5548,
+      "step": 22170
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.1108188672628444e-06,
+      "loss": 1.5124,
+      "step": 22180
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.109065404173242e-06,
+      "loss": 1.7425,
+      "step": 22190
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.107311941083641e-06,
+      "loss": 1.6523,
+      "step": 22200
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 6.105558477994039e-06,
+      "loss": 1.4788,
+      "step": 22210
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.103805014904437e-06,
+      "loss": 1.6073,
+      "step": 22220
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.102051551814835e-06,
+      "loss": 1.7168,
+      "step": 22230
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.100298088725233e-06,
+      "loss": 1.6962,
+      "step": 22240
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.098544625635631e-06,
+      "loss": 1.6731,
+      "step": 22250
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.0967911625460285e-06,
+      "loss": 1.5629,
+      "step": 22260
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 6.095037699456427e-06,
+      "loss": 1.5973,
+      "step": 22270
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.093284236366825e-06,
+      "loss": 1.6316,
+      "step": 22280
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.0915307732772235e-06,
+      "loss": 1.6893,
+      "step": 22290
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.089777310187621e-06,
+      "loss": 1.5459,
+      "step": 22300
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.088023847098019e-06,
+      "loss": 1.5398,
+      "step": 22310
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 6.086270384008417e-06,
+      "loss": 1.641,
+      "step": 22320
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.084516920918815e-06,
+      "loss": 1.6506,
+      "step": 22330
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0827634578292126e-06,
+      "loss": 1.5552,
+      "step": 22340
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.081009994739612e-06,
+      "loss": 1.6922,
+      "step": 22350
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.079256531650009e-06,
+      "loss": 1.8395,
+      "step": 22360
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.0775030685604075e-06,
+      "loss": 1.5299,
+      "step": 22370
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 6.075749605470805e-06,
+      "loss": 1.6045,
+      "step": 22380
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.073996142381203e-06,
+      "loss": 1.7557,
+      "step": 22390
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.072242679291601e-06,
+      "loss": 1.9578,
+      "step": 22400
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.070489216201999e-06,
+      "loss": 1.5137,
+      "step": 22410
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.068735753112397e-06,
+      "loss": 1.7645,
+      "step": 22420
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.066982290022796e-06,
+      "loss": 1.5173,
+      "step": 22430
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 6.065228826933194e-06,
+      "loss": 1.7633,
+      "step": 22440
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.063475363843592e-06,
+      "loss": 1.7801,
+      "step": 22450
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.06172190075399e-06,
+      "loss": 1.6121,
+      "step": 22460
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.059968437664387e-06,
+      "loss": 1.6613,
+      "step": 22470
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.058214974574786e-06,
+      "loss": 1.5641,
+      "step": 22480
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 6.056461511485183e-06,
+      "loss": 1.6404,
+      "step": 22490
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.054708048395582e-06,
+      "loss": 1.5984,
+      "step": 22500
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.05295458530598e-06,
+      "loss": 1.6687,
+      "step": 22510
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.051201122216378e-06,
+      "loss": 1.8726,
+      "step": 22520
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.049447659126776e-06,
+      "loss": 1.7498,
+      "step": 22530
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.047694196037174e-06,
+      "loss": 1.5018,
+      "step": 22540
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 6.0459407329475715e-06,
+      "loss": 1.5106,
+      "step": 22550
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.04418726985797e-06,
+      "loss": 1.7419,
+      "step": 22560
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.042433806768369e-06,
+      "loss": 1.501,
+      "step": 22570
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.0406803436787664e-06,
+      "loss": 1.5678,
+      "step": 22580
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.038926880589164e-06,
+      "loss": 1.5888,
+      "step": 22590
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.037173417499562e-06,
+      "loss": 1.7347,
+      "step": 22600
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 6.03541995440996e-06,
+      "loss": 1.8518,
+      "step": 22610
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.033666491320358e-06,
+      "loss": 1.7014,
+      "step": 22620
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.0319130282307555e-06,
+      "loss": 1.6661,
+      "step": 22630
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.030159565141154e-06,
+      "loss": 1.6047,
+      "step": 22640
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.028406102051553e-06,
+      "loss": 1.7488,
+      "step": 22650
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 6.0266526389619505e-06,
+      "loss": 1.5601,
+      "step": 22660
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.024899175872349e-06,
+      "loss": 1.496,
+      "step": 22670
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.023145712782746e-06,
+      "loss": 1.5087,
+      "step": 22680
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.021392249693145e-06,
+      "loss": 1.5525,
+      "step": 22690
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.019638786603542e-06,
+      "loss": 1.8245,
+      "step": 22700
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.0178853235139404e-06,
+      "loss": 1.6035,
+      "step": 22710
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 6.016131860424339e-06,
+      "loss": 1.6015,
+      "step": 22720
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.014378397334737e-06,
+      "loss": 1.5199,
+      "step": 22730
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.012624934245135e-06,
+      "loss": 1.6279,
+      "step": 22740
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.010871471155533e-06,
+      "loss": 1.6334,
+      "step": 22750
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.00911800806593e-06,
+      "loss": 1.713,
+      "step": 22760
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.007364544976329e-06,
+      "loss": 1.5848,
+      "step": 22770
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 6.005611081886726e-06,
+      "loss": 1.7055,
+      "step": 22780
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.003857618797125e-06,
+      "loss": 1.6382,
+      "step": 22790
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.002104155707523e-06,
+      "loss": 1.398,
+      "step": 22800
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 6.000350692617921e-06,
+      "loss": 1.6301,
+      "step": 22810
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 5.998597229528319e-06,
+      "loss": 1.6529,
+      "step": 22820
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 5.996843766438717e-06,
+      "loss": 1.667,
+      "step": 22830
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 5.9950903033491144e-06,
+      "loss": 1.8939,
+      "step": 22840
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 5.993336840259513e-06,
+      "loss": 1.8196,
+      "step": 22850
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 5.99158337716991e-06,
+      "loss": 1.6809,
+      "step": 22860
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 5.9898299140803094e-06,
+      "loss": 1.535,
+      "step": 22870
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 5.988076450990708e-06,
+      "loss": 1.6296,
+      "step": 22880
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 5.986322987901105e-06,
+      "loss": 1.5706,
+      "step": 22890
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9845695248115036e-06,
+      "loss": 1.5704,
+      "step": 22900
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.982816061721901e-06,
+      "loss": 1.4775,
+      "step": 22910
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.981062598632299e-06,
+      "loss": 1.5575,
+      "step": 22920
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.979309135542697e-06,
+      "loss": 1.5291,
+      "step": 22930
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.977555672453096e-06,
+      "loss": 1.5632,
+      "step": 22940
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 5.9758022093634935e-06,
+      "loss": 1.723,
+      "step": 22950
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.974048746273892e-06,
+      "loss": 1.5473,
+      "step": 22960
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.972295283184289e-06,
+      "loss": 1.4491,
+      "step": 22970
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.970541820094688e-06,
+      "loss": 1.7056,
+      "step": 22980
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.968788357005085e-06,
+      "loss": 1.6285,
+      "step": 22990
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.9670348939154834e-06,
+      "loss": 1.5746,
+      "step": 23000
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 5.965281430825881e-06,
+      "loss": 1.681,
+      "step": 23010
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.96352796773628e-06,
+      "loss": 1.5307,
+      "step": 23020
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.9617745046466776e-06,
+      "loss": 1.4638,
+      "step": 23030
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.960021041557076e-06,
+      "loss": 1.5175,
+      "step": 23040
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.958267578467473e-06,
+      "loss": 1.5616,
+      "step": 23050
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 5.956514115377872e-06,
+      "loss": 1.7306,
+      "step": 23060
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.954760652288269e-06,
+      "loss": 1.6493,
+      "step": 23070
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.9530071891986675e-06,
+      "loss": 1.717,
+      "step": 23080
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.951253726109067e-06,
+      "loss": 1.5387,
+      "step": 23090
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.949500263019464e-06,
+      "loss": 1.6176,
+      "step": 23100
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.9477467999298625e-06,
+      "loss": 1.6747,
+      "step": 23110
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 5.94599333684026e-06,
+      "loss": 1.6461,
+      "step": 23120
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.944239873750658e-06,
+      "loss": 1.543,
+      "step": 23130
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.942486410661056e-06,
+      "loss": 1.6541,
+      "step": 23140
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.940732947571454e-06,
+      "loss": 1.6667,
+      "step": 23150
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.938979484481852e-06,
+      "loss": 1.8277,
+      "step": 23160
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.937226021392251e-06,
+      "loss": 1.3965,
+      "step": 23170
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 5.935472558302648e-06,
+      "loss": 1.6172,
+      "step": 23180
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.9337190952130465e-06,
+      "loss": 1.6851,
+      "step": 23190
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.931965632123444e-06,
+      "loss": 1.6359,
+      "step": 23200
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.930212169033842e-06,
+      "loss": 1.6332,
+      "step": 23210
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.92845870594424e-06,
+      "loss": 1.5345,
+      "step": 23220
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 5.926705242854638e-06,
+      "loss": 1.7668,
+      "step": 23230
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.9249517797650365e-06,
+      "loss": 1.4253,
+      "step": 23240
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.923198316675435e-06,
+      "loss": 1.6387,
+      "step": 23250
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.921444853585832e-06,
+      "loss": 1.6381,
+      "step": 23260
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.919691390496231e-06,
+      "loss": 1.7422,
+      "step": 23270
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.917937927406628e-06,
+      "loss": 1.4644,
+      "step": 23280
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 5.916184464317026e-06,
+      "loss": 1.7283,
+      "step": 23290
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.914431001227424e-06,
+      "loss": 1.4324,
+      "step": 23300
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.912677538137823e-06,
+      "loss": 1.7601,
+      "step": 23310
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.910924075048221e-06,
+      "loss": 1.6025,
+      "step": 23320
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.909170611958619e-06,
+      "loss": 1.4537,
+      "step": 23330
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.907417148869017e-06,
+      "loss": 1.5648,
+      "step": 23340
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 5.905663685779415e-06,
+      "loss": 1.6183,
+      "step": 23350
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.903910222689813e-06,
+      "loss": 1.4856,
+      "step": 23360
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.9021567596002105e-06,
+      "loss": 1.609,
+      "step": 23370
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.900403296510609e-06,
+      "loss": 1.8423,
+      "step": 23380
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.898649833421007e-06,
+      "loss": 1.7945,
+      "step": 23390
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.8968963703314054e-06,
+      "loss": 1.6898,
+      "step": 23400
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 5.895142907241803e-06,
+      "loss": 1.6229,
+      "step": 23410
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.893389444152201e-06,
+      "loss": 1.5374,
+      "step": 23420
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.891635981062599e-06,
+      "loss": 1.4973,
+      "step": 23430
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.889882517972997e-06,
+      "loss": 1.6574,
+      "step": 23440
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.8881290548833945e-06,
+      "loss": 1.779,
+      "step": 23450
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 5.886375591793794e-06,
+      "loss": 1.6891,
+      "step": 23460
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.884622128704191e-06,
+      "loss": 1.6701,
+      "step": 23470
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.8828686656145895e-06,
+      "loss": 1.6705,
+      "step": 23480
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.881115202524987e-06,
+      "loss": 1.4461,
+      "step": 23490
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.879361739435385e-06,
+      "loss": 1.5811,
+      "step": 23500
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.877608276345783e-06,
+      "loss": 1.8754,
+      "step": 23510
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 5.875854813256181e-06,
+      "loss": 1.525,
+      "step": 23520
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.87410135016658e-06,
+      "loss": 1.5376,
+      "step": 23530
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.872347887076978e-06,
+      "loss": 1.7105,
+      "step": 23540
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.870594423987376e-06,
+      "loss": 1.8308,
+      "step": 23550
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.8688409608977736e-06,
+      "loss": 1.3796,
+      "step": 23560
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.867087497808172e-06,
+      "loss": 1.5451,
+      "step": 23570
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 5.865334034718569e-06,
+      "loss": 1.6496,
+      "step": 23580
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.863580571628968e-06,
+      "loss": 1.5014,
+      "step": 23590
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.861827108539365e-06,
+      "loss": 1.7103,
+      "step": 23600
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.860073645449764e-06,
+      "loss": 1.5344,
+      "step": 23610
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.858320182360162e-06,
+      "loss": 1.6186,
+      "step": 23620
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 5.85656671927056e-06,
+      "loss": 1.8106,
+      "step": 23630
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.854813256180958e-06,
+      "loss": 1.8239,
+      "step": 23640
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.853059793091356e-06,
+      "loss": 1.6505,
+      "step": 23650
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.8513063300017534e-06,
+      "loss": 1.7483,
+      "step": 23660
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.849552866912152e-06,
+      "loss": 1.5108,
+      "step": 23670
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.84779940382255e-06,
+      "loss": 1.5143,
+      "step": 23680
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 5.846045940732948e-06,
+      "loss": 1.6758,
+      "step": 23690
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.844292477643346e-06,
+      "loss": 1.5741,
+      "step": 23700
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.842539014553744e-06,
+      "loss": 1.763,
+      "step": 23710
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.840785551464142e-06,
+      "loss": 1.6409,
+      "step": 23720
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.83903208837454e-06,
+      "loss": 1.728,
+      "step": 23730
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.8372786252849375e-06,
+      "loss": 1.7528,
+      "step": 23740
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 5.835525162195336e-06,
+      "loss": 1.773,
+      "step": 23750
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.833771699105735e-06,
+      "loss": 1.6086,
+      "step": 23760
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.8320182360161325e-06,
+      "loss": 1.6055,
+      "step": 23770
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.830264772926531e-06,
+      "loss": 1.5575,
+      "step": 23780
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.828511309836928e-06,
+      "loss": 1.5958,
+      "step": 23790
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.826757846747327e-06,
+      "loss": 1.6082,
+      "step": 23800
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 5.825004383657724e-06,
+      "loss": 1.6651,
+      "step": 23810
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.823250920568122e-06,
+      "loss": 1.6213,
+      "step": 23820
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.821497457478521e-06,
+      "loss": 1.509,
+      "step": 23830
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.819743994388919e-06,
+      "loss": 1.5989,
+      "step": 23840
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.8179905312993165e-06,
+      "loss": 1.6954,
+      "step": 23850
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 5.816237068209715e-06,
+      "loss": 1.6691,
+      "step": 23860
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.814483605120112e-06,
+      "loss": 1.5584,
+      "step": 23870
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.812730142030511e-06,
+      "loss": 1.651,
+      "step": 23880
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.810976678940908e-06,
+      "loss": 1.6504,
+      "step": 23890
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.809223215851307e-06,
+      "loss": 1.5726,
+      "step": 23900
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.807469752761705e-06,
+      "loss": 1.7028,
+      "step": 23910
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 5.805716289672103e-06,
+      "loss": 1.5839,
+      "step": 23920
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.803962826582501e-06,
+      "loss": 1.6277,
+      "step": 23930
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.802209363492899e-06,
+      "loss": 1.4357,
+      "step": 23940
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.800455900403296e-06,
+      "loss": 1.6789,
+      "step": 23950
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.798702437313695e-06,
+      "loss": 1.786,
+      "step": 23960
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.796948974224092e-06,
+      "loss": 1.5486,
+      "step": 23970
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 5.795195511134491e-06,
+      "loss": 1.7472,
+      "step": 23980
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.79344204804489e-06,
+      "loss": 1.5181,
+      "step": 23990
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.791688584955287e-06,
+      "loss": 1.4256,
+      "step": 24000
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.7899351218656855e-06,
+      "loss": 1.5532,
+      "step": 24010
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.788181658776083e-06,
+      "loss": 1.6136,
+      "step": 24020
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 5.786428195686481e-06,
+      "loss": 1.8099,
+      "step": 24030
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.784674732596879e-06,
+      "loss": 1.6844,
+      "step": 24040
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.782921269507278e-06,
+      "loss": 1.7296,
+      "step": 24050
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.7811678064176755e-06,
+      "loss": 1.8963,
+      "step": 24060
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.779414343328074e-06,
+      "loss": 1.4228,
+      "step": 24070
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.777660880238471e-06,
+      "loss": 1.5557,
+      "step": 24080
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 5.77590741714887e-06,
+      "loss": 1.5625,
+      "step": 24090
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.774153954059267e-06,
+      "loss": 1.8805,
+      "step": 24100
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.772400490969665e-06,
+      "loss": 1.4768,
+      "step": 24110
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.7706470278800646e-06,
+      "loss": 1.5789,
+      "step": 24120
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.768893564790462e-06,
+      "loss": 1.5547,
+      "step": 24130
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.7671401017008595e-06,
+      "loss": 1.6696,
+      "step": 24140
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 5.765386638611258e-06,
+      "loss": 1.6584,
+      "step": 24150
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.763633175521655e-06,
+      "loss": 1.7657,
+      "step": 24160
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.761879712432054e-06,
+      "loss": 1.4762,
+      "step": 24170
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.760126249342451e-06,
+      "loss": 1.4577,
+      "step": 24180
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.7583727862528495e-06,
+      "loss": 1.6917,
+      "step": 24190
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 5.756619323163249e-06,
+      "loss": 1.7067,
+      "step": 24200
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.754865860073646e-06,
+      "loss": 1.5536,
+      "step": 24210
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.7531123969840444e-06,
+      "loss": 1.8447,
+      "step": 24220
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.751358933894442e-06,
+      "loss": 1.4774,
+      "step": 24230
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.74960547080484e-06,
+      "loss": 1.5427,
+      "step": 24240
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.747852007715238e-06,
+      "loss": 1.7886,
+      "step": 24250
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 5.746098544625636e-06,
+      "loss": 1.6093,
+      "step": 24260
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.744345081536034e-06,
+      "loss": 1.4436,
+      "step": 24270
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.742591618446433e-06,
+      "loss": 1.5763,
+      "step": 24280
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.74083815535683e-06,
+      "loss": 1.659,
+      "step": 24290
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.7390846922672285e-06,
+      "loss": 1.5719,
+      "step": 24300
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.737331229177626e-06,
+      "loss": 1.6471,
+      "step": 24310
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 5.735577766088024e-06,
+      "loss": 1.6858,
+      "step": 24320
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.733824302998422e-06,
+      "loss": 1.7393,
+      "step": 24330
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.73207083990882e-06,
+      "loss": 1.5269,
+      "step": 24340
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.7303173768192184e-06,
+      "loss": 1.4883,
+      "step": 24350
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.728563913729617e-06,
+      "loss": 1.6967,
+      "step": 24360
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.726810450640014e-06,
+      "loss": 1.4996,
+      "step": 24370
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 5.7250569875504126e-06,
+      "loss": 1.5242,
+      "step": 24380
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.72330352446081e-06,
+      "loss": 1.5316,
+      "step": 24390
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.721550061371208e-06,
+      "loss": 1.5776,
+      "step": 24400
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.719796598281606e-06,
+      "loss": 1.7986,
+      "step": 24410
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.718043135192005e-06,
+      "loss": 1.6335,
+      "step": 24420
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 5.716289672102403e-06,
+      "loss": 1.7564,
+      "step": 24430
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.714536209012801e-06,
+      "loss": 1.4445,
+      "step": 24440
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.712782745923199e-06,
+      "loss": 1.7667,
+      "step": 24450
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.711029282833597e-06,
+      "loss": 1.6312,
+      "step": 24460
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.709275819743995e-06,
+      "loss": 1.6747,
+      "step": 24470
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.7075223566543924e-06,
+      "loss": 1.6465,
+      "step": 24480
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 5.705768893564792e-06,
+      "loss": 1.5976,
+      "step": 24490
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.704015430475189e-06,
+      "loss": 1.5864,
+      "step": 24500
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.702261967385587e-06,
+      "loss": 1.6621,
+      "step": 24510
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.700508504295985e-06,
+      "loss": 1.5473,
+      "step": 24520
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.698755041206383e-06,
+      "loss": 1.6619,
+      "step": 24530
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.697001578116781e-06,
+      "loss": 1.5999,
+      "step": 24540
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 5.695248115027179e-06,
+      "loss": 1.5639,
+      "step": 24550
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.6934946519375765e-06,
+      "loss": 1.713,
+      "step": 24560
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.691741188847976e-06,
+      "loss": 1.5731,
+      "step": 24570
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.689987725758373e-06,
+      "loss": 1.7016,
+      "step": 24580
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.6882342626687715e-06,
+      "loss": 1.4362,
+      "step": 24590
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 5.686480799579169e-06,
+      "loss": 1.4878,
+      "step": 24600
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.684727336489567e-06,
+      "loss": 1.5508,
+      "step": 24610
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.682973873399965e-06,
+      "loss": 1.6345,
+      "step": 24620
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.681220410310363e-06,
+      "loss": 1.684,
+      "step": 24630
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.679466947220762e-06,
+      "loss": 1.5849,
+      "step": 24640
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.67771348413116e-06,
+      "loss": 1.5703,
+      "step": 24650
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 5.675960021041558e-06,
+      "loss": 1.647,
+      "step": 24660
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.6742065579519555e-06,
+      "loss": 1.5457,
+      "step": 24670
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.672453094862354e-06,
+      "loss": 1.5934,
+      "step": 24680
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.670699631772751e-06,
+      "loss": 1.7015,
+      "step": 24690
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.66894616868315e-06,
+      "loss": 1.5346,
+      "step": 24700
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.667192705593547e-06,
+      "loss": 1.648,
+      "step": 24710
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 5.665439242503946e-06,
+      "loss": 1.6128,
+      "step": 24720
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.663685779414344e-06,
+      "loss": 1.3878,
+      "step": 24730
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.661932316324742e-06,
+      "loss": 1.5856,
+      "step": 24740
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.66017885323514e-06,
+      "loss": 1.6694,
+      "step": 24750
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.658425390145538e-06,
+      "loss": 1.6209,
+      "step": 24760
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 5.656671927055935e-06,
+      "loss": 1.6509,
+      "step": 24770
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.654918463966334e-06,
+      "loss": 1.6302,
+      "step": 24780
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.653165000876732e-06,
+      "loss": 1.6065,
+      "step": 24790
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.65141153778713e-06,
+      "loss": 1.789,
+      "step": 24800
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.649658074697528e-06,
+      "loss": 1.7034,
+      "step": 24810
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.647904611607926e-06,
+      "loss": 1.6033,
+      "step": 24820
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 5.646151148518324e-06,
+      "loss": 1.6268,
+      "step": 24830
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.644397685428722e-06,
+      "loss": 1.8175,
+      "step": 24840
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.6426442223391195e-06,
+      "loss": 1.6278,
+      "step": 24850
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.640890759249519e-06,
+      "loss": 1.5172,
+      "step": 24860
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.639137296159917e-06,
+      "loss": 1.7332,
+      "step": 24870
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.6373838330703145e-06,
+      "loss": 1.6066,
+      "step": 24880
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 5.635630369980713e-06,
+      "loss": 1.5053,
+      "step": 24890
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.63387690689111e-06,
+      "loss": 1.728,
+      "step": 24900
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.632123443801509e-06,
+      "loss": 1.5587,
+      "step": 24910
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.630369980711906e-06,
+      "loss": 1.5059,
+      "step": 24920
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.628616517622304e-06,
+      "loss": 1.6084,
+      "step": 24930
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.626863054532703e-06,
+      "loss": 1.6802,
+      "step": 24940
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 5.625109591443101e-06,
+      "loss": 1.5548,
+      "step": 24950
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.6233561283534985e-06,
+      "loss": 1.6403,
+      "step": 24960
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.621602665263897e-06,
+      "loss": 1.5776,
+      "step": 24970
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.619849202174294e-06,
+      "loss": 1.5734,
+      "step": 24980
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.618095739084693e-06,
+      "loss": 1.4776,
+      "step": 24990
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 5.61634227599509e-06,
+      "loss": 1.5393,
+      "step": 25000
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.614588812905489e-06,
+      "loss": 1.3288,
+      "step": 25010
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.612835349815887e-06,
+      "loss": 1.5324,
+      "step": 25020
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.611081886726285e-06,
+      "loss": 1.7263,
+      "step": 25030
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.609328423636683e-06,
+      "loss": 1.5701,
+      "step": 25040
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.607574960547081e-06,
+      "loss": 1.6788,
+      "step": 25050
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 5.605821497457478e-06,
+      "loss": 1.7263,
+      "step": 25060
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.604068034367877e-06,
+      "loss": 1.5568,
+      "step": 25070
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.602314571278276e-06,
+      "loss": 1.6004,
+      "step": 25080
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.600561108188673e-06,
+      "loss": 1.7212,
+      "step": 25090
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.598807645099072e-06,
+      "loss": 1.8241,
+      "step": 25100
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.597054182009469e-06,
+      "loss": 1.5445,
+      "step": 25110
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 5.5953007189198675e-06,
+      "loss": 1.5317,
+      "step": 25120
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.593547255830265e-06,
+      "loss": 1.4916,
+      "step": 25130
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.591793792740663e-06,
+      "loss": 1.518,
+      "step": 25140
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.590040329651061e-06,
+      "loss": 1.6653,
+      "step": 25150
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.58828686656146e-06,
+      "loss": 1.5466,
+      "step": 25160
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 5.5865334034718574e-06,
+      "loss": 1.5476,
+      "step": 25170
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.584779940382256e-06,
+      "loss": 1.6372,
+      "step": 25180
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.583026477292653e-06,
+      "loss": 1.6414,
+      "step": 25190
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.5812730142030516e-06,
+      "loss": 1.6543,
+      "step": 25200
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.579519551113449e-06,
+      "loss": 1.7041,
+      "step": 25210
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.577766088023847e-06,
+      "loss": 1.8929,
+      "step": 25220
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 5.576012624934246e-06,
+      "loss": 1.5408,
+      "step": 25230
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.574259161844644e-06,
+      "loss": 1.5733,
+      "step": 25240
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.5725056987550415e-06,
+      "loss": 1.6906,
+      "step": 25250
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.57075223566544e-06,
+      "loss": 1.6932,
+      "step": 25260
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.568998772575837e-06,
+      "loss": 1.6423,
+      "step": 25270
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.567245309486236e-06,
+      "loss": 1.6135,
+      "step": 25280
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 5.565491846396633e-06,
+      "loss": 1.8203,
+      "step": 25290
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.5637383833070314e-06,
+      "loss": 1.5578,
+      "step": 25300
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.561984920217431e-06,
+      "loss": 1.4618,
+      "step": 25310
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.560231457127828e-06,
+      "loss": 1.5066,
+      "step": 25320
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.558477994038226e-06,
+      "loss": 1.6344,
+      "step": 25330
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 5.556724530948624e-06,
+      "loss": 1.6572,
+      "step": 25340
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.554971067859022e-06,
+      "loss": 1.7117,
+      "step": 25350
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.55321760476942e-06,
+      "loss": 1.756,
+      "step": 25360
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.551464141679818e-06,
+      "loss": 1.5914,
+      "step": 25370
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.549710678590216e-06,
+      "loss": 1.6267,
+      "step": 25380
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.547957215500615e-06,
+      "loss": 1.5299,
+      "step": 25390
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 5.546203752411012e-06,
+      "loss": 1.6627,
+      "step": 25400
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5444502893214105e-06,
+      "loss": 1.447,
+      "step": 25410
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.542696826231808e-06,
+      "loss": 1.5975,
+      "step": 25420
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.540943363142206e-06,
+      "loss": 1.518,
+      "step": 25430
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.539189900052604e-06,
+      "loss": 1.508,
+      "step": 25440
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.537436436963003e-06,
+      "loss": 1.6745,
+      "step": 25450
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 5.5356829738734e-06,
+      "loss": 1.7029,
+      "step": 25460
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.533929510783799e-06,
+      "loss": 1.6693,
+      "step": 25470
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.532176047694196e-06,
+      "loss": 1.6043,
+      "step": 25480
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.5304225846045945e-06,
+      "loss": 1.6073,
+      "step": 25490
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.528669121514992e-06,
+      "loss": 1.817,
+      "step": 25500
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.52691565842539e-06,
+      "loss": 1.7282,
+      "step": 25510
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 5.525162195335788e-06,
+      "loss": 1.7979,
+      "step": 25520
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.523408732246187e-06,
+      "loss": 1.7499,
+      "step": 25530
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.521655269156585e-06,
+      "loss": 1.67,
+      "step": 25540
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.519901806066983e-06,
+      "loss": 1.5459,
+      "step": 25550
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.518148342977381e-06,
+      "loss": 1.7592,
+      "step": 25560
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 5.516394879887779e-06,
+      "loss": 1.9785,
+      "step": 25570
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.514641416798177e-06,
+      "loss": 1.6566,
+      "step": 25580
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.512887953708574e-06,
+      "loss": 1.5642,
+      "step": 25590
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.511134490618974e-06,
+      "loss": 1.4359,
+      "step": 25600
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.509381027529371e-06,
+      "loss": 1.4962,
+      "step": 25610
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.507627564439769e-06,
+      "loss": 1.4751,
+      "step": 25620
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 5.505874101350167e-06,
+      "loss": 1.5166,
+      "step": 25630
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.504120638260565e-06,
+      "loss": 1.6216,
+      "step": 25640
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.502367175170963e-06,
+      "loss": 1.7827,
+      "step": 25650
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.500613712081361e-06,
+      "loss": 1.6152,
+      "step": 25660
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.4988602489917585e-06,
+      "loss": 1.6467,
+      "step": 25670
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.497106785902158e-06,
+      "loss": 1.7285,
+      "step": 25680
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 5.495353322812555e-06,
+      "loss": 1.5256,
+      "step": 25690
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.4935998597229534e-06,
+      "loss": 1.7439,
+      "step": 25700
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.491846396633351e-06,
+      "loss": 1.5584,
+      "step": 25710
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.490092933543749e-06,
+      "loss": 1.4819,
+      "step": 25720
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.488339470454147e-06,
+      "loss": 1.4883,
+      "step": 25730
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 5.486586007364545e-06,
+      "loss": 1.493,
+      "step": 25740
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.484832544274944e-06,
+      "loss": 1.4893,
+      "step": 25750
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.483079081185342e-06,
+      "loss": 1.5949,
+      "step": 25760
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.48132561809574e-06,
+      "loss": 1.5191,
+      "step": 25770
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.4795721550061375e-06,
+      "loss": 1.58,
+      "step": 25780
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.477818691916536e-06,
+      "loss": 1.5243,
+      "step": 25790
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 5.476065228826933e-06,
+      "loss": 1.7544,
+      "step": 25800
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.474311765737332e-06,
+      "loss": 1.5761,
+      "step": 25810
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.47255830264773e-06,
+      "loss": 1.8162,
+      "step": 25820
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.470804839558128e-06,
+      "loss": 1.5729,
+      "step": 25830
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.469051376468526e-06,
+      "loss": 1.6012,
+      "step": 25840
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.467297913378924e-06,
+      "loss": 1.7787,
+      "step": 25850
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 5.465544450289322e-06,
+      "loss": 1.5554,
+      "step": 25860
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.46379098719972e-06,
+      "loss": 1.6111,
+      "step": 25870
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.462037524110117e-06,
+      "loss": 1.6951,
+      "step": 25880
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.460284061020516e-06,
+      "loss": 1.5155,
+      "step": 25890
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.458530597930914e-06,
+      "loss": 1.724,
+      "step": 25900
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.456777134841312e-06,
+      "loss": 1.6635,
+      "step": 25910
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 5.45502367175171e-06,
+      "loss": 1.4564,
+      "step": 25920
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.453270208662108e-06,
+      "loss": 1.6474,
+      "step": 25930
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.451516745572506e-06,
+      "loss": 1.597,
+      "step": 25940
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.449763282482904e-06,
+      "loss": 1.7464,
+      "step": 25950
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.4480098193933014e-06,
+      "loss": 1.6197,
+      "step": 25960
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 5.446256356303701e-06,
+      "loss": 1.6068,
+      "step": 25970
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.444502893214099e-06,
+      "loss": 1.5426,
+      "step": 25980
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.4427494301244964e-06,
+      "loss": 1.7634,
+      "step": 25990
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.440995967034895e-06,
+      "loss": 1.7607,
+      "step": 26000
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.439242503945292e-06,
+      "loss": 1.5414,
+      "step": 26010
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.4374890408556906e-06,
+      "loss": 1.5784,
+      "step": 26020
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 5.435735577766088e-06,
+      "loss": 1.611,
+      "step": 26030
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.433982114676486e-06,
+      "loss": 1.527,
+      "step": 26040
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.432228651586885e-06,
+      "loss": 1.5697,
+      "step": 26050
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.430475188497283e-06,
+      "loss": 1.6733,
+      "step": 26060
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.4287217254076805e-06,
+      "loss": 1.6126,
+      "step": 26070
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.426968262318079e-06,
+      "loss": 1.5915,
+      "step": 26080
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 5.425214799228476e-06,
+      "loss": 1.6484,
+      "step": 26090
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.423461336138875e-06,
+      "loss": 1.5515,
+      "step": 26100
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.421707873049272e-06,
+      "loss": 1.5686,
+      "step": 26110
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.419954409959671e-06,
+      "loss": 1.6189,
+      "step": 26120
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.418200946870069e-06,
+      "loss": 1.5843,
+      "step": 26130
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 5.416447483780467e-06,
+      "loss": 1.5452,
+      "step": 26140
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.4146940206908646e-06,
+      "loss": 1.6645,
+      "step": 26150
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.412940557601263e-06,
+      "loss": 1.6016,
+      "step": 26160
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.41118709451166e-06,
+      "loss": 1.5053,
+      "step": 26170
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.409433631422059e-06,
+      "loss": 1.6483,
+      "step": 26180
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.407680168332458e-06,
+      "loss": 1.7008,
+      "step": 26190
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 5.405926705242855e-06,
+      "loss": 1.5493,
+      "step": 26200
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.404173242153254e-06,
+      "loss": 1.6675,
+      "step": 26210
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.402419779063651e-06,
+      "loss": 1.5608,
+      "step": 26220
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.4006663159740495e-06,
+      "loss": 1.5682,
+      "step": 26230
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.398912852884447e-06,
+      "loss": 1.6106,
+      "step": 26240
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.397159389794845e-06,
+      "loss": 1.5605,
+      "step": 26250
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 5.395405926705243e-06,
+      "loss": 1.6011,
+      "step": 26260
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.393652463615642e-06,
+      "loss": 1.6506,
+      "step": 26270
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.391899000526039e-06,
+      "loss": 1.5945,
+      "step": 26280
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.390145537436438e-06,
+      "loss": 1.6135,
+      "step": 26290
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.388392074346835e-06,
+      "loss": 1.5939,
+      "step": 26300
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.3866386112572335e-06,
+      "loss": 1.5336,
+      "step": 26310
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.384885148167631e-06,
+      "loss": 1.4841,
+      "step": 26320
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.383131685078029e-06,
+      "loss": 1.3953,
+      "step": 26330
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.381378221988428e-06,
+      "loss": 1.6136,
+      "step": 26340
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.379624758898826e-06,
+      "loss": 1.5214,
+      "step": 26350
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.3778712958092235e-06,
+      "loss": 1.5494,
+      "step": 26360
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.376117832719622e-06,
+      "loss": 1.5877,
+      "step": 26370
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.374364369630019e-06,
+      "loss": 1.4905,
+      "step": 26380
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.372610906540418e-06,
+      "loss": 1.6071,
+      "step": 26390
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.370857443450815e-06,
+      "loss": 1.4543,
+      "step": 26400
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.369103980361214e-06,
+      "loss": 1.6683,
+      "step": 26410
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.3673505172716126e-06,
+      "loss": 1.6572,
+      "step": 26420
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.36559705418201e-06,
+      "loss": 1.7517,
+      "step": 26430
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.363843591092408e-06,
+      "loss": 1.7761,
+      "step": 26440
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.362090128002806e-06,
+      "loss": 1.3925,
+      "step": 26450
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.360336664913204e-06,
+      "loss": 1.6176,
+      "step": 26460
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.358583201823602e-06,
+      "loss": 1.6951,
+      "step": 26470
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.356829738734e-06,
+      "loss": 1.4469,
+      "step": 26480
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.355076275644398e-06,
+      "loss": 1.5465,
+      "step": 26490
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.353322812554797e-06,
+      "loss": 1.6196,
+      "step": 26500
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.351569349465194e-06,
+      "loss": 1.4924,
+      "step": 26510
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.3498158863755924e-06,
+      "loss": 1.5495,
+      "step": 26520
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.34806242328599e-06,
+      "loss": 1.6056,
+      "step": 26530
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.346308960196388e-06,
+      "loss": 1.5264,
+      "step": 26540
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.344555497106786e-06,
+      "loss": 1.4997,
+      "step": 26550
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.342802034017185e-06,
+      "loss": 1.6562,
+      "step": 26560
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.341048570927582e-06,
+      "loss": 1.5179,
+      "step": 26570
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.339295107837981e-06,
+      "loss": 1.553,
+      "step": 26580
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.337541644748378e-06,
+      "loss": 1.5582,
+      "step": 26590
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.3357881816587765e-06,
+      "loss": 1.5234,
+      "step": 26600
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.334034718569174e-06,
+      "loss": 1.3989,
+      "step": 26610
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.332281255479572e-06,
+      "loss": 1.6725,
+      "step": 26620
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.33052779238997e-06,
+      "loss": 1.6547,
+      "step": 26630
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.328774329300369e-06,
+      "loss": 1.5992,
+      "step": 26640
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.327020866210767e-06,
+      "loss": 1.5056,
+      "step": 26650
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.325267403121165e-06,
+      "loss": 1.4988,
+      "step": 26660
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.323513940031563e-06,
+      "loss": 1.6821,
+      "step": 26670
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3217604769419606e-06,
+      "loss": 1.6973,
+      "step": 26680
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.320007013852359e-06,
+      "loss": 1.6061,
+      "step": 26690
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.318253550762756e-06,
+      "loss": 1.6933,
+      "step": 26700
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 5.3165000876731556e-06,
+      "loss": 1.5476,
+      "step": 26710
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.314746624583553e-06,
+      "loss": 1.6233,
+      "step": 26720
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.312993161493951e-06,
+      "loss": 1.4389,
+      "step": 26730
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.311239698404349e-06,
+      "loss": 1.5307,
+      "step": 26740
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.309486235314747e-06,
+      "loss": 1.6263,
+      "step": 26750
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.307732772225145e-06,
+      "loss": 1.6535,
+      "step": 26760
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 5.305979309135543e-06,
+      "loss": 1.4422,
+      "step": 26770
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.304225846045941e-06,
+      "loss": 1.6597,
+      "step": 26780
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.30247238295634e-06,
+      "loss": 1.6205,
+      "step": 26790
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.300718919866737e-06,
+      "loss": 1.6079,
+      "step": 26800
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.298965456777135e-06,
+      "loss": 1.5384,
+      "step": 26810
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.297211993687533e-06,
+      "loss": 1.7225,
+      "step": 26820
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 5.295458530597931e-06,
+      "loss": 1.5722,
+      "step": 26830
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.293705067508329e-06,
+      "loss": 1.4067,
+      "step": 26840
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.291951604418727e-06,
+      "loss": 1.5026,
+      "step": 26850
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.290198141329126e-06,
+      "loss": 1.4556,
+      "step": 26860
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.288444678239524e-06,
+      "loss": 1.6099,
+      "step": 26870
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 5.286691215149922e-06,
+      "loss": 1.5573,
+      "step": 26880
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.2849377520603195e-06,
+      "loss": 1.6592,
+      "step": 26890
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.283184288970718e-06,
+      "loss": 1.5498,
+      "step": 26900
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.281430825881115e-06,
+      "loss": 1.5535,
+      "step": 26910
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.279677362791514e-06,
+      "loss": 1.5648,
+      "step": 26920
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.277923899701912e-06,
+      "loss": 1.5627,
+      "step": 26930
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 5.27617043661231e-06,
+      "loss": 1.5468,
+      "step": 26940
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.274416973522708e-06,
+      "loss": 1.7215,
+      "step": 26950
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.272663510433106e-06,
+      "loss": 1.5491,
+      "step": 26960
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.2709100473435036e-06,
+      "loss": 1.4919,
+      "step": 26970
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.269156584253902e-06,
+      "loss": 1.7856,
+      "step": 26980
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.267403121164299e-06,
+      "loss": 1.5127,
+      "step": 26990
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 5.265649658074698e-06,
+      "loss": 1.6709,
+      "step": 27000
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.263896194985096e-06,
+      "loss": 1.4147,
+      "step": 27010
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.262142731895494e-06,
+      "loss": 1.553,
+      "step": 27020
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.260389268805892e-06,
+      "loss": 1.6441,
+      "step": 27030
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.25863580571629e-06,
+      "loss": 1.6217,
+      "step": 27040
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.256882342626688e-06,
+      "loss": 1.5846,
+      "step": 27050
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 5.255128879537086e-06,
+      "loss": 1.6438,
+      "step": 27060
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.253375416447483e-06,
+      "loss": 1.8263,
+      "step": 27070
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.251621953357883e-06,
+      "loss": 1.5953,
+      "step": 27080
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.249868490268281e-06,
+      "loss": 1.7981,
+      "step": 27090
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.248115027178678e-06,
+      "loss": 1.6845,
+      "step": 27100
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 5.246361564089077e-06,
+      "loss": 1.6257,
+      "step": 27110
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.244608100999474e-06,
+      "loss": 1.5622,
+      "step": 27120
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.2428546379098725e-06,
+      "loss": 1.7066,
+      "step": 27130
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.24110117482027e-06,
+      "loss": 1.4661,
+      "step": 27140
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.239347711730669e-06,
+      "loss": 1.5584,
+      "step": 27150
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.237594248641067e-06,
+      "loss": 1.5473,
+      "step": 27160
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 5.235840785551465e-06,
+      "loss": 1.5098,
+      "step": 27170
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.2340873224618625e-06,
+      "loss": 1.5278,
+      "step": 27180
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.232333859372261e-06,
+      "loss": 1.5394,
+      "step": 27190
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.230580396282658e-06,
+      "loss": 1.5964,
+      "step": 27200
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.228826933193057e-06,
+      "loss": 1.4925,
+      "step": 27210
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.227073470103454e-06,
+      "loss": 1.535,
+      "step": 27220
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 5.225320007013853e-06,
+      "loss": 1.57,
+      "step": 27230
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.223566543924251e-06,
+      "loss": 1.665,
+      "step": 27240
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.221813080834649e-06,
+      "loss": 1.5891,
+      "step": 27250
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.2200596177450465e-06,
+      "loss": 1.6486,
+      "step": 27260
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.218306154655445e-06,
+      "loss": 1.4046,
+      "step": 27270
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 5.216552691565842e-06,
+      "loss": 1.4683,
+      "step": 27280
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.214799228476241e-06,
+      "loss": 1.5156,
+      "step": 27290
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.21304576538664e-06,
+      "loss": 1.511,
+      "step": 27300
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.211292302297037e-06,
+      "loss": 1.7483,
+      "step": 27310
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.209538839207436e-06,
+      "loss": 1.7327,
+      "step": 27320
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.207785376117833e-06,
+      "loss": 1.7073,
+      "step": 27330
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 5.2060319130282314e-06,
+      "loss": 1.5093,
+      "step": 27340
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.204278449938629e-06,
+      "loss": 1.5319,
+      "step": 27350
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.202524986849027e-06,
+      "loss": 1.5638,
+      "step": 27360
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.2007715237594256e-06,
+      "loss": 1.596,
+      "step": 27370
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.199018060669824e-06,
+      "loss": 1.4417,
+      "step": 27380
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.197264597580221e-06,
+      "loss": 1.5486,
+      "step": 27390
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 5.19551113449062e-06,
+      "loss": 1.7425,
+      "step": 27400
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.193757671401017e-06,
+      "loss": 1.6035,
+      "step": 27410
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.1920042083114155e-06,
+      "loss": 1.5387,
+      "step": 27420
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.190250745221813e-06,
+      "loss": 1.5371,
+      "step": 27430
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.188497282132211e-06,
+      "loss": 1.5412,
+      "step": 27440
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 5.18674381904261e-06,
+      "loss": 1.5662,
+      "step": 27450
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.184990355953008e-06,
+      "loss": 1.7684,
+      "step": 27460
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.1832368928634054e-06,
+      "loss": 1.5344,
+      "step": 27470
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.181483429773804e-06,
+      "loss": 1.6043,
+      "step": 27480
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.179729966684201e-06,
+      "loss": 1.6163,
+      "step": 27490
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.1779765035945996e-06,
+      "loss": 1.4826,
+      "step": 27500
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 5.176223040504997e-06,
+      "loss": 1.7048,
+      "step": 27510
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.174469577415396e-06,
+      "loss": 1.6074,
+      "step": 27520
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.1727161143257945e-06,
+      "loss": 1.5035,
+      "step": 27530
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.170962651236192e-06,
+      "loss": 1.5872,
+      "step": 27540
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.16920918814659e-06,
+      "loss": 1.7233,
+      "step": 27550
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.167455725056988e-06,
+      "loss": 1.6097,
+      "step": 27560
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 5.165702261967386e-06,
+      "loss": 1.5561,
+      "step": 27570
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.163948798877784e-06,
+      "loss": 1.4967,
+      "step": 27580
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.162195335788182e-06,
+      "loss": 1.6181,
+      "step": 27590
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.16044187269858e-06,
+      "loss": 1.4308,
+      "step": 27600
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.158688409608979e-06,
+      "loss": 1.6587,
+      "step": 27610
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.156934946519376e-06,
+      "loss": 1.5972,
+      "step": 27620
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 5.155181483429774e-06,
+      "loss": 1.5869,
+      "step": 27630
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.153428020340172e-06,
+      "loss": 1.6326,
+      "step": 27640
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.15167455725057e-06,
+      "loss": 1.5298,
+      "step": 27650
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.149921094160968e-06,
+      "loss": 1.7145,
+      "step": 27660
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.148167631071367e-06,
+      "loss": 1.534,
+      "step": 27670
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 5.146414167981764e-06,
+      "loss": 1.5793,
+      "step": 27680
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.144660704892163e-06,
+      "loss": 1.6453,
+      "step": 27690
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.14290724180256e-06,
+      "loss": 1.6231,
+      "step": 27700
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1411537787129585e-06,
+      "loss": 1.6989,
+      "step": 27710
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.139400315623356e-06,
+      "loss": 1.6032,
+      "step": 27720
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.137646852533754e-06,
+      "loss": 1.5568,
+      "step": 27730
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 5.1358933894441535e-06,
+      "loss": 1.6576,
+      "step": 27740
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.134139926354551e-06,
+      "loss": 1.6104,
+      "step": 27750
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.132386463264949e-06,
+      "loss": 1.357,
+      "step": 27760
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.130633000175347e-06,
+      "loss": 1.5668,
+      "step": 27770
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.128879537085745e-06,
+      "loss": 1.484,
+      "step": 27780
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.1271260739961425e-06,
+      "loss": 1.6789,
+      "step": 27790
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 5.125372610906541e-06,
+      "loss": 1.7226,
+      "step": 27800
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.123619147816938e-06,
+      "loss": 1.613,
+      "step": 27810
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.1218656847273375e-06,
+      "loss": 1.5381,
+      "step": 27820
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.120112221637735e-06,
+      "loss": 1.5086,
+      "step": 27830
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.118358758548133e-06,
+      "loss": 1.5582,
+      "step": 27840
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 5.116605295458531e-06,
+      "loss": 1.5422,
+      "step": 27850
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.114851832368929e-06,
+      "loss": 1.577,
+      "step": 27860
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.113098369279327e-06,
+      "loss": 1.4018,
+      "step": 27870
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.111344906189725e-06,
+      "loss": 1.4735,
+      "step": 27880
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.109591443100123e-06,
+      "loss": 1.413,
+      "step": 27890
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.107837980010522e-06,
+      "loss": 1.5009,
+      "step": 27900
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 5.106084516920919e-06,
+      "loss": 1.5988,
+      "step": 27910
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.104331053831317e-06,
+      "loss": 1.5267,
+      "step": 27920
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.102577590741715e-06,
+      "loss": 1.6839,
+      "step": 27930
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.100824127652113e-06,
+      "loss": 1.6495,
+      "step": 27940
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.099070664562511e-06,
+      "loss": 1.6411,
+      "step": 27950
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.097317201472909e-06,
+      "loss": 1.4859,
+      "step": 27960
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 5.095563738383308e-06,
+      "loss": 1.46,
+      "step": 27970
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.093810275293706e-06,
+      "loss": 1.324,
+      "step": 27980
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.092056812204104e-06,
+      "loss": 1.6085,
+      "step": 27990
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.0903033491145015e-06,
+      "loss": 1.6163,
+      "step": 28000
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.0885498860249e-06,
+      "loss": 1.7457,
+      "step": 28010
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.086796422935297e-06,
+      "loss": 1.752,
+      "step": 28020
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 5.085042959845696e-06,
+      "loss": 1.6167,
+      "step": 28030
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.083289496756094e-06,
+      "loss": 1.5422,
+      "step": 28040
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.081536033666492e-06,
+      "loss": 1.6327,
+      "step": 28050
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.07978257057689e-06,
+      "loss": 1.6452,
+      "step": 28060
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.078029107487288e-06,
+      "loss": 1.6066,
+      "step": 28070
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 5.0762756443976855e-06,
+      "loss": 1.5529,
+      "step": 28080
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.074522181308084e-06,
+      "loss": 1.4682,
+      "step": 28090
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.072768718218481e-06,
+      "loss": 1.5321,
+      "step": 28100
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.0710152551288805e-06,
+      "loss": 1.5077,
+      "step": 28110
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.069261792039278e-06,
+      "loss": 1.6865,
+      "step": 28120
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.067508328949676e-06,
+      "loss": 1.5181,
+      "step": 28130
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 5.065754865860074e-06,
+      "loss": 1.4734,
+      "step": 28140
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.064001402770472e-06,
+      "loss": 1.3517,
+      "step": 28150
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.06224793968087e-06,
+      "loss": 1.5412,
+      "step": 28160
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.060494476591268e-06,
+      "loss": 1.5505,
+      "step": 28170
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.058741013501665e-06,
+      "loss": 1.5134,
+      "step": 28180
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.0569875504120646e-06,
+      "loss": 1.5201,
+      "step": 28190
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 5.055234087322463e-06,
+      "loss": 1.5685,
+      "step": 28200
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.05348062423286e-06,
+      "loss": 1.8067,
+      "step": 28210
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.051727161143259e-06,
+      "loss": 1.4606,
+      "step": 28220
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.049973698053656e-06,
+      "loss": 1.4605,
+      "step": 28230
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.0482202349640545e-06,
+      "loss": 1.6317,
+      "step": 28240
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 5.046466771874452e-06,
+      "loss": 1.8157,
+      "step": 28250
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.044713308784851e-06,
+      "loss": 1.6094,
+      "step": 28260
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.042959845695249e-06,
+      "loss": 1.6688,
+      "step": 28270
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.041206382605647e-06,
+      "loss": 1.5043,
+      "step": 28280
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.0394529195160444e-06,
+      "loss": 1.504,
+      "step": 28290
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.037699456426443e-06,
+      "loss": 1.436,
+      "step": 28300
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 5.03594599333684e-06,
+      "loss": 1.5352,
+      "step": 28310
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.0341925302472386e-06,
+      "loss": 1.3267,
+      "step": 28320
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.032439067157636e-06,
+      "loss": 1.6303,
+      "step": 28330
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.030685604068035e-06,
+      "loss": 1.6339,
+      "step": 28340
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.028932140978433e-06,
+      "loss": 1.4142,
+      "step": 28350
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.027178677888831e-06,
+      "loss": 1.5949,
+      "step": 28360
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.0254252147992285e-06,
+      "loss": 1.613,
+      "step": 28370
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.023671751709627e-06,
+      "loss": 1.451,
+      "step": 28380
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.021918288620024e-06,
+      "loss": 1.6301,
+      "step": 28390
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.020164825530423e-06,
+      "loss": 1.6105,
+      "step": 28400
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.018411362440822e-06,
+      "loss": 1.4848,
+      "step": 28410
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 5.016657899351219e-06,
+      "loss": 1.7587,
+      "step": 28420
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.014904436261618e-06,
+      "loss": 1.6141,
+      "step": 28430
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.013150973172015e-06,
+      "loss": 1.6285,
+      "step": 28440
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.011397510082413e-06,
+      "loss": 1.5534,
+      "step": 28450
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.009644046992811e-06,
+      "loss": 1.3532,
+      "step": 28460
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.007890583903209e-06,
+      "loss": 1.6093,
+      "step": 28470
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 5.0061371208136075e-06,
+      "loss": 1.311,
+      "step": 28480
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.004383657724006e-06,
+      "loss": 1.4989,
+      "step": 28490
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.002630194634403e-06,
+      "loss": 1.676,
+      "step": 28500
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 5.000876731544802e-06,
+      "loss": 1.5361,
+      "step": 28510
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.999123268455199e-06,
+      "loss": 1.7232,
+      "step": 28520
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.9973698053655975e-06,
+      "loss": 1.4287,
+      "step": 28530
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.995616342275996e-06,
+      "loss": 1.5998,
+      "step": 28540
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.993862879186393e-06,
+      "loss": 1.4362,
+      "step": 28550
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.992109416096792e-06,
+      "loss": 2.0255,
+      "step": 28560
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.990355953007189e-06,
+      "loss": 1.6738,
+      "step": 28570
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.988602489917587e-06,
+      "loss": 1.4294,
+      "step": 28580
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.986849026827986e-06,
+      "loss": 1.6893,
+      "step": 28590
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.985095563738383e-06,
+      "loss": 1.8434,
+      "step": 28600
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.983342100648782e-06,
+      "loss": 1.6592,
+      "step": 28610
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.98158863755918e-06,
+      "loss": 1.6721,
+      "step": 28620
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.979835174469577e-06,
+      "loss": 1.6668,
+      "step": 28630
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.978081711379976e-06,
+      "loss": 1.4929,
+      "step": 28640
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.976328248290374e-06,
+      "loss": 1.4568,
+      "step": 28650
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.974574785200772e-06,
+      "loss": 1.4584,
+      "step": 28660
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.97282132211117e-06,
+      "loss": 1.5314,
+      "step": 28670
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.971067859021568e-06,
+      "loss": 1.4201,
+      "step": 28680
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.9693143959319665e-06,
+      "loss": 1.3979,
+      "step": 28690
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.967560932842364e-06,
+      "loss": 1.5938,
+      "step": 28700
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.965807469752762e-06,
+      "loss": 1.5715,
+      "step": 28710
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.964054006663161e-06,
+      "loss": 1.3833,
+      "step": 28720
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.962300543573558e-06,
+      "loss": 1.3937,
+      "step": 28730
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.960547080483956e-06,
+      "loss": 1.5233,
+      "step": 28740
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.958793617394354e-06,
+      "loss": 1.4752,
+      "step": 28750
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.957040154304752e-06,
+      "loss": 1.4405,
+      "step": 28760
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.9552866912151505e-06,
+      "loss": 1.5753,
+      "step": 28770
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.953533228125548e-06,
+      "loss": 1.6243,
+      "step": 28780
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.951779765035946e-06,
+      "loss": 1.6153,
+      "step": 28790
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.950026301946345e-06,
+      "loss": 1.6256,
+      "step": 28800
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.948272838856742e-06,
+      "loss": 1.4774,
+      "step": 28810
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.9465193757671405e-06,
+      "loss": 1.4703,
+      "step": 28820
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.944765912677539e-06,
+      "loss": 1.5844,
+      "step": 28830
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.943012449587937e-06,
+      "loss": 1.5519,
+      "step": 28840
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.941258986498335e-06,
+      "loss": 1.6124,
+      "step": 28850
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.939505523408732e-06,
+      "loss": 1.7611,
+      "step": 28860
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.937752060319131e-06,
+      "loss": 1.4897,
+      "step": 28870
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.935998597229529e-06,
+      "loss": 1.534,
+      "step": 28880
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.934245134139927e-06,
+      "loss": 1.4271,
+      "step": 28890
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.9324916710503245e-06,
+      "loss": 1.5485,
+      "step": 28900
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.930738207960723e-06,
+      "loss": 1.9217,
+      "step": 28910
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.928984744871121e-06,
+      "loss": 1.5401,
+      "step": 28920
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.927231281781519e-06,
+      "loss": 1.603,
+      "step": 28930
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.925477818691917e-06,
+      "loss": 1.5914,
+      "step": 28940
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.923724355602315e-06,
+      "loss": 1.4992,
+      "step": 28950
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.921970892512713e-06,
+      "loss": 1.5037,
+      "step": 28960
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.920217429423111e-06,
+      "loss": 1.5778,
+      "step": 28970
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.9184639663335094e-06,
+      "loss": 1.5692,
+      "step": 28980
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.916710503243907e-06,
+      "loss": 1.6506,
+      "step": 28990
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.914957040154305e-06,
+      "loss": 1.485,
+      "step": 29000
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.913203577064703e-06,
+      "loss": 1.5542,
+      "step": 29010
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.911450113975101e-06,
+      "loss": 1.5913,
+      "step": 29020
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.909696650885499e-06,
+      "loss": 1.5864,
+      "step": 29030
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.907943187795897e-06,
+      "loss": 1.9161,
+      "step": 29040
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.906189724706295e-06,
+      "loss": 1.6031,
+      "step": 29050
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.9044362616166935e-06,
+      "loss": 1.6042,
+      "step": 29060
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.902682798527091e-06,
+      "loss": 1.5554,
+      "step": 29070
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.900929335437489e-06,
+      "loss": 1.5022,
+      "step": 29080
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.899175872347888e-06,
+      "loss": 1.5867,
+      "step": 29090
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.897422409258286e-06,
+      "loss": 1.7183,
+      "step": 29100
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.8956689461686834e-06,
+      "loss": 1.741,
+      "step": 29110
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.893915483079082e-06,
+      "loss": 1.548,
+      "step": 29120
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.89216201998948e-06,
+      "loss": 1.4176,
+      "step": 29130
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.8904085568998776e-06,
+      "loss": 1.5821,
+      "step": 29140
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.888655093810276e-06,
+      "loss": 1.4377,
+      "step": 29150
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.886901630720673e-06,
+      "loss": 1.5031,
+      "step": 29160
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.885148167631072e-06,
+      "loss": 1.6229,
+      "step": 29170
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.88339470454147e-06,
+      "loss": 1.4416,
+      "step": 29180
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.8816412414518675e-06,
+      "loss": 1.4422,
+      "step": 29190
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.879887778362266e-06,
+      "loss": 1.4183,
+      "step": 29200
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.878134315272664e-06,
+      "loss": 1.7316,
+      "step": 29210
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.876380852183062e-06,
+      "loss": 1.5109,
+      "step": 29220
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.87462738909346e-06,
+      "loss": 1.5344,
+      "step": 29230
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.872873926003858e-06,
+      "loss": 1.429,
+      "step": 29240
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.871120462914256e-06,
+      "loss": 1.642,
+      "step": 29250
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.869366999824654e-06,
+      "loss": 1.5856,
+      "step": 29260
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.8676135367350516e-06,
+      "loss": 1.5686,
+      "step": 29270
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.865860073645451e-06,
+      "loss": 1.4497,
+      "step": 29280
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.864106610555848e-06,
+      "loss": 1.5879,
+      "step": 29290
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.862353147466246e-06,
+      "loss": 1.5707,
+      "step": 29300
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.860599684376645e-06,
+      "loss": 1.4596,
+      "step": 29310
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.858846221287042e-06,
+      "loss": 1.5324,
+      "step": 29320
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.857092758197441e-06,
+      "loss": 1.5642,
+      "step": 29330
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.855339295107838e-06,
+      "loss": 1.5286,
+      "step": 29340
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.8535858320182365e-06,
+      "loss": 1.6952,
+      "step": 29350
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.851832368928635e-06,
+      "loss": 1.6379,
+      "step": 29360
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.850078905839032e-06,
+      "loss": 1.4118,
+      "step": 29370
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.848325442749431e-06,
+      "loss": 1.4504,
+      "step": 29380
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.846571979659829e-06,
+      "loss": 1.4415,
+      "step": 29390
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.844818516570226e-06,
+      "loss": 1.6589,
+      "step": 29400
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.843065053480625e-06,
+      "loss": 1.406,
+      "step": 29410
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.841311590391023e-06,
+      "loss": 1.5107,
+      "step": 29420
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.8395581273014205e-06,
+      "loss": 1.5698,
+      "step": 29430
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.837804664211819e-06,
+      "loss": 1.4501,
+      "step": 29440
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.836051201122216e-06,
+      "loss": 1.4354,
+      "step": 29450
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.834297738032615e-06,
+      "loss": 1.6241,
+      "step": 29460
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.832544274943013e-06,
+      "loss": 1.3984,
+      "step": 29470
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.8307908118534105e-06,
+      "loss": 1.7464,
+      "step": 29480
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.829037348763809e-06,
+      "loss": 1.4698,
+      "step": 29490
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.827283885674207e-06,
+      "loss": 1.5723,
+      "step": 29500
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.8255304225846054e-06,
+      "loss": 1.5679,
+      "step": 29510
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.823776959495003e-06,
+      "loss": 1.5729,
+      "step": 29520
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.8220234964054e-06,
+      "loss": 1.4854,
+      "step": 29530
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.8202700333157996e-06,
+      "loss": 1.6681,
+      "step": 29540
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.818516570226197e-06,
+      "loss": 1.5327,
+      "step": 29550
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.816763107136595e-06,
+      "loss": 1.5296,
+      "step": 29560
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.815009644046994e-06,
+      "loss": 1.5311,
+      "step": 29570
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.813256180957391e-06,
+      "loss": 1.7297,
+      "step": 29580
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.8115027178677895e-06,
+      "loss": 1.5641,
+      "step": 29590
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.809749254778187e-06,
+      "loss": 1.4773,
+      "step": 29600
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.807995791688585e-06,
+      "loss": 1.5482,
+      "step": 29610
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.806242328598984e-06,
+      "loss": 1.4495,
+      "step": 29620
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.804488865509381e-06,
+      "loss": 1.6052,
+      "step": 29630
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.8027354024197794e-06,
+      "loss": 1.5159,
+      "step": 29640
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.800981939330178e-06,
+      "loss": 1.6329,
+      "step": 29650
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.799228476240575e-06,
+      "loss": 1.5233,
+      "step": 29660
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.7974750131509736e-06,
+      "loss": 1.5374,
+      "step": 29670
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.795721550061372e-06,
+      "loss": 1.561,
+      "step": 29680
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.793968086971769e-06,
+      "loss": 1.5663,
+      "step": 29690
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.792214623882168e-06,
+      "loss": 1.72,
+      "step": 29700
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.790461160792565e-06,
+      "loss": 1.6227,
+      "step": 29710
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.788707697702964e-06,
+      "loss": 1.5999,
+      "step": 29720
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.786954234613362e-06,
+      "loss": 1.5547,
+      "step": 29730
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.785200771523759e-06,
+      "loss": 1.5882,
+      "step": 29740
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.783447308434158e-06,
+      "loss": 1.4569,
+      "step": 29750
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.781693845344556e-06,
+      "loss": 1.6038,
+      "step": 29760
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.779940382254954e-06,
+      "loss": 1.6168,
+      "step": 29770
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.778186919165352e-06,
+      "loss": 1.6043,
+      "step": 29780
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.77643345607575e-06,
+      "loss": 1.7236,
+      "step": 29790
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.774679992986148e-06,
+      "loss": 1.6534,
+      "step": 29800
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.772926529896546e-06,
+      "loss": 1.4315,
+      "step": 29810
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.771173066806944e-06,
+      "loss": 1.6434,
+      "step": 29820
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.7694196037173426e-06,
+      "loss": 1.7974,
+      "step": 29830
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.76766614062774e-06,
+      "loss": 1.6455,
+      "step": 29840
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.765912677538138e-06,
+      "loss": 1.4886,
+      "step": 29850
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.764159214448536e-06,
+      "loss": 1.4712,
+      "step": 29860
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.762405751358934e-06,
+      "loss": 1.5206,
+      "step": 29870
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.7606522882693325e-06,
+      "loss": 1.6667,
+      "step": 29880
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.75889882517973e-06,
+      "loss": 1.4898,
+      "step": 29890
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.757145362090128e-06,
+      "loss": 1.6104,
+      "step": 29900
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.755391899000527e-06,
+      "loss": 1.6266,
+      "step": 29910
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.753638435910924e-06,
+      "loss": 1.5251,
+      "step": 29920
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.751884972821322e-06,
+      "loss": 1.4187,
+      "step": 29930
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.750131509731721e-06,
+      "loss": 1.588,
+      "step": 29940
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.748378046642119e-06,
+      "loss": 1.5416,
+      "step": 29950
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.7466245835525166e-06,
+      "loss": 1.4969,
+      "step": 29960
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.744871120462914e-06,
+      "loss": 1.6688,
+      "step": 29970
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.743117657373313e-06,
+      "loss": 1.4995,
+      "step": 29980
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.741364194283711e-06,
+      "loss": 1.4746,
+      "step": 29990
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.739610731194109e-06,
+      "loss": 1.4178,
+      "step": 30000
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.7378572681045065e-06,
+      "loss": 1.6032,
+      "step": 30010
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.736103805014905e-06,
+      "loss": 1.6191,
+      "step": 30020
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.734350341925303e-06,
+      "loss": 1.5895,
+      "step": 30030
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.732596878835701e-06,
+      "loss": 1.523,
+      "step": 30040
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.730843415746099e-06,
+      "loss": 1.5144,
+      "step": 30050
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.729089952656497e-06,
+      "loss": 1.4387,
+      "step": 30060
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.727336489566895e-06,
+      "loss": 1.6958,
+      "step": 30070
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.725583026477293e-06,
+      "loss": 1.4305,
+      "step": 30080
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.723829563387691e-06,
+      "loss": 1.5751,
+      "step": 30090
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.722076100298089e-06,
+      "loss": 1.7727,
+      "step": 30100
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.720322637208487e-06,
+      "loss": 1.6456,
+      "step": 30110
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.718569174118885e-06,
+      "loss": 1.5236,
+      "step": 30120
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.716815711029283e-06,
+      "loss": 1.5413,
+      "step": 30130
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.715062247939681e-06,
+      "loss": 1.5825,
+      "step": 30140
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.713308784850079e-06,
+      "loss": 1.6772,
+      "step": 30150
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.711555321760478e-06,
+      "loss": 1.6934,
+      "step": 30160
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.7098018586708755e-06,
+      "loss": 1.5698,
+      "step": 30170
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.708048395581273e-06,
+      "loss": 1.4636,
+      "step": 30180
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.706294932491671e-06,
+      "loss": 1.6175,
+      "step": 30190
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.70454146940207e-06,
+      "loss": 1.4543,
+      "step": 30200
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.702788006312468e-06,
+      "loss": 1.4226,
+      "step": 30210
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.701034543222865e-06,
+      "loss": 1.5437,
+      "step": 30220
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.699281080133264e-06,
+      "loss": 1.4846,
+      "step": 30230
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.697527617043662e-06,
+      "loss": 1.5457,
+      "step": 30240
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.6957741539540595e-06,
+      "loss": 1.5413,
+      "step": 30250
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.694020690864458e-06,
+      "loss": 1.6429,
+      "step": 30260
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.692267227774856e-06,
+      "loss": 1.5038,
+      "step": 30270
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.690513764685254e-06,
+      "loss": 1.5253,
+      "step": 30280
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.688760301595652e-06,
+      "loss": 1.5399,
+      "step": 30290
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.6870068385060495e-06,
+      "loss": 1.5754,
+      "step": 30300
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.685253375416448e-06,
+      "loss": 1.4751,
+      "step": 30310
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.683499912326846e-06,
+      "loss": 1.4973,
+      "step": 30320
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.681746449237244e-06,
+      "loss": 1.3895,
+      "step": 30330
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.679992986147642e-06,
+      "loss": 1.5225,
+      "step": 30340
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.67823952305804e-06,
+      "loss": 1.7305,
+      "step": 30350
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.676486059968438e-06,
+      "loss": 1.4873,
+      "step": 30360
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.674732596878836e-06,
+      "loss": 1.4113,
+      "step": 30370
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.6729791337892335e-06,
+      "loss": 1.7004,
+      "step": 30380
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.671225670699633e-06,
+      "loss": 1.4383,
+      "step": 30390
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.66947220761003e-06,
+      "loss": 1.5451,
+      "step": 30400
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.667718744520428e-06,
+      "loss": 1.5652,
+      "step": 30410
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.665965281430827e-06,
+      "loss": 1.4814,
+      "step": 30420
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.664211818341224e-06,
+      "loss": 1.5358,
+      "step": 30430
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.662458355251623e-06,
+      "loss": 1.468,
+      "step": 30440
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.66070489216202e-06,
+      "loss": 1.4658,
+      "step": 30450
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.6589514290724184e-06,
+      "loss": 1.5021,
+      "step": 30460
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.657197965982817e-06,
+      "loss": 1.6099,
+      "step": 30470
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.655444502893214e-06,
+      "loss": 1.4408,
+      "step": 30480
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.6536910398036126e-06,
+      "loss": 1.6106,
+      "step": 30490
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.651937576714011e-06,
+      "loss": 1.6143,
+      "step": 30500
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.650184113624408e-06,
+      "loss": 1.6702,
+      "step": 30510
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.648430650534807e-06,
+      "loss": 1.5792,
+      "step": 30520
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.646677187445205e-06,
+      "loss": 1.6299,
+      "step": 30530
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.6449237243556025e-06,
+      "loss": 1.758,
+      "step": 30540
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.643170261266001e-06,
+      "loss": 1.6232,
+      "step": 30550
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.641416798176398e-06,
+      "loss": 1.4781,
+      "step": 30560
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.639663335086797e-06,
+      "loss": 1.6128,
+      "step": 30570
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.637909871997195e-06,
+      "loss": 1.5356,
+      "step": 30580
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.6361564089075924e-06,
+      "loss": 1.4767,
+      "step": 30590
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.634402945817991e-06,
+      "loss": 1.2765,
+      "step": 30600
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.632649482728389e-06,
+      "loss": 1.6906,
+      "step": 30610
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.6308960196387866e-06,
+      "loss": 1.4095,
+      "step": 30620
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.629142556549185e-06,
+      "loss": 1.6177,
+      "step": 30630
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.627389093459583e-06,
+      "loss": 1.5118,
+      "step": 30640
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.6256356303699815e-06,
+      "loss": 1.6265,
+      "step": 30650
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.623882167280379e-06,
+      "loss": 1.5529,
+      "step": 30660
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.622128704190777e-06,
+      "loss": 1.3941,
+      "step": 30670
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.620375241101176e-06,
+      "loss": 1.5197,
+      "step": 30680
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.618621778011573e-06,
+      "loss": 1.4036,
+      "step": 30690
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.6168683149219715e-06,
+      "loss": 1.6305,
+      "step": 30700
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.615114851832369e-06,
+      "loss": 1.6724,
+      "step": 30710
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.613361388742767e-06,
+      "loss": 1.461,
+      "step": 30720
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.611607925653166e-06,
+      "loss": 1.5,
+      "step": 30730
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.609854462563563e-06,
+      "loss": 1.5729,
+      "step": 30740
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.608100999473961e-06,
+      "loss": 1.7545,
+      "step": 30750
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.60634753638436e-06,
+      "loss": 1.7098,
+      "step": 30760
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.604594073294757e-06,
+      "loss": 1.5813,
+      "step": 30770
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.6028406102051555e-06,
+      "loss": 1.4226,
+      "step": 30780
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.601087147115554e-06,
+      "loss": 1.4869,
+      "step": 30790
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.599333684025951e-06,
+      "loss": 1.3897,
+      "step": 30800
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.59758022093635e-06,
+      "loss": 1.4617,
+      "step": 30810
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.595826757846747e-06,
+      "loss": 1.4245,
+      "step": 30820
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.594073294757146e-06,
+      "loss": 1.5978,
+      "step": 30830
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.592319831667544e-06,
+      "loss": 1.5336,
+      "step": 30840
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.590566368577941e-06,
+      "loss": 1.6453,
+      "step": 30850
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.58881290548834e-06,
+      "loss": 1.4344,
+      "step": 30860
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.587059442398738e-06,
+      "loss": 1.6331,
+      "step": 30870
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.585305979309136e-06,
+      "loss": 1.4925,
+      "step": 30880
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.583552516219534e-06,
+      "loss": 1.5506,
+      "step": 30890
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.581799053129932e-06,
+      "loss": 1.8352,
+      "step": 30900
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.58004559004033e-06,
+      "loss": 1.4872,
+      "step": 30910
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.578292126950728e-06,
+      "loss": 1.4634,
+      "step": 30920
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.576538663861126e-06,
+      "loss": 1.6179,
+      "step": 30930
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5747852007715245e-06,
+      "loss": 1.5605,
+      "step": 30940
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.573031737681922e-06,
+      "loss": 1.4639,
+      "step": 30950
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.57127827459232e-06,
+      "loss": 1.5326,
+      "step": 30960
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.569524811502718e-06,
+      "loss": 1.5321,
+      "step": 30970
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.567771348413116e-06,
+      "loss": 1.4178,
+      "step": 30980
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.5660178853235145e-06,
+      "loss": 1.5305,
+      "step": 30990
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.564264422233912e-06,
+      "loss": 1.5144,
+      "step": 31000
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.56251095914431e-06,
+      "loss": 1.5148,
+      "step": 31010
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.560757496054709e-06,
+      "loss": 1.4474,
+      "step": 31020
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.559004032965106e-06,
+      "loss": 1.5639,
+      "step": 31030
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.557250569875504e-06,
+      "loss": 1.5621,
+      "step": 31040
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.555497106785903e-06,
+      "loss": 1.5458,
+      "step": 31050
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.5537436436963e-06,
+      "loss": 1.6037,
+      "step": 31060
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.5519901806066985e-06,
+      "loss": 1.3442,
+      "step": 31070
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.550236717517096e-06,
+      "loss": 1.5306,
+      "step": 31080
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.548483254427495e-06,
+      "loss": 1.6873,
+      "step": 31090
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.546729791337893e-06,
+      "loss": 1.6715,
+      "step": 31100
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.544976328248291e-06,
+      "loss": 1.7701,
+      "step": 31110
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.543222865158689e-06,
+      "loss": 1.5811,
+      "step": 31120
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.541469402069087e-06,
+      "loss": 1.7386,
+      "step": 31130
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.539715938979485e-06,
+      "loss": 1.486,
+      "step": 31140
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.537962475889883e-06,
+      "loss": 1.4555,
+      "step": 31150
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.536209012800281e-06,
+      "loss": 1.5361,
+      "step": 31160
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.534455549710679e-06,
+      "loss": 1.4577,
+      "step": 31170
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.532702086621077e-06,
+      "loss": 1.5314,
+      "step": 31180
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.530948623531475e-06,
+      "loss": 1.5414,
+      "step": 31190
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.529195160441873e-06,
+      "loss": 1.4987,
+      "step": 31200
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.527441697352271e-06,
+      "loss": 1.5819,
+      "step": 31210
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.525688234262669e-06,
+      "loss": 1.5675,
+      "step": 31220
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.5239347711730675e-06,
+      "loss": 1.5404,
+      "step": 31230
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.522181308083465e-06,
+      "loss": 1.6142,
+      "step": 31240
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.520427844993863e-06,
+      "loss": 1.4934,
+      "step": 31250
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.518674381904261e-06,
+      "loss": 1.5967,
+      "step": 31260
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.51692091881466e-06,
+      "loss": 1.5517,
+      "step": 31270
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.5151674557250574e-06,
+      "loss": 1.4031,
+      "step": 31280
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.513413992635455e-06,
+      "loss": 1.3656,
+      "step": 31290
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.511660529545853e-06,
+      "loss": 1.6596,
+      "step": 31300
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.5099070664562516e-06,
+      "loss": 1.4589,
+      "step": 31310
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.50815360336665e-06,
+      "loss": 1.5647,
+      "step": 31320
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.506400140277047e-06,
+      "loss": 1.6003,
+      "step": 31330
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.504646677187446e-06,
+      "loss": 1.6502,
+      "step": 31340
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.502893214097844e-06,
+      "loss": 1.7201,
+      "step": 31350
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.5011397510082415e-06,
+      "loss": 1.6337,
+      "step": 31360
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.49938628791864e-06,
+      "loss": 1.538,
+      "step": 31370
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.497632824829038e-06,
+      "loss": 1.5516,
+      "step": 31380
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.495879361739436e-06,
+      "loss": 1.5573,
+      "step": 31390
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.494125898649834e-06,
+      "loss": 1.5307,
+      "step": 31400
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.4923724355602314e-06,
+      "loss": 1.4757,
+      "step": 31410
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.49061897247063e-06,
+      "loss": 1.5289,
+      "step": 31420
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.488865509381028e-06,
+      "loss": 1.515,
+      "step": 31430
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.4871120462914256e-06,
+      "loss": 1.5831,
+      "step": 31440
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.485358583201824e-06,
+      "loss": 1.4739,
+      "step": 31450
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.483605120112222e-06,
+      "loss": 1.646,
+      "step": 31460
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.48185165702262e-06,
+      "loss": 1.5408,
+      "step": 31470
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.480098193933018e-06,
+      "loss": 1.4125,
+      "step": 31480
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.478344730843416e-06,
+      "loss": 1.5417,
+      "step": 31490
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.476591267753815e-06,
+      "loss": 1.5035,
+      "step": 31500
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.474837804664212e-06,
+      "loss": 1.4329,
+      "step": 31510
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.47308434157461e-06,
+      "loss": 1.5015,
+      "step": 31520
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.471330878485009e-06,
+      "loss": 1.472,
+      "step": 31530
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.469577415395406e-06,
+      "loss": 1.644,
+      "step": 31540
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.467823952305805e-06,
+      "loss": 1.5082,
+      "step": 31550
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.466070489216202e-06,
+      "loss": 1.8172,
+      "step": 31560
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.4643170261266e-06,
+      "loss": 1.354,
+      "step": 31570
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.462563563036999e-06,
+      "loss": 1.5885,
+      "step": 31580
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.460810099947396e-06,
+      "loss": 1.628,
+      "step": 31590
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.4590566368577945e-06,
+      "loss": 1.5419,
+      "step": 31600
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.457303173768193e-06,
+      "loss": 1.6665,
+      "step": 31610
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.45554971067859e-06,
+      "loss": 1.4393,
+      "step": 31620
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.453796247588989e-06,
+      "loss": 1.7056,
+      "step": 31630
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.452042784499387e-06,
+      "loss": 1.5167,
+      "step": 31640
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.4502893214097845e-06,
+      "loss": 1.6355,
+      "step": 31650
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.448535858320183e-06,
+      "loss": 1.5207,
+      "step": 31660
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.44678239523058e-06,
+      "loss": 1.6283,
+      "step": 31670
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.445028932140979e-06,
+      "loss": 1.5604,
+      "step": 31680
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.443275469051377e-06,
+      "loss": 1.5933,
+      "step": 31690
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.441522005961774e-06,
+      "loss": 1.4238,
+      "step": 31700
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.439768542872173e-06,
+      "loss": 1.6753,
+      "step": 31710
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.438015079782571e-06,
+      "loss": 1.4957,
+      "step": 31720
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.4362616166929685e-06,
+      "loss": 1.573,
+      "step": 31730
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.434508153603367e-06,
+      "loss": 1.4976,
+      "step": 31740
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.432754690513765e-06,
+      "loss": 1.4682,
+      "step": 31750
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.4310012274241635e-06,
+      "loss": 1.619,
+      "step": 31760
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.429247764334561e-06,
+      "loss": 1.6544,
+      "step": 31770
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.427494301244959e-06,
+      "loss": 1.4757,
+      "step": 31780
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.425740838155358e-06,
+      "loss": 1.4819,
+      "step": 31790
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.423987375065755e-06,
+      "loss": 1.6172,
+      "step": 31800
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.4222339119761535e-06,
+      "loss": 1.5965,
+      "step": 31810
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.420480448886551e-06,
+      "loss": 1.4845,
+      "step": 31820
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.418726985796949e-06,
+      "loss": 1.575,
+      "step": 31830
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.416973522707348e-06,
+      "loss": 1.7414,
+      "step": 31840
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.415220059617745e-06,
+      "loss": 1.4835,
+      "step": 31850
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.413466596528143e-06,
+      "loss": 1.5611,
+      "step": 31860
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.411713133438542e-06,
+      "loss": 1.5873,
+      "step": 31870
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.409959670348939e-06,
+      "loss": 1.5994,
+      "step": 31880
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.4082062072593375e-06,
+      "loss": 1.5252,
+      "step": 31890
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.406452744169736e-06,
+      "loss": 1.5078,
+      "step": 31900
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.404699281080133e-06,
+      "loss": 1.5624,
+      "step": 31910
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.402945817990532e-06,
+      "loss": 1.4423,
+      "step": 31920
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.401192354900929e-06,
+      "loss": 1.6397,
+      "step": 31930
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.399438891811328e-06,
+      "loss": 1.4217,
+      "step": 31940
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.397685428721726e-06,
+      "loss": 1.3859,
+      "step": 31950
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.395931965632123e-06,
+      "loss": 1.6612,
+      "step": 31960
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.3941785025425224e-06,
+      "loss": 1.5113,
+      "step": 31970
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.39242503945292e-06,
+      "loss": 1.6408,
+      "step": 31980
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.390671576363318e-06,
+      "loss": 1.7209,
+      "step": 31990
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.388918113273716e-06,
+      "loss": 1.6609,
+      "step": 32000
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.387164650184114e-06,
+      "loss": 1.6051,
+      "step": 32010
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.385411187094512e-06,
+      "loss": 1.434,
+      "step": 32020
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.38365772400491e-06,
+      "loss": 1.4696,
+      "step": 32030
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.381904260915308e-06,
+      "loss": 1.6069,
+      "step": 32040
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.3801507978257065e-06,
+      "loss": 1.5535,
+      "step": 32050
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.378397334736104e-06,
+      "loss": 1.396,
+      "step": 32060
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.376643871646502e-06,
+      "loss": 1.5519,
+      "step": 32070
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.374890408556901e-06,
+      "loss": 1.5262,
+      "step": 32080
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.373136945467298e-06,
+      "loss": 1.4855,
+      "step": 32090
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3713834823776964e-06,
+      "loss": 1.472,
+      "step": 32100
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.369630019288094e-06,
+      "loss": 1.6057,
+      "step": 32110
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.367876556198492e-06,
+      "loss": 1.4632,
+      "step": 32120
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.3661230931088906e-06,
+      "loss": 1.5945,
+      "step": 32130
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.364369630019288e-06,
+      "loss": 1.691,
+      "step": 32140
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.362616166929686e-06,
+      "loss": 1.5623,
+      "step": 32150
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.360862703840085e-06,
+      "loss": 1.514,
+      "step": 32160
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.359109240750482e-06,
+      "loss": 1.5458,
+      "step": 32170
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.3573557776608805e-06,
+      "loss": 1.6309,
+      "step": 32180
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.355602314571278e-06,
+      "loss": 1.5278,
+      "step": 32190
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.353848851481677e-06,
+      "loss": 1.4707,
+      "step": 32200
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.352095388392075e-06,
+      "loss": 1.5208,
+      "step": 32210
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.350341925302473e-06,
+      "loss": 1.6477,
+      "step": 32220
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.348588462212871e-06,
+      "loss": 1.4418,
+      "step": 32230
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.346834999123269e-06,
+      "loss": 1.6622,
+      "step": 32240
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.345081536033667e-06,
+      "loss": 1.4894,
+      "step": 32250
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.3433280729440646e-06,
+      "loss": 1.4652,
+      "step": 32260
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.341574609854463e-06,
+      "loss": 1.5991,
+      "step": 32270
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.339821146764861e-06,
+      "loss": 1.6547,
+      "step": 32280
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.338067683675259e-06,
+      "loss": 1.7149,
+      "step": 32290
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.336314220585657e-06,
+      "loss": 1.4703,
+      "step": 32300
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.334560757496055e-06,
+      "loss": 1.4538,
+      "step": 32310
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.332807294406453e-06,
+      "loss": 1.5367,
+      "step": 32320
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.331053831316851e-06,
+      "loss": 1.5684,
+      "step": 32330
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.3293003682272495e-06,
+      "loss": 1.5359,
+      "step": 32340
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.327546905137647e-06,
+      "loss": 1.5431,
+      "step": 32350
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.325793442048045e-06,
+      "loss": 1.6122,
+      "step": 32360
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.324039978958443e-06,
+      "loss": 1.499,
+      "step": 32370
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.322286515868842e-06,
+      "loss": 1.5917,
+      "step": 32380
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.320533052779239e-06,
+      "loss": 1.6406,
+      "step": 32390
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.318779589689637e-06,
+      "loss": 1.5241,
+      "step": 32400
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.317026126600035e-06,
+      "loss": 1.5771,
+      "step": 32410
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.3152726635104335e-06,
+      "loss": 1.3432,
+      "step": 32420
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.313519200420832e-06,
+      "loss": 1.5283,
+      "step": 32430
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.311765737331229e-06,
+      "loss": 1.702,
+      "step": 32440
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.310012274241628e-06,
+      "loss": 1.3552,
+      "step": 32450
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.308258811152026e-06,
+      "loss": 1.4913,
+      "step": 32460
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.3065053480624235e-06,
+      "loss": 1.7571,
+      "step": 32470
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.304751884972822e-06,
+      "loss": 1.598,
+      "step": 32480
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.30299842188322e-06,
+      "loss": 1.4653,
+      "step": 32490
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.301244958793618e-06,
+      "loss": 1.4027,
+      "step": 32500
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.299491495704016e-06,
+      "loss": 1.4281,
+      "step": 32510
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.297738032614413e-06,
+      "loss": 1.5961,
+      "step": 32520
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.295984569524812e-06,
+      "loss": 1.888,
+      "step": 32530
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.29423110643521e-06,
+      "loss": 1.5247,
+      "step": 32540
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.2924776433456075e-06,
+      "loss": 1.6286,
+      "step": 32550
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.290724180256006e-06,
+      "loss": 1.7376,
+      "step": 32560
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.288970717166404e-06,
+      "loss": 1.5539,
+      "step": 32570
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.287217254076802e-06,
+      "loss": 1.4835,
+      "step": 32580
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.2854637909872e-06,
+      "loss": 1.5291,
+      "step": 32590
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.283710327897598e-06,
+      "loss": 1.5213,
+      "step": 32600
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.281956864807996e-06,
+      "loss": 1.3726,
+      "step": 32610
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.280203401718394e-06,
+      "loss": 1.7124,
+      "step": 32620
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.278449938628792e-06,
+      "loss": 1.5428,
+      "step": 32630
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.276696475539191e-06,
+      "loss": 1.4115,
+      "step": 32640
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.274943012449588e-06,
+      "loss": 1.5214,
+      "step": 32650
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.273189549359987e-06,
+      "loss": 1.6277,
+      "step": 32660
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.271436086270384e-06,
+      "loss": 1.4733,
+      "step": 32670
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.269682623180782e-06,
+      "loss": 1.7199,
+      "step": 32680
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.267929160091181e-06,
+      "loss": 1.5045,
+      "step": 32690
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.266175697001578e-06,
+      "loss": 1.5913,
+      "step": 32700
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.2644222339119765e-06,
+      "loss": 1.4817,
+      "step": 32710
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.262668770822375e-06,
+      "loss": 1.6013,
+      "step": 32720
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.260915307732772e-06,
+      "loss": 1.5672,
+      "step": 32730
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.259161844643171e-06,
+      "loss": 1.4405,
+      "step": 32740
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.257408381553569e-06,
+      "loss": 1.5327,
+      "step": 32750
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.2556549184639664e-06,
+      "loss": 1.8063,
+      "step": 32760
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.253901455374365e-06,
+      "loss": 1.4402,
+      "step": 32770
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.252147992284762e-06,
+      "loss": 1.5931,
+      "step": 32780
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.2503945291951606e-06,
+      "loss": 1.5517,
+      "step": 32790
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.248641066105559e-06,
+      "loss": 1.457,
+      "step": 32800
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.246887603015956e-06,
+      "loss": 1.5172,
+      "step": 32810
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.2451341399263556e-06,
+      "loss": 1.5094,
+      "step": 32820
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.243380676836753e-06,
+      "loss": 1.751,
+      "step": 32830
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.2416272137471505e-06,
+      "loss": 1.6902,
+      "step": 32840
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.239873750657549e-06,
+      "loss": 1.5455,
+      "step": 32850
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.238120287567947e-06,
+      "loss": 1.5378,
+      "step": 32860
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.2363668244783455e-06,
+      "loss": 1.4601,
+      "step": 32870
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.234613361388743e-06,
+      "loss": 1.5368,
+      "step": 32880
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.232859898299141e-06,
+      "loss": 1.814,
+      "step": 32890
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.23110643520954e-06,
+      "loss": 1.5515,
+      "step": 32900
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.229352972119937e-06,
+      "loss": 1.5608,
+      "step": 32910
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.227599509030335e-06,
+      "loss": 1.4604,
+      "step": 32920
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.225846045940734e-06,
+      "loss": 1.5054,
+      "step": 32930
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.224092582851131e-06,
+      "loss": 1.6051,
+      "step": 32940
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.2223391197615296e-06,
+      "loss": 1.3682,
+      "step": 32950
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.220585656671927e-06,
+      "loss": 1.4465,
+      "step": 32960
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.218832193582325e-06,
+      "loss": 1.7791,
+      "step": 32970
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.217078730492724e-06,
+      "loss": 1.5825,
+      "step": 32980
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.215325267403121e-06,
+      "loss": 1.6056,
+      "step": 32990
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.2135718043135195e-06,
+      "loss": 1.3463,
+      "step": 33000
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.211818341223918e-06,
+      "loss": 1.6173,
+      "step": 33010
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.210064878134315e-06,
+      "loss": 1.561,
+      "step": 33020
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.208311415044714e-06,
+      "loss": 1.572,
+      "step": 33030
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.206557951955112e-06,
+      "loss": 1.4243,
+      "step": 33040
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.20480448886551e-06,
+      "loss": 1.579,
+      "step": 33050
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.203051025775908e-06,
+      "loss": 1.6885,
+      "step": 33060
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.201297562686305e-06,
+      "loss": 1.6242,
+      "step": 33070
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.199544099596704e-06,
+      "loss": 1.4645,
+      "step": 33080
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.197790636507102e-06,
+      "loss": 1.4269,
+      "step": 33090
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.1960371734175e-06,
+      "loss": 1.4908,
+      "step": 33100
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.194283710327898e-06,
+      "loss": 1.6803,
+      "step": 33110
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.192530247238296e-06,
+      "loss": 1.5114,
+      "step": 33120
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.190776784148694e-06,
+      "loss": 1.4471,
+      "step": 33130
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.189023321059092e-06,
+      "loss": 1.6234,
+      "step": 33140
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.18726985796949e-06,
+      "loss": 1.5634,
+      "step": 33150
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.1855163948798885e-06,
+      "loss": 1.5218,
+      "step": 33160
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.183762931790286e-06,
+      "loss": 1.532,
+      "step": 33170
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.182009468700684e-06,
+      "loss": 1.4456,
+      "step": 33180
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.180256005611083e-06,
+      "loss": 1.5403,
+      "step": 33190
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.17850254252148e-06,
+      "loss": 1.519,
+      "step": 33200
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.176749079431878e-06,
+      "loss": 1.4153,
+      "step": 33210
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.174995616342276e-06,
+      "loss": 1.7524,
+      "step": 33220
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.173242153252674e-06,
+      "loss": 1.5256,
+      "step": 33230
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.1714886901630725e-06,
+      "loss": 1.5024,
+      "step": 33240
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.16973522707347e-06,
+      "loss": 1.4427,
+      "step": 33250
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.167981763983868e-06,
+      "loss": 1.5674,
+      "step": 33260
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.166228300894267e-06,
+      "loss": 1.5878,
+      "step": 33270
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.164474837804664e-06,
+      "loss": 1.6509,
+      "step": 33280
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.1627213747150625e-06,
+      "loss": 1.4865,
+      "step": 33290
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.160967911625461e-06,
+      "loss": 1.5153,
+      "step": 33300
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.159214448535859e-06,
+      "loss": 1.3914,
+      "step": 33310
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.157460985446257e-06,
+      "loss": 1.4976,
+      "step": 33320
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.155707522356655e-06,
+      "loss": 1.3985,
+      "step": 33330
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.153954059267053e-06,
+      "loss": 1.569,
+      "step": 33340
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.152200596177451e-06,
+      "loss": 1.4571,
+      "step": 33350
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.150447133087849e-06,
+      "loss": 1.4599,
+      "step": 33360
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.1486936699982465e-06,
+      "loss": 1.5807,
+      "step": 33370
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.146940206908645e-06,
+      "loss": 1.4662,
+      "step": 33380
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.145186743819043e-06,
+      "loss": 1.56,
+      "step": 33390
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.143433280729441e-06,
+      "loss": 1.363,
+      "step": 33400
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.141679817639839e-06,
+      "loss": 1.3977,
+      "step": 33410
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.139926354550237e-06,
+      "loss": 1.5602,
+      "step": 33420
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.138172891460635e-06,
+      "loss": 1.5519,
+      "step": 33430
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.136419428371033e-06,
+      "loss": 1.4664,
+      "step": 33440
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.1346659652814314e-06,
+      "loss": 1.4643,
+      "step": 33450
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.132912502191829e-06,
+      "loss": 1.574,
+      "step": 33460
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.131159039102227e-06,
+      "loss": 1.4651,
+      "step": 33470
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.129405576012625e-06,
+      "loss": 1.5919,
+      "step": 33480
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.127652112923024e-06,
+      "loss": 1.8121,
+      "step": 33490
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.125898649833421e-06,
+      "loss": 1.4637,
+      "step": 33500
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.124145186743819e-06,
+      "loss": 1.4777,
+      "step": 33510
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.122391723654218e-06,
+      "loss": 1.5751,
+      "step": 33520
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.1206382605646155e-06,
+      "loss": 1.7543,
+      "step": 33530
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.118884797475014e-06,
+      "loss": 1.468,
+      "step": 33540
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.117131334385411e-06,
+      "loss": 1.4845,
+      "step": 33550
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.11537787129581e-06,
+      "loss": 1.5638,
+      "step": 33560
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.113624408206208e-06,
+      "loss": 1.4248,
+      "step": 33570
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.1118709451166054e-06,
+      "loss": 1.7889,
+      "step": 33580
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.110117482027004e-06,
+      "loss": 1.4162,
+      "step": 33590
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.108364018937402e-06,
+      "loss": 1.466,
+      "step": 33600
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.1066105558477996e-06,
+      "loss": 1.4167,
+      "step": 33610
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.104857092758198e-06,
+      "loss": 1.6425,
+      "step": 33620
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.103103629668595e-06,
+      "loss": 1.5146,
+      "step": 33630
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.101350166578994e-06,
+      "loss": 1.5056,
+      "step": 33640
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.099596703489392e-06,
+      "loss": 1.4753,
+      "step": 33650
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.0978432403997895e-06,
+      "loss": 1.7,
+      "step": 33660
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.096089777310188e-06,
+      "loss": 1.4844,
+      "step": 33670
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.094336314220586e-06,
+      "loss": 1.6096,
+      "step": 33680
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.092582851130984e-06,
+      "loss": 1.5495,
+      "step": 33690
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.090829388041382e-06,
+      "loss": 1.5962,
+      "step": 33700
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.08907592495178e-06,
+      "loss": 1.4743,
+      "step": 33710
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.087322461862178e-06,
+      "loss": 1.5412,
+      "step": 33720
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.085568998772576e-06,
+      "loss": 1.4649,
+      "step": 33730
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.0838155356829736e-06,
+      "loss": 1.5162,
+      "step": 33740
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.082062072593373e-06,
+      "loss": 1.5025,
+      "step": 33750
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.08030860950377e-06,
+      "loss": 1.5374,
+      "step": 33760
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.0785551464141685e-06,
+      "loss": 1.524,
+      "step": 33770
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.076801683324567e-06,
+      "loss": 1.6199,
+      "step": 33780
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.075048220234964e-06,
+      "loss": 1.5223,
+      "step": 33790
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.073294757145363e-06,
+      "loss": 1.4109,
+      "step": 33800
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.07154129405576e-06,
+      "loss": 1.5732,
+      "step": 33810
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.0697878309661585e-06,
+      "loss": 1.4193,
+      "step": 33820
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.068034367876557e-06,
+      "loss": 1.6096,
+      "step": 33830
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.066280904786954e-06,
+      "loss": 1.5738,
+      "step": 33840
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.064527441697353e-06,
+      "loss": 1.5733,
+      "step": 33850
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.062773978607751e-06,
+      "loss": 1.6576,
+      "step": 33860
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.061020515518148e-06,
+      "loss": 1.478,
+      "step": 33870
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.059267052428547e-06,
+      "loss": 1.4066,
+      "step": 33880
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.057513589338945e-06,
+      "loss": 1.4289,
+      "step": 33890
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.0557601262493425e-06,
+      "loss": 1.4814,
+      "step": 33900
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.054006663159741e-06,
+      "loss": 1.5722,
+      "step": 33910
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.052253200070138e-06,
+      "loss": 1.5333,
+      "step": 33920
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.0504997369805375e-06,
+      "loss": 1.4644,
+      "step": 33930
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.048746273890935e-06,
+      "loss": 1.3323,
+      "step": 33940
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.0469928108013325e-06,
+      "loss": 1.5217,
+      "step": 33950
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.045239347711731e-06,
+      "loss": 1.5717,
+      "step": 33960
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.043485884622129e-06,
+      "loss": 1.4547,
+      "step": 33970
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.0417324215325275e-06,
+      "loss": 1.3256,
+      "step": 33980
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.039978958442925e-06,
+      "loss": 1.5223,
+      "step": 33990
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.038225495353323e-06,
+      "loss": 1.476,
+      "step": 34000
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.036472032263722e-06,
+      "loss": 1.6123,
+      "step": 34010
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.034718569174119e-06,
+      "loss": 1.6133,
+      "step": 34020
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.032965106084517e-06,
+      "loss": 1.5983,
+      "step": 34030
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.031211642994916e-06,
+      "loss": 1.5116,
+      "step": 34040
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.029458179905313e-06,
+      "loss": 1.5047,
+      "step": 34050
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.0277047168157115e-06,
+      "loss": 1.4784,
+      "step": 34060
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.025951253726109e-06,
+      "loss": 1.5104,
+      "step": 34070
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.024197790636507e-06,
+      "loss": 1.516,
+      "step": 34080
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.022444327546906e-06,
+      "loss": 1.4448,
+      "step": 34090
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.020690864457303e-06,
+      "loss": 1.4084,
+      "step": 34100
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.0189374013677015e-06,
+      "loss": 1.4359,
+      "step": 34110
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.0171839382781e-06,
+      "loss": 1.8319,
+      "step": 34120
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.015430475188497e-06,
+      "loss": 1.387,
+      "step": 34130
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.013677012098896e-06,
+      "loss": 1.7841,
+      "step": 34140
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.011923549009294e-06,
+      "loss": 1.5663,
+      "step": 34150
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.010170085919691e-06,
+      "loss": 1.5494,
+      "step": 34160
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.00841662283009e-06,
+      "loss": 1.4231,
+      "step": 34170
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.006663159740487e-06,
+      "loss": 1.6227,
+      "step": 34180
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.004909696650886e-06,
+      "loss": 1.4127,
+      "step": 34190
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.003156233561284e-06,
+      "loss": 1.5046,
+      "step": 34200
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.001402770471682e-06,
+      "loss": 1.5093,
+      "step": 34210
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.99964930738208e-06,
+      "loss": 1.6493,
+      "step": 34220
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.997895844292478e-06,
+      "loss": 1.5804,
+      "step": 34230
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.996142381202876e-06,
+      "loss": 1.6651,
+      "step": 34240
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.994388918113274e-06,
+      "loss": 1.3867,
+      "step": 34250
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.992635455023672e-06,
+      "loss": 1.4956,
+      "step": 34260
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.9908819919340704e-06,
+      "loss": 1.4776,
+      "step": 34270
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.989128528844468e-06,
+      "loss": 1.5462,
+      "step": 34280
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.987375065754866e-06,
+      "loss": 1.7588,
+      "step": 34290
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.9856216026652646e-06,
+      "loss": 1.3813,
+      "step": 34300
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.983868139575662e-06,
+      "loss": 1.524,
+      "step": 34310
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.98211467648606e-06,
+      "loss": 1.4441,
+      "step": 34320
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.980361213396458e-06,
+      "loss": 1.6159,
+      "step": 34330
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.978607750306856e-06,
+      "loss": 1.6227,
+      "step": 34340
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.9768542872172545e-06,
+      "loss": 1.5461,
+      "step": 34350
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.975100824127652e-06,
+      "loss": 1.3851,
+      "step": 34360
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.973347361038051e-06,
+      "loss": 1.8508,
+      "step": 34370
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.971593897948449e-06,
+      "loss": 1.6866,
+      "step": 34380
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.969840434858846e-06,
+      "loss": 1.509,
+      "step": 34390
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.9680869717692444e-06,
+      "loss": 1.6824,
+      "step": 34400
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.966333508679643e-06,
+      "loss": 1.613,
+      "step": 34410
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.964580045590041e-06,
+      "loss": 1.4687,
+      "step": 34420
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.9628265825004386e-06,
+      "loss": 1.5609,
+      "step": 34430
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.961073119410837e-06,
+      "loss": 1.5491,
+      "step": 34440
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.959319656321235e-06,
+      "loss": 1.471,
+      "step": 34450
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.957566193231633e-06,
+      "loss": 1.691,
+      "step": 34460
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.955812730142031e-06,
+      "loss": 1.462,
+      "step": 34470
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.9540592670524285e-06,
+      "loss": 1.5687,
+      "step": 34480
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.952305803962827e-06,
+      "loss": 1.4234,
+      "step": 34490
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.950552340873225e-06,
+      "loss": 1.4578,
+      "step": 34500
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.948798877783623e-06,
+      "loss": 1.5277,
+      "step": 34510
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.947045414694021e-06,
+      "loss": 1.5097,
+      "step": 34520
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.945291951604419e-06,
+      "loss": 1.5486,
+      "step": 34530
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.943538488514817e-06,
+      "loss": 1.6392,
+      "step": 34540
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.941785025425215e-06,
+      "loss": 1.3011,
+      "step": 34550
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.940031562335613e-06,
+      "loss": 1.4354,
+      "step": 34560
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.938278099246011e-06,
+      "loss": 1.4305,
+      "step": 34570
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.936524636156409e-06,
+      "loss": 1.6642,
+      "step": 34580
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.934771173066807e-06,
+      "loss": 1.4546,
+      "step": 34590
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.933017709977206e-06,
+      "loss": 1.5352,
+      "step": 34600
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.931264246887603e-06,
+      "loss": 1.3637,
+      "step": 34610
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.929510783798001e-06,
+      "loss": 1.6888,
+      "step": 34620
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9277573207084e-06,
+      "loss": 1.3767,
+      "step": 34630
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.9260038576187975e-06,
+      "loss": 1.5125,
+      "step": 34640
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.924250394529196e-06,
+      "loss": 1.4391,
+      "step": 34650
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.922496931439593e-06,
+      "loss": 1.429,
+      "step": 34660
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.920743468349992e-06,
+      "loss": 1.5107,
+      "step": 34670
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.91899000526039e-06,
+      "loss": 1.5224,
+      "step": 34680
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.917236542170787e-06,
+      "loss": 1.5834,
+      "step": 34690
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.915483079081186e-06,
+      "loss": 1.3609,
+      "step": 34700
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.913729615991584e-06,
+      "loss": 1.6552,
+      "step": 34710
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.9119761529019815e-06,
+      "loss": 1.4791,
+      "step": 34720
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.91022268981238e-06,
+      "loss": 1.4667,
+      "step": 34730
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.908469226722778e-06,
+      "loss": 1.4811,
+      "step": 34740
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.906715763633176e-06,
+      "loss": 1.703,
+      "step": 34750
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.904962300543574e-06,
+      "loss": 1.3726,
+      "step": 34760
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.9032088374539715e-06,
+      "loss": 1.5182,
+      "step": 34770
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.90145537436437e-06,
+      "loss": 1.4771,
+      "step": 34780
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.899701911274768e-06,
+      "loss": 1.5182,
+      "step": 34790
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.897948448185166e-06,
+      "loss": 1.41,
+      "step": 34800
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.896194985095564e-06,
+      "loss": 1.8385,
+      "step": 34810
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.894441522005962e-06,
+      "loss": 1.5675,
+      "step": 34820
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.89268805891636e-06,
+      "loss": 1.5444,
+      "step": 34830
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.890934595826758e-06,
+      "loss": 1.6387,
+      "step": 34840
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.889181132737156e-06,
+      "loss": 1.5665,
+      "step": 34850
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.887427669647555e-06,
+      "loss": 1.4969,
+      "step": 34860
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.885674206557952e-06,
+      "loss": 1.5468,
+      "step": 34870
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.8839207434683505e-06,
+      "loss": 1.442,
+      "step": 34880
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.882167280378749e-06,
+      "loss": 1.5309,
+      "step": 34890
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.880413817289146e-06,
+      "loss": 1.6298,
+      "step": 34900
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.878660354199545e-06,
+      "loss": 1.4494,
+      "step": 34910
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.876906891109942e-06,
+      "loss": 1.5528,
+      "step": 34920
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.8751534280203405e-06,
+      "loss": 1.422,
+      "step": 34930
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.873399964930739e-06,
+      "loss": 1.5412,
+      "step": 34940
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.871646501841136e-06,
+      "loss": 1.5131,
+      "step": 34950
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.869893038751535e-06,
+      "loss": 1.4256,
+      "step": 34960
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.868139575661933e-06,
+      "loss": 1.3569,
+      "step": 34970
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.86638611257233e-06,
+      "loss": 1.4036,
+      "step": 34980
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.864632649482729e-06,
+      "loss": 1.4723,
+      "step": 34990
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.862879186393127e-06,
+      "loss": 1.6137,
+      "step": 35000
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.8611257233035245e-06,
+      "loss": 1.4267,
+      "step": 35010
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.859372260213923e-06,
+      "loss": 1.522,
+      "step": 35020
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.85761879712432e-06,
+      "loss": 1.391,
+      "step": 35030
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.8558653340347195e-06,
+      "loss": 1.4972,
+      "step": 35040
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.854111870945117e-06,
+      "loss": 1.4072,
+      "step": 35050
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.8523584078555145e-06,
+      "loss": 1.5419,
+      "step": 35060
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.850604944765913e-06,
+      "loss": 1.5117,
+      "step": 35070
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.848851481676311e-06,
+      "loss": 1.6369,
+      "step": 35080
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.8470980185867094e-06,
+      "loss": 1.496,
+      "step": 35090
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.845344555497107e-06,
+      "loss": 1.5076,
+      "step": 35100
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.843591092407505e-06,
+      "loss": 1.3103,
+      "step": 35110
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.8418376293179036e-06,
+      "loss": 1.5471,
+      "step": 35120
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.840084166228301e-06,
+      "loss": 1.7179,
+      "step": 35130
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.838330703138699e-06,
+      "loss": 1.4502,
+      "step": 35140
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.836577240049098e-06,
+      "loss": 1.5826,
+      "step": 35150
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.834823776959495e-06,
+      "loss": 1.713,
+      "step": 35160
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.8330703138698935e-06,
+      "loss": 1.4587,
+      "step": 35170
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.831316850780291e-06,
+      "loss": 1.3871,
+      "step": 35180
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.829563387690689e-06,
+      "loss": 1.6378,
+      "step": 35190
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.827809924601088e-06,
+      "loss": 1.6913,
+      "step": 35200
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.826056461511485e-06,
+      "loss": 1.4757,
+      "step": 35210
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.8243029984218834e-06,
+      "loss": 1.2888,
+      "step": 35220
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.822549535332282e-06,
+      "loss": 1.5538,
+      "step": 35230
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.820796072242679e-06,
+      "loss": 1.5783,
+      "step": 35240
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.8190426091530776e-06,
+      "loss": 1.4879,
+      "step": 35250
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.817289146063476e-06,
+      "loss": 1.4289,
+      "step": 35260
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.815535682973873e-06,
+      "loss": 1.5952,
+      "step": 35270
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8137822198842717e-06,
+      "loss": 1.5264,
+      "step": 35280
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8120287567946696e-06,
+      "loss": 1.5861,
+      "step": 35290
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.810275293705068e-06,
+      "loss": 1.426,
+      "step": 35300
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.808521830615466e-06,
+      "loss": 1.5274,
+      "step": 35310
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.8067683675258637e-06,
+      "loss": 1.4909,
+      "step": 35320
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.805014904436262e-06,
+      "loss": 1.4709,
+      "step": 35330
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.80326144134666e-06,
+      "loss": 1.271,
+      "step": 35340
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.801507978257058e-06,
+      "loss": 1.5627,
+      "step": 35350
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.7997545151674558e-06,
+      "loss": 1.5557,
+      "step": 35360
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.7980010520778545e-06,
+      "loss": 1.4324,
+      "step": 35370
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.796247588988252e-06,
+      "loss": 1.4076,
+      "step": 35380
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.79449412589865e-06,
+      "loss": 1.4605,
+      "step": 35390
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7927406628090478e-06,
+      "loss": 1.6351,
+      "step": 35400
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7909871997194465e-06,
+      "loss": 1.4104,
+      "step": 35410
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7892337366298444e-06,
+      "loss": 1.4721,
+      "step": 35420
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7874802735402423e-06,
+      "loss": 1.5094,
+      "step": 35430
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.7857268104506402e-06,
+      "loss": 1.4887,
+      "step": 35440
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7839733473610386e-06,
+      "loss": 1.3069,
+      "step": 35450
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7822198842714365e-06,
+      "loss": 1.8818,
+      "step": 35460
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7804664211818344e-06,
+      "loss": 1.4266,
+      "step": 35470
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7787129580922327e-06,
+      "loss": 1.6151,
+      "step": 35480
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7769594950026306e-06,
+      "loss": 1.4079,
+      "step": 35490
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.7752060319130285e-06,
+      "loss": 1.4396,
+      "step": 35500
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7734525688234264e-06,
+      "loss": 1.3987,
+      "step": 35510
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7716991057338247e-06,
+      "loss": 1.4063,
+      "step": 35520
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7699456426442226e-06,
+      "loss": 1.541,
+      "step": 35530
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7681921795546205e-06,
+      "loss": 1.5528,
+      "step": 35540
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.7664387164650184e-06,
+      "loss": 1.6018,
+      "step": 35550
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7646852533754168e-06,
+      "loss": 1.653,
+      "step": 35560
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7629317902858147e-06,
+      "loss": 1.4553,
+      "step": 35570
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7611783271962126e-06,
+      "loss": 1.6641,
+      "step": 35580
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7594248641066113e-06,
+      "loss": 1.4299,
+      "step": 35590
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.757671401017009e-06,
+      "loss": 1.5707,
+      "step": 35600
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.7559179379274067e-06,
+      "loss": 1.3546,
+      "step": 35610
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7541644748378046e-06,
+      "loss": 1.4871,
+      "step": 35620
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7524110117482034e-06,
+      "loss": 1.4506,
+      "step": 35630
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7506575486586013e-06,
+      "loss": 1.5056,
+      "step": 35640
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.748904085568999e-06,
+      "loss": 1.4845,
+      "step": 35650
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.747150622479397e-06,
+      "loss": 1.3793,
+      "step": 35660
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.7453971593897954e-06,
+      "loss": 1.428,
+      "step": 35670
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7436436963001933e-06,
+      "loss": 1.4345,
+      "step": 35680
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.741890233210591e-06,
+      "loss": 1.3721,
+      "step": 35690
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7401367701209895e-06,
+      "loss": 1.3575,
+      "step": 35700
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7383833070313874e-06,
+      "loss": 1.535,
+      "step": 35710
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.7366298439417853e-06,
+      "loss": 1.3675,
+      "step": 35720
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7348763808521832e-06,
+      "loss": 1.3818,
+      "step": 35730
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7331229177625815e-06,
+      "loss": 1.4634,
+      "step": 35740
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7313694546729794e-06,
+      "loss": 1.5494,
+      "step": 35750
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7296159915833773e-06,
+      "loss": 1.4919,
+      "step": 35760
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7278625284937753e-06,
+      "loss": 1.447,
+      "step": 35770
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.7261090654041736e-06,
+      "loss": 1.4401,
+      "step": 35780
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7243556023145715e-06,
+      "loss": 1.5834,
+      "step": 35790
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7226021392249694e-06,
+      "loss": 1.4844,
+      "step": 35800
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.720848676135368e-06,
+      "loss": 1.6635,
+      "step": 35810
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7190952130457656e-06,
+      "loss": 1.6222,
+      "step": 35820
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7173417499561635e-06,
+      "loss": 1.4222,
+      "step": 35830
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.7155882868665614e-06,
+      "loss": 1.3217,
+      "step": 35840
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.71383482377696e-06,
+      "loss": 1.5133,
+      "step": 35850
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.712081360687358e-06,
+      "loss": 1.6268,
+      "step": 35860
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.710327897597756e-06,
+      "loss": 1.492,
+      "step": 35870
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.708574434508154e-06,
+      "loss": 1.6736,
+      "step": 35880
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.706820971418552e-06,
+      "loss": 1.5139,
+      "step": 35890
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.70506750832895e-06,
+      "loss": 1.5829,
+      "step": 35900
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.703314045239348e-06,
+      "loss": 1.3072,
+      "step": 35910
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.701560582149746e-06,
+      "loss": 1.4426,
+      "step": 35920
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.6998071190601442e-06,
+      "loss": 1.4599,
+      "step": 35930
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.698053655970542e-06,
+      "loss": 1.3497,
+      "step": 35940
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.69630019288094e-06,
+      "loss": 1.4712,
+      "step": 35950
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6945467297913384e-06,
+      "loss": 1.5206,
+      "step": 35960
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6927932667017363e-06,
+      "loss": 1.5259,
+      "step": 35970
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.691039803612134e-06,
+      "loss": 1.6085,
+      "step": 35980
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.689286340522532e-06,
+      "loss": 1.4902,
+      "step": 35990
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6875328774329304e-06,
+      "loss": 1.5598,
+      "step": 36000
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.6857794143433283e-06,
+      "loss": 1.3504,
+      "step": 36010
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.684025951253726e-06,
+      "loss": 1.4908,
+      "step": 36020
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.682272488164124e-06,
+      "loss": 1.5316,
+      "step": 36030
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6805190250745224e-06,
+      "loss": 1.5984,
+      "step": 36040
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6787655619849203e-06,
+      "loss": 1.4304,
+      "step": 36050
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.6770120988953182e-06,
+      "loss": 1.3998,
+      "step": 36060
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.675258635805717e-06,
+      "loss": 1.6545,
+      "step": 36070
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.673505172716115e-06,
+      "loss": 1.5874,
+      "step": 36080
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6717517096265128e-06,
+      "loss": 1.5339,
+      "step": 36090
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.6699982465369107e-06,
+      "loss": 1.6973,
+      "step": 36100
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.668244783447309e-06,
+      "loss": 1.5518,
+      "step": 36110
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.666491320357707e-06,
+      "loss": 1.5964,
+      "step": 36120
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.664737857268105e-06,
+      "loss": 1.5569,
+      "step": 36130
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.6629843941785027e-06,
+      "loss": 1.5435,
+      "step": 36140
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.661230931088901e-06,
+      "loss": 1.5959,
+      "step": 36150
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.659477467999299e-06,
+      "loss": 1.5033,
+      "step": 36160
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.657724004909697e-06,
+      "loss": 1.6386,
+      "step": 36170
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.655970541820095e-06,
+      "loss": 1.5928,
+      "step": 36180
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.654217078730493e-06,
+      "loss": 1.7442,
+      "step": 36190
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.652463615640891e-06,
+      "loss": 1.4585,
+      "step": 36200
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.650710152551289e-06,
+      "loss": 1.5828,
+      "step": 36210
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.648956689461687e-06,
+      "loss": 1.5576,
+      "step": 36220
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.647203226372085e-06,
+      "loss": 1.4134,
+      "step": 36230
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.645449763282483e-06,
+      "loss": 1.4581,
+      "step": 36240
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.643696300192881e-06,
+      "loss": 1.6289,
+      "step": 36250
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.6419428371032797e-06,
+      "loss": 1.4018,
+      "step": 36260
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.640189374013677e-06,
+      "loss": 1.5227,
+      "step": 36270
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.638435910924075e-06,
+      "loss": 1.5575,
+      "step": 36280
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.636682447834473e-06,
+      "loss": 1.6757,
+      "step": 36290
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6349289847448717e-06,
+      "loss": 1.6489,
+      "step": 36300
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6331755216552696e-06,
+      "loss": 1.5475,
+      "step": 36310
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6314220585656675e-06,
+      "loss": 1.519,
+      "step": 36320
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.629668595476066e-06,
+      "loss": 1.5431,
+      "step": 36330
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6279151323864637e-06,
+      "loss": 1.5751,
+      "step": 36340
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.6261616692968616e-06,
+      "loss": 1.5417,
+      "step": 36350
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6244082062072595e-06,
+      "loss": 1.4171,
+      "step": 36360
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.622654743117658e-06,
+      "loss": 1.3865,
+      "step": 36370
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6209012800280558e-06,
+      "loss": 1.5854,
+      "step": 36380
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6191478169384537e-06,
+      "loss": 1.4306,
+      "step": 36390
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.6173943538488516e-06,
+      "loss": 1.405,
+      "step": 36400
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.61564089075925e-06,
+      "loss": 1.4923,
+      "step": 36410
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.613887427669648e-06,
+      "loss": 1.4899,
+      "step": 36420
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6121339645800457e-06,
+      "loss": 1.3413,
+      "step": 36430
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.610380501490444e-06,
+      "loss": 1.5443,
+      "step": 36440
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.608627038400842e-06,
+      "loss": 1.4854,
+      "step": 36450
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.60687357531124e-06,
+      "loss": 1.4813,
+      "step": 36460
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.6051201122216377e-06,
+      "loss": 1.4611,
+      "step": 36470
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.6033666491320365e-06,
+      "loss": 1.4884,
+      "step": 36480
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.601613186042434e-06,
+      "loss": 1.6745,
+      "step": 36490
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.599859722952832e-06,
+      "loss": 1.4697,
+      "step": 36500
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.5981062598632298e-06,
+      "loss": 1.5529,
+      "step": 36510
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.5963527967736285e-06,
+      "loss": 1.442,
+      "step": 36520
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5945993336840264e-06,
+      "loss": 1.4533,
+      "step": 36530
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5928458705944243e-06,
+      "loss": 1.5178,
+      "step": 36540
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5910924075048226e-06,
+      "loss": 1.4854,
+      "step": 36550
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5893389444152205e-06,
+      "loss": 1.5862,
+      "step": 36560
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5875854813256184e-06,
+      "loss": 1.2678,
+      "step": 36570
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.5858320182360163e-06,
+      "loss": 1.479,
+      "step": 36580
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5840785551464147e-06,
+      "loss": 1.5894,
+      "step": 36590
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5823250920568126e-06,
+      "loss": 1.5023,
+      "step": 36600
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5805716289672105e-06,
+      "loss": 1.5145,
+      "step": 36610
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5788181658776084e-06,
+      "loss": 1.4649,
+      "step": 36620
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5770647027880067e-06,
+      "loss": 1.4134,
+      "step": 36630
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.5753112396984046e-06,
+      "loss": 1.4785,
+      "step": 36640
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5735577766088025e-06,
+      "loss": 1.6213,
+      "step": 36650
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.571804313519201e-06,
+      "loss": 1.4342,
+      "step": 36660
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5700508504295987e-06,
+      "loss": 1.4686,
+      "step": 36670
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5682973873399966e-06,
+      "loss": 1.4649,
+      "step": 36680
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.5665439242503945e-06,
+      "loss": 1.4253,
+      "step": 36690
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5647904611607933e-06,
+      "loss": 1.4176,
+      "step": 36700
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5630369980711908e-06,
+      "loss": 1.686,
+      "step": 36710
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5612835349815887e-06,
+      "loss": 1.4428,
+      "step": 36720
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5595300718919866e-06,
+      "loss": 1.6201,
+      "step": 36730
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5577766088023853e-06,
+      "loss": 1.4444,
+      "step": 36740
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.5560231457127832e-06,
+      "loss": 1.5355,
+      "step": 36750
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.554269682623181e-06,
+      "loss": 1.5779,
+      "step": 36760
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.552516219533579e-06,
+      "loss": 1.5416,
+      "step": 36770
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5507627564439774e-06,
+      "loss": 1.5113,
+      "step": 36780
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5490092933543753e-06,
+      "loss": 1.6346,
+      "step": 36790
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.547255830264773e-06,
+      "loss": 1.5579,
+      "step": 36800
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.5455023671751715e-06,
+      "loss": 1.4833,
+      "step": 36810
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5437489040855694e-06,
+      "loss": 1.3973,
+      "step": 36820
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5419954409959673e-06,
+      "loss": 1.33,
+      "step": 36830
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.540241977906365e-06,
+      "loss": 1.6048,
+      "step": 36840
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5384885148167635e-06,
+      "loss": 1.4365,
+      "step": 36850
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.5367350517271614e-06,
+      "loss": 1.4387,
+      "step": 36860
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5349815886375593e-06,
+      "loss": 1.3038,
+      "step": 36870
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5332281255479572e-06,
+      "loss": 1.4741,
+      "step": 36880
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5314746624583555e-06,
+      "loss": 1.4916,
+      "step": 36890
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5297211993687535e-06,
+      "loss": 1.7711,
+      "step": 36900
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.5279677362791514e-06,
+      "loss": 1.4034,
+      "step": 36910
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.52621427318955e-06,
+      "loss": 1.4712,
+      "step": 36920
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5244608100999476e-06,
+      "loss": 1.5342,
+      "step": 36930
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5227073470103455e-06,
+      "loss": 1.4595,
+      "step": 36940
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.5209538839207434e-06,
+      "loss": 1.4511,
+      "step": 36950
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.519200420831142e-06,
+      "loss": 1.6928,
+      "step": 36960
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.51744695774154e-06,
+      "loss": 1.5328,
+      "step": 36970
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.515693494651938e-06,
+      "loss": 1.3505,
+      "step": 36980
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.513940031562336e-06,
+      "loss": 1.6172,
+      "step": 36990
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.512186568472734e-06,
+      "loss": 1.6756,
+      "step": 37000
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.510433105383132e-06,
+      "loss": 1.6499,
+      "step": 37010
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.50867964229353e-06,
+      "loss": 1.5496,
+      "step": 37020
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.5069261792039283e-06,
+      "loss": 1.494,
+      "step": 37030
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.505172716114326e-06,
+      "loss": 1.365,
+      "step": 37040
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.503419253024724e-06,
+      "loss": 1.5391,
+      "step": 37050
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.501665789935122e-06,
+      "loss": 1.4212,
+      "step": 37060
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.4999123268455203e-06,
+      "loss": 1.7056,
+      "step": 37070
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.4981588637559182e-06,
+      "loss": 1.6208,
+      "step": 37080
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.496405400666316e-06,
+      "loss": 1.6167,
+      "step": 37090
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.494651937576714e-06,
+      "loss": 1.4487,
+      "step": 37100
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4928984744871124e-06,
+      "loss": 1.4548,
+      "step": 37110
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4911450113975103e-06,
+      "loss": 1.512,
+      "step": 37120
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.489391548307908e-06,
+      "loss": 1.3902,
+      "step": 37130
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.487638085218307e-06,
+      "loss": 1.421,
+      "step": 37140
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.4858846221287044e-06,
+      "loss": 1.4026,
+      "step": 37150
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4841311590391023e-06,
+      "loss": 1.608,
+      "step": 37160
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4823776959495e-06,
+      "loss": 1.4521,
+      "step": 37170
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.480624232859899e-06,
+      "loss": 1.5294,
+      "step": 37180
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.478870769770297e-06,
+      "loss": 1.3866,
+      "step": 37190
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4771173066806948e-06,
+      "loss": 1.4567,
+      "step": 37200
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.4753638435910927e-06,
+      "loss": 1.5352,
+      "step": 37210
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.473610380501491e-06,
+      "loss": 1.5212,
+      "step": 37220
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.471856917411889e-06,
+      "loss": 1.5258,
+      "step": 37230
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4701034543222868e-06,
+      "loss": 1.4114,
+      "step": 37240
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.4683499912326847e-06,
+      "loss": 1.6007,
+      "step": 37250
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.466596528143083e-06,
+      "loss": 1.476,
+      "step": 37260
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.464843065053481e-06,
+      "loss": 1.5162,
+      "step": 37270
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.463089601963879e-06,
+      "loss": 1.5238,
+      "step": 37280
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.461336138874277e-06,
+      "loss": 1.461,
+      "step": 37290
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.459582675784675e-06,
+      "loss": 1.4878,
+      "step": 37300
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.457829212695073e-06,
+      "loss": 1.4335,
+      "step": 37310
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.456075749605471e-06,
+      "loss": 1.4004,
+      "step": 37320
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.454322286515869e-06,
+      "loss": 1.5214,
+      "step": 37330
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.452568823426267e-06,
+      "loss": 1.3508,
+      "step": 37340
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.450815360336665e-06,
+      "loss": 1.6633,
+      "step": 37350
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.449061897247063e-06,
+      "loss": 1.648,
+      "step": 37360
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.447308434157461e-06,
+      "loss": 1.4541,
+      "step": 37370
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.445554971067859e-06,
+      "loss": 1.6134,
+      "step": 37380
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.443801507978257e-06,
+      "loss": 1.4116,
+      "step": 37390
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4420480448886558e-06,
+      "loss": 1.4349,
+      "step": 37400
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4402945817990537e-06,
+      "loss": 1.4797,
+      "step": 37410
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4385411187094516e-06,
+      "loss": 1.4384,
+      "step": 37420
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.4367876556198495e-06,
+      "loss": 1.5708,
+      "step": 37430
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.435034192530248e-06,
+      "loss": 1.6503,
+      "step": 37440
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4332807294406457e-06,
+      "loss": 1.5323,
+      "step": 37450
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4315272663510436e-06,
+      "loss": 1.4339,
+      "step": 37460
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4297738032614415e-06,
+      "loss": 1.3893,
+      "step": 37470
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.42802034017184e-06,
+      "loss": 1.4944,
+      "step": 37480
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.4262668770822377e-06,
+      "loss": 1.5298,
+      "step": 37490
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4245134139926356e-06,
+      "loss": 1.508,
+      "step": 37500
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.422759950903034e-06,
+      "loss": 1.5787,
+      "step": 37510
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.421006487813432e-06,
+      "loss": 1.5528,
+      "step": 37520
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4192530247238298e-06,
+      "loss": 1.5647,
+      "step": 37530
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.4174995616342277e-06,
+      "loss": 1.5387,
+      "step": 37540
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.415746098544626e-06,
+      "loss": 1.5838,
+      "step": 37550
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.413992635455024e-06,
+      "loss": 1.4074,
+      "step": 37560
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.412239172365422e-06,
+      "loss": 1.7443,
+      "step": 37570
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.4104857092758197e-06,
+      "loss": 1.4572,
+      "step": 37580
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.408732246186218e-06,
+      "loss": 1.5128,
+      "step": 37590
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.406978783096616e-06,
+      "loss": 1.6752,
+      "step": 37600
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.405225320007014e-06,
+      "loss": 1.5208,
+      "step": 37610
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4034718569174126e-06,
+      "loss": 1.7878,
+      "step": 37620
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.4017183938278105e-06,
+      "loss": 1.4601,
+      "step": 37630
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3999649307382084e-06,
+      "loss": 1.6556,
+      "step": 37640
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3982114676486063e-06,
+      "loss": 1.5932,
+      "step": 37650
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3964580045590046e-06,
+      "loss": 1.5095,
+      "step": 37660
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3947045414694025e-06,
+      "loss": 1.4549,
+      "step": 37670
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3929510783798004e-06,
+      "loss": 1.624,
+      "step": 37680
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3911976152901983e-06,
+      "loss": 1.3585,
+      "step": 37690
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3894441522005966e-06,
+      "loss": 1.394,
+      "step": 37700
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3876906891109945e-06,
+      "loss": 1.5496,
+      "step": 37710
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3859372260213924e-06,
+      "loss": 1.2775,
+      "step": 37720
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3841837629317903e-06,
+      "loss": 1.5341,
+      "step": 37730
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3824302998421887e-06,
+      "loss": 1.4751,
+      "step": 37740
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3806768367525866e-06,
+      "loss": 1.5914,
+      "step": 37750
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3789233736629845e-06,
+      "loss": 1.3926,
+      "step": 37760
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.377169910573383e-06,
+      "loss": 1.4998,
+      "step": 37770
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3754164474837807e-06,
+      "loss": 1.6062,
+      "step": 37780
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3736629843941786e-06,
+      "loss": 1.4883,
+      "step": 37790
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3719095213045765e-06,
+      "loss": 1.457,
+      "step": 37800
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.370156058214975e-06,
+      "loss": 1.3447,
+      "step": 37810
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3684025951253727e-06,
+      "loss": 1.4682,
+      "step": 37820
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3666491320357706e-06,
+      "loss": 1.42,
+      "step": 37830
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3648956689461685e-06,
+      "loss": 1.4551,
+      "step": 37840
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3631422058565673e-06,
+      "loss": 1.3806,
+      "step": 37850
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.361388742766965e-06,
+      "loss": 1.6194,
+      "step": 37860
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.359635279677363e-06,
+      "loss": 1.3968,
+      "step": 37870
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3578818165877614e-06,
+      "loss": 1.3429,
+      "step": 37880
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3561283534981593e-06,
+      "loss": 1.5634,
+      "step": 37890
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3543748904085572e-06,
+      "loss": 1.4824,
+      "step": 37900
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.352621427318955e-06,
+      "loss": 1.5132,
+      "step": 37910
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3508679642293535e-06,
+      "loss": 1.441,
+      "step": 37920
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3491145011397514e-06,
+      "loss": 1.6077,
+      "step": 37930
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3473610380501493e-06,
+      "loss": 1.2739,
+      "step": 37940
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.345607574960547e-06,
+      "loss": 1.6236,
+      "step": 37950
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3438541118709455e-06,
+      "loss": 1.5239,
+      "step": 37960
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3421006487813434e-06,
+      "loss": 1.4003,
+      "step": 37970
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3403471856917413e-06,
+      "loss": 1.7158,
+      "step": 37980
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3385937226021396e-06,
+      "loss": 1.5127,
+      "step": 37990
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3368402595125375e-06,
+      "loss": 1.4275,
+      "step": 38000
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3350867964229354e-06,
+      "loss": 1.5406,
+      "step": 38010
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.5363,
+      "step": 38020
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.331579870243732e-06,
+      "loss": 1.4623,
+      "step": 38030
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3298264071541296e-06,
+      "loss": 1.3038,
+      "step": 38040
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3280729440645275e-06,
+      "loss": 1.4088,
+      "step": 38050
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3263194809749254e-06,
+      "loss": 1.5131,
+      "step": 38060
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.324566017885324e-06,
+      "loss": 1.6532,
+      "step": 38070
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.322812554795722e-06,
+      "loss": 1.4521,
+      "step": 38080
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.32105909170612e-06,
+      "loss": 1.5733,
+      "step": 38090
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.319305628616518e-06,
+      "loss": 1.6671,
+      "step": 38100
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.317552165526916e-06,
+      "loss": 1.4466,
+      "step": 38110
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.315798702437314e-06,
+      "loss": 1.5922,
+      "step": 38120
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.314045239347712e-06,
+      "loss": 1.4795,
+      "step": 38130
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3122917762581103e-06,
+      "loss": 1.6505,
+      "step": 38140
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.310538313168508e-06,
+      "loss": 1.5706,
+      "step": 38150
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.308784850078906e-06,
+      "loss": 1.49,
+      "step": 38160
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.307031386989304e-06,
+      "loss": 1.4692,
+      "step": 38170
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3052779238997023e-06,
+      "loss": 1.4795,
+      "step": 38180
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3035244608101e-06,
+      "loss": 1.3399,
+      "step": 38190
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.301770997720498e-06,
+      "loss": 1.4532,
+      "step": 38200
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.300017534630896e-06,
+      "loss": 1.6146,
+      "step": 38210
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.2982640715412943e-06,
+      "loss": 1.6158,
+      "step": 38220
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.2965106084516922e-06,
+      "loss": 1.5121,
+      "step": 38230
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.29475714536209e-06,
+      "loss": 1.6141,
+      "step": 38240
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.293003682272489e-06,
+      "loss": 1.777,
+      "step": 38250
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2912502191828864e-06,
+      "loss": 1.6326,
+      "step": 38260
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.2894967560932843e-06,
+      "loss": 1.547,
+      "step": 38270
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.287743293003682e-06,
+      "loss": 1.6033,
+      "step": 38280
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.285989829914081e-06,
+      "loss": 1.5203,
+      "step": 38290
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.284236366824479e-06,
+      "loss": 1.5949,
+      "step": 38300
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2824829037348767e-06,
+      "loss": 1.5617,
+      "step": 38310
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2807294406452746e-06,
+      "loss": 1.3488,
+      "step": 38320
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.278975977555673e-06,
+      "loss": 1.4581,
+      "step": 38330
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.277222514466071e-06,
+      "loss": 1.7499,
+      "step": 38340
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.2754690513764688e-06,
+      "loss": 1.4506,
+      "step": 38350
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.273715588286867e-06,
+      "loss": 1.5076,
+      "step": 38360
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.271962125197265e-06,
+      "loss": 1.4933,
+      "step": 38370
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.270208662107663e-06,
+      "loss": 1.5224,
+      "step": 38380
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.268455199018061e-06,
+      "loss": 1.5332,
+      "step": 38390
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.266701735928459e-06,
+      "loss": 1.6548,
+      "step": 38400
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.264948272838857e-06,
+      "loss": 1.6109,
+      "step": 38410
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.263194809749255e-06,
+      "loss": 1.5579,
+      "step": 38420
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.261441346659653e-06,
+      "loss": 1.5015,
+      "step": 38430
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.259687883570051e-06,
+      "loss": 1.5074,
+      "step": 38440
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.257934420480449e-06,
+      "loss": 1.6704,
+      "step": 38450
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.256180957390847e-06,
+      "loss": 1.5974,
+      "step": 38460
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2544274943012457e-06,
+      "loss": 1.6506,
+      "step": 38470
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.252674031211643e-06,
+      "loss": 1.5325,
+      "step": 38480
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.250920568122041e-06,
+      "loss": 1.48,
+      "step": 38490
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.249167105032439e-06,
+      "loss": 1.463,
+      "step": 38500
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2474136419428377e-06,
+      "loss": 1.4884,
+      "step": 38510
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.2456601788532356e-06,
+      "loss": 1.5113,
+      "step": 38520
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2439067157636335e-06,
+      "loss": 1.6203,
+      "step": 38530
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2421532526740314e-06,
+      "loss": 1.5631,
+      "step": 38540
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2403997895844298e-06,
+      "loss": 1.4348,
+      "step": 38550
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2386463264948277e-06,
+      "loss": 1.4441,
+      "step": 38560
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2368928634052256e-06,
+      "loss": 1.431,
+      "step": 38570
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.2351394003156235e-06,
+      "loss": 1.5514,
+      "step": 38580
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.233385937226022e-06,
+      "loss": 1.4277,
+      "step": 38590
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2316324741364197e-06,
+      "loss": 1.3899,
+      "step": 38600
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.2298790110468176e-06,
+      "loss": 1.3098,
+      "step": 38610
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.228125547957216e-06,
+      "loss": 1.4424,
+      "step": 38620
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.226372084867614e-06,
+      "loss": 1.5533,
+      "step": 38630
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2246186217780117e-06,
+      "loss": 1.5817,
+      "step": 38640
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2228651586884096e-06,
+      "loss": 1.5636,
+      "step": 38650
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.221111695598808e-06,
+      "loss": 1.3485,
+      "step": 38660
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.219358232509206e-06,
+      "loss": 1.534,
+      "step": 38670
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2176047694196038e-06,
+      "loss": 1.46,
+      "step": 38680
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.2158513063300017e-06,
+      "loss": 1.5056,
+      "step": 38690
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2140978432404e-06,
+      "loss": 1.5903,
+      "step": 38700
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.212344380150798e-06,
+      "loss": 1.4955,
+      "step": 38710
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.210590917061196e-06,
+      "loss": 1.5397,
+      "step": 38720
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2088374539715945e-06,
+      "loss": 1.3607,
+      "step": 38730
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2070839908819924e-06,
+      "loss": 1.3974,
+      "step": 38740
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.2053305277923904e-06,
+      "loss": 1.4218,
+      "step": 38750
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2035770647027883e-06,
+      "loss": 1.6828,
+      "step": 38760
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2018236016131866e-06,
+      "loss": 1.5712,
+      "step": 38770
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.2000701385235845e-06,
+      "loss": 1.6535,
+      "step": 38780
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.1983166754339824e-06,
+      "loss": 1.4976,
+      "step": 38790
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.1965632123443803e-06,
+      "loss": 1.4608,
+      "step": 38800
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1948097492547786e-06,
+      "loss": 1.4027,
+      "step": 38810
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1930562861651765e-06,
+      "loss": 1.4807,
+      "step": 38820
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1913028230755744e-06,
+      "loss": 1.5332,
+      "step": 38830
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1895493599859727e-06,
+      "loss": 1.5988,
+      "step": 38840
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1877958968963706e-06,
+      "loss": 1.5292,
+      "step": 38850
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.1860424338067685e-06,
+      "loss": 1.3481,
+      "step": 38860
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1842889707171664e-06,
+      "loss": 1.3678,
+      "step": 38870
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1825355076275648e-06,
+      "loss": 1.5308,
+      "step": 38880
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1807820445379627e-06,
+      "loss": 1.2686,
+      "step": 38890
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1790285814483606e-06,
+      "loss": 1.5808,
+      "step": 38900
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.1772751183587585e-06,
+      "loss": 1.504,
+      "step": 38910
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.175521655269157e-06,
+      "loss": 1.3614,
+      "step": 38920
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1737681921795547e-06,
+      "loss": 1.4339,
+      "step": 38930
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1720147290899526e-06,
+      "loss": 1.4713,
+      "step": 38940
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1702612660003514e-06,
+      "loss": 1.4994,
+      "step": 38950
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.1685078029107493e-06,
+      "loss": 1.5965,
+      "step": 38960
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.166754339821147e-06,
+      "loss": 1.5673,
+      "step": 38970
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.165000876731545e-06,
+      "loss": 1.6054,
+      "step": 38980
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1632474136419434e-06,
+      "loss": 1.5595,
+      "step": 38990
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1614939505523413e-06,
+      "loss": 1.4622,
+      "step": 39000
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.159740487462739e-06,
+      "loss": 1.5399,
+      "step": 39010
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.157987024373137e-06,
+      "loss": 1.5634,
+      "step": 39020
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.1562335612835354e-06,
+      "loss": 1.707,
+      "step": 39030
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1544800981939333e-06,
+      "loss": 1.3804,
+      "step": 39040
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1527266351043312e-06,
+      "loss": 1.4884,
+      "step": 39050
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.150973172014729e-06,
+      "loss": 1.4576,
+      "step": 39060
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1492197089251275e-06,
+      "loss": 1.6101,
+      "step": 39070
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1474662458355254e-06,
+      "loss": 1.3954,
+      "step": 39080
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.1457127827459233e-06,
+      "loss": 1.4509,
+      "step": 39090
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1439593196563216e-06,
+      "loss": 1.3263,
+      "step": 39100
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1422058565667195e-06,
+      "loss": 1.4335,
+      "step": 39110
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1404523934771174e-06,
+      "loss": 1.4572,
+      "step": 39120
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1386989303875153e-06,
+      "loss": 1.4705,
+      "step": 39130
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1369454672979136e-06,
+      "loss": 1.4517,
+      "step": 39140
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.1351920042083115e-06,
+      "loss": 1.5545,
+      "step": 39150
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1334385411187094e-06,
+      "loss": 1.4004,
+      "step": 39160
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.1316850780291073e-06,
+      "loss": 1.4442,
+      "step": 39170
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.129931614939506e-06,
+      "loss": 1.3894,
+      "step": 39180
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.128178151849904e-06,
+      "loss": 1.5847,
+      "step": 39190
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.126424688760302e-06,
+      "loss": 1.5493,
+      "step": 39200
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1246712256707e-06,
+      "loss": 1.4428,
+      "step": 39210
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.122917762581098e-06,
+      "loss": 1.5474,
+      "step": 39220
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.121164299491496e-06,
+      "loss": 1.5023,
+      "step": 39230
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.119410836401894e-06,
+      "loss": 1.3816,
+      "step": 39240
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.1176573733122922e-06,
+      "loss": 1.6211,
+      "step": 39250
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.11590391022269e-06,
+      "loss": 1.5713,
+      "step": 39260
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.114150447133088e-06,
+      "loss": 1.4256,
+      "step": 39270
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.112396984043486e-06,
+      "loss": 1.4093,
+      "step": 39280
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1106435209538843e-06,
+      "loss": 1.4607,
+      "step": 39290
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.108890057864282e-06,
+      "loss": 1.5585,
+      "step": 39300
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.10713659477468e-06,
+      "loss": 1.5374,
+      "step": 39310
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.1053831316850784e-06,
+      "loss": 1.4703,
+      "step": 39320
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.1036296685954763e-06,
+      "loss": 1.513,
+      "step": 39330
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.101876205505874e-06,
+      "loss": 1.4679,
+      "step": 39340
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.100122742416272e-06,
+      "loss": 1.4497,
+      "step": 39350
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.0983692793266704e-06,
+      "loss": 1.3316,
+      "step": 39360
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.0966158162370683e-06,
+      "loss": 1.5265,
+      "step": 39370
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0948623531474662e-06,
+      "loss": 1.4337,
+      "step": 39380
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.093108890057864e-06,
+      "loss": 1.5885,
+      "step": 39390
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.091355426968263e-06,
+      "loss": 1.4214,
+      "step": 39400
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.089601963878661e-06,
+      "loss": 1.4369,
+      "step": 39410
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.0878485007890587e-06,
+      "loss": 1.3248,
+      "step": 39420
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.086095037699457e-06,
+      "loss": 1.6034,
+      "step": 39430
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.084341574609855e-06,
+      "loss": 1.5617,
+      "step": 39440
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.082588111520253e-06,
+      "loss": 1.4003,
+      "step": 39450
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.0808346484306507e-06,
+      "loss": 1.433,
+      "step": 39460
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.079081185341049e-06,
+      "loss": 1.3881,
+      "step": 39470
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.077327722251447e-06,
+      "loss": 1.3822,
+      "step": 39480
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.075574259161845e-06,
+      "loss": 1.6155,
+      "step": 39490
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.0738207960722428e-06,
+      "loss": 1.45,
+      "step": 39500
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.072067332982641e-06,
+      "loss": 1.5101,
+      "step": 39510
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.070313869893039e-06,
+      "loss": 1.488,
+      "step": 39520
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.068560406803437e-06,
+      "loss": 1.4539,
+      "step": 39530
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.066806943713835e-06,
+      "loss": 1.5915,
+      "step": 39540
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.065053480624233e-06,
+      "loss": 1.4716,
+      "step": 39550
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.063300017534631e-06,
+      "loss": 1.373,
+      "step": 39560
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.061546554445029e-06,
+      "loss": 1.4952,
+      "step": 39570
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.0597930913554272e-06,
+      "loss": 1.5219,
+      "step": 39580
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.058039628265825e-06,
+      "loss": 1.6459,
+      "step": 39590
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.056286165176223e-06,
+      "loss": 1.5778,
+      "step": 39600
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.054532702086621e-06,
+      "loss": 1.48,
+      "step": 39610
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0527792389970197e-06,
+      "loss": 1.6759,
+      "step": 39620
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0510257759074176e-06,
+      "loss": 1.3096,
+      "step": 39630
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0492723128178155e-06,
+      "loss": 1.4633,
+      "step": 39640
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0475188497282134e-06,
+      "loss": 1.429,
+      "step": 39650
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.0457653866386117e-06,
+      "loss": 1.3866,
+      "step": 39660
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0440119235490096e-06,
+      "loss": 1.4865,
+      "step": 39670
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0422584604594075e-06,
+      "loss": 1.5141,
+      "step": 39680
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.040504997369806e-06,
+      "loss": 1.5872,
+      "step": 39690
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0387515342802038e-06,
+      "loss": 1.4072,
+      "step": 39700
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0369980711906017e-06,
+      "loss": 1.4209,
+      "step": 39710
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.0352446081009996e-06,
+      "loss": 1.484,
+      "step": 39720
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.033491145011398e-06,
+      "loss": 1.4741,
+      "step": 39730
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.031737681921796e-06,
+      "loss": 1.7064,
+      "step": 39740
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0299842188321937e-06,
+      "loss": 1.4257,
+      "step": 39750
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.0282307557425916e-06,
+      "loss": 1.5052,
+      "step": 39760
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.02647729265299e-06,
+      "loss": 1.4403,
+      "step": 39770
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.024723829563388e-06,
+      "loss": 1.5045,
+      "step": 39780
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0229703664737857e-06,
+      "loss": 1.6475,
+      "step": 39790
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0212169033841845e-06,
+      "loss": 1.4491,
+      "step": 39800
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.019463440294582e-06,
+      "loss": 1.4246,
+      "step": 39810
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.01770997720498e-06,
+      "loss": 1.4102,
+      "step": 39820
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.0159565141153778e-06,
+      "loss": 1.6105,
+      "step": 39830
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0142030510257765e-06,
+      "loss": 1.5595,
+      "step": 39840
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0124495879361744e-06,
+      "loss": 1.4477,
+      "step": 39850
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0106961248465723e-06,
+      "loss": 1.5618,
+      "step": 39860
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0089426617569702e-06,
+      "loss": 1.4681,
+      "step": 39870
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0071891986673685e-06,
+      "loss": 1.6129,
+      "step": 39880
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.0054357355777665e-06,
+      "loss": 1.5973,
+      "step": 39890
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0036822724881644e-06,
+      "loss": 1.472,
+      "step": 39900
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0019288093985627e-06,
+      "loss": 1.477,
+      "step": 39910
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.0001753463089606e-06,
+      "loss": 1.3592,
+      "step": 39920
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.9984218832193585e-06,
+      "loss": 1.4839,
+      "step": 39930
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 2.9966684201297564e-06,
+      "loss": 1.372,
+      "step": 39940
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9949149570401547e-06,
+      "loss": 1.5975,
+      "step": 39950
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9931614939505526e-06,
+      "loss": 1.4737,
+      "step": 39960
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9914080308609505e-06,
+      "loss": 1.5334,
+      "step": 39970
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9896545677713484e-06,
+      "loss": 1.5277,
+      "step": 39980
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9879011046817467e-06,
+      "loss": 1.4781,
+      "step": 39990
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 2.9861476415921446e-06,
+      "loss": 1.5582,
+      "step": 40000
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9843941785025425e-06,
+      "loss": 1.5165,
+      "step": 40010
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9826407154129405e-06,
+      "loss": 1.5304,
+      "step": 40020
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9808872523233388e-06,
+      "loss": 1.3589,
+      "step": 40030
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9791337892337367e-06,
+      "loss": 1.5744,
+      "step": 40040
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9773803261441346e-06,
+      "loss": 1.4035,
+      "step": 40050
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 2.9756268630545333e-06,
+      "loss": 1.502,
+      "step": 40060
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9738733999649312e-06,
+      "loss": 1.6074,
+      "step": 40070
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.972119936875329e-06,
+      "loss": 1.4143,
+      "step": 40080
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.970366473785727e-06,
+      "loss": 1.346,
+      "step": 40090
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9686130106961254e-06,
+      "loss": 1.4947,
+      "step": 40100
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.9668595476065233e-06,
+      "loss": 1.542,
+      "step": 40110
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 2.965106084516921e-06,
+      "loss": 1.5652,
+      "step": 40120
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.963352621427319e-06,
+      "loss": 1.4653,
+      "step": 40130
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9615991583377174e-06,
+      "loss": 1.6217,
+      "step": 40140
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9598456952481153e-06,
+      "loss": 1.4702,
+      "step": 40150
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.958092232158513e-06,
+      "loss": 1.7315,
+      "step": 40160
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 2.9563387690689115e-06,
+      "loss": 1.4175,
+      "step": 40170
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9545853059793094e-06,
+      "loss": 1.3805,
+      "step": 40180
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9528318428897073e-06,
+      "loss": 1.4628,
+      "step": 40190
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9510783798001052e-06,
+      "loss": 1.503,
+      "step": 40200
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9493249167105036e-06,
+      "loss": 1.433,
+      "step": 40210
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9475714536209015e-06,
+      "loss": 1.3959,
+      "step": 40220
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 2.9458179905312994e-06,
+      "loss": 1.4452,
+      "step": 40230
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9440645274416973e-06,
+      "loss": 1.5547,
+      "step": 40240
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9423110643520956e-06,
+      "loss": 1.5265,
+      "step": 40250
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9405576012624935e-06,
+      "loss": 1.3552,
+      "step": 40260
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.9388041381728914e-06,
+      "loss": 1.4138,
+      "step": 40270
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.93705067508329e-06,
+      "loss": 1.3314,
+      "step": 40280
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 2.935297211993688e-06,
+      "loss": 1.2945,
+      "step": 40290
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.933543748904086e-06,
+      "loss": 1.4983,
+      "step": 40300
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.931790285814484e-06,
+      "loss": 1.5723,
+      "step": 40310
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.930036822724882e-06,
+      "loss": 1.6058,
+      "step": 40320
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.92828335963528e-06,
+      "loss": 1.57,
+      "step": 40330
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 2.926529896545678e-06,
+      "loss": 1.4554,
+      "step": 40340
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.924776433456076e-06,
+      "loss": 1.4027,
+      "step": 40350
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.923022970366474e-06,
+      "loss": 1.5855,
+      "step": 40360
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.921269507276872e-06,
+      "loss": 1.4848,
+      "step": 40370
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.91951604418727e-06,
+      "loss": 1.4456,
+      "step": 40380
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.917762581097668e-06,
+      "loss": 1.3722,
+      "step": 40390
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 2.9160091180080662e-06,
+      "loss": 1.4267,
+      "step": 40400
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.914255654918464e-06,
+      "loss": 1.5437,
+      "step": 40410
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.912502191828862e-06,
+      "loss": 1.5698,
+      "step": 40420
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9107487287392604e-06,
+      "loss": 1.5121,
+      "step": 40430
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.9089952656496583e-06,
+      "loss": 1.3548,
+      "step": 40440
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.907241802560056e-06,
+      "loss": 1.4356,
+      "step": 40450
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 2.905488339470454e-06,
+      "loss": 1.5418,
+      "step": 40460
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9037348763808524e-06,
+      "loss": 1.6742,
+      "step": 40470
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.9019814132912503e-06,
+      "loss": 1.5066,
+      "step": 40480
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.900227950201648e-06,
+      "loss": 1.4348,
+      "step": 40490
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.898474487112046e-06,
+      "loss": 1.6361,
+      "step": 40500
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 2.896721024022445e-06,
+      "loss": 1.4236,
+      "step": 40510
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8949675609328428e-06,
+      "loss": 1.4436,
+      "step": 40520
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8932140978432407e-06,
+      "loss": 1.5319,
+      "step": 40530
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.891460634753639e-06,
+      "loss": 1.4328,
+      "step": 40540
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.889707171664037e-06,
+      "loss": 1.4103,
+      "step": 40550
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.887953708574435e-06,
+      "loss": 1.4556,
+      "step": 40560
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 2.8862002454848327e-06,
+      "loss": 1.4063,
+      "step": 40570
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.884446782395231e-06,
+      "loss": 1.8354,
+      "step": 40580
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.882693319305629e-06,
+      "loss": 1.6524,
+      "step": 40590
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.880939856216027e-06,
+      "loss": 1.5388,
+      "step": 40600
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.8791863931264247e-06,
+      "loss": 1.4653,
+      "step": 40610
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.877432930036823e-06,
+      "loss": 1.5375,
+      "step": 40620
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 2.875679466947221e-06,
+      "loss": 1.5154,
+      "step": 40630
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.873926003857619e-06,
+      "loss": 1.4326,
+      "step": 40640
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.872172540768017e-06,
+      "loss": 1.536,
+      "step": 40650
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.870419077678415e-06,
+      "loss": 1.4909,
+      "step": 40660
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.868665614588813e-06,
+      "loss": 1.7301,
+      "step": 40670
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.866912151499211e-06,
+      "loss": 1.4753,
+      "step": 40680
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 2.8651586884096092e-06,
+      "loss": 1.4903,
+      "step": 40690
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.863405225320007e-06,
+      "loss": 1.3572,
+      "step": 40700
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.861651762230405e-06,
+      "loss": 1.5126,
+      "step": 40710
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.859898299140803e-06,
+      "loss": 1.5031,
+      "step": 40720
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8581448360512017e-06,
+      "loss": 1.5401,
+      "step": 40730
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 2.8563913729615996e-06,
+      "loss": 1.5128,
+      "step": 40740
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8546379098719975e-06,
+      "loss": 1.541,
+      "step": 40750
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.852884446782396e-06,
+      "loss": 1.4761,
+      "step": 40760
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8511309836927937e-06,
+      "loss": 1.5509,
+      "step": 40770
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8493775206031916e-06,
+      "loss": 1.5906,
+      "step": 40780
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.8476240575135895e-06,
+      "loss": 1.3068,
+      "step": 40790
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 2.845870594423988e-06,
+      "loss": 1.4175,
+      "step": 40800
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8441171313343857e-06,
+      "loss": 1.4109,
+      "step": 40810
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8423636682447836e-06,
+      "loss": 1.5855,
+      "step": 40820
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8406102051551815e-06,
+      "loss": 1.2784,
+      "step": 40830
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.83885674206558e-06,
+      "loss": 1.4253,
+      "step": 40840
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8371032789759778e-06,
+      "loss": 1.3371,
+      "step": 40850
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 2.8353498158863757e-06,
+      "loss": 1.4078,
+      "step": 40860
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8335963527967736e-06,
+      "loss": 1.5095,
+      "step": 40870
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.831842889707172e-06,
+      "loss": 1.5699,
+      "step": 40880
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.83008942661757e-06,
+      "loss": 1.4394,
+      "step": 40890
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.8283359635279677e-06,
+      "loss": 1.4529,
+      "step": 40900
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 2.826582500438366e-06,
+      "loss": 1.6316,
+      "step": 40910
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.824829037348764e-06,
+      "loss": 1.3273,
+      "step": 40920
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.823075574259162e-06,
+      "loss": 1.5005,
+      "step": 40930
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8213221111695597e-06,
+      "loss": 1.4557,
+      "step": 40940
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8195686480799585e-06,
+      "loss": 1.3919,
+      "step": 40950
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8178151849903564e-06,
+      "loss": 1.4202,
+      "step": 40960
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 2.8160617219007543e-06,
+      "loss": 1.4383,
+      "step": 40970
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.814308258811152e-06,
+      "loss": 1.4596,
+      "step": 40980
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8125547957215505e-06,
+      "loss": 1.8452,
+      "step": 40990
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8108013326319484e-06,
+      "loss": 1.4664,
+      "step": 41000
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8090478695423463e-06,
+      "loss": 1.5255,
+      "step": 41010
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8072944064527447e-06,
+      "loss": 1.435,
+      "step": 41020
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 2.8055409433631426e-06,
+      "loss": 1.3129,
+      "step": 41030
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8037874802735405e-06,
+      "loss": 1.5057,
+      "step": 41040
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8020340171839384e-06,
+      "loss": 1.3932,
+      "step": 41050
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.8002805540943367e-06,
+      "loss": 1.7861,
+      "step": 41060
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.7985270910047346e-06,
+      "loss": 1.483,
+      "step": 41070
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.7967736279151325e-06,
+      "loss": 1.6223,
+      "step": 41080
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 2.7950201648255304e-06,
+      "loss": 1.4435,
+      "step": 41090
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7932667017359287e-06,
+      "loss": 1.4685,
+      "step": 41100
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7915132386463266e-06,
+      "loss": 1.4352,
+      "step": 41110
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7897597755567245e-06,
+      "loss": 1.4244,
+      "step": 41120
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.788006312467123e-06,
+      "loss": 1.4795,
+      "step": 41130
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 2.7862528493775207e-06,
+      "loss": 1.4906,
+      "step": 41140
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7844993862879187e-06,
+      "loss": 1.2737,
+      "step": 41150
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7827459231983166e-06,
+      "loss": 1.3616,
+      "step": 41160
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.7809924601087153e-06,
+      "loss": 1.476,
+      "step": 41170
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.779238997019113e-06,
+      "loss": 1.5345,
+      "step": 41180
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.777485533929511e-06,
+      "loss": 1.5056,
+      "step": 41190
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 2.775732070839909e-06,
+      "loss": 1.45,
+      "step": 41200
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7739786077503073e-06,
+      "loss": 1.7209,
+      "step": 41210
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7722251446607052e-06,
+      "loss": 1.4485,
+      "step": 41220
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.770471681571103e-06,
+      "loss": 1.3865,
+      "step": 41230
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7687182184815015e-06,
+      "loss": 1.3944,
+      "step": 41240
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7669647553918994e-06,
+      "loss": 1.5172,
+      "step": 41250
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 2.7652112923022973e-06,
+      "loss": 1.4343,
+      "step": 41260
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.763457829212695e-06,
+      "loss": 1.5521,
+      "step": 41270
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7617043661230935e-06,
+      "loss": 1.4383,
+      "step": 41280
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7599509030334914e-06,
+      "loss": 1.4662,
+      "step": 41290
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.7581974399438893e-06,
+      "loss": 1.4596,
+      "step": 41300
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 2.756443976854287e-06,
+      "loss": 1.3434,
+      "step": 41310
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7546905137646855e-06,
+      "loss": 1.5646,
+      "step": 41320
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7529370506750834e-06,
+      "loss": 1.4008,
+      "step": 41330
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7511835875854813e-06,
+      "loss": 1.3857,
+      "step": 41340
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7494301244958792e-06,
+      "loss": 1.457,
+      "step": 41350
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7476766614062776e-06,
+      "loss": 1.6602,
+      "step": 41360
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 2.7459231983166755e-06,
+      "loss": 1.3637,
+      "step": 41370
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.7441697352270734e-06,
+      "loss": 1.5705,
+      "step": 41380
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.742416272137472e-06,
+      "loss": 1.4533,
+      "step": 41390
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.74066280904787e-06,
+      "loss": 1.4371,
+      "step": 41400
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.738909345958268e-06,
+      "loss": 1.5888,
+      "step": 41410
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.737155882868666e-06,
+      "loss": 1.5302,
+      "step": 41420
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 2.735402419779064e-06,
+      "loss": 1.408,
+      "step": 41430
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.733648956689462e-06,
+      "loss": 1.2932,
+      "step": 41440
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.73189549359986e-06,
+      "loss": 1.5308,
+      "step": 41450
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.730142030510258e-06,
+      "loss": 1.4124,
+      "step": 41460
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.728388567420656e-06,
+      "loss": 1.482,
+      "step": 41470
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 2.726635104331054e-06,
+      "loss": 1.4898,
+      "step": 41480
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.724881641241452e-06,
+      "loss": 1.5536,
+      "step": 41490
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7231281781518503e-06,
+      "loss": 1.5472,
+      "step": 41500
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7213747150622482e-06,
+      "loss": 1.559,
+      "step": 41510
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.719621251972646e-06,
+      "loss": 1.4975,
+      "step": 41520
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.717867788883044e-06,
+      "loss": 1.3799,
+      "step": 41530
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 2.7161143257934423e-06,
+      "loss": 1.391,
+      "step": 41540
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7143608627038402e-06,
+      "loss": 1.4347,
+      "step": 41550
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.712607399614238e-06,
+      "loss": 1.5153,
+      "step": 41560
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.710853936524636e-06,
+      "loss": 1.4137,
+      "step": 41570
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7091004734350344e-06,
+      "loss": 1.4659,
+      "step": 41580
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.7073470103454323e-06,
+      "loss": 1.4324,
+      "step": 41590
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 2.70559354725583e-06,
+      "loss": 1.4569,
+      "step": 41600
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.703840084166229e-06,
+      "loss": 1.332,
+      "step": 41610
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.702086621076627e-06,
+      "loss": 1.4845,
+      "step": 41620
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.7003331579870247e-06,
+      "loss": 1.3123,
+      "step": 41630
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.6985796948974226e-06,
+      "loss": 1.4434,
+      "step": 41640
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.696826231807821e-06,
+      "loss": 1.5024,
+      "step": 41650
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 2.695072768718219e-06,
+      "loss": 1.6276,
+      "step": 41660
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6933193056286168e-06,
+      "loss": 1.2948,
+      "step": 41670
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.6915658425390147e-06,
+      "loss": 1.5944,
+      "step": 41680
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.689812379449413e-06,
+      "loss": 1.5568,
+      "step": 41690
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.688058916359811e-06,
+      "loss": 1.4104,
+      "step": 41700
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 2.686305453270209e-06,
+      "loss": 1.4466,
+      "step": 41710
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.684551990180607e-06,
+      "loss": 1.4724,
+      "step": 41720
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.682798527091005e-06,
+      "loss": 1.2171,
+      "step": 41730
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.681045064001403e-06,
+      "loss": 1.4983,
+      "step": 41740
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.679291600911801e-06,
+      "loss": 1.4582,
+      "step": 41750
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.677538137822199e-06,
+      "loss": 1.3686,
+      "step": 41760
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 2.675784674732597e-06,
+      "loss": 1.4225,
+      "step": 41770
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.674031211642995e-06,
+      "loss": 1.5329,
+      "step": 41780
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.672277748553393e-06,
+      "loss": 1.5221,
+      "step": 41790
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.670524285463791e-06,
+      "loss": 1.4754,
+      "step": 41800
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.668770822374189e-06,
+      "loss": 1.496,
+      "step": 41810
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.667017359284587e-06,
+      "loss": 1.5227,
+      "step": 41820
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 2.665263896194985e-06,
+      "loss": 1.4325,
+      "step": 41830
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6635104331053836e-06,
+      "loss": 1.6816,
+      "step": 41840
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6617569700157815e-06,
+      "loss": 1.5646,
+      "step": 41850
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6600035069261794e-06,
+      "loss": 1.5833,
+      "step": 41860
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6582500438365778e-06,
+      "loss": 1.3941,
+      "step": 41870
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 2.6564965807469757e-06,
+      "loss": 1.5184,
+      "step": 41880
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6547431176573736e-06,
+      "loss": 1.4565,
+      "step": 41890
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6529896545677715e-06,
+      "loss": 1.5019,
+      "step": 41900
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.65123619147817e-06,
+      "loss": 1.4891,
+      "step": 41910
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6494827283885677e-06,
+      "loss": 1.4795,
+      "step": 41920
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6477292652989656e-06,
+      "loss": 1.6439,
+      "step": 41930
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 2.6459758022093635e-06,
+      "loss": 1.5238,
+      "step": 41940
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.644222339119762e-06,
+      "loss": 1.5148,
+      "step": 41950
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6424688760301597e-06,
+      "loss": 1.5865,
+      "step": 41960
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6407154129405576e-06,
+      "loss": 1.3178,
+      "step": 41970
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.638961949850956e-06,
+      "loss": 1.5687,
+      "step": 41980
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.637208486761354e-06,
+      "loss": 1.5771,
+      "step": 41990
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 2.6354550236717518e-06,
+      "loss": 1.4309,
+      "step": 42000
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6337015605821497e-06,
+      "loss": 1.4385,
+      "step": 42010
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.631948097492548e-06,
+      "loss": 1.5166,
+      "step": 42020
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.630194634402946e-06,
+      "loss": 1.4969,
+      "step": 42030
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.628441171313344e-06,
+      "loss": 1.4236,
+      "step": 42040
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 2.6266877082237417e-06,
+      "loss": 1.6246,
+      "step": 42050
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6249342451341405e-06,
+      "loss": 1.5282,
+      "step": 42060
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6231807820445384e-06,
+      "loss": 1.4153,
+      "step": 42070
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6214273189549363e-06,
+      "loss": 1.5201,
+      "step": 42080
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6196738558653346e-06,
+      "loss": 1.5488,
+      "step": 42090
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6179203927757325e-06,
+      "loss": 1.4788,
+      "step": 42100
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 2.6161669296861304e-06,
+      "loss": 1.3021,
+      "step": 42110
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6144134665965283e-06,
+      "loss": 1.4559,
+      "step": 42120
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6126600035069266e-06,
+      "loss": 1.4607,
+      "step": 42130
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6109065404173245e-06,
+      "loss": 1.4087,
+      "step": 42140
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6091530773277224e-06,
+      "loss": 1.534,
+      "step": 42150
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6073996142381203e-06,
+      "loss": 1.5641,
+      "step": 42160
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 2.6056461511485187e-06,
+      "loss": 1.5943,
+      "step": 42170
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6038926880589166e-06,
+      "loss": 1.6205,
+      "step": 42180
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6021392249693145e-06,
+      "loss": 1.3916,
+      "step": 42190
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.6003857618797128e-06,
+      "loss": 1.4484,
+      "step": 42200
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.5986322987901107e-06,
+      "loss": 1.4039,
+      "step": 42210
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.5968788357005086e-06,
+      "loss": 1.6565,
+      "step": 42220
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 2.5951253726109065e-06,
+      "loss": 1.5098,
+      "step": 42230
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.593371909521305e-06,
+      "loss": 1.3858,
+      "step": 42240
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5916184464317027e-06,
+      "loss": 1.399,
+      "step": 42250
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5898649833421006e-06,
+      "loss": 1.591,
+      "step": 42260
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5881115202524985e-06,
+      "loss": 1.5213,
+      "step": 42270
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 2.5863580571628973e-06,
+      "loss": 1.4338,
+      "step": 42280
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.584604594073295e-06,
+      "loss": 1.5051,
+      "step": 42290
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.582851130983693e-06,
+      "loss": 1.5003,
+      "step": 42300
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.581097667894091e-06,
+      "loss": 1.3868,
+      "step": 42310
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.5793442048044893e-06,
+      "loss": 1.6241,
+      "step": 42320
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.577590741714887e-06,
+      "loss": 1.4751,
+      "step": 42330
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 2.575837278625285e-06,
+      "loss": 1.4672,
+      "step": 42340
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5740838155356834e-06,
+      "loss": 1.5258,
+      "step": 42350
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5723303524460813e-06,
+      "loss": 1.421,
+      "step": 42360
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5705768893564792e-06,
+      "loss": 1.4499,
+      "step": 42370
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.568823426266877e-06,
+      "loss": 1.4193,
+      "step": 42380
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5670699631772755e-06,
+      "loss": 1.4206,
+      "step": 42390
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 2.5653165000876734e-06,
+      "loss": 1.3791,
+      "step": 42400
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5635630369980713e-06,
+      "loss": 1.5387,
+      "step": 42410
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.561809573908469e-06,
+      "loss": 1.5573,
+      "step": 42420
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5600561108188675e-06,
+      "loss": 1.4937,
+      "step": 42430
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5583026477292654e-06,
+      "loss": 1.4179,
+      "step": 42440
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 2.5565491846396633e-06,
+      "loss": 1.3395,
+      "step": 42450
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5547957215500616e-06,
+      "loss": 1.3925,
+      "step": 42460
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5530422584604595e-06,
+      "loss": 1.3705,
+      "step": 42470
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5512887953708574e-06,
+      "loss": 1.4106,
+      "step": 42480
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.5495353322812553e-06,
+      "loss": 1.3951,
+      "step": 42490
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.547781869191654e-06,
+      "loss": 1.5842,
+      "step": 42500
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 2.546028406102052e-06,
+      "loss": 1.3861,
+      "step": 42510
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.54427494301245e-06,
+      "loss": 1.5251,
+      "step": 42520
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.542521479922848e-06,
+      "loss": 1.3633,
+      "step": 42530
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.540768016833246e-06,
+      "loss": 1.3451,
+      "step": 42540
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.539014553743644e-06,
+      "loss": 1.576,
+      "step": 42550
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.537261090654042e-06,
+      "loss": 1.3869,
+      "step": 42560
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 2.5355076275644402e-06,
+      "loss": 1.427,
+      "step": 42570
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.533754164474838e-06,
+      "loss": 1.4551,
+      "step": 42580
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.532000701385236e-06,
+      "loss": 1.4531,
+      "step": 42590
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.530247238295634e-06,
+      "loss": 1.4723,
+      "step": 42600
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.5284937752060323e-06,
+      "loss": 1.447,
+      "step": 42610
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 2.52674031211643e-06,
+      "loss": 1.4295,
+      "step": 42620
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.524986849026828e-06,
+      "loss": 1.6286,
+      "step": 42630
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.523233385937226e-06,
+      "loss": 1.4515,
+      "step": 42640
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5214799228476243e-06,
+      "loss": 1.5239,
+      "step": 42650
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.5197264597580222e-06,
+      "loss": 1.3664,
+      "step": 42660
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.51797299666842e-06,
+      "loss": 1.681,
+      "step": 42670
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 2.516219533578818e-06,
+      "loss": 1.374,
+      "step": 42680
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5144660704892163e-06,
+      "loss": 1.6481,
+      "step": 42690
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5127126073996142e-06,
+      "loss": 1.3341,
+      "step": 42700
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.510959144310012e-06,
+      "loss": 1.6354,
+      "step": 42710
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.509205681220411e-06,
+      "loss": 1.4162,
+      "step": 42720
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.507452218130809e-06,
+      "loss": 1.4142,
+      "step": 42730
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 2.5056987550412067e-06,
+      "loss": 1.3922,
+      "step": 42740
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.5039452919516046e-06,
+      "loss": 1.5845,
+      "step": 42750
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.502191828862003e-06,
+      "loss": 1.6402,
+      "step": 42760
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.500438365772401e-06,
+      "loss": 1.4195,
+      "step": 42770
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.4986849026827987e-06,
+      "loss": 1.5039,
+      "step": 42780
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.4969314395931966e-06,
+      "loss": 1.3766,
+      "step": 42790
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 2.4951779765035945e-06,
+      "loss": 1.5826,
+      "step": 42800
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.493424513413993e-06,
+      "loss": 1.3755,
+      "step": 42810
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.491671050324391e-06,
+      "loss": 1.3179,
+      "step": 42820
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.4899175872347887e-06,
+      "loss": 1.4021,
+      "step": 42830
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.488164124145187e-06,
+      "loss": 1.4777,
+      "step": 42840
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 2.486410661055585e-06,
+      "loss": 1.3847,
+      "step": 42850
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4846571979659832e-06,
+      "loss": 1.5184,
+      "step": 42860
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.482903734876381e-06,
+      "loss": 1.3845,
+      "step": 42870
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.481150271786779e-06,
+      "loss": 1.323,
+      "step": 42880
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.479396808697177e-06,
+      "loss": 1.3979,
+      "step": 42890
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.4776433456075753e-06,
+      "loss": 1.4984,
+      "step": 42900
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 2.475889882517973e-06,
+      "loss": 1.5454,
+      "step": 42910
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.474136419428371e-06,
+      "loss": 1.4485,
+      "step": 42920
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4723829563387694e-06,
+      "loss": 1.5801,
+      "step": 42930
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4706294932491673e-06,
+      "loss": 1.6487,
+      "step": 42940
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4688760301595656e-06,
+      "loss": 1.376,
+      "step": 42950
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4671225670699635e-06,
+      "loss": 1.5547,
+      "step": 42960
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 2.4653691039803614e-06,
+      "loss": 1.4494,
+      "step": 42970
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4636156408907593e-06,
+      "loss": 1.5217,
+      "step": 42980
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4618621778011576e-06,
+      "loss": 1.4091,
+      "step": 42990
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4601087147115556e-06,
+      "loss": 1.4227,
+      "step": 43000
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4583552516219535e-06,
+      "loss": 1.5177,
+      "step": 43010
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 2.4566017885323514e-06,
+      "loss": 1.3539,
+      "step": 43020
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4548483254427497e-06,
+      "loss": 1.4613,
+      "step": 43030
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4530948623531476e-06,
+      "loss": 1.4953,
+      "step": 43040
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4513413992635455e-06,
+      "loss": 1.4927,
+      "step": 43050
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.449587936173944e-06,
+      "loss": 1.5542,
+      "step": 43060
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.4478344730843417e-06,
+      "loss": 1.5344,
+      "step": 43070
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 2.44608100999474e-06,
+      "loss": 1.5225,
+      "step": 43080
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.444327546905138e-06,
+      "loss": 1.578,
+      "step": 43090
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.442574083815536e-06,
+      "loss": 1.5247,
+      "step": 43100
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.4408206207259337e-06,
+      "loss": 1.4479,
+      "step": 43110
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.439067157636332e-06,
+      "loss": 1.6037,
+      "step": 43120
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.43731369454673e-06,
+      "loss": 1.5656,
+      "step": 43130
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 2.435560231457128e-06,
+      "loss": 1.3924,
+      "step": 43140
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4338067683675258e-06,
+      "loss": 1.5512,
+      "step": 43150
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.432053305277924e-06,
+      "loss": 1.4391,
+      "step": 43160
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4302998421883224e-06,
+      "loss": 1.5584,
+      "step": 43170
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4285463790987203e-06,
+      "loss": 1.5373,
+      "step": 43180
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.4267929160091182e-06,
+      "loss": 1.4764,
+      "step": 43190
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 2.425039452919516e-06,
+      "loss": 1.4176,
+      "step": 43200
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4232859898299145e-06,
+      "loss": 1.5127,
+      "step": 43210
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4215325267403124e-06,
+      "loss": 1.6974,
+      "step": 43220
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4197790636507103e-06,
+      "loss": 1.5658,
+      "step": 43230
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.418025600561108e-06,
+      "loss": 1.3934,
+      "step": 43240
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 2.4162721374715065e-06,
+      "loss": 1.6103,
+      "step": 43250
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4145186743819044e-06,
+      "loss": 1.4067,
+      "step": 43260
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4127652112923027e-06,
+      "loss": 1.5379,
+      "step": 43270
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4110117482027e-06,
+      "loss": 1.5377,
+      "step": 43280
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4092582851130985e-06,
+      "loss": 1.4518,
+      "step": 43290
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.407504822023497e-06,
+      "loss": 1.5801,
+      "step": 43300
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 2.4057513589338948e-06,
+      "loss": 1.4578,
+      "step": 43310
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4039978958442927e-06,
+      "loss": 1.9791,
+      "step": 43320
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.4022444327546906e-06,
+      "loss": 1.5324,
+      "step": 43330
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.400490969665089e-06,
+      "loss": 1.4301,
+      "step": 43340
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.3987375065754868e-06,
+      "loss": 1.3766,
+      "step": 43350
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.3969840434858847e-06,
+      "loss": 1.5824,
+      "step": 43360
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 2.3952305803962826e-06,
+      "loss": 1.5733,
+      "step": 43370
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.393477117306681e-06,
+      "loss": 1.4414,
+      "step": 43380
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.391723654217079e-06,
+      "loss": 1.3427,
+      "step": 43390
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.389970191127477e-06,
+      "loss": 1.5297,
+      "step": 43400
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.388216728037875e-06,
+      "loss": 1.3533,
+      "step": 43410
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 2.386463264948273e-06,
+      "loss": 1.6643,
+      "step": 43420
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3847098018586713e-06,
+      "loss": 1.4191,
+      "step": 43430
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.382956338769069e-06,
+      "loss": 1.4742,
+      "step": 43440
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.381202875679467e-06,
+      "loss": 1.6615,
+      "step": 43450
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.379449412589865e-06,
+      "loss": 1.5177,
+      "step": 43460
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.3776959495002633e-06,
+      "loss": 1.4933,
+      "step": 43470
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 2.375942486410661e-06,
+      "loss": 1.4426,
+      "step": 43480
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3741890233210595e-06,
+      "loss": 1.5642,
+      "step": 43490
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.372435560231457e-06,
+      "loss": 1.4335,
+      "step": 43500
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3706820971418553e-06,
+      "loss": 1.2795,
+      "step": 43510
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3689286340522532e-06,
+      "loss": 1.6796,
+      "step": 43520
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3671751709626516e-06,
+      "loss": 1.3776,
+      "step": 43530
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 2.3654217078730495e-06,
+      "loss": 1.4967,
+      "step": 43540
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3636682447834474e-06,
+      "loss": 1.4664,
+      "step": 43550
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3619147816938457e-06,
+      "loss": 1.401,
+      "step": 43560
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3601613186042436e-06,
+      "loss": 1.5371,
+      "step": 43570
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3584078555146415e-06,
+      "loss": 1.3524,
+      "step": 43580
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 2.3566543924250394e-06,
+      "loss": 1.3851,
+      "step": 43590
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3549009293354377e-06,
+      "loss": 1.4099,
+      "step": 43600
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3531474662458356e-06,
+      "loss": 1.6796,
+      "step": 43610
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.351394003156234e-06,
+      "loss": 1.4483,
+      "step": 43620
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.349640540066632e-06,
+      "loss": 1.4438,
+      "step": 43630
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.3478870769770298e-06,
+      "loss": 1.5147,
+      "step": 43640
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 2.346133613887428e-06,
+      "loss": 1.4518,
+      "step": 43650
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.344380150797826e-06,
+      "loss": 1.3039,
+      "step": 43660
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.342626687708224e-06,
+      "loss": 1.3655,
+      "step": 43670
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.340873224618622e-06,
+      "loss": 1.4884,
+      "step": 43680
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.33911976152902e-06,
+      "loss": 1.3446,
+      "step": 43690
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.337366298439418e-06,
+      "loss": 1.4506,
+      "step": 43700
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 2.3356128353498164e-06,
+      "loss": 1.3966,
+      "step": 43710
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.333859372260214e-06,
+      "loss": 1.5162,
+      "step": 43720
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.332105909170612e-06,
+      "loss": 1.5589,
+      "step": 43730
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.33035244608101e-06,
+      "loss": 1.5786,
+      "step": 43740
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3285989829914084e-06,
+      "loss": 1.5234,
+      "step": 43750
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.3268455199018063e-06,
+      "loss": 1.5274,
+      "step": 43760
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 2.325092056812204e-06,
+      "loss": 1.4872,
+      "step": 43770
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3233385937226025e-06,
+      "loss": 1.6169,
+      "step": 43780
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3215851306330004e-06,
+      "loss": 1.3019,
+      "step": 43790
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3198316675433983e-06,
+      "loss": 1.3348,
+      "step": 43800
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3180782044537962e-06,
+      "loss": 1.6354,
+      "step": 43810
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 2.3163247413641945e-06,
+      "loss": 1.3861,
+      "step": 43820
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3145712782745924e-06,
+      "loss": 1.625,
+      "step": 43830
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3128178151849908e-06,
+      "loss": 1.2665,
+      "step": 43840
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3110643520953887e-06,
+      "loss": 1.5624,
+      "step": 43850
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3093108890057866e-06,
+      "loss": 1.5067,
+      "step": 43860
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.3075574259161845e-06,
+      "loss": 1.53,
+      "step": 43870
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 2.305803962826583e-06,
+      "loss": 1.4757,
+      "step": 43880
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3040504997369807e-06,
+      "loss": 1.6477,
+      "step": 43890
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.3022970366473786e-06,
+      "loss": 1.4166,
+      "step": 43900
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.300543573557777e-06,
+      "loss": 1.5195,
+      "step": 43910
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.298790110468175e-06,
+      "loss": 1.444,
+      "step": 43920
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.297036647378573e-06,
+      "loss": 1.2869,
+      "step": 43930
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 2.2952831842889706e-06,
+      "loss": 1.7213,
+      "step": 43940
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.293529721199369e-06,
+      "loss": 1.4968,
+      "step": 43950
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.291776258109767e-06,
+      "loss": 1.5578,
+      "step": 43960
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.290022795020165e-06,
+      "loss": 1.4891,
+      "step": 43970
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.288269331930563e-06,
+      "loss": 1.3081,
+      "step": 43980
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 2.286515868840961e-06,
+      "loss": 1.5679,
+      "step": 43990
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.284762405751359e-06,
+      "loss": 1.4847,
+      "step": 44000
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2830089426617572e-06,
+      "loss": 1.4793,
+      "step": 44010
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.281255479572155e-06,
+      "loss": 1.48,
+      "step": 44020
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.279502016482553e-06,
+      "loss": 1.5107,
+      "step": 44030
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2777485533929514e-06,
+      "loss": 1.4851,
+      "step": 44040
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 2.2759950903033493e-06,
+      "loss": 1.4297,
+      "step": 44050
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2742416272137476e-06,
+      "loss": 1.4637,
+      "step": 44060
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2724881641241455e-06,
+      "loss": 1.3628,
+      "step": 44070
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2707347010345434e-06,
+      "loss": 1.4769,
+      "step": 44080
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2689812379449413e-06,
+      "loss": 1.4208,
+      "step": 44090
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2672277748553396e-06,
+      "loss": 1.4672,
+      "step": 44100
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 2.2654743117657375e-06,
+      "loss": 1.4153,
+      "step": 44110
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2637208486761354e-06,
+      "loss": 1.4091,
+      "step": 44120
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2619673855865337e-06,
+      "loss": 1.4066,
+      "step": 44130
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2602139224969317e-06,
+      "loss": 1.5167,
+      "step": 44140
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.25846045940733e-06,
+      "loss": 1.3575,
+      "step": 44150
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 2.2567069963177275e-06,
+      "loss": 1.51,
+      "step": 44160
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2549535332281258e-06,
+      "loss": 1.3487,
+      "step": 44170
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2532000701385237e-06,
+      "loss": 1.4994,
+      "step": 44180
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.251446607048922e-06,
+      "loss": 1.3393,
+      "step": 44190
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.24969314395932e-06,
+      "loss": 1.3682,
+      "step": 44200
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.247939680869718e-06,
+      "loss": 1.3665,
+      "step": 44210
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 2.2461862177801157e-06,
+      "loss": 1.5185,
+      "step": 44220
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.244432754690514e-06,
+      "loss": 1.6056,
+      "step": 44230
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.242679291600912e-06,
+      "loss": 1.4444,
+      "step": 44240
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.24092582851131e-06,
+      "loss": 1.5252,
+      "step": 44250
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.239172365421708e-06,
+      "loss": 1.5038,
+      "step": 44260
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.237418902332106e-06,
+      "loss": 1.5074,
+      "step": 44270
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 2.2356654392425044e-06,
+      "loss": 1.4561,
+      "step": 44280
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2339119761529023e-06,
+      "loss": 1.6361,
+      "step": 44290
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2321585130633e-06,
+      "loss": 1.3271,
+      "step": 44300
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.230405049973698e-06,
+      "loss": 1.348,
+      "step": 44310
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2286515868840964e-06,
+      "loss": 1.3334,
+      "step": 44320
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2268981237944943e-06,
+      "loss": 1.255,
+      "step": 44330
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 2.2251446607048922e-06,
+      "loss": 1.4408,
+      "step": 44340
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.22339119761529e-06,
+      "loss": 1.5577,
+      "step": 44350
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2216377345256885e-06,
+      "loss": 1.456,
+      "step": 44360
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2198842714360864e-06,
+      "loss": 1.5114,
+      "step": 44370
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2181308083464843e-06,
+      "loss": 1.5375,
+      "step": 44380
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 2.2163773452568826e-06,
+      "loss": 1.4983,
+      "step": 44390
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2146238821672805e-06,
+      "loss": 1.4782,
+      "step": 44400
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.212870419077679e-06,
+      "loss": 1.3679,
+      "step": 44410
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2111169559880767e-06,
+      "loss": 1.5444,
+      "step": 44420
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2093634928984746e-06,
+      "loss": 1.404,
+      "step": 44430
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.2076100298088725e-06,
+      "loss": 1.4513,
+      "step": 44440
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 2.205856566719271e-06,
+      "loss": 1.4823,
+      "step": 44450
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2041031036296688e-06,
+      "loss": 1.3037,
+      "step": 44460
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2023496405400667e-06,
+      "loss": 1.4512,
+      "step": 44470
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.2005961774504646e-06,
+      "loss": 1.4629,
+      "step": 44480
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.198842714360863e-06,
+      "loss": 1.3974,
+      "step": 44490
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.1970892512712612e-06,
+      "loss": 1.4325,
+      "step": 44500
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 2.195335788181659e-06,
+      "loss": 1.4535,
+      "step": 44510
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.193582325092057e-06,
+      "loss": 1.3622,
+      "step": 44520
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.191828862002455e-06,
+      "loss": 1.3812,
+      "step": 44530
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.1900753989128532e-06,
+      "loss": 1.574,
+      "step": 44540
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.188321935823251e-06,
+      "loss": 1.5112,
+      "step": 44550
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 2.186568472733649e-06,
+      "loss": 1.5682,
+      "step": 44560
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.184815009644047e-06,
+      "loss": 1.4895,
+      "step": 44570
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1830615465544453e-06,
+      "loss": 1.6069,
+      "step": 44580
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.181308083464843e-06,
+      "loss": 1.6033,
+      "step": 44590
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.179554620375241e-06,
+      "loss": 1.4902,
+      "step": 44600
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.177801157285639e-06,
+      "loss": 1.4589,
+      "step": 44610
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 2.1760476941960373e-06,
+      "loss": 1.7119,
+      "step": 44620
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1742942311064356e-06,
+      "loss": 1.5146,
+      "step": 44630
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1725407680168335e-06,
+      "loss": 1.3798,
+      "step": 44640
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1707873049272314e-06,
+      "loss": 1.558,
+      "step": 44650
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1690338418376293e-06,
+      "loss": 1.4781,
+      "step": 44660
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1672803787480277e-06,
+      "loss": 1.472,
+      "step": 44670
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 2.1655269156584256e-06,
+      "loss": 1.5523,
+      "step": 44680
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1637734525688235e-06,
+      "loss": 1.572,
+      "step": 44690
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1620199894792214e-06,
+      "loss": 1.3818,
+      "step": 44700
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1602665263896197e-06,
+      "loss": 1.5035,
+      "step": 44710
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.1585130633000176e-06,
+      "loss": 1.6829,
+      "step": 44720
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.156759600210416e-06,
+      "loss": 1.4643,
+      "step": 44730
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 2.155006137120814e-06,
+      "loss": 1.2757,
+      "step": 44740
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1532526740312117e-06,
+      "loss": 1.5287,
+      "step": 44750
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.15149921094161e-06,
+      "loss": 1.483,
+      "step": 44760
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.149745747852008e-06,
+      "loss": 1.4377,
+      "step": 44770
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.147992284762406e-06,
+      "loss": 1.507,
+      "step": 44780
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 2.1462388216728038e-06,
+      "loss": 1.6978,
+      "step": 44790
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.144485358583202e-06,
+      "loss": 1.3414,
+      "step": 44800
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.1427318954936e-06,
+      "loss": 1.4471,
+      "step": 44810
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.140978432403998e-06,
+      "loss": 1.4213,
+      "step": 44820
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.139224969314396e-06,
+      "loss": 1.5877,
+      "step": 44830
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.137471506224794e-06,
+      "loss": 1.4958,
+      "step": 44840
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 2.135718043135192e-06,
+      "loss": 1.454,
+      "step": 44850
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1339645800455904e-06,
+      "loss": 1.4496,
+      "step": 44860
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1322111169559883e-06,
+      "loss": 1.4043,
+      "step": 44870
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.130457653866386e-06,
+      "loss": 1.4927,
+      "step": 44880
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1287041907767845e-06,
+      "loss": 1.3037,
+      "step": 44890
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1269507276871824e-06,
+      "loss": 1.2571,
+      "step": 44900
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 2.1251972645975803e-06,
+      "loss": 1.3755,
+      "step": 44910
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.123443801507978e-06,
+      "loss": 1.2982,
+      "step": 44920
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1216903384183765e-06,
+      "loss": 1.4388,
+      "step": 44930
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1199368753287744e-06,
+      "loss": 1.3644,
+      "step": 44940
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1181834122391727e-06,
+      "loss": 1.4964,
+      "step": 44950
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 2.1164299491495706e-06,
+      "loss": 1.4682,
+      "step": 44960
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1146764860599685e-06,
+      "loss": 1.4304,
+      "step": 44970
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.112923022970367e-06,
+      "loss": 1.4284,
+      "step": 44980
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1111695598807648e-06,
+      "loss": 1.4392,
+      "step": 44990
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1094160967911627e-06,
+      "loss": 1.239,
+      "step": 45000
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.1076626337015606e-06,
+      "loss": 1.2977,
+      "step": 45010
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 2.105909170611959e-06,
+      "loss": 1.6826,
+      "step": 45020
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.104155707522357e-06,
+      "loss": 1.4377,
+      "step": 45030
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.102402244432755e-06,
+      "loss": 1.3872,
+      "step": 45040
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.1006487813431526e-06,
+      "loss": 1.5275,
+      "step": 45050
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.098895318253551e-06,
+      "loss": 1.4863,
+      "step": 45060
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.097141855163949e-06,
+      "loss": 1.439,
+      "step": 45070
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 2.095388392074347e-06,
+      "loss": 1.4519,
+      "step": 45080
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.093634928984745e-06,
+      "loss": 1.3979,
+      "step": 45090
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.091881465895143e-06,
+      "loss": 1.49,
+      "step": 45100
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.0901280028055413e-06,
+      "loss": 1.6803,
+      "step": 45110
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.088374539715939e-06,
+      "loss": 1.4634,
+      "step": 45120
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 2.086621076626337e-06,
+      "loss": 1.4087,
+      "step": 45130
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.084867613536735e-06,
+      "loss": 1.4457,
+      "step": 45140
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0831141504471333e-06,
+      "loss": 1.5768,
+      "step": 45150
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0813606873575312e-06,
+      "loss": 1.3254,
+      "step": 45160
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0796072242679296e-06,
+      "loss": 1.3485,
+      "step": 45170
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0778537611783275e-06,
+      "loss": 1.4843,
+      "step": 45180
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 2.0761002980887254e-06,
+      "loss": 1.2597,
+      "step": 45190
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0743468349991233e-06,
+      "loss": 1.4602,
+      "step": 45200
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0725933719095216e-06,
+      "loss": 1.4187,
+      "step": 45210
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0708399088199195e-06,
+      "loss": 1.5729,
+      "step": 45220
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0690864457303174e-06,
+      "loss": 1.472,
+      "step": 45230
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0673329826407157e-06,
+      "loss": 1.385,
+      "step": 45240
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 2.0655795195511136e-06,
+      "loss": 1.4294,
+      "step": 45250
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.063826056461512e-06,
+      "loss": 1.7955,
+      "step": 45260
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0620725933719094e-06,
+      "loss": 1.4978,
+      "step": 45270
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0603191302823078e-06,
+      "loss": 1.4538,
+      "step": 45280
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.0585656671927057e-06,
+      "loss": 1.3969,
+      "step": 45290
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.056812204103104e-06,
+      "loss": 1.5494,
+      "step": 45300
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 2.055058741013502e-06,
+      "loss": 1.5161,
+      "step": 45310
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0533052779238998e-06,
+      "loss": 1.5456,
+      "step": 45320
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.0515518148342977e-06,
+      "loss": 1.5308,
+      "step": 45330
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.049798351744696e-06,
+      "loss": 1.3638,
+      "step": 45340
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.048044888655094e-06,
+      "loss": 1.563,
+      "step": 45350
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 2.046291425565492e-06,
+      "loss": 1.5936,
+      "step": 45360
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.04453796247589e-06,
+      "loss": 1.4812,
+      "step": 45370
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.042784499386288e-06,
+      "loss": 1.5333,
+      "step": 45380
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0410310362966864e-06,
+      "loss": 1.3354,
+      "step": 45390
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.0392775732070843e-06,
+      "loss": 1.4412,
+      "step": 45400
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.037524110117482e-06,
+      "loss": 1.3304,
+      "step": 45410
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 2.03577064702788e-06,
+      "loss": 1.3835,
+      "step": 45420
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0340171839382784e-06,
+      "loss": 1.505,
+      "step": 45430
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0322637208486763e-06,
+      "loss": 1.5367,
+      "step": 45440
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.030510257759074e-06,
+      "loss": 1.4143,
+      "step": 45450
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0287567946694725e-06,
+      "loss": 1.4028,
+      "step": 45460
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0270033315798704e-06,
+      "loss": 1.3345,
+      "step": 45470
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 2.0252498684902688e-06,
+      "loss": 1.4015,
+      "step": 45480
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0234964054006662e-06,
+      "loss": 1.535,
+      "step": 45490
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0217429423110646e-06,
+      "loss": 1.4642,
+      "step": 45500
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0199894792214625e-06,
+      "loss": 1.4577,
+      "step": 45510
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.018236016131861e-06,
+      "loss": 1.4232,
+      "step": 45520
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 2.0164825530422587e-06,
+      "loss": 1.3748,
+      "step": 45530
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0147290899526566e-06,
+      "loss": 1.5097,
+      "step": 45540
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0129756268630545e-06,
+      "loss": 1.37,
+      "step": 45550
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.011222163773453e-06,
+      "loss": 1.4344,
+      "step": 45560
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0094687006838507e-06,
+      "loss": 1.495,
+      "step": 45570
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.0077152375942486e-06,
+      "loss": 1.4743,
+      "step": 45580
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 2.005961774504647e-06,
+      "loss": 1.5554,
+      "step": 45590
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.004208311415045e-06,
+      "loss": 1.4316,
+      "step": 45600
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.002454848325443e-06,
+      "loss": 1.5056,
+      "step": 45610
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 2.000701385235841e-06,
+      "loss": 1.4558,
+      "step": 45620
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 1.998947922146239e-06,
+      "loss": 1.4437,
+      "step": 45630
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 1.997194459056637e-06,
+      "loss": 1.5245,
+      "step": 45640
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 1.9954409959670352e-06,
+      "loss": 1.4357,
+      "step": 45650
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.993687532877433e-06,
+      "loss": 1.4771,
+      "step": 45660
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.991934069787831e-06,
+      "loss": 1.4161,
+      "step": 45670
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.990180606698229e-06,
+      "loss": 1.4226,
+      "step": 45680
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.9884271436086273e-06,
+      "loss": 1.3246,
+      "step": 45690
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 1.9866736805190256e-06,
+      "loss": 1.3817,
+      "step": 45700
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.984920217429423e-06,
+      "loss": 1.5707,
+      "step": 45710
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.9831667543398214e-06,
+      "loss": 1.3473,
+      "step": 45720
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.9814132912502193e-06,
+      "loss": 1.507,
+      "step": 45730
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.9796598281606176e-06,
+      "loss": 1.3412,
+      "step": 45740
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.9779063650710155e-06,
+      "loss": 1.6343,
+      "step": 45750
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 1.9761529019814134e-06,
+      "loss": 1.4386,
+      "step": 45760
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9743994388918113e-06,
+      "loss": 1.602,
+      "step": 45770
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9726459758022096e-06,
+      "loss": 1.3918,
+      "step": 45780
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9708925127126075e-06,
+      "loss": 1.3973,
+      "step": 45790
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9691390496230054e-06,
+      "loss": 1.4622,
+      "step": 45800
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9673855865334033e-06,
+      "loss": 1.4496,
+      "step": 45810
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 1.9656321234438017e-06,
+      "loss": 1.5255,
+      "step": 45820
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.9638786603542e-06,
+      "loss": 1.3913,
+      "step": 45830
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.962125197264598e-06,
+      "loss": 1.3727,
+      "step": 45840
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.960371734174996e-06,
+      "loss": 1.3364,
+      "step": 45850
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.9586182710853937e-06,
+      "loss": 1.5362,
+      "step": 45860
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.956864807995792e-06,
+      "loss": 1.4976,
+      "step": 45870
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 1.95511134490619e-06,
+      "loss": 1.5606,
+      "step": 45880
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.953357881816588e-06,
+      "loss": 1.3358,
+      "step": 45890
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.9516044187269857e-06,
+      "loss": 1.4415,
+      "step": 45900
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.949850955637384e-06,
+      "loss": 1.306,
+      "step": 45910
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.948097492547782e-06,
+      "loss": 1.5039,
+      "step": 45920
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 1.94634402945818e-06,
+      "loss": 1.6321,
+      "step": 45930
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.944590566368578e-06,
+      "loss": 1.444,
+      "step": 45940
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.942837103278976e-06,
+      "loss": 1.3642,
+      "step": 45950
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.9410836401893744e-06,
+      "loss": 1.4747,
+      "step": 45960
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.9393301770997723e-06,
+      "loss": 1.5181,
+      "step": 45970
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.9375767140101702e-06,
+      "loss": 1.4822,
+      "step": 45980
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 1.935823250920568e-06,
+      "loss": 1.618,
+      "step": 45990
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.9340697878309665e-06,
+      "loss": 1.3645,
+      "step": 46000
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.9323163247413644e-06,
+      "loss": 1.6331,
+      "step": 46010
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.9305628616517623e-06,
+      "loss": 1.4816,
+      "step": 46020
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.92880939856216e-06,
+      "loss": 1.4917,
+      "step": 46030
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.9270559354725585e-06,
+      "loss": 1.4765,
+      "step": 46040
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 1.9253024723829564e-06,
+      "loss": 1.3549,
+      "step": 46050
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.9235490092933547e-06,
+      "loss": 1.4858,
+      "step": 46060
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.9217955462037526e-06,
+      "loss": 1.461,
+      "step": 46070
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.9200420831141505e-06,
+      "loss": 1.3731,
+      "step": 46080
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.918288620024549e-06,
+      "loss": 1.3792,
+      "step": 46090
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 1.9165351569349467e-06,
+      "loss": 1.4147,
+      "step": 46100
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.9147816938453446e-06,
+      "loss": 1.465,
+      "step": 46110
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.9130282307557426e-06,
+      "loss": 1.5124,
+      "step": 46120
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.911274767666141e-06,
+      "loss": 1.4136,
+      "step": 46130
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.9095213045765388e-06,
+      "loss": 1.5035,
+      "step": 46140
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.9077678414869367e-06,
+      "loss": 1.4985,
+      "step": 46150
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 1.9060143783973348e-06,
+      "loss": 1.3129,
+      "step": 46160
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.904260915307733e-06,
+      "loss": 1.5307,
+      "step": 46170
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.902507452218131e-06,
+      "loss": 1.4356,
+      "step": 46180
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.900753989128529e-06,
+      "loss": 1.4966,
+      "step": 46190
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.8990005260389273e-06,
+      "loss": 1.4613,
+      "step": 46200
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.897247062949325e-06,
+      "loss": 1.399,
+      "step": 46210
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 1.8954935998597233e-06,
+      "loss": 1.3601,
+      "step": 46220
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.8937401367701212e-06,
+      "loss": 1.7016,
+      "step": 46230
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.8919866736805193e-06,
+      "loss": 1.5832,
+      "step": 46240
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.8902332105909172e-06,
+      "loss": 1.5418,
+      "step": 46250
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.8884797475013153e-06,
+      "loss": 1.3975,
+      "step": 46260
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 1.8867262844117132e-06,
+      "loss": 1.3566,
+      "step": 46270
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8849728213221113e-06,
+      "loss": 1.4043,
+      "step": 46280
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8832193582325092e-06,
+      "loss": 1.4402,
+      "step": 46290
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8814658951429073e-06,
+      "loss": 1.3396,
+      "step": 46300
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8797124320533057e-06,
+      "loss": 1.5212,
+      "step": 46310
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8779589689637033e-06,
+      "loss": 1.4478,
+      "step": 46320
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 1.8762055058741017e-06,
+      "loss": 1.3802,
+      "step": 46330
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8744520427844996e-06,
+      "loss": 1.4371,
+      "step": 46340
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8726985796948977e-06,
+      "loss": 1.3834,
+      "step": 46350
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8709451166052956e-06,
+      "loss": 1.3951,
+      "step": 46360
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8691916535156937e-06,
+      "loss": 1.4839,
+      "step": 46370
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8674381904260916e-06,
+      "loss": 1.4539,
+      "step": 46380
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 1.8656847273364897e-06,
+      "loss": 1.4942,
+      "step": 46390
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.8639312642468876e-06,
+      "loss": 1.499,
+      "step": 46400
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.8621778011572857e-06,
+      "loss": 1.4315,
+      "step": 46410
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.860424338067684e-06,
+      "loss": 1.6094,
+      "step": 46420
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.8586708749780818e-06,
+      "loss": 1.4201,
+      "step": 46430
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.85691741188848e-06,
+      "loss": 1.5118,
+      "step": 46440
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 1.855163948798878e-06,
+      "loss": 1.5381,
+      "step": 46450
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.853410485709276e-06,
+      "loss": 1.4762,
+      "step": 46460
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.851657022619674e-06,
+      "loss": 1.3231,
+      "step": 46470
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.8499035595300721e-06,
+      "loss": 1.4583,
+      "step": 46480
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.84815009644047e-06,
+      "loss": 1.5116,
+      "step": 46490
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 1.8463966333508681e-06,
+      "loss": 1.5251,
+      "step": 46500
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.844643170261266e-06,
+      "loss": 1.4933,
+      "step": 46510
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.8428897071716641e-06,
+      "loss": 1.505,
+      "step": 46520
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.841136244082062e-06,
+      "loss": 1.3232,
+      "step": 46530
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.8393827809924602e-06,
+      "loss": 1.3944,
+      "step": 46540
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.8376293179028585e-06,
+      "loss": 1.4396,
+      "step": 46550
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 1.8358758548132564e-06,
+      "loss": 1.38,
+      "step": 46560
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8341223917236545e-06,
+      "loss": 1.5006,
+      "step": 46570
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8323689286340524e-06,
+      "loss": 1.4151,
+      "step": 46580
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8306154655444505e-06,
+      "loss": 1.4619,
+      "step": 46590
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8288620024548484e-06,
+      "loss": 1.3839,
+      "step": 46600
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8271085393652465e-06,
+      "loss": 1.652,
+      "step": 46610
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 1.8253550762756444e-06,
+      "loss": 1.3731,
+      "step": 46620
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.8236016131860426e-06,
+      "loss": 1.3175,
+      "step": 46630
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.8218481500964405e-06,
+      "loss": 1.5403,
+      "step": 46640
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.8200946870068386e-06,
+      "loss": 1.4675,
+      "step": 46650
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.8183412239172365e-06,
+      "loss": 1.6359,
+      "step": 46660
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 1.8165877608276348e-06,
+      "loss": 1.2527,
+      "step": 46670
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.814834297738033e-06,
+      "loss": 1.4204,
+      "step": 46680
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.8130808346484308e-06,
+      "loss": 1.4409,
+      "step": 46690
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.811327371558829e-06,
+      "loss": 1.3131,
+      "step": 46700
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.8095739084692268e-06,
+      "loss": 1.5501,
+      "step": 46710
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.807820445379625e-06,
+      "loss": 1.4313,
+      "step": 46720
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 1.8060669822900228e-06,
+      "loss": 1.3605,
+      "step": 46730
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.804313519200421e-06,
+      "loss": 1.505,
+      "step": 46740
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.8025600561108189e-06,
+      "loss": 1.4986,
+      "step": 46750
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.800806593021217e-06,
+      "loss": 1.5104,
+      "step": 46760
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.7990531299316149e-06,
+      "loss": 1.7373,
+      "step": 46770
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.7972996668420132e-06,
+      "loss": 1.3865,
+      "step": 46780
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 1.7955462037524113e-06,
+      "loss": 1.6687,
+      "step": 46790
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7937927406628092e-06,
+      "loss": 1.4522,
+      "step": 46800
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7920392775732073e-06,
+      "loss": 1.4778,
+      "step": 46810
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7902858144836052e-06,
+      "loss": 1.4407,
+      "step": 46820
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7885323513940034e-06,
+      "loss": 1.4055,
+      "step": 46830
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7867788883044013e-06,
+      "loss": 1.403,
+      "step": 46840
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 1.7850254252147994e-06,
+      "loss": 1.4194,
+      "step": 46850
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.7832719621251973e-06,
+      "loss": 1.3448,
+      "step": 46860
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.7815184990355954e-06,
+      "loss": 1.2657,
+      "step": 46870
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.7797650359459933e-06,
+      "loss": 1.4486,
+      "step": 46880
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.7780115728563916e-06,
+      "loss": 1.4208,
+      "step": 46890
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 1.7762581097667895e-06,
+      "loss": 1.4335,
+      "step": 46900
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7745046466771876e-06,
+      "loss": 1.5151,
+      "step": 46910
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7727511835875857e-06,
+      "loss": 1.3574,
+      "step": 46920
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7709977204979836e-06,
+      "loss": 1.3833,
+      "step": 46930
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7692442574083818e-06,
+      "loss": 1.5581,
+      "step": 46940
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7674907943187797e-06,
+      "loss": 1.5167,
+      "step": 46950
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 1.7657373312291778e-06,
+      "loss": 1.5038,
+      "step": 46960
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.7639838681395757e-06,
+      "loss": 1.43,
+      "step": 46970
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.7622304050499738e-06,
+      "loss": 1.3632,
+      "step": 46980
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.7604769419603717e-06,
+      "loss": 1.7489,
+      "step": 46990
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.75872347887077e-06,
+      "loss": 1.5665,
+      "step": 47000
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.756970015781168e-06,
+      "loss": 1.2171,
+      "step": 47010
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 1.755216552691566e-06,
+      "loss": 1.4937,
+      "step": 47020
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.7534630896019641e-06,
+      "loss": 1.357,
+      "step": 47030
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.751709626512362e-06,
+      "loss": 1.5017,
+      "step": 47040
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.7499561634227602e-06,
+      "loss": 1.3245,
+      "step": 47050
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.748202700333158e-06,
+      "loss": 1.3932,
+      "step": 47060
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 1.7464492372435562e-06,
+      "loss": 1.5549,
+      "step": 47070
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.744695774153954e-06,
+      "loss": 1.2859,
+      "step": 47080
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.7429423110643522e-06,
+      "loss": 1.4541,
+      "step": 47090
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.74118884797475e-06,
+      "loss": 1.4476,
+      "step": 47100
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.7394353848851484e-06,
+      "loss": 1.4829,
+      "step": 47110
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.7376819217955463e-06,
+      "loss": 1.5556,
+      "step": 47120
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 1.7359284587059444e-06,
+      "loss": 1.368,
+      "step": 47130
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7341749956163423e-06,
+      "loss": 1.5331,
+      "step": 47140
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7324215325267405e-06,
+      "loss": 1.4156,
+      "step": 47150
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7306680694371386e-06,
+      "loss": 1.3404,
+      "step": 47160
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7289146063475365e-06,
+      "loss": 1.4797,
+      "step": 47170
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7271611432579346e-06,
+      "loss": 1.4274,
+      "step": 47180
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 1.7254076801683325e-06,
+      "loss": 1.4549,
+      "step": 47190
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.7236542170787306e-06,
+      "loss": 1.1848,
+      "step": 47200
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.7219007539891285e-06,
+      "loss": 1.533,
+      "step": 47210
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.7201472908995268e-06,
+      "loss": 1.6301,
+      "step": 47220
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.7183938278099247e-06,
+      "loss": 1.4027,
+      "step": 47230
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 1.7166403647203228e-06,
+      "loss": 1.6526,
+      "step": 47240
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.7148869016307208e-06,
+      "loss": 1.6092,
+      "step": 47250
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.7131334385411189e-06,
+      "loss": 1.4516,
+      "step": 47260
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.711379975451517e-06,
+      "loss": 1.6498,
+      "step": 47270
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.7096265123619149e-06,
+      "loss": 1.4107,
+      "step": 47280
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.707873049272313e-06,
+      "loss": 1.4951,
+      "step": 47290
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 1.706119586182711e-06,
+      "loss": 1.3797,
+      "step": 47300
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.704366123093109e-06,
+      "loss": 1.5269,
+      "step": 47310
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.702612660003507e-06,
+      "loss": 1.3996,
+      "step": 47320
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.7008591969139052e-06,
+      "loss": 1.3501,
+      "step": 47330
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.6991057338243031e-06,
+      "loss": 1.5262,
+      "step": 47340
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.6973522707347013e-06,
+      "loss": 1.4178,
+      "step": 47350
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 1.6955988076450992e-06,
+      "loss": 1.47,
+      "step": 47360
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6938453445554973e-06,
+      "loss": 1.4162,
+      "step": 47370
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6920918814658952e-06,
+      "loss": 1.478,
+      "step": 47380
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6903384183762933e-06,
+      "loss": 1.2781,
+      "step": 47390
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6885849552866914e-06,
+      "loss": 1.5187,
+      "step": 47400
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6868314921970893e-06,
+      "loss": 1.6421,
+      "step": 47410
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 1.6850780291074874e-06,
+      "loss": 1.4628,
+      "step": 47420
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6833245660178853e-06,
+      "loss": 1.4142,
+      "step": 47430
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6815711029282836e-06,
+      "loss": 1.6091,
+      "step": 47440
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6798176398386815e-06,
+      "loss": 1.3373,
+      "step": 47450
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6780641767490797e-06,
+      "loss": 1.3926,
+      "step": 47460
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 1.6763107136594776e-06,
+      "loss": 1.4315,
+      "step": 47470
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.6745572505698757e-06,
+      "loss": 1.3483,
+      "step": 47480
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.6728037874802736e-06,
+      "loss": 1.3551,
+      "step": 47490
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.6710503243906717e-06,
+      "loss": 1.4971,
+      "step": 47500
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.6692968613010698e-06,
+      "loss": 1.4505,
+      "step": 47510
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.6675433982114677e-06,
+      "loss": 1.4841,
+      "step": 47520
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 1.665789935121866e-06,
+      "loss": 1.4876,
+      "step": 47530
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.6640364720322637e-06,
+      "loss": 1.3759,
+      "step": 47540
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.662283008942662e-06,
+      "loss": 1.3261,
+      "step": 47550
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.66052954585306e-06,
+      "loss": 1.4182,
+      "step": 47560
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.658776082763458e-06,
+      "loss": 1.5856,
+      "step": 47570
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.657022619673856e-06,
+      "loss": 1.4331,
+      "step": 47580
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 1.655269156584254e-06,
+      "loss": 1.2598,
+      "step": 47590
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.653515693494652e-06,
+      "loss": 1.4571,
+      "step": 47600
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.65176223040505e-06,
+      "loss": 1.4874,
+      "step": 47610
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.650008767315448e-06,
+      "loss": 1.4865,
+      "step": 47620
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.6482553042258461e-06,
+      "loss": 1.2916,
+      "step": 47630
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 1.6465018411362444e-06,
+      "loss": 1.5504,
+      "step": 47640
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6447483780466421e-06,
+      "loss": 1.3657,
+      "step": 47650
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6429949149570405e-06,
+      "loss": 1.4793,
+      "step": 47660
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6412414518674384e-06,
+      "loss": 1.4509,
+      "step": 47670
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6394879887778365e-06,
+      "loss": 1.4686,
+      "step": 47680
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6377345256882344e-06,
+      "loss": 1.4491,
+      "step": 47690
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 1.6359810625986325e-06,
+      "loss": 1.3728,
+      "step": 47700
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6342275995090304e-06,
+      "loss": 1.3749,
+      "step": 47710
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6324741364194285e-06,
+      "loss": 1.6916,
+      "step": 47720
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6307206733298264e-06,
+      "loss": 1.4461,
+      "step": 47730
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6289672102402245e-06,
+      "loss": 1.4155,
+      "step": 47740
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6272137471506229e-06,
+      "loss": 1.5299,
+      "step": 47750
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 1.6254602840610205e-06,
+      "loss": 1.462,
+      "step": 47760
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.6237068209714189e-06,
+      "loss": 1.2761,
+      "step": 47770
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.6219533578818168e-06,
+      "loss": 1.5732,
+      "step": 47780
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.6201998947922149e-06,
+      "loss": 1.4188,
+      "step": 47790
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.6184464317026128e-06,
+      "loss": 1.4053,
+      "step": 47800
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 1.616692968613011e-06,
+      "loss": 1.3356,
+      "step": 47810
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.6149395055234088e-06,
+      "loss": 1.4532,
+      "step": 47820
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.613186042433807e-06,
+      "loss": 1.401,
+      "step": 47830
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.6114325793442048e-06,
+      "loss": 1.2812,
+      "step": 47840
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.609679116254603e-06,
+      "loss": 1.6089,
+      "step": 47850
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.6079256531650008e-06,
+      "loss": 1.5092,
+      "step": 47860
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 1.606172190075399e-06,
+      "loss": 1.3702,
+      "step": 47870
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.6044187269857973e-06,
+      "loss": 1.4029,
+      "step": 47880
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.6026652638961952e-06,
+      "loss": 1.4166,
+      "step": 47890
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.6009118008065933e-06,
+      "loss": 1.4061,
+      "step": 47900
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.5991583377169912e-06,
+      "loss": 1.3247,
+      "step": 47910
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.5974048746273893e-06,
+      "loss": 1.5097,
+      "step": 47920
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 1.5956514115377872e-06,
+      "loss": 1.5297,
+      "step": 47930
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5938979484481853e-06,
+      "loss": 1.514,
+      "step": 47940
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5921444853585832e-06,
+      "loss": 1.3886,
+      "step": 47950
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5903910222689813e-06,
+      "loss": 1.4293,
+      "step": 47960
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5886375591793792e-06,
+      "loss": 1.6194,
+      "step": 47970
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5868840960897774e-06,
+      "loss": 1.7169,
+      "step": 47980
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 1.5851306330001757e-06,
+      "loss": 1.2988,
+      "step": 47990
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.5833771699105736e-06,
+      "loss": 1.4239,
+      "step": 48000
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.5816237068209717e-06,
+      "loss": 1.3736,
+      "step": 48010
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.5798702437313696e-06,
+      "loss": 1.6756,
+      "step": 48020
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.5781167806417677e-06,
+      "loss": 1.5263,
+      "step": 48030
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 1.5763633175521656e-06,
+      "loss": 1.294,
+      "step": 48040
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5746098544625637e-06,
+      "loss": 1.2938,
+      "step": 48050
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5728563913729616e-06,
+      "loss": 1.5707,
+      "step": 48060
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5711029282833597e-06,
+      "loss": 1.4424,
+      "step": 48070
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5693494651937576e-06,
+      "loss": 1.3932,
+      "step": 48080
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5675960021041558e-06,
+      "loss": 1.4747,
+      "step": 48090
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 1.5658425390145537e-06,
+      "loss": 1.344,
+      "step": 48100
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.564089075924952e-06,
+      "loss": 1.4679,
+      "step": 48110
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.56233561283535e-06,
+      "loss": 1.6455,
+      "step": 48120
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.560582149745748e-06,
+      "loss": 1.4002,
+      "step": 48130
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.5588286866561461e-06,
+      "loss": 1.4375,
+      "step": 48140
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.557075223566544e-06,
+      "loss": 1.5099,
+      "step": 48150
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 1.5553217604769421e-06,
+      "loss": 1.3968,
+      "step": 48160
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.55356829738734e-06,
+      "loss": 1.3439,
+      "step": 48170
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.5518148342977382e-06,
+      "loss": 1.4123,
+      "step": 48180
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.550061371208136e-06,
+      "loss": 1.4942,
+      "step": 48190
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.5483079081185342e-06,
+      "loss": 1.3266,
+      "step": 48200
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 1.546554445028932e-06,
+      "loss": 1.3897,
+      "step": 48210
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5448009819393304e-06,
+      "loss": 1.5789,
+      "step": 48220
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5430475188497285e-06,
+      "loss": 1.5788,
+      "step": 48230
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5412940557601264e-06,
+      "loss": 1.4339,
+      "step": 48240
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5395405926705245e-06,
+      "loss": 1.4856,
+      "step": 48250
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5377871295809224e-06,
+      "loss": 1.3514,
+      "step": 48260
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 1.5360336664913205e-06,
+      "loss": 1.5582,
+      "step": 48270
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5342802034017184e-06,
+      "loss": 1.2978,
+      "step": 48280
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5325267403121166e-06,
+      "loss": 1.331,
+      "step": 48290
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5307732772225145e-06,
+      "loss": 1.4059,
+      "step": 48300
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5290198141329126e-06,
+      "loss": 1.5109,
+      "step": 48310
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5272663510433105e-06,
+      "loss": 1.437,
+      "step": 48320
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 1.5255128879537088e-06,
+      "loss": 1.3883,
+      "step": 48330
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.5237594248641067e-06,
+      "loss": 1.4585,
+      "step": 48340
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.5220059617745048e-06,
+      "loss": 1.2876,
+      "step": 48350
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.520252498684903e-06,
+      "loss": 1.3343,
+      "step": 48360
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.5184990355953008e-06,
+      "loss": 1.4454,
+      "step": 48370
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 1.516745572505699e-06,
+      "loss": 1.5872,
+      "step": 48380
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.5149921094160969e-06,
+      "loss": 1.3226,
+      "step": 48390
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.513238646326495e-06,
+      "loss": 1.4671,
+      "step": 48400
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.5114851832368929e-06,
+      "loss": 1.5055,
+      "step": 48410
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.509731720147291e-06,
+      "loss": 1.4063,
+      "step": 48420
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.5079782570576889e-06,
+      "loss": 1.3842,
+      "step": 48430
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 1.5062247939680872e-06,
+      "loss": 1.5357,
+      "step": 48440
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.5044713308784851e-06,
+      "loss": 1.4414,
+      "step": 48450
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.5027178677888832e-06,
+      "loss": 1.4845,
+      "step": 48460
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.5009644046992813e-06,
+      "loss": 1.4398,
+      "step": 48470
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.4992109416096792e-06,
+      "loss": 1.3187,
+      "step": 48480
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.4974574785200774e-06,
+      "loss": 1.3055,
+      "step": 48490
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 1.4957040154304753e-06,
+      "loss": 1.4784,
+      "step": 48500
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4939505523408734e-06,
+      "loss": 1.4597,
+      "step": 48510
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4921970892512713e-06,
+      "loss": 1.427,
+      "step": 48520
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4904436261616694e-06,
+      "loss": 1.2338,
+      "step": 48530
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4886901630720673e-06,
+      "loss": 1.3844,
+      "step": 48540
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4869366999824656e-06,
+      "loss": 1.3688,
+      "step": 48550
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 1.4851832368928635e-06,
+      "loss": 1.4785,
+      "step": 48560
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4834297738032616e-06,
+      "loss": 1.4022,
+      "step": 48570
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4816763107136595e-06,
+      "loss": 1.4346,
+      "step": 48580
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4799228476240576e-06,
+      "loss": 1.3965,
+      "step": 48590
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4781693845344558e-06,
+      "loss": 1.4147,
+      "step": 48600
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 1.4764159214448537e-06,
+      "loss": 1.4612,
+      "step": 48610
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.4746624583552518e-06,
+      "loss": 1.3742,
+      "step": 48620
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.4729089952656497e-06,
+      "loss": 1.4324,
+      "step": 48630
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.4711555321760478e-06,
+      "loss": 1.5955,
+      "step": 48640
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.4694020690864457e-06,
+      "loss": 1.3771,
+      "step": 48650
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.467648605996844e-06,
+      "loss": 1.4566,
+      "step": 48660
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 1.465895142907242e-06,
+      "loss": 1.5915,
+      "step": 48670
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.46414167981764e-06,
+      "loss": 1.2732,
+      "step": 48680
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.462388216728038e-06,
+      "loss": 1.5604,
+      "step": 48690
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.460634753638436e-06,
+      "loss": 1.4856,
+      "step": 48700
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.458881290548834e-06,
+      "loss": 1.5465,
+      "step": 48710
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.457127827459232e-06,
+      "loss": 1.3831,
+      "step": 48720
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 1.4553743643696302e-06,
+      "loss": 1.3508,
+      "step": 48730
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.453620901280028e-06,
+      "loss": 1.3283,
+      "step": 48740
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.4518674381904262e-06,
+      "loss": 1.5597,
+      "step": 48750
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.450113975100824e-06,
+      "loss": 1.5133,
+      "step": 48760
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.4483605120112224e-06,
+      "loss": 1.4073,
+      "step": 48770
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 1.4466070489216203e-06,
+      "loss": 1.5154,
+      "step": 48780
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4448535858320184e-06,
+      "loss": 1.3012,
+      "step": 48790
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4431001227424163e-06,
+      "loss": 1.4267,
+      "step": 48800
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4413466596528145e-06,
+      "loss": 1.2951,
+      "step": 48810
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4395931965632124e-06,
+      "loss": 1.5447,
+      "step": 48820
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4378397334736105e-06,
+      "loss": 1.4013,
+      "step": 48830
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 1.4360862703840086e-06,
+      "loss": 1.3769,
+      "step": 48840
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4343328072944065e-06,
+      "loss": 1.5367,
+      "step": 48850
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4325793442048046e-06,
+      "loss": 1.4682,
+      "step": 48860
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4308258811152025e-06,
+      "loss": 1.3708,
+      "step": 48870
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4290724180256008e-06,
+      "loss": 1.4224,
+      "step": 48880
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4273189549359987e-06,
+      "loss": 1.3239,
+      "step": 48890
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 1.4255654918463969e-06,
+      "loss": 1.3749,
+      "step": 48900
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.4238120287567948e-06,
+      "loss": 1.4676,
+      "step": 48910
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.4220585656671929e-06,
+      "loss": 1.4659,
+      "step": 48920
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.4203051025775908e-06,
+      "loss": 1.7267,
+      "step": 48930
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.4185516394879889e-06,
+      "loss": 1.3131,
+      "step": 48940
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.4167981763983868e-06,
+      "loss": 1.4098,
+      "step": 48950
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 1.415044713308785e-06,
+      "loss": 1.5169,
+      "step": 48960
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.413291250219183e-06,
+      "loss": 1.3772,
+      "step": 48970
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.411537787129581e-06,
+      "loss": 1.3971,
+      "step": 48980
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.4097843240399792e-06,
+      "loss": 1.4854,
+      "step": 48990
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.4080308609503771e-06,
+      "loss": 1.4256,
+      "step": 49000
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 1.4062773978607753e-06,
+      "loss": 1.4317,
+      "step": 49010
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.4045239347711732e-06,
+      "loss": 1.3745,
+      "step": 49020
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.4027704716815713e-06,
+      "loss": 1.5857,
+      "step": 49030
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.4010170085919692e-06,
+      "loss": 1.4623,
+      "step": 49040
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.3992635455023673e-06,
+      "loss": 1.5125,
+      "step": 49050
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.3975100824127652e-06,
+      "loss": 1.5319,
+      "step": 49060
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 1.3957566193231633e-06,
+      "loss": 1.3634,
+      "step": 49070
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3940031562335614e-06,
+      "loss": 1.409,
+      "step": 49080
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3922496931439593e-06,
+      "loss": 1.4702,
+      "step": 49090
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3904962300543577e-06,
+      "loss": 1.4949,
+      "step": 49100
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3887427669647556e-06,
+      "loss": 1.5521,
+      "step": 49110
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3869893038751537e-06,
+      "loss": 1.454,
+      "step": 49120
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 1.3852358407855516e-06,
+      "loss": 1.4697,
+      "step": 49130
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.3834823776959497e-06,
+      "loss": 1.5634,
+      "step": 49140
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.3817289146063476e-06,
+      "loss": 1.3743,
+      "step": 49150
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.3799754515167457e-06,
+      "loss": 1.3302,
+      "step": 49160
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.3782219884271436e-06,
+      "loss": 1.5895,
+      "step": 49170
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 1.3764685253375417e-06,
+      "loss": 1.4191,
+      "step": 49180
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.3747150622479396e-06,
+      "loss": 1.3711,
+      "step": 49190
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.3729615991583377e-06,
+      "loss": 1.5518,
+      "step": 49200
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.371208136068736e-06,
+      "loss": 1.5532,
+      "step": 49210
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.369454672979134e-06,
+      "loss": 1.5898,
+      "step": 49220
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.367701209889532e-06,
+      "loss": 1.4733,
+      "step": 49230
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 1.36594774679993e-06,
+      "loss": 1.4538,
+      "step": 49240
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.364194283710328e-06,
+      "loss": 1.5433,
+      "step": 49250
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.362440820620726e-06,
+      "loss": 1.4076,
+      "step": 49260
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.3606873575311241e-06,
+      "loss": 1.5641,
+      "step": 49270
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.358933894441522e-06,
+      "loss": 1.5444,
+      "step": 49280
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.3571804313519201e-06,
+      "loss": 1.5682,
+      "step": 49290
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 1.355426968262318e-06,
+      "loss": 1.506,
+      "step": 49300
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.3536735051727161e-06,
+      "loss": 1.5099,
+      "step": 49310
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.3519200420831145e-06,
+      "loss": 1.6332,
+      "step": 49320
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.3501665789935124e-06,
+      "loss": 1.6256,
+      "step": 49330
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.3484131159039105e-06,
+      "loss": 1.4089,
+      "step": 49340
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 1.3466596528143084e-06,
+      "loss": 1.429,
+      "step": 49350
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3449061897247065e-06,
+      "loss": 1.5532,
+      "step": 49360
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3431527266351044e-06,
+      "loss": 1.4841,
+      "step": 49370
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3413992635455025e-06,
+      "loss": 1.3618,
+      "step": 49380
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3396458004559004e-06,
+      "loss": 1.4546,
+      "step": 49390
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3378923373662985e-06,
+      "loss": 1.2958,
+      "step": 49400
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 1.3361388742766964e-06,
+      "loss": 1.4308,
+      "step": 49410
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.3343854111870945e-06,
+      "loss": 1.5935,
+      "step": 49420
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.3326319480974924e-06,
+      "loss": 1.3778,
+      "step": 49430
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.3308784850078908e-06,
+      "loss": 1.5057,
+      "step": 49440
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.3291250219182889e-06,
+      "loss": 1.4396,
+      "step": 49450
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.3273715588286868e-06,
+      "loss": 1.5275,
+      "step": 49460
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 1.325618095739085e-06,
+      "loss": 1.4422,
+      "step": 49470
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.3238646326494828e-06,
+      "loss": 1.4518,
+      "step": 49480
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.322111169559881e-06,
+      "loss": 1.5518,
+      "step": 49490
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.3203577064702788e-06,
+      "loss": 1.5081,
+      "step": 49500
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.318604243380677e-06,
+      "loss": 1.4702,
+      "step": 49510
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.3168507802910748e-06,
+      "loss": 1.3484,
+      "step": 49520
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 1.315097317201473e-06,
+      "loss": 1.3567,
+      "step": 49530
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.3133438541118709e-06,
+      "loss": 1.7638,
+      "step": 49540
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.3115903910222692e-06,
+      "loss": 1.32,
+      "step": 49550
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.3098369279326673e-06,
+      "loss": 1.4364,
+      "step": 49560
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.3080834648430652e-06,
+      "loss": 1.5511,
+      "step": 49570
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 1.3063300017534633e-06,
+      "loss": 1.4711,
+      "step": 49580
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.3045765386638612e-06,
+      "loss": 1.4676,
+      "step": 49590
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.3028230755742593e-06,
+      "loss": 1.315,
+      "step": 49600
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.3010696124846572e-06,
+      "loss": 1.377,
+      "step": 49610
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.2993161493950553e-06,
+      "loss": 1.3766,
+      "step": 49620
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.2975626863054532e-06,
+      "loss": 1.5639,
+      "step": 49630
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 1.2958092232158514e-06,
+      "loss": 1.4508,
+      "step": 49640
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2940557601262493e-06,
+      "loss": 1.6528,
+      "step": 49650
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2923022970366476e-06,
+      "loss": 1.5075,
+      "step": 49660
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2905488339470455e-06,
+      "loss": 1.4773,
+      "step": 49670
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2887953708574436e-06,
+      "loss": 1.4431,
+      "step": 49680
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2870419077678417e-06,
+      "loss": 1.6304,
+      "step": 49690
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 1.2852884446782396e-06,
+      "loss": 1.3397,
+      "step": 49700
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.2835349815886377e-06,
+      "loss": 1.3487,
+      "step": 49710
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.2817815184990356e-06,
+      "loss": 1.5255,
+      "step": 49720
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.2800280554094338e-06,
+      "loss": 1.4704,
+      "step": 49730
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.2782745923198317e-06,
+      "loss": 1.4662,
+      "step": 49740
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 1.2765211292302298e-06,
+      "loss": 1.484,
+      "step": 49750
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.2747676661406277e-06,
+      "loss": 1.3274,
+      "step": 49760
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.273014203051026e-06,
+      "loss": 1.4089,
+      "step": 49770
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.271260739961424e-06,
+      "loss": 1.4836,
+      "step": 49780
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.269507276871822e-06,
+      "loss": 1.4677,
+      "step": 49790
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.2677538137822201e-06,
+      "loss": 1.2653,
+      "step": 49800
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 1.266000350692618e-06,
+      "loss": 1.4616,
+      "step": 49810
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.2642468876030161e-06,
+      "loss": 1.3776,
+      "step": 49820
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.262493424513414e-06,
+      "loss": 1.3353,
+      "step": 49830
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.2607399614238122e-06,
+      "loss": 1.3948,
+      "step": 49840
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.25898649833421e-06,
+      "loss": 1.3717,
+      "step": 49850
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.2572330352446082e-06,
+      "loss": 1.3115,
+      "step": 49860
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 1.255479572155006e-06,
+      "loss": 1.3278,
+      "step": 49870
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.2537261090654044e-06,
+      "loss": 1.421,
+      "step": 49880
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.2519726459758023e-06,
+      "loss": 1.5282,
+      "step": 49890
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.2502191828862004e-06,
+      "loss": 1.3656,
+      "step": 49900
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.2484657197965983e-06,
+      "loss": 1.4303,
+      "step": 49910
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 1.2467122567069964e-06,
+      "loss": 1.4583,
+      "step": 49920
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2449587936173943e-06,
+      "loss": 1.6769,
+      "step": 49930
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2432053305277924e-06,
+      "loss": 1.4496,
+      "step": 49940
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2414518674381906e-06,
+      "loss": 1.4674,
+      "step": 49950
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2396984043485885e-06,
+      "loss": 1.4344,
+      "step": 49960
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2379449412589866e-06,
+      "loss": 1.3543,
+      "step": 49970
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 1.2361914781693847e-06,
+      "loss": 1.5587,
+      "step": 49980
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2344380150797828e-06,
+      "loss": 1.3632,
+      "step": 49990
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2326845519901807e-06,
+      "loss": 1.5095,
+      "step": 50000
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2309310889005788e-06,
+      "loss": 1.503,
+      "step": 50010
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2291776258109767e-06,
+      "loss": 1.4377,
+      "step": 50020
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2274241627213748e-06,
+      "loss": 1.5475,
+      "step": 50030
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 1.2256706996317727e-06,
+      "loss": 1.566,
+      "step": 50040
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.2239172365421709e-06,
+      "loss": 1.4661,
+      "step": 50050
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.222163773452569e-06,
+      "loss": 1.4887,
+      "step": 50060
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.2204103103629669e-06,
+      "loss": 1.5409,
+      "step": 50070
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.218656847273365e-06,
+      "loss": 1.3737,
+      "step": 50080
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.2169033841837629e-06,
+      "loss": 1.4146,
+      "step": 50090
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 1.2151499210941612e-06,
+      "loss": 1.4767,
+      "step": 50100
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.2133964580045591e-06,
+      "loss": 1.3393,
+      "step": 50110
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.2116429949149572e-06,
+      "loss": 1.5742,
+      "step": 50120
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.2098895318253551e-06,
+      "loss": 1.313,
+      "step": 50130
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.2081360687357532e-06,
+      "loss": 1.5464,
+      "step": 50140
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 1.2063826056461514e-06,
+      "loss": 1.4384,
+      "step": 50150
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.2046291425565493e-06,
+      "loss": 1.5065,
+      "step": 50160
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.2028756794669474e-06,
+      "loss": 1.5277,
+      "step": 50170
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.2011222163773453e-06,
+      "loss": 1.2195,
+      "step": 50180
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.1993687532877434e-06,
+      "loss": 1.2903,
+      "step": 50190
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.1976152901981413e-06,
+      "loss": 1.3813,
+      "step": 50200
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 1.1958618271085394e-06,
+      "loss": 1.4471,
+      "step": 50210
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1941083640189375e-06,
+      "loss": 1.5249,
+      "step": 50220
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1923549009293356e-06,
+      "loss": 1.4207,
+      "step": 50230
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1906014378397335e-06,
+      "loss": 1.4853,
+      "step": 50240
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1888479747501317e-06,
+      "loss": 1.4097,
+      "step": 50250
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1870945116605298e-06,
+      "loss": 1.4308,
+      "step": 50260
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 1.1853410485709277e-06,
+      "loss": 1.4222,
+      "step": 50270
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1835875854813258e-06,
+      "loss": 1.5016,
+      "step": 50280
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1818341223917237e-06,
+      "loss": 1.4203,
+      "step": 50290
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1800806593021218e-06,
+      "loss": 1.4237,
+      "step": 50300
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1783271962125197e-06,
+      "loss": 1.8307,
+      "step": 50310
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 1.1765737331229178e-06,
+      "loss": 1.3508,
+      "step": 50320
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.174820270033316e-06,
+      "loss": 1.817,
+      "step": 50330
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.173066806943714e-06,
+      "loss": 1.269,
+      "step": 50340
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.171313343854112e-06,
+      "loss": 1.4444,
+      "step": 50350
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.16955988076451e-06,
+      "loss": 1.4957,
+      "step": 50360
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.1678064176749082e-06,
+      "loss": 1.5156,
+      "step": 50370
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 1.166052954585306e-06,
+      "loss": 1.3302,
+      "step": 50380
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.1642994914957042e-06,
+      "loss": 1.3901,
+      "step": 50390
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.162546028406102e-06,
+      "loss": 1.6357,
+      "step": 50400
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.1607925653165002e-06,
+      "loss": 1.5719,
+      "step": 50410
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.1590391022268981e-06,
+      "loss": 1.4533,
+      "step": 50420
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.1572856391372962e-06,
+      "loss": 1.4766,
+      "step": 50430
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 1.1555321760476943e-06,
+      "loss": 1.453,
+      "step": 50440
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1537787129580922e-06,
+      "loss": 1.3562,
+      "step": 50450
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1520252498684904e-06,
+      "loss": 1.4568,
+      "step": 50460
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1502717867788885e-06,
+      "loss": 1.5299,
+      "step": 50470
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1485183236892866e-06,
+      "loss": 1.4342,
+      "step": 50480
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1467648605996845e-06,
+      "loss": 1.4765,
+      "step": 50490
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 1.1450113975100826e-06,
+      "loss": 1.4454,
+      "step": 50500
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.1432579344204805e-06,
+      "loss": 1.4597,
+      "step": 50510
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.1415044713308786e-06,
+      "loss": 1.3547,
+      "step": 50520
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.1397510082412765e-06,
+      "loss": 1.3852,
+      "step": 50530
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.1379975451516746e-06,
+      "loss": 1.4386,
+      "step": 50540
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 1.1362440820620727e-06,
+      "loss": 1.3611,
+      "step": 50550
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.1344906189724706e-06,
+      "loss": 1.3651,
+      "step": 50560
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.1327371558828688e-06,
+      "loss": 1.3205,
+      "step": 50570
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.1309836927932669e-06,
+      "loss": 1.5234,
+      "step": 50580
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.129230229703665e-06,
+      "loss": 1.4074,
+      "step": 50590
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.1274767666140629e-06,
+      "loss": 1.5351,
+      "step": 50600
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 1.125723303524461e-06,
+      "loss": 1.4184,
+      "step": 50610
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.123969840434859e-06,
+      "loss": 1.5224,
+      "step": 50620
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.122216377345257e-06,
+      "loss": 1.4573,
+      "step": 50630
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.120462914255655e-06,
+      "loss": 1.8162,
+      "step": 50640
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.118709451166053e-06,
+      "loss": 1.3375,
+      "step": 50650
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.1169559880764512e-06,
+      "loss": 1.2747,
+      "step": 50660
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 1.115202524986849e-06,
+      "loss": 1.4523,
+      "step": 50670
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.1134490618972472e-06,
+      "loss": 1.5361,
+      "step": 50680
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.111695598807645e-06,
+      "loss": 1.3894,
+      "step": 50690
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.1099421357180432e-06,
+      "loss": 1.3916,
+      "step": 50700
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.1081886726284413e-06,
+      "loss": 1.4055,
+      "step": 50710
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 1.1064352095388394e-06,
+      "loss": 1.3932,
+      "step": 50720
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.1046817464492373e-06,
+      "loss": 1.3696,
+      "step": 50730
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.1029282833596354e-06,
+      "loss": 1.3602,
+      "step": 50740
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.1011748202700333e-06,
+      "loss": 1.3781,
+      "step": 50750
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.0994213571804314e-06,
+      "loss": 1.357,
+      "step": 50760
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.0976678940908296e-06,
+      "loss": 1.4829,
+      "step": 50770
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 1.0959144310012275e-06,
+      "loss": 1.4718,
+      "step": 50780
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0941609679116256e-06,
+      "loss": 1.541,
+      "step": 50790
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0924075048220235e-06,
+      "loss": 1.475,
+      "step": 50800
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0906540417324216e-06,
+      "loss": 1.4718,
+      "step": 50810
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0889005786428195e-06,
+      "loss": 1.3728,
+      "step": 50820
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0871471155532178e-06,
+      "loss": 1.4559,
+      "step": 50830
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 1.0853936524636157e-06,
+      "loss": 1.3763,
+      "step": 50840
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.0836401893740138e-06,
+      "loss": 1.4384,
+      "step": 50850
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.0818867262844117e-06,
+      "loss": 1.4924,
+      "step": 50860
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.0801332631948099e-06,
+      "loss": 1.5705,
+      "step": 50870
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.078379800105208e-06,
+      "loss": 1.3436,
+      "step": 50880
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 1.0766263370156059e-06,
+      "loss": 1.2841,
+      "step": 50890
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.074872873926004e-06,
+      "loss": 1.3793,
+      "step": 50900
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.0731194108364019e-06,
+      "loss": 1.4682,
+      "step": 50910
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.0713659477468e-06,
+      "loss": 1.4891,
+      "step": 50920
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.069612484657198e-06,
+      "loss": 1.452,
+      "step": 50930
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.067859021567596e-06,
+      "loss": 1.4422,
+      "step": 50940
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 1.0661055584779941e-06,
+      "loss": 1.423,
+      "step": 50950
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0643520953883922e-06,
+      "loss": 1.2304,
+      "step": 50960
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0625986322987901e-06,
+      "loss": 1.4646,
+      "step": 50970
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0608451692091883e-06,
+      "loss": 1.4887,
+      "step": 50980
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0590917061195864e-06,
+      "loss": 1.4294,
+      "step": 50990
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0573382430299843e-06,
+      "loss": 1.527,
+      "step": 51000
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 1.0555847799403824e-06,
+      "loss": 1.3207,
+      "step": 51010
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0538313168507803e-06,
+      "loss": 1.4077,
+      "step": 51020
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0520778537611784e-06,
+      "loss": 1.4475,
+      "step": 51030
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0503243906715763e-06,
+      "loss": 1.5971,
+      "step": 51040
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0485709275819744e-06,
+      "loss": 1.5006,
+      "step": 51050
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0468174644923725e-06,
+      "loss": 1.4779,
+      "step": 51060
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 1.0450640014027706e-06,
+      "loss": 1.467,
+      "step": 51070
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.0433105383131686e-06,
+      "loss": 1.2813,
+      "step": 51080
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.0415570752235667e-06,
+      "loss": 1.3833,
+      "step": 51090
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.0398036121339648e-06,
+      "loss": 1.4545,
+      "step": 51100
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.0380501490443627e-06,
+      "loss": 1.5915,
+      "step": 51110
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 1.0362966859547608e-06,
+      "loss": 1.2866,
+      "step": 51120
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.0345432228651587e-06,
+      "loss": 1.4286,
+      "step": 51130
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.0327897597755568e-06,
+      "loss": 1.5652,
+      "step": 51140
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.0310362966859547e-06,
+      "loss": 1.5224,
+      "step": 51150
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.0292828335963528e-06,
+      "loss": 1.3849,
+      "step": 51160
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.027529370506751e-06,
+      "loss": 1.4271,
+      "step": 51170
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 1.0257759074171488e-06,
+      "loss": 1.4105,
+      "step": 51180
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.024022444327547e-06,
+      "loss": 1.5004,
+      "step": 51190
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.022268981237945e-06,
+      "loss": 1.4256,
+      "step": 51200
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.0205155181483432e-06,
+      "loss": 1.4208,
+      "step": 51210
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.018762055058741e-06,
+      "loss": 1.4298,
+      "step": 51220
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.0170085919691392e-06,
+      "loss": 1.6347,
+      "step": 51230
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 1.015255128879537e-06,
+      "loss": 1.4456,
+      "step": 51240
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.0135016657899352e-06,
+      "loss": 1.2881,
+      "step": 51250
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.0117482027003331e-06,
+      "loss": 1.5208,
+      "step": 51260
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.0099947396107312e-06,
+      "loss": 1.5431,
+      "step": 51270
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.0082412765211293e-06,
+      "loss": 1.3295,
+      "step": 51280
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 1.0064878134315273e-06,
+      "loss": 1.5095,
+      "step": 51290
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.0047343503419254e-06,
+      "loss": 1.4234,
+      "step": 51300
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.0029808872523235e-06,
+      "loss": 1.2585,
+      "step": 51310
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 1.0012274241627216e-06,
+      "loss": 1.4683,
+      "step": 51320
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.994739610731195e-07,
+      "loss": 1.5351,
+      "step": 51330
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.977204979835176e-07,
+      "loss": 1.3625,
+      "step": 51340
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.959670348939155e-07,
+      "loss": 1.3363,
+      "step": 51350
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.942135718043136e-07,
+      "loss": 1.4349,
+      "step": 51360
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.924601087147115e-07,
+      "loss": 1.3973,
+      "step": 51370
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.907066456251096e-07,
+      "loss": 1.4275,
+      "step": 51380
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.889531825355078e-07,
+      "loss": 1.4451,
+      "step": 51390
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.871997194459057e-07,
+      "loss": 1.3896,
+      "step": 51400
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.854462563563038e-07,
+      "loss": 1.4232,
+      "step": 51410
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.836927932667017e-07,
+      "loss": 1.6046,
+      "step": 51420
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.819393301771e-07,
+      "loss": 1.3118,
+      "step": 51430
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.80185867087498e-07,
+      "loss": 1.3511,
+      "step": 51440
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.78432403997896e-07,
+      "loss": 1.5289,
+      "step": 51450
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.76678940908294e-07,
+      "loss": 1.4844,
+      "step": 51460
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.74925477818692e-07,
+      "loss": 1.3922,
+      "step": 51470
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.7317201472909e-07,
+      "loss": 1.4668,
+      "step": 51480
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.71418551639488e-07,
+      "loss": 1.4463,
+      "step": 51490
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.696650885498862e-07,
+      "loss": 1.4269,
+      "step": 51500
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.67911625460284e-07,
+      "loss": 1.4805,
+      "step": 51510
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.661581623706822e-07,
+      "loss": 1.4377,
+      "step": 51520
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.6440469928108e-07,
+      "loss": 1.7121,
+      "step": 51530
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.626512361914782e-07,
+      "loss": 1.4414,
+      "step": 51540
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.608977731018763e-07,
+      "loss": 1.4162,
+      "step": 51550
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.591443100122744e-07,
+      "loss": 1.457,
+      "step": 51560
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.573908469226723e-07,
+      "loss": 1.4121,
+      "step": 51570
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.556373838330704e-07,
+      "loss": 1.364,
+      "step": 51580
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.538839207434683e-07,
+      "loss": 1.4947,
+      "step": 51590
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.521304576538665e-07,
+      "loss": 1.4134,
+      "step": 51600
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.503769945642645e-07,
+      "loss": 1.404,
+      "step": 51610
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.486235314746625e-07,
+      "loss": 1.2704,
+      "step": 51620
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.468700683850606e-07,
+      "loss": 1.4253,
+      "step": 51630
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.451166052954586e-07,
+      "loss": 1.3464,
+      "step": 51640
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.433631422058566e-07,
+      "loss": 1.5018,
+      "step": 51650
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.416096791162546e-07,
+      "loss": 1.3615,
+      "step": 51660
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.398562160266528e-07,
+      "loss": 1.4183,
+      "step": 51670
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.381027529370508e-07,
+      "loss": 1.4754,
+      "step": 51680
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.363492898474488e-07,
+      "loss": 1.5855,
+      "step": 51690
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.345958267578469e-07,
+      "loss": 1.509,
+      "step": 51700
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.328423636682449e-07,
+      "loss": 1.6846,
+      "step": 51710
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.310889005786429e-07,
+      "loss": 1.5175,
+      "step": 51720
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.293354374890409e-07,
+      "loss": 1.4387,
+      "step": 51730
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.27581974399439e-07,
+      "loss": 1.5318,
+      "step": 51740
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.25828511309837e-07,
+      "loss": 1.4988,
+      "step": 51750
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.24075048220235e-07,
+      "loss": 1.3426,
+      "step": 51760
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.22321585130633e-07,
+      "loss": 1.614,
+      "step": 51770
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.20568122041031e-07,
+      "loss": 1.4657,
+      "step": 51780
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.188146589514292e-07,
+      "loss": 1.6028,
+      "step": 51790
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.170611958618273e-07,
+      "loss": 1.4973,
+      "step": 51800
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.153077327722253e-07,
+      "loss": 1.5275,
+      "step": 51810
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.135542696826233e-07,
+      "loss": 1.6025,
+      "step": 51820
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.118008065930213e-07,
+      "loss": 1.5883,
+      "step": 51830
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.100473435034193e-07,
+      "loss": 1.6238,
+      "step": 51840
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.082938804138174e-07,
+      "loss": 1.5432,
+      "step": 51850
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.065404173242154e-07,
+      "loss": 1.5668,
+      "step": 51860
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 9.047869542346134e-07,
+      "loss": 1.49,
+      "step": 51870
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 9.030334911450114e-07,
+      "loss": 1.3858,
+      "step": 51880
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 9.012800280554094e-07,
+      "loss": 1.4878,
+      "step": 51890
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 8.995265649658074e-07,
+      "loss": 1.3491,
+      "step": 51900
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 8.977731018762057e-07,
+      "loss": 1.4838,
+      "step": 51910
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 8.960196387866037e-07,
+      "loss": 1.4769,
+      "step": 51920
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.942661756970017e-07,
+      "loss": 1.4463,
+      "step": 51930
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.925127126073997e-07,
+      "loss": 1.5833,
+      "step": 51940
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.907592495177977e-07,
+      "loss": 1.216,
+      "step": 51950
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.890057864281958e-07,
+      "loss": 1.557,
+      "step": 51960
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.872523233385938e-07,
+      "loss": 1.5892,
+      "step": 51970
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 8.854988602489918e-07,
+      "loss": 1.5297,
+      "step": 51980
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.837453971593898e-07,
+      "loss": 1.4046,
+      "step": 51990
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.819919340697878e-07,
+      "loss": 1.3093,
+      "step": 52000
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.802384709801858e-07,
+      "loss": 1.4302,
+      "step": 52010
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.78485007890584e-07,
+      "loss": 1.3948,
+      "step": 52020
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 8.767315448009821e-07,
+      "loss": 1.3327,
+      "step": 52030
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.749780817113801e-07,
+      "loss": 1.5346,
+      "step": 52040
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.732246186217781e-07,
+      "loss": 1.368,
+      "step": 52050
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.714711555321761e-07,
+      "loss": 1.4097,
+      "step": 52060
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.697176924425742e-07,
+      "loss": 1.4316,
+      "step": 52070
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.679642293529722e-07,
+      "loss": 1.3589,
+      "step": 52080
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 8.662107662633702e-07,
+      "loss": 1.4207,
+      "step": 52090
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.644573031737682e-07,
+      "loss": 1.3166,
+      "step": 52100
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.627038400841662e-07,
+      "loss": 1.5257,
+      "step": 52110
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.609503769945643e-07,
+      "loss": 1.4217,
+      "step": 52120
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.591969139049624e-07,
+      "loss": 1.3419,
+      "step": 52130
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.574434508153604e-07,
+      "loss": 1.3927,
+      "step": 52140
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 8.556899877257585e-07,
+      "loss": 1.3944,
+      "step": 52150
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.539365246361565e-07,
+      "loss": 1.6174,
+      "step": 52160
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.521830615465545e-07,
+      "loss": 1.3163,
+      "step": 52170
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.504295984569526e-07,
+      "loss": 1.4634,
+      "step": 52180
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.486761353673506e-07,
+      "loss": 1.3932,
+      "step": 52190
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.469226722777486e-07,
+      "loss": 1.383,
+      "step": 52200
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 8.451692091881466e-07,
+      "loss": 1.4982,
+      "step": 52210
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.434157460985447e-07,
+      "loss": 1.4887,
+      "step": 52220
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.416622830089427e-07,
+      "loss": 1.3935,
+      "step": 52230
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.399088199193408e-07,
+      "loss": 1.4104,
+      "step": 52240
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.381553568297388e-07,
+      "loss": 1.4185,
+      "step": 52250
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 8.364018937401368e-07,
+      "loss": 1.5997,
+      "step": 52260
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.346484306505349e-07,
+      "loss": 1.6145,
+      "step": 52270
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.32894967560933e-07,
+      "loss": 1.3553,
+      "step": 52280
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.31141504471331e-07,
+      "loss": 1.4057,
+      "step": 52290
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.29388041381729e-07,
+      "loss": 1.4439,
+      "step": 52300
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.27634578292127e-07,
+      "loss": 1.3665,
+      "step": 52310
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 8.25881115202525e-07,
+      "loss": 1.5393,
+      "step": 52320
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.241276521129231e-07,
+      "loss": 1.4907,
+      "step": 52330
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.223741890233211e-07,
+      "loss": 1.3437,
+      "step": 52340
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.206207259337192e-07,
+      "loss": 1.4123,
+      "step": 52350
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.188672628441172e-07,
+      "loss": 1.2332,
+      "step": 52360
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.171137997545152e-07,
+      "loss": 1.3511,
+      "step": 52370
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 8.153603366649132e-07,
+      "loss": 1.4219,
+      "step": 52380
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.136068735753114e-07,
+      "loss": 1.3341,
+      "step": 52390
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.118534104857094e-07,
+      "loss": 1.316,
+      "step": 52400
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.100999473961074e-07,
+      "loss": 1.363,
+      "step": 52410
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.083464843065054e-07,
+      "loss": 1.3653,
+      "step": 52420
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 8.065930212169035e-07,
+      "loss": 1.313,
+      "step": 52430
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 8.048395581273015e-07,
+      "loss": 1.4161,
+      "step": 52440
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 8.030860950376995e-07,
+      "loss": 1.4284,
+      "step": 52450
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 8.013326319480976e-07,
+      "loss": 1.4287,
+      "step": 52460
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 7.995791688584956e-07,
+      "loss": 1.4831,
+      "step": 52470
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 7.978257057688936e-07,
+      "loss": 1.3943,
+      "step": 52480
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 7.960722426792916e-07,
+      "loss": 1.6008,
+      "step": 52490
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.943187795896896e-07,
+      "loss": 1.4713,
+      "step": 52500
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.925653165000878e-07,
+      "loss": 1.4291,
+      "step": 52510
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.908118534104858e-07,
+      "loss": 1.4036,
+      "step": 52520
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.890583903208839e-07,
+      "loss": 1.3585,
+      "step": 52530
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.873049272312819e-07,
+      "loss": 1.55,
+      "step": 52540
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 7.855514641416799e-07,
+      "loss": 1.38,
+      "step": 52550
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.837980010520779e-07,
+      "loss": 1.5177,
+      "step": 52560
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.82044537962476e-07,
+      "loss": 1.5216,
+      "step": 52570
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.80291074872874e-07,
+      "loss": 1.4768,
+      "step": 52580
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.78537611783272e-07,
+      "loss": 1.4442,
+      "step": 52590
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.7678414869367e-07,
+      "loss": 1.6114,
+      "step": 52600
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 7.75030685604068e-07,
+      "loss": 1.4317,
+      "step": 52610
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.73277222514466e-07,
+      "loss": 1.2718,
+      "step": 52620
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.715237594248643e-07,
+      "loss": 1.3666,
+      "step": 52630
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.697702963352623e-07,
+      "loss": 1.5324,
+      "step": 52640
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.680168332456603e-07,
+      "loss": 1.4541,
+      "step": 52650
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 7.662633701560583e-07,
+      "loss": 1.3261,
+      "step": 52660
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.645099070664563e-07,
+      "loss": 1.3187,
+      "step": 52670
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.627564439768544e-07,
+      "loss": 1.3955,
+      "step": 52680
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.610029808872524e-07,
+      "loss": 1.3645,
+      "step": 52690
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.592495177976504e-07,
+      "loss": 1.4405,
+      "step": 52700
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.574960547080484e-07,
+      "loss": 1.4019,
+      "step": 52710
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 7.557425916184464e-07,
+      "loss": 1.5667,
+      "step": 52720
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.539891285288444e-07,
+      "loss": 1.3979,
+      "step": 52730
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.522356654392426e-07,
+      "loss": 1.4966,
+      "step": 52740
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.504822023496407e-07,
+      "loss": 1.3459,
+      "step": 52750
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.487287392600387e-07,
+      "loss": 1.3546,
+      "step": 52760
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.469752761704367e-07,
+      "loss": 1.3103,
+      "step": 52770
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 7.452218130808347e-07,
+      "loss": 1.4859,
+      "step": 52780
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.434683499912328e-07,
+      "loss": 1.4991,
+      "step": 52790
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.417148869016308e-07,
+      "loss": 1.255,
+      "step": 52800
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.399614238120288e-07,
+      "loss": 1.5178,
+      "step": 52810
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.382079607224268e-07,
+      "loss": 1.3891,
+      "step": 52820
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 7.364544976328248e-07,
+      "loss": 1.3469,
+      "step": 52830
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.347010345432228e-07,
+      "loss": 1.4042,
+      "step": 52840
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.32947571453621e-07,
+      "loss": 1.5545,
+      "step": 52850
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.31194108364019e-07,
+      "loss": 1.4148,
+      "step": 52860
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.29440645274417e-07,
+      "loss": 1.2801,
+      "step": 52870
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.276871821848151e-07,
+      "loss": 1.2547,
+      "step": 52880
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 7.259337190952131e-07,
+      "loss": 1.4317,
+      "step": 52890
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.241802560056112e-07,
+      "loss": 1.5235,
+      "step": 52900
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.224267929160092e-07,
+      "loss": 1.3545,
+      "step": 52910
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.206733298264072e-07,
+      "loss": 1.4241,
+      "step": 52920
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.189198667368052e-07,
+      "loss": 1.374,
+      "step": 52930
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.171664036472032e-07,
+      "loss": 1.4908,
+      "step": 52940
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 7.154129405576013e-07,
+      "loss": 1.5048,
+      "step": 52950
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 7.136594774679994e-07,
+      "loss": 1.5311,
+      "step": 52960
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 7.119060143783974e-07,
+      "loss": 1.3443,
+      "step": 52970
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 7.101525512887954e-07,
+      "loss": 1.3245,
+      "step": 52980
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 7.083990881991934e-07,
+      "loss": 1.2594,
+      "step": 52990
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 7.066456251095915e-07,
+      "loss": 1.4541,
+      "step": 53000
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 7.048921620199896e-07,
+      "loss": 1.4903,
+      "step": 53010
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 7.031386989303876e-07,
+      "loss": 1.4171,
+      "step": 53020
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 7.013852358407856e-07,
+      "loss": 1.3035,
+      "step": 53030
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 6.996317727511836e-07,
+      "loss": 1.446,
+      "step": 53040
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 6.978783096615817e-07,
+      "loss": 1.3376,
+      "step": 53050
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 6.961248465719797e-07,
+      "loss": 1.2928,
+      "step": 53060
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.943713834823778e-07,
+      "loss": 1.4227,
+      "step": 53070
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.926179203927758e-07,
+      "loss": 1.4158,
+      "step": 53080
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.908644573031738e-07,
+      "loss": 1.5014,
+      "step": 53090
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.891109942135718e-07,
+      "loss": 1.4705,
+      "step": 53100
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.873575311239698e-07,
+      "loss": 1.3519,
+      "step": 53110
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 6.85604068034368e-07,
+      "loss": 1.3713,
+      "step": 53120
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.83850604944766e-07,
+      "loss": 1.5808,
+      "step": 53130
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.82097141855164e-07,
+      "loss": 1.4399,
+      "step": 53140
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.803436787655621e-07,
+      "loss": 1.4306,
+      "step": 53150
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.785902156759601e-07,
+      "loss": 1.4331,
+      "step": 53160
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.768367525863581e-07,
+      "loss": 1.3715,
+      "step": 53170
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 6.750832894967562e-07,
+      "loss": 1.3577,
+      "step": 53180
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.733298264071542e-07,
+      "loss": 1.3916,
+      "step": 53190
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.715763633175522e-07,
+      "loss": 1.3679,
+      "step": 53200
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.698229002279502e-07,
+      "loss": 1.4636,
+      "step": 53210
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.680694371383482e-07,
+      "loss": 1.4654,
+      "step": 53220
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 6.663159740487462e-07,
+      "loss": 1.4688,
+      "step": 53230
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.645625109591444e-07,
+      "loss": 1.3389,
+      "step": 53240
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.628090478695425e-07,
+      "loss": 1.3641,
+      "step": 53250
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.610555847799405e-07,
+      "loss": 1.4005,
+      "step": 53260
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.593021216903385e-07,
+      "loss": 1.3376,
+      "step": 53270
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.575486586007365e-07,
+      "loss": 1.3307,
+      "step": 53280
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 6.557951955111346e-07,
+      "loss": 1.1979,
+      "step": 53290
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.540417324215326e-07,
+      "loss": 1.3387,
+      "step": 53300
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.522882693319306e-07,
+      "loss": 1.5235,
+      "step": 53310
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.505348062423286e-07,
+      "loss": 1.2995,
+      "step": 53320
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.487813431527266e-07,
+      "loss": 1.2993,
+      "step": 53330
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.470278800631246e-07,
+      "loss": 1.2558,
+      "step": 53340
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 6.452744169735227e-07,
+      "loss": 1.5786,
+      "step": 53350
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.435209538839209e-07,
+      "loss": 1.4816,
+      "step": 53360
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.417674907943189e-07,
+      "loss": 1.46,
+      "step": 53370
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.400140277047169e-07,
+      "loss": 1.554,
+      "step": 53380
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.382605646151149e-07,
+      "loss": 1.4472,
+      "step": 53390
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 6.36507101525513e-07,
+      "loss": 1.3726,
+      "step": 53400
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.34753638435911e-07,
+      "loss": 1.4014,
+      "step": 53410
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.33000175346309e-07,
+      "loss": 1.3707,
+      "step": 53420
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.31246712256707e-07,
+      "loss": 1.387,
+      "step": 53430
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.29493249167105e-07,
+      "loss": 1.4745,
+      "step": 53440
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.27739786077503e-07,
+      "loss": 1.498,
+      "step": 53450
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 6.259863229879012e-07,
+      "loss": 1.4452,
+      "step": 53460
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.242328598982992e-07,
+      "loss": 1.483,
+      "step": 53470
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.224793968086972e-07,
+      "loss": 1.4239,
+      "step": 53480
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.207259337190953e-07,
+      "loss": 1.5784,
+      "step": 53490
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.189724706294933e-07,
+      "loss": 1.3494,
+      "step": 53500
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.172190075398914e-07,
+      "loss": 1.693,
+      "step": 53510
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 6.154655444502894e-07,
+      "loss": 1.4765,
+      "step": 53520
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.137120813606874e-07,
+      "loss": 1.5446,
+      "step": 53530
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.119586182710854e-07,
+      "loss": 1.5301,
+      "step": 53540
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.102051551814834e-07,
+      "loss": 1.3502,
+      "step": 53550
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.084516920918814e-07,
+      "loss": 1.4142,
+      "step": 53560
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 6.066982290022796e-07,
+      "loss": 1.4651,
+      "step": 53570
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 6.049447659126776e-07,
+      "loss": 1.3542,
+      "step": 53580
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 6.031913028230757e-07,
+      "loss": 1.4839,
+      "step": 53590
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 6.014378397334737e-07,
+      "loss": 1.3834,
+      "step": 53600
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 5.996843766438717e-07,
+      "loss": 1.4689,
+      "step": 53610
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 5.979309135542697e-07,
+      "loss": 1.4507,
+      "step": 53620
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 5.961774504646678e-07,
+      "loss": 1.416,
+      "step": 53630
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.944239873750658e-07,
+      "loss": 1.3861,
+      "step": 53640
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.926705242854638e-07,
+      "loss": 1.3382,
+      "step": 53650
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.909170611958618e-07,
+      "loss": 1.4603,
+      "step": 53660
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.891635981062599e-07,
+      "loss": 1.4198,
+      "step": 53670
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.87410135016658e-07,
+      "loss": 1.3432,
+      "step": 53680
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 5.85656671927056e-07,
+      "loss": 1.4672,
+      "step": 53690
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.839032088374541e-07,
+      "loss": 1.4517,
+      "step": 53700
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.821497457478521e-07,
+      "loss": 1.4297,
+      "step": 53710
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.803962826582501e-07,
+      "loss": 1.3479,
+      "step": 53720
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.786428195686481e-07,
+      "loss": 1.2685,
+      "step": 53730
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.768893564790461e-07,
+      "loss": 1.5918,
+      "step": 53740
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 5.751358933894442e-07,
+      "loss": 1.3508,
+      "step": 53750
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 5.733824302998422e-07,
+      "loss": 1.4595,
+      "step": 53760
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 5.716289672102403e-07,
+      "loss": 1.4366,
+      "step": 53770
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 5.698755041206383e-07,
+      "loss": 1.3503,
+      "step": 53780
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 5.681220410310364e-07,
+      "loss": 1.3207,
+      "step": 53790
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 5.663685779414344e-07,
+      "loss": 1.4948,
+      "step": 53800
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.646151148518325e-07,
+      "loss": 1.4573,
+      "step": 53810
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.628616517622305e-07,
+      "loss": 1.5376,
+      "step": 53820
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.611081886726285e-07,
+      "loss": 1.5347,
+      "step": 53830
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.593547255830265e-07,
+      "loss": 1.4814,
+      "step": 53840
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.576012624934245e-07,
+      "loss": 1.4667,
+      "step": 53850
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 5.558477994038225e-07,
+      "loss": 1.3564,
+      "step": 53860
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.540943363142206e-07,
+      "loss": 1.3524,
+      "step": 53870
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.523408732246187e-07,
+      "loss": 1.5038,
+      "step": 53880
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.505874101350167e-07,
+      "loss": 1.4927,
+      "step": 53890
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.488339470454148e-07,
+      "loss": 1.2869,
+      "step": 53900
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.470804839558128e-07,
+      "loss": 1.702,
+      "step": 53910
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 5.453270208662108e-07,
+      "loss": 1.6569,
+      "step": 53920
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.435735577766089e-07,
+      "loss": 1.5514,
+      "step": 53930
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.418200946870069e-07,
+      "loss": 1.501,
+      "step": 53940
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.400666315974049e-07,
+      "loss": 1.341,
+      "step": 53950
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.383131685078029e-07,
+      "loss": 1.3244,
+      "step": 53960
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 5.365597054182009e-07,
+      "loss": 1.4052,
+      "step": 53970
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.34806242328599e-07,
+      "loss": 1.4191,
+      "step": 53980
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.330527792389971e-07,
+      "loss": 1.4312,
+      "step": 53990
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.312993161493951e-07,
+      "loss": 1.3965,
+      "step": 54000
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.295458530597932e-07,
+      "loss": 1.3759,
+      "step": 54010
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.277923899701912e-07,
+      "loss": 1.4837,
+      "step": 54020
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 5.260389268805892e-07,
+      "loss": 1.4376,
+      "step": 54030
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.242854637909872e-07,
+      "loss": 1.4653,
+      "step": 54040
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.225320007013853e-07,
+      "loss": 1.5097,
+      "step": 54050
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.207785376117833e-07,
+      "loss": 1.433,
+      "step": 54060
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.190250745221813e-07,
+      "loss": 1.5378,
+      "step": 54070
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.172716114325793e-07,
+      "loss": 1.5005,
+      "step": 54080
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 5.155181483429774e-07,
+      "loss": 1.4121,
+      "step": 54090
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 5.137646852533755e-07,
+      "loss": 1.2627,
+      "step": 54100
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 5.120112221637735e-07,
+      "loss": 1.4281,
+      "step": 54110
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 5.102577590741716e-07,
+      "loss": 1.7764,
+      "step": 54120
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 5.085042959845696e-07,
+      "loss": 1.5093,
+      "step": 54130
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 5.067508328949676e-07,
+      "loss": 1.5831,
+      "step": 54140
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 5.049973698053656e-07,
+      "loss": 1.3238,
+      "step": 54150
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 5.032439067157636e-07,
+      "loss": 1.4387,
+      "step": 54160
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 5.014904436261617e-07,
+      "loss": 1.4088,
+      "step": 54170
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.997369805365597e-07,
+      "loss": 1.4466,
+      "step": 54180
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.979835174469578e-07,
+      "loss": 1.4183,
+      "step": 54190
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 4.962300543573558e-07,
+      "loss": 1.4837,
+      "step": 54200
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.944765912677539e-07,
+      "loss": 1.4153,
+      "step": 54210
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.927231281781519e-07,
+      "loss": 1.4171,
+      "step": 54220
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.9096966508855e-07,
+      "loss": 1.5172,
+      "step": 54230
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.89216201998948e-07,
+      "loss": 1.5696,
+      "step": 54240
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.87462738909346e-07,
+      "loss": 1.3727,
+      "step": 54250
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 4.85709275819744e-07,
+      "loss": 1.5253,
+      "step": 54260
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.83955812730142e-07,
+      "loss": 1.4433,
+      "step": 54270
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.8220234964054e-07,
+      "loss": 1.5398,
+      "step": 54280
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.804488865509382e-07,
+      "loss": 1.448,
+      "step": 54290
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.786954234613362e-07,
+      "loss": 1.444,
+      "step": 54300
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.769419603717342e-07,
+      "loss": 1.6705,
+      "step": 54310
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 4.7518849728213223e-07,
+      "loss": 1.3021,
+      "step": 54320
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.734350341925303e-07,
+      "loss": 1.5438,
+      "step": 54330
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.716815711029283e-07,
+      "loss": 1.6623,
+      "step": 54340
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.699281080133264e-07,
+      "loss": 1.3054,
+      "step": 54350
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.681746449237244e-07,
+      "loss": 1.4415,
+      "step": 54360
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 4.6642118183412243e-07,
+      "loss": 1.3884,
+      "step": 54370
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.6466771874452044e-07,
+      "loss": 1.4192,
+      "step": 54380
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.629142556549185e-07,
+      "loss": 1.4696,
+      "step": 54390
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.611607925653165e-07,
+      "loss": 1.3856,
+      "step": 54400
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.594073294757146e-07,
+      "loss": 1.278,
+      "step": 54410
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.5765386638611263e-07,
+      "loss": 1.3782,
+      "step": 54420
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 4.5590040329651064e-07,
+      "loss": 1.3624,
+      "step": 54430
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.541469402069087e-07,
+      "loss": 1.4204,
+      "step": 54440
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.523934771173067e-07,
+      "loss": 1.3875,
+      "step": 54450
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.506400140277047e-07,
+      "loss": 1.5409,
+      "step": 54460
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.4888655093810283e-07,
+      "loss": 1.4289,
+      "step": 54470
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.4713308784850084e-07,
+      "loss": 1.4138,
+      "step": 54480
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 4.4537962475889885e-07,
+      "loss": 1.3344,
+      "step": 54490
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.436261616692969e-07,
+      "loss": 1.365,
+      "step": 54500
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.418726985796949e-07,
+      "loss": 1.4379,
+      "step": 54510
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.401192354900929e-07,
+      "loss": 1.5203,
+      "step": 54520
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.3836577240049104e-07,
+      "loss": 1.5146,
+      "step": 54530
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 4.3661230931088905e-07,
+      "loss": 1.4912,
+      "step": 54540
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.348588462212871e-07,
+      "loss": 1.4799,
+      "step": 54550
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.331053831316851e-07,
+      "loss": 1.4748,
+      "step": 54560
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.313519200420831e-07,
+      "loss": 1.5221,
+      "step": 54570
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.295984569524812e-07,
+      "loss": 1.4259,
+      "step": 54580
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.2784499386287924e-07,
+      "loss": 1.5688,
+      "step": 54590
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 4.2609153077327725e-07,
+      "loss": 1.5052,
+      "step": 54600
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.243380676836753e-07,
+      "loss": 1.378,
+      "step": 54610
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.225846045940733e-07,
+      "loss": 1.3395,
+      "step": 54620
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.2083114150447133e-07,
+      "loss": 1.4714,
+      "step": 54630
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.190776784148694e-07,
+      "loss": 1.4107,
+      "step": 54640
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.1732421532526745e-07,
+      "loss": 1.329,
+      "step": 54650
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 4.155707522356655e-07,
+      "loss": 1.4105,
+      "step": 54660
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.138172891460635e-07,
+      "loss": 1.4331,
+      "step": 54670
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.1206382605646153e-07,
+      "loss": 1.4568,
+      "step": 54680
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.103103629668596e-07,
+      "loss": 1.518,
+      "step": 54690
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.085568998772576e-07,
+      "loss": 1.3614,
+      "step": 54700
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.068034367876557e-07,
+      "loss": 1.3892,
+      "step": 54710
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 4.050499736980537e-07,
+      "loss": 1.582,
+      "step": 54720
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.0329651060845173e-07,
+      "loss": 1.5618,
+      "step": 54730
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 4.0154304751884974e-07,
+      "loss": 1.379,
+      "step": 54740
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 3.997895844292478e-07,
+      "loss": 1.4603,
+      "step": 54750
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 3.980361213396458e-07,
+      "loss": 1.7065,
+      "step": 54760
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 3.962826582500439e-07,
+      "loss": 1.5127,
+      "step": 54770
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.9452919516044193e-07,
+      "loss": 1.5186,
+      "step": 54780
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.9277573207083994e-07,
+      "loss": 1.419,
+      "step": 54790
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.91022268981238e-07,
+      "loss": 1.4637,
+      "step": 54800
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.89268805891636e-07,
+      "loss": 1.6021,
+      "step": 54810
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.87515342802034e-07,
+      "loss": 1.5056,
+      "step": 54820
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.8576187971243213e-07,
+      "loss": 1.3784,
+      "step": 54830
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.8400841662283014e-07,
+      "loss": 1.3956,
+      "step": 54840
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.8225495353322814e-07,
+      "loss": 1.4806,
+      "step": 54850
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.805014904436262e-07,
+      "loss": 1.4172,
+      "step": 54860
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.787480273540242e-07,
+      "loss": 1.4474,
+      "step": 54870
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.769945642644222e-07,
+      "loss": 1.5446,
+      "step": 54880
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.7524110117482034e-07,
+      "loss": 1.4683,
+      "step": 54890
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.7348763808521834e-07,
+      "loss": 1.3677,
+      "step": 54900
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.717341749956164e-07,
+      "loss": 1.5423,
+      "step": 54910
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.699807119060144e-07,
+      "loss": 1.3528,
+      "step": 54920
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.682272488164124e-07,
+      "loss": 1.4668,
+      "step": 54930
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.664737857268105e-07,
+      "loss": 1.4509,
+      "step": 54940
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.647203226372085e-07,
+      "loss": 1.6704,
+      "step": 54950
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.6296685954760655e-07,
+      "loss": 1.4682,
+      "step": 54960
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.612133964580046e-07,
+      "loss": 1.4205,
+      "step": 54970
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.594599333684026e-07,
+      "loss": 1.5351,
+      "step": 54980
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.5770647027880063e-07,
+      "loss": 1.4123,
+      "step": 54990
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.559530071891987e-07,
+      "loss": 1.3569,
+      "step": 55000
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.541995440995967e-07,
+      "loss": 1.3322,
+      "step": 55010
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.524460810099948e-07,
+      "loss": 1.4881,
+      "step": 55020
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.506926179203928e-07,
+      "loss": 1.3306,
+      "step": 55030
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.4893915483079083e-07,
+      "loss": 1.4244,
+      "step": 55040
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.471856917411889e-07,
+      "loss": 1.4936,
+      "step": 55050
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.454322286515869e-07,
+      "loss": 1.4342,
+      "step": 55060
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.436787655619849e-07,
+      "loss": 1.4996,
+      "step": 55070
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.41925302472383e-07,
+      "loss": 1.3189,
+      "step": 55080
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.4017183938278103e-07,
+      "loss": 1.2583,
+      "step": 55090
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.3841837629317903e-07,
+      "loss": 1.2622,
+      "step": 55100
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.366649132035771e-07,
+      "loss": 1.3673,
+      "step": 55110
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.349114501139751e-07,
+      "loss": 1.5621,
+      "step": 55120
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.331579870243731e-07,
+      "loss": 1.4751,
+      "step": 55130
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.314045239347712e-07,
+      "loss": 1.6227,
+      "step": 55140
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.2965106084516923e-07,
+      "loss": 1.3006,
+      "step": 55150
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.278975977555673e-07,
+      "loss": 1.2731,
+      "step": 55160
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.261441346659653e-07,
+      "loss": 1.5199,
+      "step": 55170
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.243906715763633e-07,
+      "loss": 1.4381,
+      "step": 55180
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.2263720848676137e-07,
+      "loss": 1.5883,
+      "step": 55190
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.2088374539715943e-07,
+      "loss": 1.3823,
+      "step": 55200
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.1913028230755744e-07,
+      "loss": 1.4517,
+      "step": 55210
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.173768192179555e-07,
+      "loss": 1.5606,
+      "step": 55220
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.156233561283535e-07,
+      "loss": 1.4116,
+      "step": 55230
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.138698930387515e-07,
+      "loss": 1.5937,
+      "step": 55240
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.121164299491496e-07,
+      "loss": 1.4964,
+      "step": 55250
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.1036296685954764e-07,
+      "loss": 1.3554,
+      "step": 55260
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.086095037699457e-07,
+      "loss": 1.5209,
+      "step": 55270
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.068560406803437e-07,
+      "loss": 1.2984,
+      "step": 55280
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.051025775907417e-07,
+      "loss": 1.4042,
+      "step": 55290
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 3.033491145011398e-07,
+      "loss": 1.4854,
+      "step": 55300
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 3.0159565141153784e-07,
+      "loss": 1.4509,
+      "step": 55310
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.9984218832193585e-07,
+      "loss": 1.5483,
+      "step": 55320
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.980887252323339e-07,
+      "loss": 1.3876,
+      "step": 55330
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.963352621427319e-07,
+      "loss": 1.3664,
+      "step": 55340
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.945817990531299e-07,
+      "loss": 1.461,
+      "step": 55350
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.92828335963528e-07,
+      "loss": 1.3391,
+      "step": 55360
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.9107487287392605e-07,
+      "loss": 1.2771,
+      "step": 55370
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.8932140978432406e-07,
+      "loss": 1.431,
+      "step": 55380
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.875679466947221e-07,
+      "loss": 1.337,
+      "step": 55390
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.858144836051201e-07,
+      "loss": 1.5431,
+      "step": 55400
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.840610205155182e-07,
+      "loss": 1.4779,
+      "step": 55410
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.8230755742591625e-07,
+      "loss": 1.3937,
+      "step": 55420
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.8055409433631426e-07,
+      "loss": 1.4442,
+      "step": 55430
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7880063124671226e-07,
+      "loss": 1.263,
+      "step": 55440
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.770471681571103e-07,
+      "loss": 1.3503,
+      "step": 55450
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.7529370506750833e-07,
+      "loss": 1.7572,
+      "step": 55460
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.735402419779064e-07,
+      "loss": 1.5193,
+      "step": 55470
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.7178677888830445e-07,
+      "loss": 1.3491,
+      "step": 55480
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.7003331579870246e-07,
+      "loss": 1.3224,
+      "step": 55490
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.6827985270910047e-07,
+      "loss": 1.3096,
+      "step": 55500
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.6652638961949853e-07,
+      "loss": 1.456,
+      "step": 55510
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.647729265298966e-07,
+      "loss": 1.2873,
+      "step": 55520
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.630194634402946e-07,
+      "loss": 1.3914,
+      "step": 55530
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.6126600035069266e-07,
+      "loss": 1.3875,
+      "step": 55540
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.5951253726109067e-07,
+      "loss": 1.4358,
+      "step": 55550
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.577590741714887e-07,
+      "loss": 1.3867,
+      "step": 55560
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.5600561108188674e-07,
+      "loss": 1.2881,
+      "step": 55570
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.542521479922848e-07,
+      "loss": 1.4048,
+      "step": 55580
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.524986849026828e-07,
+      "loss": 1.3976,
+      "step": 55590
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.5074522181308087e-07,
+      "loss": 1.5491,
+      "step": 55600
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.489917587234789e-07,
+      "loss": 1.4848,
+      "step": 55610
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.4723829563387694e-07,
+      "loss": 1.4699,
+      "step": 55620
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.45484832544275e-07,
+      "loss": 1.2848,
+      "step": 55630
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.43731369454673e-07,
+      "loss": 1.4958,
+      "step": 55640
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.41977906365071e-07,
+      "loss": 1.4811,
+      "step": 55650
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.402244432754691e-07,
+      "loss": 1.5206,
+      "step": 55660
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.384709801858671e-07,
+      "loss": 1.2531,
+      "step": 55670
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.3671751709626515e-07,
+      "loss": 1.4288,
+      "step": 55680
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.349640540066632e-07,
+      "loss": 1.3809,
+      "step": 55690
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.3321059091706122e-07,
+      "loss": 1.4756,
+      "step": 55700
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.3145712782745925e-07,
+      "loss": 1.3946,
+      "step": 55710
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.297036647378573e-07,
+      "loss": 1.575,
+      "step": 55720
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.2795020164825532e-07,
+      "loss": 1.4005,
+      "step": 55730
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.2619673855865335e-07,
+      "loss": 1.3294,
+      "step": 55740
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.2444327546905141e-07,
+      "loss": 1.3933,
+      "step": 55750
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.2268981237944942e-07,
+      "loss": 1.3974,
+      "step": 55760
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.2093634928984746e-07,
+      "loss": 1.4661,
+      "step": 55770
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.1918288620024552e-07,
+      "loss": 1.3155,
+      "step": 55780
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.1742942311064355e-07,
+      "loss": 1.3114,
+      "step": 55790
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.1567596002104156e-07,
+      "loss": 1.6061,
+      "step": 55800
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.1392249693143962e-07,
+      "loss": 1.4328,
+      "step": 55810
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.1216903384183766e-07,
+      "loss": 1.3963,
+      "step": 55820
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.1041557075223567e-07,
+      "loss": 1.4066,
+      "step": 55830
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.0866210766263373e-07,
+      "loss": 1.3456,
+      "step": 55840
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.0690864457303176e-07,
+      "loss": 1.2592,
+      "step": 55850
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.051551814834298e-07,
+      "loss": 1.4435,
+      "step": 55860
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.0340171839382786e-07,
+      "loss": 1.5905,
+      "step": 55870
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.0164825530422586e-07,
+      "loss": 1.3251,
+      "step": 55880
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 1.998947922146239e-07,
+      "loss": 1.4571,
+      "step": 55890
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 1.9814132912502196e-07,
+      "loss": 1.4442,
+      "step": 55900
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 1.9638786603541997e-07,
+      "loss": 1.483,
+      "step": 55910
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.94634402945818e-07,
+      "loss": 1.5173,
+      "step": 55920
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.9288093985621606e-07,
+      "loss": 1.3737,
+      "step": 55930
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.9112747676661407e-07,
+      "loss": 1.3043,
+      "step": 55940
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.893740136770121e-07,
+      "loss": 1.5362,
+      "step": 55950
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.8762055058741017e-07,
+      "loss": 1.2547,
+      "step": 55960
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 1.858670874978082e-07,
+      "loss": 1.4626,
+      "step": 55970
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.841136244082062e-07,
+      "loss": 1.4392,
+      "step": 55980
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.8236016131860424e-07,
+      "loss": 1.3048,
+      "step": 55990
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.806066982290023e-07,
+      "loss": 1.3775,
+      "step": 56000
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.7885323513940031e-07,
+      "loss": 1.4866,
+      "step": 56010
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.7709977204979835e-07,
+      "loss": 1.5872,
+      "step": 56020
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 1.753463089601964e-07,
+      "loss": 1.4374,
+      "step": 56030
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.7359284587059444e-07,
+      "loss": 1.5489,
+      "step": 56040
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.7183938278099245e-07,
+      "loss": 1.3843,
+      "step": 56050
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.7008591969139051e-07,
+      "loss": 1.4125,
+      "step": 56060
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.6833245660178855e-07,
+      "loss": 1.6306,
+      "step": 56070
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 1.6657899351218656e-07,
+      "loss": 1.3708,
+      "step": 56080
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.6482553042258462e-07,
+      "loss": 1.585,
+      "step": 56090
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.6307206733298265e-07,
+      "loss": 1.4543,
+      "step": 56100
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.6131860424338069e-07,
+      "loss": 1.398,
+      "step": 56110
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.5956514115377872e-07,
+      "loss": 1.4019,
+      "step": 56120
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.5781167806417676e-07,
+      "loss": 1.3009,
+      "step": 56130
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 1.560582149745748e-07,
+      "loss": 1.4872,
+      "step": 56140
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.5430475188497285e-07,
+      "loss": 1.5406,
+      "step": 56150
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.5255128879537086e-07,
+      "loss": 1.6007,
+      "step": 56160
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.5079782570576892e-07,
+      "loss": 1.3706,
+      "step": 56170
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.4904436261616695e-07,
+      "loss": 1.4472,
+      "step": 56180
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.4729089952656496e-07,
+      "loss": 1.4418,
+      "step": 56190
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 1.4553743643696302e-07,
+      "loss": 1.339,
+      "step": 56200
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.4378397334736106e-07,
+      "loss": 1.5329,
+      "step": 56210
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.420305102577591e-07,
+      "loss": 1.4495,
+      "step": 56220
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.4027704716815713e-07,
+      "loss": 1.4646,
+      "step": 56230
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.3852358407855516e-07,
+      "loss": 1.4713,
+      "step": 56240
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.367701209889532e-07,
+      "loss": 1.5927,
+      "step": 56250
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 1.3501665789935123e-07,
+      "loss": 1.5331,
+      "step": 56260
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.3326319480974927e-07,
+      "loss": 1.3473,
+      "step": 56270
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.315097317201473e-07,
+      "loss": 1.4492,
+      "step": 56280
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.2975626863054534e-07,
+      "loss": 1.4146,
+      "step": 56290
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.2800280554094337e-07,
+      "loss": 1.4289,
+      "step": 56300
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 1.262493424513414e-07,
+      "loss": 1.5269,
+      "step": 56310
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.2449587936173944e-07,
+      "loss": 1.3585,
+      "step": 56320
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.227424162721375e-07,
+      "loss": 1.4577,
+      "step": 56330
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.209889531825355e-07,
+      "loss": 1.3125,
+      "step": 56340
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.1923549009293354e-07,
+      "loss": 1.3575,
+      "step": 56350
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.174820270033316e-07,
+      "loss": 1.4226,
+      "step": 56360
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 1.1572856391372963e-07,
+      "loss": 1.3561,
+      "step": 56370
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.1397510082412766e-07,
+      "loss": 1.3279,
+      "step": 56380
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.1222163773452571e-07,
+      "loss": 1.3282,
+      "step": 56390
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.1046817464492373e-07,
+      "loss": 1.4353,
+      "step": 56400
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.0871471155532178e-07,
+      "loss": 1.2003,
+      "step": 56410
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.0696124846571981e-07,
+      "loss": 1.3686,
+      "step": 56420
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 1.0520778537611783e-07,
+      "loss": 1.3498,
+      "step": 56430
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 1.0345432228651588e-07,
+      "loss": 1.4436,
+      "step": 56440
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 1.0170085919691393e-07,
+      "loss": 1.2733,
+      "step": 56450
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 9.994739610731195e-08,
+      "loss": 1.4338,
+      "step": 56460
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 9.819393301770998e-08,
+      "loss": 1.7698,
+      "step": 56470
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 9.644046992810803e-08,
+      "loss": 1.3726,
+      "step": 56480
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 9.468700683850605e-08,
+      "loss": 1.4454,
+      "step": 56490
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 9.29335437489041e-08,
+      "loss": 1.5059,
+      "step": 56500
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 9.118008065930212e-08,
+      "loss": 1.531,
+      "step": 56510
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 8.942661756970016e-08,
+      "loss": 1.5258,
+      "step": 56520
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 8.76731544800982e-08,
+      "loss": 1.5797,
+      "step": 56530
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 8.591969139049623e-08,
+      "loss": 1.3917,
+      "step": 56540
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 8.416622830089427e-08,
+      "loss": 1.6648,
+      "step": 56550
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 8.241276521129231e-08,
+      "loss": 1.4092,
+      "step": 56560
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 8.065930212169034e-08,
+      "loss": 1.3776,
+      "step": 56570
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 7.890583903208838e-08,
+      "loss": 1.3508,
+      "step": 56580
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 7.715237594248643e-08,
+      "loss": 1.5816,
+      "step": 56590
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 7.539891285288446e-08,
+      "loss": 1.4211,
+      "step": 56600
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 7.364544976328248e-08,
+      "loss": 1.3621,
+      "step": 56610
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 7.189198667368053e-08,
+      "loss": 1.4777,
+      "step": 56620
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 7.013852358407856e-08,
+      "loss": 1.3707,
+      "step": 56630
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 6.83850604944766e-08,
+      "loss": 1.431,
+      "step": 56640
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 6.663159740487463e-08,
+      "loss": 1.3034,
+      "step": 56650
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.487813431527267e-08,
+      "loss": 1.3131,
+      "step": 56660
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.31246712256707e-08,
+      "loss": 1.354,
+      "step": 56670
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 6.137120813606875e-08,
+      "loss": 1.3454,
+      "step": 56680
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 5.961774504646677e-08,
+      "loss": 1.5125,
+      "step": 56690
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 5.786428195686481e-08,
+      "loss": 1.6611,
+      "step": 56700
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 5.6110818867262854e-08,
+      "loss": 1.4072,
+      "step": 56710
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 5.435735577766089e-08,
+      "loss": 1.4185,
+      "step": 56720
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 5.2603892688058916e-08,
+      "loss": 1.4345,
+      "step": 56730
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 5.0850429598456964e-08,
+      "loss": 1.4444,
+      "step": 56740
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 4.909696650885499e-08,
+      "loss": 1.3128,
+      "step": 56750
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 4.7343503419253027e-08,
+      "loss": 1.4268,
+      "step": 56760
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 4.559004032965106e-08,
+      "loss": 1.5134,
+      "step": 56770
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.38365772400491e-08,
+      "loss": 1.2621,
+      "step": 56780
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.208311415044714e-08,
+      "loss": 1.4789,
+      "step": 56790
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 4.032965106084517e-08,
+      "loss": 1.3811,
+      "step": 56800
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 3.857618797124321e-08,
+      "loss": 1.3546,
+      "step": 56810
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 3.682272488164124e-08,
+      "loss": 1.4494,
+      "step": 56820
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 3.506926179203928e-08,
+      "loss": 1.318,
+      "step": 56830
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 3.3315798702437316e-08,
+      "loss": 1.4429,
+      "step": 56840
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 3.156233561283535e-08,
+      "loss": 1.5105,
+      "step": 56850
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.9808872523233386e-08,
+      "loss": 1.5096,
+      "step": 56860
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.8055409433631427e-08,
+      "loss": 1.3179,
+      "step": 56870
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.6301946344029458e-08,
+      "loss": 1.3616,
+      "step": 56880
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.4548483254427496e-08,
+      "loss": 1.3592,
+      "step": 56890
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.279502016482553e-08,
+      "loss": 1.3939,
+      "step": 56900
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.104155707522357e-08,
+      "loss": 1.4041,
+      "step": 56910
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.9288093985621606e-08,
+      "loss": 1.3942,
+      "step": 56920
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.753463089601964e-08,
+      "loss": 1.4087,
+      "step": 56930
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 1.5781167806417676e-08,
+      "loss": 1.4626,
+      "step": 56940
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.4027704716815713e-08,
+      "loss": 1.2961,
+      "step": 56950
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.2274241627213748e-08,
+      "loss": 1.4794,
+      "step": 56960
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 1.0520778537611784e-08,
+      "loss": 1.5063,
+      "step": 56970
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 8.76731544800982e-09,
+      "loss": 1.3976,
+      "step": 56980
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 7.013852358407857e-09,
+      "loss": 1.3433,
+      "step": 56990
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 5.260389268805892e-09,
+      "loss": 1.3999,
+      "step": 57000
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 57030,
+  "num_train_epochs": 10,
+  "save_steps": 200,
+  "total_flos": 2.1789204128989043e+20,
+  "trial_name": null,
+  "trial_params": null
+}